(De Wekker&snyder) 2013 Mountain Weather Research and Forecasting PDF

Mountain Weather Research and Forecasting
Springer Atmospheric Sciences
For further volumes:

http://www.springer.com/series/10176
Fotini Katopodes Chow • Stephan F.J. De Wekker
Bradley J. Snyder
Editors
Mountain Weather Research

and Forecasting
Recent Progress and Current Challenges
123
Editors
Fotini Katopodes Chow Stephan F.J. De Wekker
Department of Civil and Environmental Department of Environmental Sciences
Engineering, MC 1710 University of Virginia
University of California, Berkeley McCormick Rd. 291
Berkeley, CA 94720-1710 Charlottesville, Virginia
USA USA
Bradley J. Snyder
Meteorological Service of Canada
#201 401 Burrard Street
Vancouver, BC, Canada
ISBN 978-94-007-4097-6 ISBN 978-94-007-4098-3 (eBook)

DOI 10.1007/978-94-007-4098-3
Springer Dordrecht Heidelberg New York London
Library of Congress Control Number: 2012940730
© Springer Science+Business Media B.V. 2013

This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of
the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation,
broadcasting, reproduction on microfilms or in any other physical way, and transmission or information
storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology
now known or hereafter developed. Exempted from this legal reservation are brief excerpts in connection
with reviews or scholarly analysis or material supplied specifically for the purpose of being entered
and executed on a computer system, for exclusive use by the purchaser of the work. Duplication of
this publication or parts thereof is permitted only under the provisions of the Copyright Law of the
Publisher’s location, in its current version, and permission for use must always be obtained from Springer.
Permissions for use may be obtained through RightsLink at the Copyright Clearance Center. Violations
are liable to prosecution under the respective Copyright Law.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication
does not imply, even in the absence of a specific statement, that such names are exempt from the relevant
protective laws and regulations and therefore free for general use.
While the advice and information in this book are believed to be true and accurate at the date of
publication, neither the authors nor the editors nor the publisher can accept any legal responsibility for
any errors or omissions that may be made. The publisher makes no warranty, express or implied, with
respect to the material contained herein.
Printed on acid-free paper
Springer is part of Springer Science+Business Media (www.springer.com)

Preface
This book is the result of a multiyear effort that began with the organization of a
workshop designed to bring researchers and forecasters together to discuss current
progress and challenges in mountain weather. The chapters herein represent the
topics from this Mountain Weather Workshop, which took place in Whistler, British
Columbia, Canada, 5–8 August 2008. The inspiration for the workshop and book
arose under the guidance of the American Meteorological Society (AMS) Mountain
Meteorology Committee.
One of the main goals of the workshop was to bridge the gap between the
research and forecasting communities by providing a forum for extended discussion
and joint education. The workshop consisted of lectures given by 13 distinguished
speakers, several discussion opportunities in small groups, and a day of laboratory
exercises designed for forecaster training for the 2010 Winter Olympics in Vancou-
ver. The lectures provided a detailed overview of important and emerging topics
in mountain meteorology. About 100 participants attended, roughly evenly split
among forecasters, researchers, and graduate students (see Fig. 12.2 for a picture
of participants). One of the highlights of the week was a group activity to design
the best observation and modeling system to nowcast for the Olympic ski jump
event; this was an excellent opportunity for researchers and operational forecasters
to work together and “bridge the gap.” The lectures from the workshop can be
accessed online in the COMET MetEd tutorial collection (http://www.meted.ucar.
edu/training module.php?id=878).
The chapters in this book have been written with the intent to provide a thorough
overview of each topic with an emphasis on recent research and progress in the field,
especially since the last collection of topics in mountain meteorology was published
more than two decades ago (Blumen 1990). It is our hope that this new offering
will be used extensively in mountain weather courses at universities and forecast
offices and also used as a general reference book for researchers, forecasters, and
students. Readers will be provided with a broad understanding of the fundamental
principles driving flow over complex terrain, including historical context for recent
developments and future directions for researchers and forecasters. For academic
v
vi Preface
Fig. P1.1 View of the PEAK 2 PEAK gondola connecting Whistler and Blackcomb mountains,
looking across toward Blackcomb. Whistler village is on the far left (© James Dunning. Reprinted
with permission)
researchers, the book will provide some insight into issues important to the
forecasting community. For the forecasting community, we hope the book will
provide training on fundamentals of flows specific to mountainous regions which
are notoriously difficult to predict, understanding of current research challenges,
and an opportunity to learn about the latest contributions and advancements to the
field. Our goal of bridging the gap between research and forecasting with this book
is aptly captured in the image below showing Whistler and Blackcomb mountains,
connected by the new PEAK 2 PEAK gondola, built for the 2010 Winter Olympics,
bridging the gap between the two mountains (Fig. P1.1).
The first chapter provides an overview of mountain weather and forecasting
challenges specific to complex terrain. This is followed by chapters that focus on
diurnal mountain/valley flows that develop under calm conditions (Chap. 2) and
dynamically driven winds under strong forcing (Chap. 3). The focus then shifts to
other specific phenomena that are difficult to understand and predict in mountain
regions: Alpine foehn (Chap. 4) and boundary layer phenomena and air quality
(Chap. 5). The following two chapters address processes that bring wet mountain
weather, in the form of rain, snow, or other hydrometeors, with a discussion of
specific orographic precipitation processes (Chap. 6) and the details of microphysics
parameterizations (Chap. 7). Having covered the major physical processes, the book
shifts to observation and modeling techniques used in mountain regions. First, a
detailed discussion of field measurements in complex terrain is given (Chap. 8).
Preface vii
Then, the following three chapters describe the basics of mesoscale numerical
modeling (Chap. 9), model configuration and physical parameterizations such as
turbulence (Chap. 10), and model applications in operational forecasting (Chap. 11).
The book concludes with a chapter that discusses the current state of research and
forecasting in complex terrain, including a vision of how to bridge the gap in the
future (Chap. 12).
We are quite fortunate to have a set of conscientious and thorough authors who
have contributed their knowledge and expertise to create this book, largely in their
spare time. We are also extremely grateful to the many reviewers who were involved
in ensuring the quality of this book. Given the length of some of the chapters, we
were particularly impressed by the care they took to thoroughly review the chapter
content, from comments on overall structure to details on style and formatting.
Funding to support the publication of this book and for student travel to the
workshop was provided by the National Science Foundation (NSF) (award ATM-
0810090). Funding for the workshop was provided by the American Meteorological
Society (AMS), the University Corporation for Atmospheric Research (UCAR)
acting on behalf of the Cooperative Program for Operational Meteorology, Edu-
cation and Training (COMET), and the Meteorological Service of Canada (MSC).
The workshop was mainly organized by us (editors of this book) with the help of
many others on the AMS Mountain Meteorology Committee, in addition to Cara
Campbell at AMS. We thank our colleagues who were AMS Mountain Meteorology
Committee members with us over the years (Brian Colle, Lisa Darby, Mike Meyers,
Stephen Mobbs, Greg Poulos, Heather Reeves, Alex Reinecke, Simon Vosper, Doug
Wesley, and David Whiteman) and members of the AMS publications department
(Peter Lamb, Sarah Jane Shangraw, and Ken Heideman) for their support of this
effort.
Contents
1 Mountain Weather Prediction: Phenomenological

Challenges and Forecast Methodology .. . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 1
Michael P. Meyers and W. James Steenburgh
2 Diurnal Mountain Wind Systems . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 35
Dino Zardi and C. David Whiteman
3 Dynamically-Driven Winds . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 121
Peter L. Jackson, Georg Mayr, and Simon Vosper
4 Understanding and Forecasting Alpine Foehn . . . . . .. . . . . . . . . . . . . . . . . . . . 219
Hans Richner and Patrick Hächler
5 Boundary Layers and Air Quality in Mountainous Terrain . . . . . . . . . . . 261
Douw G. Steyn, Stephan F.J. De Wekker, Meinolf Kossmann,
and Alberto Martilli
6 Theory, Observations, and Predictions of Orographic
Precipitation .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 291
Brian A. Colle, Ronald B. Smith, and Douglas A. Wesley
7 Microphysical Processes Within Winter Orographic
Cloud and Precipitation Systems . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 345
Mark T. Stoelinga, Ronald E. Stewart, Gregory Thompson,
and Julie M. Thériault
8 Observational Techniques: Sampling the Mountain Atmosphere.. . . . 409
Robert M. Banta, C.M. Shun, Daniel C. Law, William Brown,
Roger F. Reinking, R. Michael Hardesty, Christoph J. Senff,
W. Alan Brewer, M.J. Post, and Lisa S. Darby
ix
x Contents
9 Mesoscale Modeling over Complex Terrain: Numerical

and Predictability Perspectives. . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 531
James D. Doyle, Craig C. Epifanio, Anders Persson,
Patrick A. Reinecke, and Günther Zängl
10 Meso- and Fine-Scale Modeling over Complex Terrain:
Parameterizations and Applications.. . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 591
Shiyuan Zhong and Fotini Katopodes Chow
11 Numerical Weather Prediction and Weather Forecasting
in Complex Terrain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 655
Brad Colman, Kirby Cook, and Bradley J. Snyder
12 Bridging the Gap Between Operations and Research to
Improve Weather Prediction in Mountainous Regions .. . . . . . . . . . . . . . . . 693
W. James Steenburgh, David M. Schultz, Bradley J. Snyder,
and Michael P. Meyers
Author Index.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 717
Subject Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 725

Contributors
Robert M. Banta NOAA Earth System Research Laboratory, Boulder, CO, USA
W. Alan Brewer NOAA Earth System Research Laboratory, Boulder, CO, USA
William Brown National Center for Atmospheric Research, Boulder, CO, USA
Fotini Katopodes Chow Department of Civil and Environmental Engineering,
University of California, Berkeley, CA, USA
Brian A. Colle School of Marine and Atmospheric Sciences, Stony Brook Univer-
sity/SUNY, Stony Brook, NY, USA
Brad Colman NOAA/National Weather Service, Seattle-Tacoma, WA, USA
Kirby Cook NOAA/National Weather Service, Seattle-Tacoma, WA, USA
Lisa S. Darby NOAA Earth System Research Laboratory, Boulder, CO, USA
Stephan F.J. De Wekker Department of Environmental Sciences, University of
Virginia, Charlottesville, VA, USA
James D. Doyle Naval Research Laboratory, Marine Meteorology Division, Mon-
terey, CA, USA
Craig C. Epifanio Texas A&M University, Texas, USA
Patrick Hächler Federal Office of Meteorology and Climatology MeteoSwiss,
Zurich, Switzerland
R. Michael Hardesty NOAA Earth System Research Laboratory, Boulder, CO,
USA
Peter L. Jackson University of Northern British Columbia, Prince George, BC,
Canada
Meinolf Kossmann Climate and Environment Consultancy, Deutscher Wetterdi-
enst, Offenbach am Main, Germany
xi
xii Contributors
Daniel C. Law NOAA Earth System Research Laboratory, Boulder, CO, USA
Alberto Martilli CIEMAT Unidad de Contaminacion Atmosferice, Madrid, Spain
Georg Mayr University of Innsbruck, Innsbruck, Austria
Michael P. Meyers NOAA/National Weather Service, Grand Junction, CO, USA
Anders Persson UK MetOffice, SMHI, Norrköping, Sweden
M.J. Post NOAA Earth System Research Laboratory, Boulder, CO, USA
Patrick A. Reinecke Naval Research Laboratory, Marine Meteorology Division,
Monterey, CA, USA
Roger F. Reinking NOAA Earth System Research Laboratory, Boulder, CO, USA
Hans Richner Institute for Atmospheric and Climate Science (IACETH), ETH,
Zurich, Switzerland
David M. Schultz Division of Atmospheric Sciences, Department of Physics,
University of Helsinki, Finland
Finnish Meteorological Institute, Helsinki, Finland
Centre for Atmospheric Science, School of Earth, Atmospheric and Environmental
Sciences, University of Manchester, Manchester, United Kingdom
Christoph J. Senff NOAA Earth System Research Laboratory, Boulder, CO, USA
Cooperative Institute for Research in the Environmental Sciences, Boulder, CO,
USA
C.M. Shun Hong Kong Observatory, Kowloon, Hong Kong
Ronald B. Smith Geology and Geophysics Department, Yale University, New
Haven, CT, USA
Bradley J. Snyder Meteorological Services of Canada, Vancouver, BC, Canada
W. James Steenburgh Department of Atmospheric Sciences, University of Utah,
Salt Lake City, UT, USA
Ronald E. Stewart Department of Environment and Geography, University of
Manitoba, Winnipeg, MB, Canada
Douw G. Steyn Department of Earth and Ocean Sciences, The University of
British Columbia, Vancouver, BC, Canada
Mark T. Stoelinga 3TIER, Inc, Seattle, WA, USA
Julie M. Thériault National Center for Atmospheric Research, Boulder, CO, USA
Gregory Thompson National Center for Atmospheric Research, Boulder, CO,
USA
Contributors xiii
Simon Vosper Met Office, Exeter, UK

Douglas A. Wesley Compass Wind, Denver, CO, USA
C. David Whiteman Atmospheric Sciences Department, University of Utah, Salt
Lake City, UT, USA
Dino Zardi Atmospheric Physics Group, Department of Civil and Environmental
Engineering, University of Trento, Italy
Günther Zängl Deutscher Wetterdienst, Offenbach, Germany
Shiyuan Zhong Department of Geography, Michigan State University, East Lans-
ing, MI, USA
Chapter 1
Mountain Weather Prediction:
Phenomenological Challenges and Forecast
Methodology
Michael P. Meyers and W. James Steenburgh
Abstract This chapter summarizes the modern practice of weather analysis and
forecasting in complex terrain with special emphasis placed on the role of humans.
Weather in areas of complex terrain affects roughly half of the world’s land
surface, population, and surface runoff, and frequently poses a threat to lives and
property. Mountain weather phenomena also impact a diverse group of users, which
may have both beneficial and detrimental implications on societal and economic
levels.
Advances in forecast skill derive not only from advances in numerical weather
prediction, geophysical observations, and cyber infrastructure, but also improve-
ments in the utilization of these advances by operational weather forecasters.
Precipitation skill scores during the past two decades, for example, show that
operational weather forecasters have maintained a consistent threat score advantage
over numerical precipitation forecasts. Although the role of human forecasters is
evolving, for many applications, the so-called “human-machine mix” continues to
provide an improved product over what can be produced by automated systems
alone. To produce the best forecasts possible for the benefit of society, it is crucial for
the mountain meteorologist to possess an in-depth knowledge of mountain weather
phenomena and the tools and techniques used for atmospheric observations and
prediction in complex terrain.
M.P. Meyers ()

National Weather Service, 2844 Aviators Way, Grand Junction, CO 81506, USA
e-mail: mike.meyers@noaa.gov
W.J. Steenburgh
Department of Atmospheric Sciences, University of Utah, Salt Lake City, UT, USA
F. Chow et al. (eds.), Mountain Weather Research and Forecasting, 1

Springer Atmospheric Sciences, DOI 10.1007/978-94-007-4098-3 1,
© Springer ScienceCBusiness Media B.V. 2013
2 M.P. Meyers and W.J. Steenburgh
1.1 Introduction
Contemporary mountain weather forecasting involves the integration of geophysical

observations, numerical and statistical analysis and modeling, and human cognition
to meet the challenges posed by a diverse range of terrain-induced phenomena.
This integration, known as the “human-machine mix” (Snellman 1977), produces
significantly better forecasts than can be produced by automated systems alone, with
the value added by human cognition representing a 5–10 year advance in numerical
weather prediction skill (Bosart 2003; Steenburgh et al. 2012, Chap. 12). The
human-machine mix is only effective, however, when operational meteorologists
possess in-depth knowledge of mountain weather phenomena and the tools and
techniques used for atmospheric observation and prediction in complex terrain.
In this chapter we provide a review of the major phenomenological challenges
confronting mountain meteorologists and qualitatively describe the contemporary
forecast process, with emphasis on the human element over complex terrain. Our
goal is to provide a foundation for subsequent chapters that focus on specific
mountain weather phenomena or forecast tools and techniques, including numerical
weather prediction, which ultimately must be integrated to produce societally
relevant forecasts. We conclude with a discussion of ongoing forecast applications
in areas of complex terrain.
1.2 Phenomenological Challenges in Complex Terrain
Mountains cover 25% of the Earth’s land surface, contain 26% of the global
population, and produce 32% of the surface runoff (Meybeck et al. 2001). Hills
and plateaus account for another 21% of the land surface, 20% of the population,
and 19% of the runoff. Thus, the weather in areas of complex terrain affects roughly
half of the world’s land surface, population, and surface runoff. The numbers are
greater if one considers the remote effects of mountains on the general circulation,
storm tracks, moisture transport, and river runoff.
The protection of lives and property from high impact events is a forecast
priority; in addition, accurate forecasts of day-to-day mountain weather variability
benefit commerce and the general public. For instance, many mountain recreation-
alists are impacted by mountain weather. In the United States, the total number of
people who participated in outdoor activities in 2007 is estimated at 217 million
(Cordell 2008). Outdoor recreation (camping, snow sports, rafting, hiking, hunting
and fishing, etc.) contributes $730 billion to the economy annually and supports
6.5 million jobs (1 in 20 U.S. jobs) according to the Outdoor Industry Association
(http://www.outdoorindustry.org). In addition to the general population, numerous
other industries are dependent on weather that occurs over complex terrain.
The primary phenomenological challenges confronting mountain meteorologists
include: (a) snow and (b) ice storms produced by orographic precipitation and/or
1 Mountain Weather Prediction: Phenomenological Challenges. . . 3
terrain-induced cold advection and cold-air damming; (c) floods, landslides, and
debris flows generated by orographic rainfall and/or terrain-induced deep con-
vection; (d) droughts; (e) extreme wildfire spread and behavior driven by fuels,
topography, and weather; (f) severe local windstorms created by high-amplitude
mountain waves and gap flows; (g) severe convective storms; and (h) cold-air
pools and associated air quality hazards. In addition to loss of life, high-impact
weather events generated by these phenomena can produce staggering economic
losses, often requiring participation from government entities to address them. To
adequately meet these phenomenological challenges, forecasters need not only a
strong foundation in mountain meteorology (also see: Whiteman 2000), but also
knowledge in areas such as climatology, hydrology, ecology, land-surface processes,
and societal vulnerability.
1.2.1 Snowstorms
Snowstorms (also see Chaps. 6 and 7) exert a heavy toll on public safety and
transportation. The average annual cost of snow removal for public roadways in
the United States exceeds US $2 billion (Doesken and Judson 1997; National
Research Council (NRC) 2004). Thornes (2000) estimated that the United Kingdom
spends over US $2 billion annually on direct and indirect costs related to winter
maintenance of roadways and road traffic delays. Airport delays and closures cost
US $3 billion annually for US carriers and produce adverse sociological impacts for
the travelers. The potential benefits from improved forecasting of snow and icing
diagnostics at U.S. airports exceed US $600 million annually (Adams et al. 2004).
In mountainous regions, the economic losses due to snow-related highway
closures are considerable (Fig. 1.1). For example, in Europe, during a prolonged
snowy period in February 1999, more than 40 tourist resorts in the Swiss Alps
were cut off from the outside world due to road closures for up to 14 consecutive
days, resulting in indirect costs of US $200 million (Nöthiger and Elsasser 2004).
In the United States, losses produced by the snow-related closure of Interstate
90, the major highway bisecting the Cascade Mountains of Washington State,
are estimated at US $700,000 per hour, and similar shut downs of Interstates 70
and 80 through Colorado and Wyoming, respectively, approach US $1 million
per hour. Poor driving conditions due to weather cause over 1.5 million vehicular
accidents in the United States annually with total economic costs of US $42 billion
dollars. These vehicular accidents result in 800,000 injuries with 7,400 fatalities
indirectly related to poor weather conditions (National Research Council 2004). A
conservative estimate of the annual costs of weather-related vehicular accidents in
Canada is US $1.1 billion dollars (Andrey et al. 2001).
Snowstorms seriously impact urban corridors adjacent to mountain locations.
Personal claims from the March 2003 Colorado Front Range Blizzard exceeded
US $93 million. However, with the bad comes the good: during the blizzard the
Fig. 1.1 Snow removal in the southwestern Colorado Mountains (Courtesy of the Colorado
Avalanche Information Center [CAIC] and the Colorado Department of Transportation [CDOT])
snowpack in the Colorado Rockies went from inadequate to above 100% of average.
A senior agriculturist for the Western Sugar Cooperative in eastern Colorado called
it a billion-dollar storm due to its positive impact on the snowpack (Kohler 2003).
Avalanches are another potential impact of winter storms in complex ter-
rain (Fig. 1.2). Recent major weather related avalanche disasters have killed 14
in Súðavı́k, Iceland in January 1995; 20 in Flateyri, Iceland in October 1995
(Jóhannesson and Arnalds 2001); and 55 in Swiss and Austrian villages in February
1999 (Keiler et al. 2005). Avalanche disasters are not restricted to mountainous
regions. On New Year’s Day 1999, nine were killed in Kangiqsualujjuaq, Quebec
when an avalanche ran down a steep hill and hit the local school (Branswell 1999).
Conversely, mountain snowstorms can be a winter recreationalist’s dream
(Fig. 1.3) and provide benefits for the winter recreational economy. There are
approximately 6,000 ski resorts with nearly 400 million skier days per year (i.e.,
1 day of downhill skiing or snowboarding with a pass/ticket) globally (Hudson
2002; Skistar 2009). Europe is the largest market with 200 million skier days
per year, followed by North America with 80 million skier days per year (http://
www.nsaa.org/nsaa/press). The ski industry in the United States has annual revenue
of approximately US $12 billion (Scott 2006). Additionally, in North America,
snowmobilers spend more than US $28 billion annually on equipment, clothing,
Fig. 1.2 Avalanche over the southwestern Colorado Mountains (Courtesy of CAIC and CDOT)
accessories and vacations (US $6 billion in Canada) (International Snowmobile

Manufacturers Association http://www.snowmobile.org). Avalanches can, however,
pose a hazard for these recreationists. From 1991 to 2001, the International
Commission of Avalanche Rescue (http://www.ikar-cisa.org) reports nearly 1,500
fatalities due to avalanches with France and the United States having the two highest
percentages of deaths.
1.2.2 Ice-Storms
The advection and/or entrenchment of cold air by cold-air damming (Forbes et al.
1987; Bell and Bosart 1988) and orographic channeling affect the locations of rain-
snow transition zones in winter storms (also see: Chaps. 6 and 7). The January 1998
ice storm that devastated the northeastern United States and southeastern Canada
Fig. 1.3 Recreational skiing in Colorado (Courtesy of Meyers)
produced 80–100 mm of freezing rain over a 5 day period, causing more than
US $4 billion in economic damage, including US $3 billion in Canada (Reagan
1998; Roebber and Gyakum 2003). Cold-air channeling within the St. Lawrence,
Ottawa River, and Lake Champlain valleys enabled the persistence of low-level
cold air within the precipitating region, controlled the position of the surface-
based freezing line, and enhanced precipitation rates through frontogenesis. Such
orographic channeling can also lower freezing levels and snow levels in interior
basins and mountain passes, such as the Columbia River Basin, Snake River Plain,
Columbia River Gorge, Snoqualmie Pass, and the Frazier River Valley in the
Cascade Mountains and Coast Range of western North America (e.g., Decker 1979;
Ferber et al. 1993; Steenburgh et al. 1997).
1.2.3 Floods, Flash Floods and Debris Flows
Floods are among the most common of geologic hazards worldwide. Typically, most
river systems flood (i.e., leave their confining channels and flow outward onto the
adjacent floodplain) every year or two. There are two types of floods: regional floods
that can last for several weeks or months, and flash floods that last for minutes to
hours. Both are dangerous and capable of adversely impacting lives and property
(National Research Council 2005; http://www.azgs.state.az.us/).
Flooding produced by orographic precipitation is responsible for many of the
weather-related natural disasters in mountainous regions. Of the 13 weather-related
disasters observed in the western United States since 1980 with damages and costs
exceeding US $1 billion, four were produced by heavy and/or persistent orographic
precipitation and associated surface snowmelt (Lott and Ross 2006). In the United
States, floods cost upwards of US $6 billion and about 140 people are killed by
floods each year (Knutson 2001). In Europe, losses from the Italian Piedmont floods
of 1994 included 64 casualties and US $9 billion in property damage (Linnerooth-
Bayer and Amendola 2003; Barredo 2007). The damage and fatalities produced
by these events frequently extend over the flood plains well removed from the
orography responsible for the precipitation enhancement.
Orography also contributes to localized but extremely hazardous flash floods
by influencing the formation and movement of deep convection and mesoscale
convective systems. Well documented examples include the Rapid City Flash Flood
of June 1972, which killed 238 and produced US $100 million in property damage
in South Dakota; the Big Thompson Canyon Flood of July 1976, which killed 145
and produced US $25.5 million in property damage in Colorado; the Vaison-La-
Romaine Flash Flood of September 1992, which killed 46 and produced US $460
million in property damage in southeastern France (Maddox et al. 1978; Sénési
et al. 1996; Barredo 2007); and the September 2002 severe flood event in the
western Mediterranean mountainous region of southern France (Nuissier et al. 2008;
Ducrocq et al. 2008) which killed 24 people and produced an economic damage
estimated at nearly US $2 billion (Huet et al. 2003).
Debris flows can occur on steep slopes where loose, unconsolidated earthen
materials, such as soils and rocks, experience gravitational acceleration during
heavy rains, glacial melt or snowmelt (Iverson 1997). Debris flows can move
downslope rapidly, at speeds of greater than 10 m s1 ; their less viscous, fine-
grained relative, mudflows, have been clocked traveling at 40 m s1 in steep
mountain canyons. Wildfires may potentially increase the risks for debris flow
development by destroying vegetation and making soils more hydrophobic (Cannon
et al. 1998, 2001, 2003; Cannon and Reneau 2000). The major hazards of debris
flows are from the impact of earthen materials, such as boulders and rocks, and
being buried or carried away by the flow. Debris flows can be devastating to
life and property. In the United States, damage estimates due to debris flows are
close to US $3 billion annually (Restrepo et al. 2008). During December of 1999
exceptionally heavy rain triggered catastrophic floods and landslides along portions
of the mountainous coastal region of northern Venezuela (Lyon 2003). Over 10,000
fatalities were reported and the cost of reconstruction was estimated at nearly US $2
billion.
1.2.4 Droughts
One long duration weather phenomenon which may impact short term decision
making for a mountain meteorologist, including hydrologists and fire weather
meteorologists, is drought. Droughts come in various forms, which may impact
society with varying intensities and durations. By definition, a drought is “a period
of abnormally dry weather sufficiently prolonged for the lack of water to cause
serious hydrologic imbalance (i.e. crop damage, water-supply shortage, etc.) in
the affected area” (American Meteorological Society 1986). On an annual basis,
average losses and costs in the United States due to drought are estimated to exceed
US $8 billion (Knutson 2001).
The main water source for over 30 million people in the mainly arid climate
of the southwestern United States is the Colorado River. Snowmelt from mountain
snowpack provides over 70% of the water supply for this region (Chang et al. 1987;
Christensen et al. 2004), and the estimated benefit of water storage exceeds US
$350 billion dollars annually (Adams et al. 2004). Water is also the driving force
behind the agricultural industry of the southwestern United States. In California,
the agriculture industry accounts for nearly US $150 billion annually according
to the California Department of Food and Agriculture (http://www.cdfa.ca.gov/).
The economic impacts on the agricultural community due to droughts can be
tremendous. The University of California, Davis, estimated that US $2.8 billion
in agriculture-related wages, and as many as 95,000 jobs across the valley were
potentially lost due to the 2008 drought (Howitt et al. 2009).
Droughts can adversely impact the winter recreational industry, such as skiing
and snowmobiling, which depends on mountain snowpack. For example, many
skiers and snowboarders tend to favor lower-density, abundant snow (Steenburgh
and Alcott 2008), and low quality or meager snow can affect the demand for ski days
(Englin and Moeltner 2004). Summertime recreational use is not immune from low
snowpack. Low runoff during drought years can significantly reduce white-water
rafting revenue, which in Colorado alone, attracts 540,000 visitors and generates
US $150 million annually. Recreational fishing and hunting may also be affected by
drought. The impact on fish and other aquatic life due to drought may be significant
(Matthews and Marsh-Matthews 2003). Drought-depleted ecosystems and wetlands
would have a drastic effect on other wildlife.
Drought is not as visually obvious as other weather phenomena in the mountains,
but it is perhaps the most devastating to the ecological state because it can signif-
icantly weaken a forest’s defenses against insect infestation and wildfires (Morris
and Walls 2009). Recently, various populations of bark beetles are impacting
the western United States and western Canada with unprecedented levels of tree
mortality (Fettig et al. 2007). For instance, beetles have devastated several million
acres of trees in Colorado and Wyoming by the end of 2008 (Robbins 2008).
Extreme drought conditions in mountainous regions may also lead to increased
wildfire activity, as was the case in 1988 (Yellowstone National Park Fire) and in
2002 (Hayman, Missionary Ridge and Coal Seam fires in Colorado).
Fig. 1.4 2002 Big Fish Wildfire in Colorado (Photo by Mike Chamberlain; courtesy of the
NOAA/NWS/GJT)
1.2.5 Wildfire Behavior
Prescribed and unplanned wildfires (Fig. 1.4) impact tens of millions of acres
annually around the world (also see Chap. 2). In the western United States, where
the wealth and development of communities has migrated to the wildlands or the
wildland-urban interface over the past few decades, there have been eight wildfires
since 1980 that have produced more than US $1 billion in damage (Lott and Ross
2006). Of particular safety concern for firefighters are the conditions that lead to
fire “blowups”, which cause rapid fire spread. Wildfires over complex terrain are
especially susceptible to these blowups, due the diurnal changes of surface wind
direction and speed, as well as the surface mixing of synoptic and convectively-
driven winds.
One of the most devastating wildfires, the Big Burn of 1910, killed at least 78
firefighters and burned millions of acres in northern Idaho and western Montana
(National Wildfire Coordination Group 1997). This wildfire raised political aware-
ness of the economic and human impacts produced by wildfires. More recently,
from 1990 to 2006, there were 310 fatalities during western US wildland fire
operations, including the 1994 Storm King wildfire (14 fatalities) (National Wildfire
Coordination Group 2007). The number one cause of death during this time period
was burnovers (40%), “a situation where personnel or equipment is caught in an

advancing flamefront” (National Wildfire Coordination Group 2007), and secondly
due to aircraft and vehicular accidents (30%). Wildfires often have an adverse
impact on recreation and tourism. After the devastating wildfires in Yellowstone
National Park in 1988, visits to the park dropped 15% the following year (National
Park Service 2009). High wildfire danger may result in the closure of wildland areas,
which negatively impact recreation and logging.
1.2.6 Severe Windstorms Created by High-Amplitude

Mountain Waves and Gap Flows
Terrain-forced flows such as downslope windstorms and gap winds can produce
severe winds, many of which have exotic names like the Chinook along the eastern
slopes of the Rocky Mountains, the Mistral of southeast France, the Bora of
Slovenia, Croatia, and Bosnia, the Zonda along the east slopes of the Andes in South
America, the Taku in Alaska and the Föehn of central Europe (also see: Chaps. 3
and 4). Surface winds during extreme downslope windstorms can exceed hurricane
force (>33 m s1 ) and associated turbulence, rotors, and aircraft icing present a
threat to aviation safety (Nance and Colman 2000). Downslope windstorms are also
associated with the rapid spread of wildfires (e.g. Santa Ana [southwest United
States], Föehn, Chinook), making some regions which experience these winds
particularly prone to extreme fire weather behavior. Terrain-forced flows are also
a concern for maritime travel and commerce in and near mountainous coastal zones
around the world. It even has been suggested that these downslope windstorms are
often associated with illnesses ranging from migraines to psychosis (Soyka 1983).
Less frequent downslope windstorms sometimes occur in a synoptically un-
common flow pattern and represent a difficult forecast problem given their low
frequency. In October 1997, while a blizzard was occurring over the Front Range
of Colorado (Poulos et al. 2002), an easterly downslope windstorm was occurring
over the Mt. Zirkel Wilderness in the north central Colorado Rockies (Meyers
et al. 2003). This easterly downslope wind event had devastating ecological
consequences, resulting in 13,000 acres of forest blowdown in the Routt National
Forest.
Although orography also produces thermally driven winds, they are not typically
severe. Exceptions include large-scale katabatic flows such as those that occur
along the coast of Antarctica, which can become violent, particularly if they are
accelerated through interactions with local topography or enhanced by the large-
scale pressure gradient (e.g., Parish and Bromwich 1998).
1.2.7 Severe Convective Storms
Mountainous terrain can have an indirect impact on tornado development by

modifying airmasses downstream over the plains. One scenario occurs when the
mountains modify the plains environment by providing an elevated mixed layer
which results in a higher severe weather potential for tornado development (Lanicci
and Warner 1991). Over the mountains, tornadoes are more infrequent than their
plains counterparts, but can be significant when they do form. Tornadoes and
funnel clouds occur occasionally over the Rocky Mountains during the late spring
and summer (Bluestein and Golden 1993). These tornadoes usually develop in
non-supercell storms because the vertical shear is usually too weak for supercell
formation. Similar type storms have been observed in Switzerland (Linder and
Schmid 1996). However, mountain tornadoes can potentially be devastating. Fujita
(1989) documented an F4 tornado that crossed the Continental Divide within
Yellowstone National Park in Wyoming in 1987. This tornado traveled 24 miles
and leveled 15,000 acres of mature pine forest. On 11 August 1999 an F2 tornado
developed southwest of downtown Salt Lake City, Utah, and moved directly through
the city. This tornado resulted in one fatality, more than 100 injuries and US $170
million in damages (Dunn and Vasiloff 2001). Bosart et al. (2006) analyzed a long-
lived supercell that became tornadic over complex terrain in Massachusetts, on 29
May 1995. The F3 tornado left a 50–1,000-m-wide damage path that stretched for
50 km. Other documented supercell tornadoes over complex terrain include those
over the hilly terrain of the upper Rhine Valley of Germany (Hannesen et al. 2000)
and over the Colorado Mountains (Bluestein 2000).
Another potentially devastating severe wind that occurs in proximity to mountain
locations is the downburst wind or the smaller scale microburst (<4 km). Downburst
winds in arid mountain areas like the western United States are frequently dry and
develop in an environment with a deep, nearly dry-adiabatic subcloud layer; a shal-
low moist mid-layer near 500 mb; weak synoptic-scale forcing with only moderate
(<25 m s1 ) winds aloft; and weak instability [lifted index (LI) usually >2 K]
(Wakimoto 1985). The exact locations of these downburst winds are difficult to
forecast since they are often formed by seemingly benign-looking clouds or reflec-
tivity signatures (Mielke and Carle 1987; Meyers et al. 2006). Downbursts can be
extremely hazardous for aircraft operations, especially during takeoffs and landings.
Hail and graupel (soft hail) is often found in convective precipitation over
mountainous regions, in part, due to the relatively low freezing level above the
ground compared to lower elevations. Severe hail can often occur adjacent to
mountain locations such as the High Plains to the lee of the Rocky Mountains
(Doswell 1980). For example, a hailstorm caused $350 million dollars in damage
over the Front Range, in Denver, Colorado on 13 June 1984 (Blanchard and Howard
1986). Numerous severe hailstorms have also been documented to the east of the
Canadian Rockies in Alberta, Canada (Wojtiw 1975; Smith and Yau 1987) and in
central Switzerland near the Jura Mountains to the north and the Alps to the south
(Houze et al. 1993).
1.2.8 Cold-Air Pools and Air Quality
Cold-air pools are not typically considered a “severe weather phenomenon”, but
their persistence can lead to poor air quality episodes in basins and valleys (also see
Chaps. 2 and 5). The World Health Organization recently estimated that 800,000
deaths per year worldwide could be attributed to urban outdoor air pollution and
the economic impact from air pollution-related illness is estimated at US $150
billion per year (World Health Organization 2002). Often these cold pools form
in basins or within valleys with terrain constrictions that allow cold air to build
up behind the constriction. The development of cold pools reduces the dispersion
of air pollutants and adversely affects air quality. These poor air quality episodes
are not restricted to large urban areas, but can occur anywhere where emissions
are concentrated, such as in the lower end of the Colorado Gore Valley west of
Vail or along the Mossau and Finenbach Valleys in Germany (Geiger et al. 1995;
Whiteman 2000). Some of the highest particulate matter concentrations observed in
the United States occur episodically in Logan, Utah, a mid-sized metropolitan area
with a population of about 125,000 that is in the topographically confined Cache
Valley (Malek et al. 2006). More recently, in the vicinity of the Jonah–Pinedale
Anticline natural gas field, in the rural Upper Green River Basin of Wyoming,
air quality instrumentation measured 8-h averaged ozone concentrations above the
Environmental Protection Agency’s threshold of 75 parts per billion (ppb). This
criteria, which is more typical of summertime events, was exceeded on 14 days and
resulted in the first ever wintertime ozone advisories in Wyoming (Schnell et al.
2009). The formation of diurnal cold-air pools and valley inversions has a potential
secondary effect. It can also be a complicating factor in frost events, which can be
problematic for agriculture at critical times of the growing season.
Degraded air quality due to mountain haze and pollutant transport can impact the
tourism industry including the mountain vistas in national parks. Poor air quality
degrades the majestic views on public lands, but also can negatively impact the
long-term health of plants, trees and animals.
1.3 The Contemporary Forecast Process in Complex Terrain
At its core, weather forecasting is a scientific endeavor involving hypothesis

formulation, hypothesis testing, and prediction (Roebber et al. 2004), and applying
this scientific method is especially crucial over complex terrain. The forecaster
must develop a conceptual understanding of the past and present weather, formu-
late hypotheses about why and how the atmosphere is evolving, and then seek
evidence to confirm or reject the hypotheses. This iterative process continues and
the hypothesis is refined until a prediction is made. Given the huge volume of
data, analyses, and numerical forecasts available, the forecaster must employ rapid
cognition, make quick decisions, and make judgments in the face of uncertainty
(Doswell 2004). When faced with a “firehose” of data, the ability of a forecaster
to skillfully determine what is important is an example of what Gladwell (2005)
described as “thin slicing”.
Klein Associates has studied the cognitive and psychological aspects of U.S.
military meteorologists (Klein 1998; Stuart et al. 2007). They found distinct
differences between inexperienced or non-engaged forecasters and the experienced
or expert forecasters. The inexperienced forecasters typically relied too much on
computer models and tended to be reactive with the forecasts. The experienced
forecasters had a more global perspective and they tended to be more flexible
with their tools and procedures, often relying on conceptual models in the forecast
process. They typically employed a recognition-primed decision model (Klein
1998) which combines both analysis and intuitive methods (Doswell 2004) that
allows the forecaster to absorb incomplete information under time constraints and
arrive at proper forecasts. The expert forecasters draw from their vast forecast
experience to arrive at their forecast decisions.
Bosart (2003) argues that the forecast process is most effective when the
forecaster addresses six critical questions: (1) What happened? (2) Why did it
happen? (3) What is happening? (4) Why is it happening? (5) What is going
to happen? (6) Why is it going to happen? In an era of increasingly skillful
numerical weather prediction models, it is easy for forecasters to concentrate only
on question 5. Nevertheless, knowledge of the antecedent conditions and underlying
physical processes is extremely valuable, particularly when the numerical guidance
“goes awry” or is unable to resolve critical orographic effects (Bosart 2003; Dunn
2003).
Although there are a variety of forecasting styles (Pliske et al. 2004), skillful
forecasting in mountainous regions typically requires: (1) a core understanding of
synoptic scale and orographic processes, (2) careful evaluation of the evolving syn-
optic setting and flow interaction with the terrain, (3) knowledge of the advantages
and limitations of the objective tools of forecasting over complex terrain, and (4)
the subjective integration of these tools by the forecaster.
1.3.1 Scale Interaction: The Forecast Funnel
Forecasters commonly use the so-called forecast funnel to evaluate the synoptic
setting and flow interaction with terrain (Snellman 1982; Horel et al. 1988;
Steenburgh 2002; Dunn 2003). As illustrated conceptually by Fig. 1.5, the forecaster
begins at the global or planetary scale, focuses attention on progressively smaller
scales, and ultimately builds in the orographic effects at the local scale. The
forecaster considers how the interaction of the large-scale or synoptic flow with the
regional and local orography will influence weather locally. An important premise
of the forecast funnel is that processes on each scale are dependent upon those at
other scales. For example, in the case of orographic precipitation, the forecaster
typically begins by evaluating the past, current, and future synoptic setting and
Fig. 1.5 The forecast funnel (Courtesy of COMET) (The source of this material is the COMET®
Website at http://meted.ucar.edu/ of the University Corporation for Atmospheric Research
[UCAR], sponsored in part through cooperative agreement(s) with the National Oceanic and
Atmospheric Administration [NOAA], U.S. Department of Commerce [DOC]. ©1997–2010
University Corporation for Atmospheric Research. All Rights Reserved)
the large-scale characteristics of the airmass interacting with the topography. The
forecaster will determine his/her confidence in the large scale forecast and if it
is low, it will adversely affect his/her confidence in the forecast guidance on the
smaller scales. As the forecaster examines the regional scales, parameters such
as stability, temperature, humidity and wind and their influence on the terrain-
induced flow and precipitation dynamics are considered. Finally, the forecaster will
scrutinize the local scale to determine how the topography has influenced and will
influence winds, moisture and precipitation distribution at a specific location. Local
orographic and microphysical effects are considered at this level. This scale is the
most difficult to address in an operational environment and requires reliance on
pattern recognition and knowledge of the local climatology.
The forecast funnel in complex terrain requires a sound understanding of
synoptic, mesoscale and mountain weather processes. It also requires the forecaster
to recognize the advantages and limitations of the available observational and
numerical tools and apply the proper forecasting techniques. Ultimately, the forecast
funnel enables the forecaster to prioritize and assimilate the massive volume of geo-
physical and numerical data and to identify what is important on a day-to-day basis.
1.3.2 Objective Tools
The forecaster utilizes a number of objective analysis and forecast tools including
in situ and remotely-sensed observations (also see Chap. 8) and numerical weather
prediction models (also see Chaps. 9, 10, 11). Unique aspects of the use of these
tools in complex terrain are described below.
1.3.2.1 Surface-Based Observations
A major challenge for weather analysis over complex terrain is the need for
high-density surface observations to resolve the fine-scale gradients in surface
weather produced by topographic forcing. As a result, forecasters desire as much
observational data as possible and are willing to compromise, often relying on
observations from heterogeneous networks with differing sensor types, biases and
reliabilities (e.g., Horel et al. 2002). Some of the advantages of this approach
include higher data density, higher frequency observations at some locations,
and observations from non-conventional locations. Some of the disadvantages
include non-uniform data and siting characteristics, stations that do not report the
full suite of data, an increased need for quality control by the forecaster, and
varying instrumentation with inconsistent averaging intervals. Another critical issue
regarding surface observations in the mountains is the siting or microclimate where
instrumentation is placed. For example, many Remote Automated Weather Systems
(RAWS) (Myrick and Horel 2008) which are geared to fire weather observations
are sited on south aspects to capture worst case scenarios with regard to fire
weather applications. Geiger et al. (1995) showed that temperatures on south aspects
in the midlatitudes are typically 3ı C warmer than northern slopes due to sun
angle. Most Snowpack Telemetry (SNOTEL) instrumentation, run by the USDA’s
National Resources Conservation Service, is sited on aspects more conducive for
deeper snowpack and hydrological applications (Dressler et al. 2006). A failure to
understand these types of differences can lead to unrealistic biases in analyses and
forecasts. Additionally, incorrect or imprecise specification of the location of the
instrumentation can negatively impact resolution of fine-scale features and model
verification scores in complex terrain (Ludwig et al. 2006). The bottom line is that
forecasters need to make sure their observational data is quality controlled and siting
biases are taken into account.
Surface-based observations provide an added benefit after the event as well.
Observations allow the forecaster to do post-analysis for verification and to conduct
post-mortems of the event. This process is vital to understanding the physical
controls of a particular storm or event. Observational data are also used as a “first
forecast” to produce weather fields based on the previous day’s observations through
bias-corrected statistics.
1.3.2.2 Radar
Radars (also see Chap. 8) are an important tool in weather analysis and forecasting
over complex terrain. However, one must understand some of the limitations of
radars in the mountains such as strong ground clutter and beam blockage by terrain
(Germann and Joss 2004). MeteoSwiss found that one algorithm improvement
may lead to a degradation of other algorithm statistics. They took a systematic
approach with a combination of refinements to eliminate residual ground clutter and
shielding from mountains, hardware calibration, and minimize biases to improve
quantitative precipitation estimates from radar (Germann and Joss 2004). Typically,
mountain radars in the western United States are situated at higher elevations to
optimize scan strategies and to limit shielding or beam blockage (Wood et al. 2003)
which potentially creates another problem due to overshooting of precipitation. This
problem is more apparent in wintertime orographic precipitation events and also
shallow monsoonal or tropical precipitation. Breidenbach et al. (1999, 2001) showed
the effective radar coverage for summer and winter seasons in the Pacific Northwest
(Fig. 1.6). These coverages are based on a climatology analysis of radar-derived
precipitation computed from WSR-88D data (1996–2000) during the warm and
cold seasons. It shows that radars located at high elevations in mountainous terrain
are impacted by range effects (beam overshooting precipitation) and severe beam
blockage issues and these impacts are exacerbated during the winter season. Rea-
sonable precipitation estimates can be obtained only within the effective coverage
area. Wetzel et al. (2004) has documented cases where the Grand Junction, Colorado
WSR-88D radar (GJX) detected little or no precipitation signal over the Park Range
(elevations 2,150–3,550 m) of north-central Colorado (200 km from GJX and
7,300 m MSL at 0.5ı ) when, in fact, shallow orographically-forced heavy snowfall
was occurring over these mountains. Additionally, the spacing between WSR-88D
radars across the western third of the United States (Fig. 1.6) is much greater than
the remainder of the country which increases issues with detection (Westrick et al.
1999; Maddox et al. 2002). Vasiloff et al. (2011) has examined the benefits of using
mobile dual-polarization radar over the beam blocked complex terrain areas of the
central Colorado Mountains. These deficiencies of the radar coverage over complex
terrain must be considered by the forecaster.
1.3.2.3 Satellite
Satellite data provides a significant benefit over complex terrain for the mountain
meteorologist, providing a diversity of data for analysis and forecast applications.
Satellite fog products are very useful in valley and basin locations to detect and
forecast low level clouds and fog which have important implications for aviation
and ground transportation. Water vapor imagery is an excellent tool for diagnosing
mountain wave signatures. As was discussed in the previous section, radar coverage
can be seriously compromised over complex terrain. However, satellite products
supplement analysis of data sparse regions which may be neglected by the radar
Fig. 1.6 WSR-88D coverage

over the northwestern United
States during summer (top)
and winter (bottom) months
(Breidenbach et al. 1999,
2001). These coverages are
based on a climatology
analysis of radar-derived
precipitation computed from
WSR-88D data (1996–2000)
during the warm and cold
seasons (© American
Meteorological Society.
Reprinted with permission)
coverage, especially at farther distances from the radar. Visible satellite imagery
is especially useful during wintertime orographic snowfall where radar detection
may overshoot the shallow orographic clouds. Similarly, visible satellite imagery
and other satellite applications provide supplemental detection of the shallower
convective events such as monsoonal surges in the mountainous terrain, as well as
wildfire detection.
1.3.2.4 Lightning
Lightning data is also crucial for the forecaster in the mountains (Holle and Lopez
1993). Lightning observations are often used by forecasters in complex terrain to
provide another layer of detection of convection cells, particularly in areas with poor
radar coverage. Lightning analysis is often overlaid on radar or satellite imagery to
complement these observational tools. Lightning detection adds in the diagnosis
of location, storm strength, and ice processes and the potential for hail. In some
situations, lightning may provide a secondary method of storm detection when radar
analysis is inadequate.
1.3.2.5 Numerical Weather Prediction (NWP)
Numerical models (also see Chap. 11) are arguably the most used tool in a
forecaster’s toolbox. However, it is imperative that the forecaster understands the
limitations of numerical models over complex terrain. After all, as well stated
by Doswell (1986), “when guidance is needed the most, it is generally the least
useful.” Topography may enhance predictability of certain flow types, but can also
exacerbate large-scale forecast errors.
Evaluating the value added utility of a high-resolution NWP forecast is the
ultimate challenge for the forecaster. Application of the forecast funnel helps with
this component of the forecast process, is especially important when the large-scale
forecast is not representative of reality (McMurdie and Mass 2004), and is essential
for predicting phenomena that are unresolved or poorly simulated with existing
models and physics parameterizations. If the large scale is not predicted well, the
forecaster may have low confidence in the mesoscale NWP forecast and must adjust
the numerical guidance substantially. On the other hand, forecasters can lean toward
the high-resolution model solution when confidence in both the large-scale forecast
and the mesoscale model forecast are high.
Ensemble forecast systems are now widely utilized by weather forecasters
(Roebber et al. 2004; Novak et al. 2008). These ensemble forecast systems may
incorporate different models, systematically varying initial conditions, physical
schemes, and/or boundary conditions to produce separate runs which are valid at
a specific time (Brooks and Doswell 1993; Sivillo et al. 1997). Ensemble forecast
systems have recently been used for short term prediction which is useful in
topographic settings (Arribas et al. 2005; Yuan et al. 2007; Reinecke and Durran
2009). These ensemble solutions help determine pattern evolution and model
consistency.
Mountain meteorologists often rely on post processing techniques, such as Model
Output Statistics (MOS) or other statistical applications as an objective tool for
forecasting specific surface parameters at site specific locations (Glahn and Lowry
1972; Cheng and Steenburgh 2007). Although it is a very useful tool, mountain
weather forecasters need to be careful when applying these statistical methods over
gridded forecasts of varying degrees of complex terrain. The forecaster can possibly
fall into the trap of using MOS too much without using other available forecasting
tools. MOS is a convenient and quick way to provide a forecast; however, it may
come with potentially dire consequences. As Bosart (2003) mentioned, “Forecasters
who grow accustomed to letting MOS and the models do their thinking for them on
a regular basis during the course of their daily activities are at high risk of ‘going
down in flames’ when the atmosphere is in an outlier mode”.
1.3.3 Subjective Tools: The Human Element
1.3.3.1 Geographic Familiarization
Forecasting in complex terrain requires a familiarization of the geography to provide

a basis for climatology and conceptual models that are useful in the mountains.
Roebber et al. (1996) speculated that regional knowledge is especially useful for
mountain forecasting skill. Forecasters who do not have a strong knowledge of the
local topography will have a difficult time determining forcing at the local scale of
the forecast funnel. The best way for the forecaster to achieve this understanding
is to travel around the forecast area and familiarize themselves with the geography
and local topography. If this is not possible, using a Geographic Information System
(GIS) or Google Earth application to evaluate the terrain is helpful. The populations
of mountain towns and resorts have a strong seasonal dependency, and it is important
to know the potential weather impacts on these communities. Furthermore, it is vital
to know the potentially problematic regions for hazardous weather. For instance, in
the southwestern United States, narrow slotted canyon regions will have a much
higher flash flood potential than other topographic areas. It is critical for the
forecaster to have a firm understanding of these diverse geographic regions, as well
as the potential vulnerabilities for hazardous weather in the areas.
1.3.3.2 Rules of Thumb
A “rule of thumb” is an easy-to-remember forecasting method based on experience

or statistics. Ideally, a “rule of thumb” is validated over a statistically significant
sample to be useful for the forecaster. Two examples: 3ı C at 700 hPa corresponds
to a 7,500 ft (2,286 m) rain-snow line or an inch (2.5 cm) of precipitable water
is required before any flash flooding will occur. A rule of thumb usually provides
the forecaster with a first guess, but it may be fine-tuned based on the conditions
of the given day. For instance, in the rain-snow example above, the snow level
may be adjusted lower in a convective environment. “Rules of thumb” have
drawbacks which can limit the forecaster. They can be limited by a particular
model configuration, where the effectiveness is lost after a model upgrade or other
modification. Often “rules of thumb” are a black box to the user, where the physical
basis is not known or is non-existent. “Rules of thumb” can be helpful tools, but the
forecaster needs to understand the limitations and the need for sound scientific basis
behind these approaches.
1.3.3.3 Pattern Recognition and Climatology
Despite dramatic advances in numerical weather prediction over the past few
decades, pattern recognition and climatology continue to play an essential role in the
forecast process. Knowledge of the precipitation climatology is particularly valuable
in complex terrain. Despite the limited density of observing stations, generally reli-
able mean precipitation climatologies are now available for the contiguous United
States (Fig. 1.7; Daly et al. 1994, 2008; http://prism.oregonstate.edu/; Johnson
et al. 2000) and have been developed for other mountainous regions, including the
European Alps (e.g., Frei and Schär 1998; Schwarb 2000; Schwarb et al. 2001).
These climatologies are useful for downscaling or refining precipitation forecasts
from numerical models that are unable to fully resolve the local topographic effects.
Forecasters use conditional climatologies, numerical model forecasts, and pattern
recognition as tools to identify situations where the precipitation distribution can
potentially deviate significantly from the mean, such as when (1) the precipitation-
altitude relationship is unusually large or small, (2) orographic effects alter the
mesoscale distribution of precipitation, or (3) the synoptic forcing is not uniform
(e.g., frontal precipitation is expected to the south but not the north). For example,
the distribution and intensity of orographically-driven precipitation is dependent
on flow direction, which controls where orographic ascent and descent occur. Even
a relatively small shift in large-scale wind direction can have dramatic effects on
precipitation distribution (Ralph et al. 2003).
Pattern recognition is also useful for adjusting systematic model biases. For
example, the mesoscale modeling system used for the 2002 Olympic Winter Games,
which was based on the fifth-generation Pennsylvania State University–National
Center for Atmospheric Research (PSU–NCAR) Mesoscale Model (MM5), featured
a prominent wet bias on the windward slopes of the Wasatch Mountains and a dry
bias in the leeward valleys (Fig. 1.8; Hart et al. 2005). Similar biases have been
identified in MM5 simulations over the Cascade Mountains by Colle et al. (2000).
Knowledge of these biases enables forecasters to make appropriate adjustments.
Furthermore, model biases are often dependent on the synoptic regime (and flow
direction). With this knowledge, forecasters should be able to make adjustments
based on experience.
Fig. 1.7 Annual precipitation climatology for the United States generated by the PRISM model
([Daly et al. 1994, 2008]. Copyright © 2011, PRISM Climate Group, Oregon State University,
Reprinted with permission. http://prism.oregonstate.edu. Map created Jun 16 2006)
Fig. 1.8 Analysis of the 24-h precipitation PB scores for the (a) 12- and (b) 4-km domains.
Contours drawn at 40% intervals from 80% to 200%. Bold lines indicate maxima (solid) and
minima (dashed). Inset shows the 24-h precipitation PB scores versus model elevation bias (m).
Elevation scale at lower left (From Hart et al. 2005. © American Meteorological Society. Reprinted
with permission)
Fig. 1.9 Conceptual model based on the 100-h snowstorm over the Wasatch Mountains (From
Steenburgh 2003. © American Meteorological Society. Reprinted with permission)
1.3.3.4 Conceptual Models
A conceptual model is a physically-based model detailing a simplified represen-

tation of the physical phenomena. Conceptual models are a higher-order hybrid
approach using both the objective and the human element. These models are
based on information received through experimental and observational data, theory,
and climatology to ensure a proper physical basis. Conceptual models are widely
used by meteorologists as an important diagnostic tool and learned through training.
The conceptual model may help the forecaster determine whether a critical threat
exists, or may enhance the situational awareness of an event already occurring
or imminent. An example of a conceptual model over western United States is
shown in Fig. 1.9. In this conceptual model, forecasters are provided a basis for
this type of storm evolution. An example from the Alps is shown in Fig. 1.10, which
details a blocked-stable case versus an unblocked-unstable precipitation case. In
both these scenarios the forecaster is provided with a forecast foundation based on
the conceptual model. Often the storm does not duplicate the details of the model,
however a subjective assessment based on proper training, past experience and an
intuitive forecast process (Doswell 2004; Stuart et al. 2007) will enable him/her to
improve upon the conceptual model and anticipate the future outcome.
Fig. 1.10 Conceptual model of (a) an unblocked and unstable case, (b) a blocked and stable
case over the Alps (From Medina and Houze 2003; Rotunno and Houze 2007. © American
Meteorological Society. Reprinted with permission)
The ability to effectively “thin slice” as described by Gladwell (2005) and formu-
late and test hypotheses about the evolving atmosphere requires accurate conceptual
models that describe the structure and physical processes controlling synoptic,
mesoscale, and mountain weather systems. Conceptual models with application
to complex terrain are especially lacking compared to existing conceptual models
developed for the Plains and for severe weather (Andra et al. 2002). Conceptual
models are one application which would benefit from input of both the research and
operational forecasting community, providing the forecasters a mental model of the
theory. This interaction would require the forecasters and researchers to bridge the
gap and work together to improve the science of meteorology over complex terrain
as discussed later in this book (see Chap. 12).
1.4 Forecasting in Complex Terrain
1.4.1 Complexity of Mountain Weather Forecasting
The main priority for any forecaster, including the mountain meteorologist is to
protect lives and property during severe weather events. The phenomenological
challenges which confront the mountain weather forecaster are detailed in Sect. 1.2.
It is critical that the mountain meteorologist have a strong foundation in mountain
meteorology and understand the advantages and limitations of the subjective and
objective forecasting tools used in mountainous terrain as discussed in the previous

section. The mountain meteorologist must understand that hazardous weather may
occur during synoptically-benign weather. For instance, hot and dry conditions may
lead to snowmelt flooding during the spring or significant fire weather issues during
the summer. In severe weather events, the situational awareness of the forecaster
must be dynamic, especially since severe weather over complex terrain may be
more isolated and develop in more benign environments than at lower elevations.
In complex terrain mountain weather forecasters need to be flexible and multi-task
oriented, often forecasting for multiple severe weather types of events (e.g. winter
weather, convective weather) that may occur simultaneously across the forecast area
of responsibility (Meyers et al. 2006). The forecaster must also understand how their
users, such as the media and emergency managers, will assimilate the hazardous
forecasts (Morss et al. 2005; Morss and Ralph 2007).
Accurate predictions of day-to-day non-severe weather variability in regions of
complex terrain are essential for commerce and the public good. The daily challenge
for the mountain weather forecaster is to produce temporally and spatially detailed
forecasts that are accurate, reliable, and well communicated for a diverse user com-
munity in regions with extreme topographic, meteorological, and in many instances
socio-economic contrasts (American Meteorological Society (AMS) 2007). Further,
a dilemma exists which fuels our forecast challenge; a good impact for some may
have an adverse impact on another.
People who live in the mountains understand the complexity of mountain
weather. There is a saying, “if you don’t like the weather in the mountains, just
wait an hour”. However, they do not seem to be as forgiving, when it comes to
missed forecasts. Typically, the majority of users are not adversely affected by false
alarms (Barnes et al. 2007). However, there is a group of users in the mountain
snow community that relies on adverse weather for their business or pleasure and
false alarms may not be viewed favorably by this sector. A forecaster must be
cognizant of this community that relies on snow to make a living or to satisfy their
recreational desires. For example, there may be situations when heavy snowfall is
forecasted and it occurs at one ski area while another ski area 20 km away receives
little or none. The forecast may be a hit for one area but a miss for the other and
the area that did not receive the snow will often let you know about the missed
forecast. Mountain weather forecasters cannot be overly-sensitive when dealing
with these forecast inconsistencies and the potentially negative feedback from the
customer.
1.4.2 Specialized Forecast Examples
Many forecast consumers need a level of specificity beyond that of the general
weather forecast (AMS 2007). Businesses need very specific information relevant
to their operations, while the public is interested in how forecasts impact their
activities. For example, the Tyrolean ski industry in Austria invests nearly US $100
million a year in snowmaking (Steiger and Mayer 2008). This industry relies on
accurate temperature and humidity forecasts to determine wet bulb temperatures
below 5ı C when good snow quality can be produced. Accurate forecasts also
impact the winter Olympic Nordic ski performance. The correct choice of ski wax
used by the athletes depends on accurate forecasts of the snow surface conditions
on the race course. The collection of snow measuring and forecasting techniques for
the 2010 winter Olympics are detailed by Wagner and Horel (2008). Several more
examples of specialized forecasts in complex terrain are discussed next.
1.4.2.1 Fire Weather Forecasts
Fire weather forecasters have unique concerns that typically focus on diverse
microscale interactions over complex terrain. For this reason, fire weather forecasts
are very demanding and stressful due to the high impact of the forecast on lives or
property. A fire weather forecaster must understand the fire environment, which is
dependent on the fuel moisture, topography and weather. How these components
interact determines the fire behavior. A fire weather forecaster should be well-
informed of the fuel conditions in a particular forecast area. To understand fuel
conditions the forecaster must understand the antecedent seasonal conditions that
have occurred for the forecast area, including how drought may have impacted a
particular region. The forecaster must also understand the fine-scale topography
and how it can modify the diurnal evolution of the surface and boundary layer
structures. This knowledge aids in temperature, wind, and humidity forecasts, which
can vary tremendously over small horizontal distances in mountainous regions. The
fire weather forecaster scrutinizes local observations in the complex terrain and
must also predict thunderstorm activity, which is a critical source for new fire starts
when fuels are dry. Additionally, the fire weather forecaster must understand smoke
management issues which are complicated along slopes. The parameters associated
with a smoke management forecast are dependent on an accurate forecast of the
mixed boundary layer, as well as transport winds which are complicated along
slopes where in-situ measurements may not be representative of the entire area.
Fire weather forecasters must clearly understand the potential risks of smoke on
population centers near these fires.
A critical element of fire weather forecasts includes vital site-specific spot
forecasts for wildfire and prescribed fire projects. Requests for spot forecasts have
been increasing dramatically, with the NWS alone providing nearly 20,000 spots
every year. Wildfire mitigation through prescribed burns can be an efficient and
effective tool in land management. Accurate weather forecasts are essential for the
effective and safe use of prescribed fires, and also to ensure the safety of firefighters
who frequently operate in areas of rugged and remote terrain during wildfires and
prescribed fire projects.
1.4.2.2 Hydrological Runoff Forecasts
Hydrologists have a variety of responsibilities over complex terrain. One of their

main concerns, especially during the winter and spring months is to determine the
condition of the snowpack for runoff prediction. This task is especially difficult
to determine due to the sparse network of observations over the mountains. In the
western United States hydrologists depend on SNOTEL observations which provide
a network of snow water equivalent (SWE) measurements over the higher terrain.
SWE is a better indicator of the amount of liquid within the snowpack than snow
depth or accumulated liquid equivalent precipitation, a portion of which potentially
can fall as rain. SWE also will account for losses due to sublimation or melting.
They must determine the representativeness of these observations, which may be
compromised by anomalous precipitation-elevation distributions, wind and drifting
issues, and other measurement inconsistencies. The forecasters employ a variety of
hydrologic and hydraulic models to determine how the snowpack will runoff. These
models consider subsurface soil moisture conditions, soil type and permeability,
losses due to evapotranspiration, stream and river channel characteristics, and the
current state of the snowpack. Equally important, the hydrologists must maintain
a daily weather watch of the current weather conditions where snowpack runoff
is critically dependent on temperature, wind, moisture and precipitation. Rain-on-
snow events, where significant storm rainfall combined with snowmelt can lead to
devastating flooding and damage in mountain locations (Kattelmann 1997; Graybeal
and Leathers 2006). There have also been recent studies (Barrett et al. 2008)
documenting the impact of dust on snow events. These dust layers decrease the
albedo of the snowpack and accelerate snowmelt which has significant implications
for snowpack runoff and avalanche potential.
Another responsibility of the hydrologists is to maintain a heightened awareness
of the dams and reservoirs. This awareness includes an assessment of the integrity of
the dam as well as the constant monitoring of critical levels of dam capacity, inflow
rates, and other factors which may lead to a dam break. These dam breaks (e.g. Teton
Dam Break) can potentially have devastating human and economic consequences.
1.4.2.3 Avalanche Forecasts
Similar to hydrologists, avalanche forecasters must maintain an active weather

watch and understanding of the current surface conditions. Avalanche forecasters
are interested in the amount of snowfall, but pay close attention to the snow density
or the percentage of water in the snowfall. They are concerned with periods when
atmospheric conditions lead to snow metamorphism or wind transport. Other issues
such as ‘rain on snow’ events can dramatically alter the structure and stability of the
snowpack.
Avalanche forecasters spend a great deal of time outside evaluating the snow-
pack. They need to have a strong understanding of the topographic variability
and knowledge of the avalanche-prone regions. Avalanche forecasters often ski
to representative locations and dig a snow pit to evaluate the thermal, mois-
ture and snow characteristics of the snow core to determine its integrity for
avalanche potential. Avalanche forecast users include governmental transportation
agencies which use their expertise to keep transportation corridors safe. Another
group of users include outdoor recreationalists such as skiers and snowboarders,
snowmobilers, and snowshoers. These groups sometimes seek adventure during
dangerous situations. The avalanche forecasters mitigate the potential danger by
educating these user groups, and issuing forecasts including warnings of critical
conditions which potentially could lead to avalanches.
1.4.2.4 Aviation Forecasts
Aviation customers in the mountains include commercial industry, private pilots,

hot air balloonists, glider pilots, hang-glider pilots, and others. Aviation forecasters
in complex terrain have a multitude of issues which they must deal with to
provide an accurate and timely forecast for these customers. In the absence of
synoptic or convectively driven winds, diurnal winds in the vicinity of mountains
dominates the typical wind structure with upslope/upvalley winds developing during
the late morning hours and downslope/downvalley winds developing after sunset.
Climatology of the wind evolution for specific airport locations is a useful tool for
the forecaster. It is also beneficial for the forecaster to visit the airport to understand
the complexity of the terrain surrounding the airport. Even though the diurnal
evolution of the winds of each airport location is easily known, it is important to
forecast the magnitude and transition times of this diurnal evolution for aviation
purposes. Typically the environment is complicated by synoptic or convective winds
that can overwhelm the diurnal winds. The aviation forecaster needs to keep an
active weather watch since even subtle changes in the synoptic or convective forcing
can have a dramatic impact on the surface wind patterns in the mountains.
Aviation forecasters must know airport thresholds and minimums for takeoffs
and landings. Each airport has different criteria based on their local terrain and
requirements. The forecaster must know these thresholds when forecasting ceiling
and visibility degradation due to weather obscuration. A forecast with ceilings
and visibilities in the Marginal Visual Flight Rules (MVFR) or Instrument Flight
Rules (IFR) categories could require commercial aircraft to carry more fuel or
possibly restrict flights until conditions improve. The fact that the precipitation
may be showery, with varying intensities of precipitation over short periods of time,
exacerbates this problem.
The aviation industry is especially concerned with wind shear, turbulence, and
icing potential at the ground and during flights through the mountains. These
potential hazards are prevalent due to the complex winds and areas of local lifting
that occur in the vicinity of mountains. In extreme cases, with certain stability and
flow, downslope windstorms may develop. These extreme wind events can produce
strong gradients, extreme wind shear, rotors, (also see Chap. 3) turbulence and icing
conditions.
One additional issue which the aviation industry must address especially during
the warm season is density altitude. Density altitude is the corrected altitude for non-
standard atmospheric conditions. Pressure, humidity and temperature affect density
altitude, potentially resulting in a higher altitude than the actual altitude. In these
situations, aircraft performance is potentially compromised and density altitude
is especially sensitive at higher elevations. The implication of density altitude is
especially critical to smaller aircraft which are used in fire weather and emergency
response.
1.5 Conclusions
This chapter has summarized some of the phenomenological challenges of mountain

weather and the societal and economic impacts associated with them. These
challenges demonstrate the importance of forecasting in mountainous regions. The
core forecast process includes a careful evaluation of the evolving scales and flow
interactions on the local terrain as well as an understanding of the objective and
subjective forecasting tools and techniques used in mountain weather analysis
and forecasting. In summary, the mountain meteorologist should have a core
understanding of the synoptic scale and orographic processes that are detailed in
the subsequent chapters of this book.
Acknowledgments We thank the participants in the 2008 AMS/COMET/MSC Mountain Weather

Workshop: Bridging the Gap between Research and Forecasting for 4 days of lectures and
discussion that stimulated this chapter, as well as our editors and chapter coauthors and their
contributions to this book. Thanks are also extended to the Grand Junction NWS staff, Jeffrey
Manion, and three anonymous reviewers for their contributions to the manuscript. Contributing
author Steenburgh acknowledges the support of the National Science Foundation under grant
ATM-0627937 and the National Weather Service under a series of grants provided by the C-STAR
program. Any opinions, findings, and conclusions or recommendations expressed in this material
are those of the authors and do not necessarily reflect the views of the National Science Foundation
or National Weather Service.
References
Adams, R. M., L. J. Houston, R. F. Weiher, 2004: The value of snow and snow information
services. Report prepared for NOAA’s National Operational Hydrological Remote Sensing
Center, Chanhassen, MN, under contract DG1330-03-SE-1097.
American Meteorology Society, 1986: Glossary of Meteorology. Edited by Ralph E. Huschke.
Boston, MA, AMS. 638 pp.
American Meteorological Society, 2007: Weather analysis and forecasting: An information
statement of the American Meteorological Society. Bull. Amer. Meteor. Soc., 88, 1655–1659.
Andra, D.L., E.M. Quoetone, and W.F. Bunting, 2002: Warning Decision Making: The Relative
Roles of Conceptual Models, Technology, Strategy, and Forecaster Expertise on 3 May 1999.
Wea. Forecasting, 17, 559–566.
Andrey, J., B. Mills, and J. Vandermolen, 2001: Weather information and road safety. ICLR Paper
Series No. 15, Institute for Catastrophic Loss Reduction, London, Ontario. 36 pp.
Arribas, A., K.B. Robertson, and K.R. Mylne, 2005: Test of a Poor Man’s Ensemble Prediction
System for Short-Range Probability Forecasting. Mon. Wea. Rev., 133, 1825–1839.
Barnes, L.R., E.C. Gruntfest, M.H. Hayden, D.M. Schultz, and C. Benight, 2007: False Alarms
and Close Calls: A Conceptual Model of Warning Accuracy. Wea. Forecasting, 22, 1140–1147.
Barredo, J. I., 2007: Major flood disasters in Europe: 1950–2005. Nat. Hazards, 42, 125–148.
Barrett, A. P., T. H. Painter and C. C. Landry, 2008: Desert dust enhancement of moun-
tain snowmelt. Intermountain West Climate Summary. Boulder, CO, Western Water
Assessment. 4:2.
Bell, G.D., and L.F. Bosart, 1988: Appalachian Cold-Air Damming. Mon. Wea. Rev., 116, 137–161.
Blanchard, D.O., and K.W. Howard, 1986: The Denver Hailstorm of 13 June 1984. Bull. Amer.
Meteor. Soc., 67, 1123–1131.
Bluestein, H. B. and J. H. Golden, 1993: A review of tornado observations. The Tornado: Its
Structure, Dynamics, Prediction, and Hazards, Geophys. Monogr., No. 79, Amer. Geophys.
Union, 319–352.
Bluestein, H. B., 2000: A tornadic supercell over elevated, complex terrain: The Divide, Colorado,
storm of 12 July 1996. Mon. Wea. Rev.,128, 795–809.
Bosart, L.F., 2003: Whither the Weather Analysis and Forecasting Process? Wea. Forecasting, 18,
520–529.
Bosart, L.F., A. Seimon, K.D. LaPenta, and M.J. Dickinson, 2006: Supercell Tornadogenesis
over Complex Terrain: The Great Barrington, Massachusetts, Tornado on 29 May 1995. Wea.
Forecasting, 21, 897–922.
Branswell, B., 1999: Avalanche in Quebec. Maclean’s Magazine. Maclean Hunter Publishing Ltd.
Breidenbach, J. P., D.J. Seo, P. Tilles and K. Roy, 1999: Accounting for radar beam blockage
patterns in radar-derived precipitation mosaics for River Forecast Centers. Preprints, 15th Int.
Conference on Interactive Information Processing Systems (IIPS) for Meteorology, Oceanog-
raphy, and Hydrology, Dallas, TX, AMS 5.22. [Available online at http://ams.confex.com/ams/
older/99annual/abstracts/1699.htm.].
Breidenbach, J.P., D.J. Seo, P. Tilles and C. Pham, 2001: Seasonal variation in multi-radar
coverage for WSR-88D precipitation estimation in mountainous regions. Preprints. Symposium
on Precipitation extremes: Prediction, Impacts, and Responses, Albuquerque, NM, AMS.
[Available online at http://ams.confex.com/ams/annual2001/techprogram/paper 18668.htm.].
Brooks, H. E., and C. A. Doswell III 1993: New technology and numerical weather prediction: A
wasted opportunity? Weather, 48, 173–177.
Cannon, S. H., P. S. Powers, and W. Z. Savage, 1998: Fire-related hyperconcentrated and debris
flows on Storm King Mountain, Glenwood Springs, Colorado, USA. Environ. Geol., 35 (2–3),
210–218.
Cannon, S. H., and S. L. Reneau, 2000: Conditions for generation of fire-related debris flows,
Capulin Canyon, New Mexico. Earth Surf. Processes Landforms, 25, 1103–1121.
Cannon, S. H., E. R. Bigio, and E. Mine, 2001: A process for fire-related debris flow initiation,
Cerro Grande fire, New Mexico. Hydrol. Processes, 15, 3011–3023.
Cannon, S. H., J. E. Gartner, A. Holland-Sears, B. M. Thurston, and J. A. Gleason, 2003: Debris-
flow response of basins burned by the 2002 Coal Seam and Missionary Ridge fires, Colorado.
Engineering Geology in Colorado—Contributions, Trends, and Case Histories, D. D. Boyer,
P. M. Santi, and W. P. Rogers, Eds., Association of Engineering Geologists Special Publication
14, 1–31.
Chang, A. T. C., J. L. Foster, P. Gloersen, W. J. Campbell, E. G. Josberger, A. Rango and Z. F.
Danes. 1987. Estimating snowpack parameters in the Colorado River basin. In: Proc. Large
Scale Effects of Seasonal Snow Cover, IAHS Pub. No. 166, 343–353.
Cheng, W.Y.Y., and W.J. Steenburgh, 2007: Strengths and Weaknesses of MOS, Running-Mean
Bias Removal, and Kalman Filter Techniques for Improving Model Forecasts over the Western
United States. Wea. Forecasting, 22, 1304–1318.
Christensen, N.S., A.W. Wood, N. Voisin, D.P. Lettenmaier, and R.N. Palmer. 2004: The Effects of
Climate Change on the Hydrology and Water Resources of the Colorado River Basin. Climatic
Change, 62, 377–363.
Colle, B. A., C. F. Mass and K. J. Westrick, 2000: MM5 precipitation verification over the Pacific
Northwest during the1997–1999 cool seasons. Wea. Forecasting, 15, 730–744.
Cordell, H.K., 2008: The latest on trends in nature-based outdoor recreation. Forest History
Today, Spring, pp 4–10. Last retrieved on October 20, 2008, from http://www.foresthistory.
org/Publications/FHT/FHTSpring2008/Cordell.pdf.
Daly, C., R.P. Neilson, and D.L. Phillips, 1994: A Statistical-Topographic Model for Mapping
Climatological Precipitation over Mountainous Terrain. J. Appl. Meteor., 33, 140–158.
Daly, C., Halbleib, M., Smith, J.I., Gibson, W.P., Doggett, M.K., Taylor, G.H., Curtis, J., and
Pasteris, P.A. 2008. Physiographically-sensitive mapping of temperature and precipitation
across the conterminous United States. Int. J. Climatol., 28, 2031–2064.
Decker, F. W., 1979: Oregon’s silver thaw. Weatherwise, 32, 76–78.
Doesken, N. J. and A. Judson, 1997. “The Snow Booklet: A Guide to the Science, Climatology,
and Measurement of Snow in the United States”. Colorado Climate Center, Department of
Atmospheric Science, Solorado State University, Fort Collins, CO. p. 86.
Doswell, C.A., 1980: Synoptic-Scale Environments Associated with High Plains Severe Thunder-
storms. Bull. Amer. Meteor. Soc., 61, 1388–1400.
Doswell, C.A. III 1986: The Human Element in Weather Forecasting, Nat. Wea. Dig., 11, 6–17.
Doswell, C.A. III, 2004: Weather Forecasting by Humans—Heuristics and Decision Making. Wea.
Dressler, K.A., S.R. Fassnacht, and R.C. Bales, 2006: A Comparison of Snow Telemetry and Snow
Course Measurements in the Colorado River Basin. J. Hydrometeor., 7, 705–712.
Ducrocq, V., O. Nuissier, D. Ricard, C. Lebeaupin, and T. Thouvenin, 2008: A numerical study
of three catastrophic precipitating events over southern France. II: Mesoscale triggering and
stationarity factors. Quart. J. Roy. Meteor. Soc., 134, 131–145.
Dunn, L. B., and S. V. Vasiloff, 2001: Tornadogenesis and operational considerations of the
11 August 1999 Salt Lake City Tornado as seen from two different doppler radars. Wea.
Dunn, L. B., 2003: Modern weather forecasting. Handbook of Weather, Climate, and Water, T. D.
Potter and B. R. Colman, Eds., Wiley-Interscience, 677–688.
Englin, J., and K. Moeltner. 2004. The Value of Snowfall to Skiers and Boarders. Environmental
and Resource Economics,29,123–136.
Ferber, G. K., C. F. Mass, G. M. Lackmann, and M. W. Patnoe, 1993: Snowstorms over the Puget
Sound lowlands. Wea. Forecasting, 8, 481–504.
Forbes, G.S., R.A. Anthes, and D.W. Thomson, 1987: Synoptic and Mesoscale Aspects of an
Appalachian Ice Storm Associated with Cold-Air Damming. Mon. Wea. Rev., 115, 564–591.
Fettig, C.J., K.D. Klepzig, R.F. Billings, A.S. Munson, T.E. Nebeker, J.F. Negron, and J.T. Nowak.
2007: The Effectiveness of Vegetation Management Practices for Prevention and Control of
Bark Beetle Infestations in Coniferous Forests of the Western and Southern United States.
Forest Ecology and Management, 238, 24–53.
Frei, C., and C. Schär, 1998: A precipitation climatology of the Alps from high-resolution rain-
gauge observations. Int. J. Climatol., 18, 873–900.
Fujita, T.T., 1989: The Teton–Yellowstone tornado of 21 July 1987. Mon. Wea. Rev., 117,
1913–1940.
Geiger, R., R. H. Aron, and P. Todhunter: 1995, The Climate Near the Ground, Vieweg Publishing,
528 pp.
Germann, U. and J. Joss, 2004: Operational measurement of precipitation in mountainous terrain.
In P.Meischner, editor, Weather Radar: Principles and Advanced Applications, vol XVII of
Physics of Earth and Space Environment, chap. 2, 52–77, Springer Verlag.
Gladwell, M., 2005: Blink: The Power of Thinking without Thinking. Little, Brown and Company,
288 pp.
Glahn, H.R., and D.A. Lowry, 1972: The Use of Model Output Statistics (MOS) in Objective
Weather Forecasting. J. Appl. Meteor., 11, 1203–1211.
Graybeal, D.Y. and D.J. Leathers, 2006: Snowmelt-related flood risk in Appalachia: First estimates
from a historical snow climatology. J. Appl. Meteor. Climatol., 45, 178–193.
Hannesen, R., N. Dotzek, and J. Handwerker, 2000: Radar analysis of a tornado over hilly terrain
on 23 July 1996. Phys. Chem. Earth B 25, 1079–1084.
Hart, K. A., W. J. Steenburgh, and D. J. Onton, 2005: Model forecast improvements with decreased
horizontal grid spacing over fine-scale Intermountain orography during the 2002 Olympic
Winter Games. Wea. Forecasting, 20, 558–576.
Holle, R.L, and R.E. Lopez, 1993: Overview of Real-Time Lighting Detection Systems and Their
Meteorological Uses, NOAA Tech. Memo. ERL NSSL-102, 68 pp.
Horel, John D., Staley, Lloyd R., Barker, Timothy W. 1988: The University of Utah Interactive
Dynamics Program-One Approach to Interactive Access and Storage of Meteorological Data.
Bull. Amer. Meteor. Soc., 69, 1321–1327.
Horel, J., M. Splitt, L. Dunn, J. Pechmann, B. White, C. Ciliberti, S. Lazarus, J. Slemmer, D. Zaff,
J. Burks, 2002: MesoWest: Cooperative Mesonets in the Western United States. Bull. Amer.
Meteor. Soc., 83, 211–226.
Houze, R., W. Schmid, R. Fovell, and H.H. Schiesser, 1993: Hailstorms in Switzerland: Left
Movers, Right Movers, and False Hooks. Mon. Wea. Rev., 121, 3345–3370.
Howitt, R. E., D. MacEwan, and J. Medellin-Azuara, 2009: Economic impacts of reductions
in delta exports on Central Valley agriculture. Agriculture and Resource Economics update.
Giannini Foundation of Agricultural Economics, University of California. 4 pp.
Hudson, S., 2002: Sport and adventure tourism. Binghamton, NY, Haworth Press. 324 pp.
Huet, P., X. Martin, J.-L. Prime, P. Foin, C. Laurain, and P. Cannard, 2003: Retour d’expérience
des crues de septembre 2002 dans les departments du Gard, de l’Hèrault, du Vaucluse, des
Bouches-du-Rhône, de l’Ardèche et de la Drôme. Technical report, Ministère del’Ecologie et
du Développement Durable, Republique Franc¸aise, 133 pp.
Iverson, R.M., 1997: The physics of debris flows. Reviews of Geophysics, 35, 245–296.
Jóhannesson, T., and Þ. Arnalds, 2001: Accidents and economic damage due to snow avalanches
and landslides in Iceland. Jökull, 50, 81–94.
Johnson, G. L., C. Daly, G.H. Taylor, and C.L. Hanson, 2000: Spatial Variability and Interpolation
of Stochastic Weather Simulation Model Parameters. J. Appl. Meteor., 39, 778–796.
Kattelmann, R., 1997: Flooding from rain-on-snow events in the Sierra Nevada. Destructive Water:
Water-Caused Natural Disasters, Their Abatement and Control, G. H. Leavesley et al., Eds.,
IAHS Publication 239, 59–65.
Klein, G., 1998: Sources of Power: How People Make Decisions. The MIT Press, 330 pp.
Keiler, M., A. Zischg, S. Fuchs, M. Hama, and J. Stötter, 2005: Avalanche related damage
potential – changes of persons and mobile values since the mid-twentieth century, case study
Galtür. Natural Hazards and Earth System Sciences, 5, 49–58.
Knutson, C. 2001. “A comparison of droughts, floods and hurricanes in the U.S”. National Drought
Mitigation Center, Lincoln, NE, USA.
Kohler, J., 2003: Colorado, Wyoming recovering from worst blizzard in 90 years. Associated Press.
21 March 2003.
Lanicci, J.M., and T. T. Warner, 1991: A Synoptic Climatology of the Elevated Mixed-Layer
Inversion over the Southern Great Plains in Spring. Part I: Structure, Dynamics, and Seasonal
Evolution. Wea. Forecasting, 6, 198–213.
Linder, W., and W. Schmid, 1996: A tornadic thunderstorm in Switzerland exhibiting a radar-
detectable low-level vortex. Preprints, 12th Conf. on Clouds and Precipitation, Zurich,
Switzerland, International Commission on Cloud Physics and WMO, 577–580.
Linnerooth-Bayer, J. and A. Amendola, 2003: Introduction to special issue on flood risks in Europe.
Risk Analysis, 23, 537–543.
Lott, N., and T. Ross, 2006: Tracking and evaluating U.S. billion dollar weather disasters,
1980–2005. Preprints, AMS Forum: Environmental Risk and Impacts on Society: Successes
and Challenges. Atlanta, GA, Amer. Meteor. Soc.
Ludwig, F. L., D. K. Miller and S. G. Gallaher, 2006: Evaluating a hybrid prognostic/diagnostic

model that improves wind forecast resolution in complex coastal topography. J. Appl. Meteorol.
and Climatol., 45, 155–177.
Lyon, B., 2003: Enhanced Seasonal Rainfall in Northern Venezuela and the Extreme Events of
December 1999. J. Climate, 16, 2302–2306.
Maddox, R.A., L.R. Hoxit, C.F. Chappell, and F. Caracena, 1978: Comparison of Meteorological
Aspects of the Big Thompson and Rapid City Flash Floods. Mon. Wea. Rev., 106, 375–389.
Maddox, R. A., J. Z. Zhang, J. J. Gourley, and K. W. Howard, 2002: Weather radar coverage over
the contiguous United States. Wea. Forecasting, 17, 927–934.
Malek, E., T. Davis, R. S. Martin, and P. J. Silva, 2006: Meteorological and environmental aspects
of one of the worst national air pollution episodes (January 2004) in Logan, Cache Valley, Utah,
USA. Atmos. Res., 79, 108–122.
Matthews, W.J. and E. Marsh-Matthews. 2003. Effects of drought on fish across axes of space,
time, and ecological complexity. Freshwater Biology, 48, 1232–1253.
McMurdie, L. and C. Mass, 2004: Major numerical forecast failures over the northeast Pacific.
Medina, S., and R. A. Jr. Houze, 2003: Air motions and precipitation growth in Alpine storms.
Q. J. R. Meteorol. Soc., 129, 345–372.
Meybeck, M., P. Green, and C. Vörösmarty, 2001: A new typology for mountains and other
relief classes: an application to global continental water resources and population distribution.
Mountain Research and Development 21, 34–45.
Meyers, M.P., J.S. Snook, D.A. Wesley, and G.S. Poulos, 2003: A Rocky Mountain storm. Part II:
The forest blowdown over the West Slope of the northern Colorado mountains – Observations,
analysis, and modeling. Wea. Forecasting, 18, 662–674.
Meyers, M. P., J. Ladue, J. Pringle, and C. Cuoco, 2006: A Forecaster challenge: convectively
enhanced winds in a strongly forced synoptic environment in complex terrain, 12th Conference
on Mountain Meteorology, Santa Fe, NM, AMS.
Mielke, K.B., and E.R. Carle, 1987: An Early Morning Dry Microburst in the Great Basin. Wea.
Morris, D. and M. Walls, 2009: Climate change and outdoor recreation resources. Washington,
DC, Resources for the future.
Morss, R. E., O. V. Wilhelmi, M. W. Downton, and E. Gruntfest, 2005: Flood risk, uncertainty,
and scientific information for decision making: Lessons from an interdisciplinary project. Bull.
Amer. Meteor. Soc., 86, 1593–1601.
Morss, R.E., and F.M. Ralph, 2007: Use of Information by National Weather Service Fore-
casters and Emergency Managers during CALJET and PACJET-2001. Wea. Forecasting, 22,
539–555.
Myrick, D.T., and J.D. Horel, 2008: Sensitivity of Surface Analyses over the Western United States
to RAWS Observations. Wea. Forecasting, 23, 145–158.
Nance, L. B., and B. R. Colman, 2000: Evaluating the use of a nonlinear two-dimensional model
in downslope wind forecasts. Wea. Forecasting, 15, 715–729.
National Park Service. 2009. NPS Stats. National Park Service Public Use Statistics Office. http://
www.nature.nps.gov/stats/park.cfm.
National Research Council, 2004: Where the Weather Meets the Road: A Research Agenda for
Improving Road Weather Services. National Academies Press, Washington, D. C., 188 pp.
National Research Council, 2005: Flash Flood Forecasting over Complex Terrain. National
Academies Press, Washington, D. C., 206 pp.
National Wildfire Coordinating Group. 1997: Historical wildland firefighter fatalities: 1910–1996.
NFES 1849. Boise, ID: National Interagency Fire Center. 29 pp.
National Wildfire Coordinating Group, 2007: Wildland Firefighter Fatalities in the United States:
1990–2006. Boise, ID, National Interagency Fire Center. 30 pp.
Nöthiger, C., and H. Elsasser, 2004: Natural hazards and tourism: New findings on the European
Alps. Mountain Research and Development, 24, 24–27. Publisher: International Mountain
Society.
Novak, D.R., D.R. Bright, and M.J. Brennan, 2008: Operational Forecaster Uncertainty Needs and
Future Roles. Wea. Forecasting, 23, 1069–1084.
Nuissier, O., V. Ducrocq, D. Ricard, C. Lebeaupin, and S. Anquetin, 2008: A numerical study
of three catastrophic precipitating events over southern France. I: Numerical framework and
synoptic ingredients. Quart. J. Roy. Meteor. Soc., 134, 111–130.
Parish, T. R., and D. H. Bromwich, 1998: A case study of Antarctic katabatic wind interaction with
large-scale forcing. Mon. Wea. Rev., 126, 199–209.
Pliske, R., B. Crandall, and G. Klein, 2004: Competence in weather forecasting. Psychological
Investigations of Competent Decision Making. J. Schanteau, P. Johnson, and K. Smith, Eds.,
Cambridge University Press.
Poulos, G. S., D. A. Wesley, J. S. Snook, M. P. Meyers, 2002: A Rocky Mountain Storm. Part I:
The Blizzard – kinematic evolution and the potential for high-resolution numerical forecasting
of snowfall. Wea. Forecasting, 17, 955–970.
Ralph, F. M., P. J. Neiman, D. E. Kingsmill, P. O. G. Persson, A. B. White, E. T. Strem, E. D.
Andrews, R. C. Antweiler, 2003: The impact of a prominent rain shadow on flooding in
California’s Santa Cruz Mountains: A CALJET case study and sensitivity to the ENSO cycle.
Journal of Hydrometeorology, 4, 1243–1264.
Reagan , M., 1998: Canadian ice storm 1998. World Meteorological Organization Bulletin, Vol. 47,
No. 3, 250–256.
Reinecke P.A., D. R. Durran, 2009: Initial-condition sensitivities and the predictability of
downslope winds. J. Atmos. Sci., 66, 3401–3418.
Restrepo, P., D.P. Jorgensen, S.H. Cannon, J. Costa, J. Laber, J. Major, B. Martner, J. Purpura,
and K. Werner, 2008: Joint NOAA/NWS/USGS Prototype Debris Flow Warning System for
Recently Burned Areas in Southern California. Bull. Amer. Meteor. Soc., 89, 1845–1851.
Robbins, J., 2008: Bark Beetles Kill Millions of Acres of Trees in West. New York Times, November
18, D3.
Roebber, P.J., L.F. Bosart, and G.S. Forbes, 1996: Does Distance from the Forecast Site Affect
Skill? Wea. Forecasting, 11, 582–589.
Roebber, P. J., and J. R. Gyakum, 2003: Orographic influences on the mesoscale structure of the
1998 ice storm. Mon. Wea. Rev., 131, 27–50.
Roebber, P. J., D. M. Schultz, B. A. Colle, and D. J. Stensrud, 2004: Toward improved prediction:
High-resolution and ensemble modeling in operations. Wea. Forecasting, 19, 936–949.
Rotunno, R., Houze, R. A., 2007: Lessons on orographic precipitation from the Mesoscale Alpine
Programme. Quarterly Journal of the Royal Meteorological Society, 133, 811–830.
Schnell, R. C., S. J. Oltmans, R. R. Neely, M. S. Endres, J. V. Molenar & A. B. White, 2009: Rapid
photochemical production of ozone at high concentrations in a rural site during winter. Nature
Geoscience, 2(2), 120–122.
Schwarb, M., 2000: The Alpine precipitation climate: Evaluation of a high-resolution analysis
scheme using comprehensive rain-gauge data. Dissertation, Swiss Federal Institute of Technol-
ogy Zurich, 131 pp.
Schwarb, M., C. Daly, C. Frey, and C. Schär, 2001: Mean annual precipitation throughout the
European Alps 1971–1990. Hydrological Atlas of Switzerland, plate 2.6.
Scott, D. 2006. Ski Industry Adaptation to Climate Change. In: Tourism and Global Environmental
Change. S. Gossling and M. Hall (eds). London: Routledge. 262–285.
Sénési, S., P. Bougeault, J.-L. Chèze, P. Cosentino, and R.-M. Thepenier, 1996: The Vaison-
La Romaine Flash Flood: Mesoscale analysis and predictability issues. Wea. Forecasting, 4,
417–442.
Sivillo, J.K., J.E. Ahlquist, and Z. Toth, 1997: An Ensemble Forecasting Primer. Wea. Forecasting,
12, 809–818.
Skistar, 2009. The Alpine world market. http://corporate.skistar.com.
Smith, S. B. and M. K. Yau, 1987: The mesoscale effect of topography on the genesis of Alberta
hailstorms. Beitr. Phys, Atmos., 60, 371–392.
Snellman, L.W., 1977: Operational Forecasting Using Automated Guidance. Bull. Amer. Meteor.
Soc., 58, 1036–1044.
Snellman, L. W., 1982: Impact of AFOS on Operational Forecasting, Reprints, 9th Conference on
Weather Forecasting and Analysis, AMS, 13–16.
Soyka, F., 1983: The ion effect. New York, NY. Bantam Books.
Steiger, R. and M. Mayer, 2008: Snowmaking and climate change: Future options for snow
production in Tyrolean ski resorts. Mountain Research and Development, 28, 292–298.
Steenburgh, W. J., 2002: Using real-time mesoscale modeling in undergraduate education. Bull.
Amer. Meteor. Soc., 83, 1447–1451.
Steenburgh, W. J., 2003: One hundred inches in one hundred hours: Evolution of a Wasatch
Mountain winter storm cycle. Wea. Forecasting, 18, 1018–1036.
Steenburgh W. J., C. F. Mass, and S. A. Ferguson, 1997: The influence of terrain-induced
circulations on wintertime temperature and snow level in the Washington Cascades. Wea.
Steenburgh, W.J., and T.I. Alcott, 2008: Secrets of the “Greatest Snow on Earth”. Bull. Amer.
Meteor. Soc., 89, 1285–1293.
Stuart, N.A., D.M. Schultz, and G. Klein, 2007: Maintaining the Role of Humans in the Forecast
Process: Analyzing the Psyche of Expert Forecasters. Bull. Amer. Meteor. Soc., 88, 1893–1898.
Thornes, J.E., 2000: Road Salting – an international benefit/cost review. Proceedings of the 8th
World Salt Symposium, 787–792, Elsevier, Amsterdam.
Vasiloff, S. K., J. Busto, M. P. Meyers, D. J. Gochis, and K. Friedrich, 2011: Flood hazard
monitoring in Colorado mountains with mobile Doppler radars [INVITED]. 25th Conference
on Hydrology. Seattle, WA. AMS.
Wagner, W. and J. D. Horel, 2008: Observing and forecasting snow surface temperatures for Nordic
ski race courses at the 2010 Winter Olympic games in Vancouver, BC. Proceedings for the
International Snow Science Workshop, Whistler, BC.
Wakimoto, R. M., 1985: Forecasting dry microburst activity over the high plains. Mon. Wea. Rev.,
113, 1131–1143.
Westrick, K. L., C. F. Mass, and B. A. Colle, 1999: The limitations of the WSR-88D radar network
for quanititive precipitation measurement over the coastal western United States. Bull. Amer.
Meteor. Soc., 80, 2289–2298.
Wetzel, M., M. Meyers, R. Borys, R. McAnelly, W. Cotton, A. Rossi, P. Frisbie, D.Nadler, D.
Lowenthal, S. Cohn, and W. Brown, 2004: Mesoscale snowfall prediction and verification in
mountainous terrain. Wea. Forecasting, 19, 806–828.
Whiteman, C. D., 2000: Mountain meteorology: Fundamentals and applications. Oxford University
Press, New York and Oxford. 355 pp.
Wojtiw, L., 1975: Climate summaries of hailfall in central Alberta (1957–73). Alberta Research
Council, 102 pp. [Available from Alberta Research Council, Edmonton, Alberta, T6H 5X2.]
Wood, T. V., R. A. Brown, and S. V. Vasiloff, 2003: Improved detection using negative elevation
angles for mountaintop WSR-88Ds. Part II: Simulations of the three radars covering Utah. Wea.
World Health Organization, 2002: The world health report: Reducing risks, promoting healthy life.
Geneva, Switzerland. World Health Organization.
Yuan, H., S.L. Mullen, X. Gao, S. Sorooshian, J. Du, and H.M.H. Juang, 2007: Short-Range
Probabilistic Quantitative Precipitation Forecasts over the Southwest United States by the RSM
Ensemble System. Mon. Wea. Rev., 135, 1685–1698.
Chapter 2
Diurnal Mountain Wind Systems
Dino Zardi and C. David Whiteman
Abstract Diurnal mountain wind systems are local thermally driven wind systems
that form over mountainous terrain and are produced by the buoyancy effects
associated with the diurnal cycle of heating and cooling of the lower atmospheric
layers. This chapter reviews the present scientific understanding of diurnal mountain
wind systems, focusing on research findings published since 1988. Slope flows
are examined first, as they provide a good introduction to the many factors
affecting diurnal mountain wind systems. The energy budgets governing slope
flows; the effects of turbulence, slope angle, ambient stability, background flows
and slope inhomogeneities on slope flows; and the methods used to simulate slope
flows are examined. Then, valley winds are reviewed in a similar manner and the
diurnal phases of valley and slope winds and their interactions are summarized.
Recent research on large-scale mountain-plain wind systems is reviewed, with an
emphasis on the Rocky Mountains and the Alps. Winds occurring in closed basins
and over plateaus are then discussed, and analogies between the two wind systems
are outlined. This is followed by a discussion of forecasting considerations for
diurnal mountain wind systems. Finally, the chapter concludes with a summary of
open questions and productive areas for further research.
D. Zardi ()
Atmospheric Physics Group, Department of Civil and Environmental Engineering,
University of Trento, Via Mesiano, 77, Trento I-38123, Italy
e-mail: dino.zardi@ing.unitn.it
C.D. Whiteman
Atmospheric Sciences Department, University of Utah, Salt Lake City, UT, USA

36 D. Zardi and C.D. Whiteman
2.1 Introduction
Diurnal mountain winds develop, typically under fair weather conditions, over
complex topography of all scales, from small hills to large mountain massifs, and
are characterized by a reversal of wind direction twice per day. As a rule, upslope,
up-valley and plain-to-mountain flows occur during daytime and downslope, down-
valley and mountain-to-plain flows occur during nighttime.
Because diurnal mountain winds are produced by heating of atmospheric layers
during daytime and cooling during nighttime, they are also called thermally driven
winds. Horizontal temperature differences develop daily in complex terrain when
cold or warm boundary layers form immediately above sloping surfaces. Air at
a given point in the sloping boundary layer is colder or warmer than the air just
outside the boundary layer at the same elevation. These temperature differences
result in pressure differences and thus winds that blow from areas with lower
temperatures and higher pressures toward areas with higher temperatures and lower
pressures. The boundary layer flows and return, or compensation, flows higher in
the atmosphere form closed circulations.
The atmosphere above and around a mountain massif is composed of three
distinct regions or sloping layers in which the thermal structure undergoes diurnal
variations and in which diurnal winds develop: the slope atmosphere, the valley
atmosphere and the mountain atmosphere (Ekhart 1948; Fig. 2.1). The slope
atmosphere is the domain of the slope flows, the valley atmosphere is the domain
of the valley flows and the mountain atmosphere is the domain of the mountain-
plain flows. It is difficult to observe any one component of the diurnal mountain
wind system in its pure form, because each interacts with the others and each can
be affected by larger scale flows aloft.
Indeed well-organized thermally driven flows can be identified over a broad spec-
trum of spatial scales, ranging from the dimension of the largest mountain chains
to the smallest local landforms, such as, for instance, “farm-scale” topographic
features (Dixit and Chen 2011; Bodine et al. 2009). All of these flows are induced
by local factors, and in particular by the combination of landforms and surface
energy budgets, under synoptic scale situations allowing for their development, and
eventually conditioning them “top-down” by acting as boundary conditions. On the
other hand the same flows may in turn affect large-scale phenomena. This is the case,
for instance, of diurnal flows over sloping terrain forcing the Great Plains low-level
jet (cf. Holton 1966; Parish and Oolman 2010) and of some organized convective
systems triggered by thermally induced flows in the southern part of the Himalayas
(Egger 1987; Yang et al. 2004; Bhatt and Nakamura 2006; Liu et al. 2009).
The regular evolution of the diurnal mountain wind systems exhibits four
phases that are closely connected to the formation and dissipation of temperature
inversions. The nighttime phase, characterized by the presence of a ground-based
temperature inversion, or stable layer, and winds flowing down the terrain, is
followed by a morning transition phase in which the inversion is destroyed and the
2 Diurnal Mountain Wind Systems 37
Fig. 2.1 Diagram of the structure of the atmosphere above a mountain ridge (Adapted from Ekhart
1948, © Société Météorologique de France. Used with permission)
winds undergo a reversal to the daytime winds flowing up the terrain. The evening
transition phase completes the cycle as the inversion rebuilds and the winds once
again reverse to the nighttime direction.
Diurnal mountain winds are a key feature of the climatology of mountainous
regions (Whiteman 1990, 2000; Sturman et al. 1999). In fact, they are so consistent
that they often appear prominently in long-term climatic averages (Martı́nez et al.
2008). They are particularly prevalent in anticyclonic synoptic weather conditions
where background winds are weak and skies are clear, allowing maximum incoming
solar radiation during daytime and maximum outgoing longwave emission from
the ground during nighttime. Diurnal wind systems are usually better developed in
summer than in winter, because of the stronger day-night heating contrasts. The
local wind field patterns and their timing are usually quite similar from day to day
under anticyclonic weather conditions (see, e.g., Guardans and Palomino 1995). The
characteristic diurnal reversal of the slope, valley, and mountain-plain wind systems
is also seen under partly cloudy or wind-disturbed conditions, though it is sometimes
weakened by the reduced energy input or modified by winds aloft.
The speed, depth, duration and onset times of diurnal wind systems vary from
place to place, depending on many factors including terrain characteristics, ground
cover, soil moisture, exposure to insolation, local shading and surface energy budget
(Zängl 2004). Many of these factors have a strong seasonal dependence. Indeed, the
amplitude of horizontal pressure gradients driving valley winds displays appreciable
seasonal variations (see e.g. Cogliati and Mazzeo 2005).
The scientific study of diurnal mountain winds has important practical

applications, ranging from operational weather forecasting in the mountains, to the
climatological characterization of mountain areas (e.g. for agricultural purposes),
to the impact assessments of proposed new settlements or infrastructures. Indeed
diurnal winds affect the distribution of air temperature in complex terrain (Lindkvist
and Lindqvist 1997; Gustavsson et al. 1998; Lindkvist et al. 2000; Mahrt 2006),
the transport and diffusion of smoke and other air pollutants, the formation and
dissipation of fogs (Cuxart and Jiménez 2011) and low clouds, frost damage in
vineyards and orchards, surface transportation safety from risks due to frost and
fog, flight assistance (e.g. through the issuance of Terminal Aerodrome Forecasts
and the evaluation of flying conditions for soaring or motorized flights), long-term
weathering of structures, and the behavior of wildland and prescribed fires.
Diurnal mountain winds have been studied scientifically since the nineteenth
century. Wagner (1938) laid the foundation for our current understanding of diurnal
mountain wind systems. He provided a systematic overview and synthesis of the
findings from field measurements and theoretical investigations gained at that time.
Indeed much of the early research was published in German or French. This
is evident in the first comprehensive English-language review of early research
provided by Hawkes (1947). Whiteman and Dreiseitl (1984) published English
translations of seminal papers by Wagner, Ekhart and F. Defant. Additional reviews
of research on diurnal mountain wind systems include those of Defant (1951),
Vergeiner and Dreiseitl (1987), Whiteman (1990), Egger (1990), Simpson (1999),
and Poulos and Zhong (2008). Additionally, chapters on diurnal mountain wind
systems are available in textbooks by Yoshino (1975), Atkinson (1981), Stull (1988),
Whiteman (2000), Barry (2008), and Geiger et al. (2009).
A fundamental obstacle to rapid progress in mountain meteorology is that
there are almost infinitely many possible terrain configurations. So, any field
measurement or numerical experiment that is valid for a specific situation does not
automatically have greater significance beyond that case. Nonetheless, significant
progress has been achieved by gathering real cases into groups according to similar
landforms, and generalizing results from these groups with respect to the inherent
important physical processes or mechanisms. The organization of this chapter
benefits from this approach, with a special section on basin and plateau circulations.
Another useful approach has been to focus conceptual or numerical simulations
on simple or idealized topographies to investigate the essential meteorological
processes. Other obstacles to progress should also be mentioned. Because most
studies of diurnal mountain wind systems have been conducted on fine weather days
with weak synoptic forcing, relatively little is known about other conditions. Further,
the effects of forest canopies and other types of varying and patchy surface cover
on diurnal wind systems have been largely neglected in much of the experimental
and modeling work to date. Finally, while diurnal mountain wind systems are
thought to be complete or closed circulations, little experimental evidence has been
accumulated on the difficult-to-measure “return” or upper-branch circulations.
This review aims to provide information on the basic physics of diurnal
mountain wind systems and to summarize new research findings since the American
Meteorological Society’s 1988 workshop on mountain meteorology, as later

published in an AMS book (Blumen 1990). Priority in the referenced material
is given to peer-reviewed papers published in English since 1988. Other chapters
in the present book also deal peripherally with diurnal mountain wind systems.
Chapters 3 and 4 discuss the effects on valley winds of dynamically induced
channeling from larger scale flows. Chapter 5 treats atmospheric boundary layer
phenomena in mountain areas that affect pollutant transport. Chapters 6 and 7 deal
with orographic precipitation, including triggering from thermally forced flows.
Chapter 8 treats observational resources and strategies for mountain atmosphere
processes, including diurnal mountain winds. Chapters 9, 10, and 11 discuss
numerical model simulations and operational weather forecasting.
2.2 The Slope Wind System
The slope wind system is a diurnal thermally driven wind system that blows up
or down the slopes of a valley sidewall or an isolated hill or mountain, with
upslope flows during daytime and downslope flows during nighttime. The up- and
downslope flows are the lower branch of a closed circulation produced by inclined
cold or warm boundary layers that form above the slopes. During daytime, the
heated air in the boundary layer above a slope rises up the slope while continuing to
gain heat from the underlying surface. During nighttime, air in the cooled boundary
layer flows down the slope while continuing to lose heat to the underlying surface.
Upslope and downslope winds are alternatively called anabatic and katabatic winds,
respectively. These terms, however, are also used more comprehensively to refer to
any flows that run up or down the terrain. Further, the term katabatic is typically used
to describe the large-scale slope flows on the ice domes of Antarctica and Greenland.
The term drainage wind may be used to refer collectively to downslope and down-
valley winds, and may also include mountain-to-plain flows. To reduce confusion,
we will use the terms upslope and downslope, and will omit from consideration
the flows over extensive ice surfaces which, in contrast to the smaller-scale mid-
latitude flows discussed here, are affected by Coriolis forces (see, e.g., Kavčič and
Grisogono 2007; Shapiro and Fedorovich 2008). An example of typical nighttime
and daytime wind and temperature profiles through the slope flow layer is given
in Fig. 2.2. The biggest daytime temperature excesses and nighttime temperature
deficits occur near the ground, where radiative processes heat and cool the surface.
It is here that the buoyancy forces that drive the slope flows are strongest. The wind
profiles, however, have their peak speeds above the surface, since surface friction
causes the speeds to decrease to zero at the ground. This gives both the daytime and
nighttime wind speed profiles their characteristic jet-like shape in which the peak
wind speed occurs above the surface.
Typical characteristics of the equilibrium steady-state wind and temperature
profiles for mid-latitude downslope winds (Fig. 2.2a) include the strength (3–7ı C)
and depth (1–20 m) of the temperature deficit layer, the maximum wind speed
Fig. 2.2 Typical wind and a

temperature profiles through
(a) downslope and (b)
upslope flows on valley
sidewalls. Shown are the
typical depths and strengths
of the temperature deficits or
excesses, the heights and flow
strengths of the wind
maxima, and the slope flow
depths. The TKE profile in
(a) follows that of
Skyllingstad (2003) (Adapted
from Whiteman 2000)
(1–4 m s1 ), its height above ground (1–15 m), and the depth of the downslope
flow layer (3–100 m). Downslope flows on uniform slopes often increase in both
depth and strength with distance down the slope. A useful guideline for field studies
is that the depth of the temperature deficit layer at a point on a slope is about 5% of
the vertical drop from the top of the slope (Horst and Doran 1986). Characteristic
values for upslope flows (Fig. 2.2b) include the strength (2–7ıC) and depth (10–
100 m) of the temperature excess layer, the maximum wind speed (1–5 m s1 ),
its height above ground (10–50 m), and the depth of the upslope flow layer (20–
200 m). Upslope flows on uniform slopes often increase in both depth and speed
with distance up the slope.
Not covered in this review are mid-latitude “skin flows” (Manins and Sawford
1979a) – shallow (<1 or 2 m), weak (perhaps <1 m s1 ), non-turbulent katabatic
flows that have been reported by several investigators (Thompson 1986; Manins
1992; Mahrt et al. 2001; Soler et al. 2002; Clements et al. 2003). These have not yet
been studied systematically.
Our discussion of slope flows begins with a simplified set of Reynolds-
averaged Navier–Stokes (RANS) equations to guide understanding. Then different
approaches to modeling slope flows are discussed and the state of scientific
knowledge on upslope and downslope flows, as supported by both field and
modeling investigations, are summarized.
2.2.1 Slope Flow Models
2.2.1.1 Reynolds-Averaged Navier–Stokes Equations
A reduced set of RANS equations can be used heuristically to gain an understanding

of the key physics of slope flows. This is conveniently done using a coordinate
system that is oriented along (s) and perpendicular (n) to an infinitely extended
slope of uniform inclination angle ˛. n is perpendicular to the slope and increases
upwards; s is parallel to the slope and increases in the downslope direction. Slope
flows result from the cooling or heating of an air layer over the slope relative to air
at the same elevation away from the slope. The potential temperature profile over
the slope is expressed as
D 0 C z C d.s; n; t/ (2.1)
where 0 is a reference value for the potential temperature at the ground in the
absence of surface cooling or heating, is the ambient vertical gradient of potential
temperature, z is the vertical coordinate (z D n cos ˛ s sin ˛), and d is the
potential temperature perturbation in the vicinity of the slope, negative when the air
is colder than in the unperturbed state. The along-slope and slope perpendicular
equations of motion, the thermodynamic energy equation, and the continuity
equation are then written as
@u @u @u 1 @.p pa / d @u0 w0
Cu Cw D g sin ˛ (2.2)
@t @s @n 0 @s 0 @n
@w @w @w 1 @.p pa / d
Cu Cw D C g cos ˛ (2.3)
@t @s @n 0 @n 0
@ @ @ 1 @R @w0 0
Cu Cw D (2.4)
@t @s @n 0 cp @n @n
and
@u @w
C D0 (2.5)
@s @n
where u and w are the velocity components parallel to s and n, respectively. These
equations, from Horst and Doran (1986), were derived by Manins and Sawford
(1979b) using the Boussinesq approximation. They considered the flow to be two-
dimensional, with wind and temperature independent of the cross-slope direction,
with the ambient air at rest, and with negligible Coriolis force. 0 is a reference value
of density in the absence of surface cooling, g is the gravitational acceleration, cp
is the specific heat of air at constant pressure, R is the upward radiative flux, and
u0 w0 and w0 0 are the kinematic turbulent Reynolds fluxes of momentum and heat.
Note that, consistent with the assumption of a shallow layer, this set of reduced
equations includes only the slope-normal turbulent fluxes in (2.2) and (2.4); other
turbulent fluxes are ignored. For simplicity, latent heat fluxes are not included here,
although they can sometimes play a crucial role in the overall energy budgets. The
ambient pressure pa is determined from hydrostatic balance @pa =@z D a g, and,
through the ideal gas law and the definition of potential temperature, is a function of
the ambient potential temperature distribution a D 0 C z. Accelerations normal
to the slope are generally considered negligible so that the atmosphere normal to
the slope is in quasi-hydrostatic balance and the RHS of (2.3) is practically zero.
The coordinate system rotation needed to get to this set of equations can lead to
interpretation errors if not done properly, as explained by Haiden (2003). Mahrt
(1982) provided further details on the coordinate system and an overview of the
classification of gravity flows in terms of the relative magnitudes of the individual
terms of the along-slope momentum equation (2.2).
This system of equations can be solved numerically after the turbulent fluxes
are parameterized in terms of mean flow variables. The parameterization of the
momentum flux normal to the slope is usually expressed in terms of the wind
shear and turbulence length and velocity scales. This type of parameterization was
originally developed using knowledge of turbulence profiles over flat ground in
flows that were driven by large-scale pressure gradients that changed little with
height outside the surface layer. In contrast, slope flows are shallow flows in which
the largest forcing is at the ground. The suitability of such parameterizations is thus
called into question (Grisogono et al. 2007).
It is instructive to consider which terms in these equations are most important
for understanding slope flow physics. The key term that drives slope flows is the
buoyancy force caused by the temperature excess or deficit that forms over the slope
(second term on the right hand side of (2.2)). The temperature excess or deficit itself
is produced by radiative and sensible heat flux convergences or divergences (the first
and second terms on the RHS of (2.4)) that warm or heat a layer of air above the
slope. Considering the along-slope momentum equation (2.2) for the downslope
flow situation, a steady state flow may be retarded by advection of slower moving
air into the flow (two advection terms on the LHS), by the thickening or cooling of
the layer with downslope distance which causes an adverse pressure gradient (first
term on the RHS, see Princevac et al. 2008), and by the divergence of turbulent
momentum flux (last term on the RHS).
Mesoscale numerical models that use the full set of RANS equations (see
Chaps. 9 and 10) have been used to simulate a variety of meteorological phenomena.
These models, where turbulence is parameterized with closure assumptions, have
been applied to complex terrain and have proven very useful in gaining an under-
standing of diurnal mountain wind systems (see, e.g., Luhar and Rao 1993; Denby
1999). Extensive experience with mesoscale models of this type, however, has
shown that shallow slope flows cannot be resolved adequately in a mesoscale setting
for which the turbulence parameterizations are most appropriate. Simulated downs-
lope flows are often too deep and strong, temperature deficits over the slope are too
deep and weak, and it has been a challenge to get both the wind and temperature
fields to agree with observations (see Chap. 10). In instances where the wind and
temperature fields are in rough agreement with observations, these models have been
used to investigate the relative roles of individual terms in the momentum and heat
budget equations. Comparisons of slope flow simulations with multiple mesoscale
models have provided valuable information on the effects of different parameteriza-
tions, resolutions, coordinate systems, and boundary conditions on model outputs.
Zhong and Fast (2003) compared simulations of flows in Utah’s Salt Lake Valley
using the RAMS, MM5 and Meso-Eta models and compared the simulations to
slope flow data. Forecast errors with these models were surprisingly similar, despite
different coordinate systems, numerical algorithms and physical parameterizations.
The models exhibited a cold bias, with weaker than observed inversion strengths,
especially near the surface. All models successfully simulated the diurnally revers-
ing slope flows but there were significant discrepancies in depths and strengths of
wind circulations and temperature structure. Surprisingly, it is only recently that
mesoscale numerical models have added parameterizations of slope shadowing from
surrounding topography (Colette et al. 2003; Hauge and Hole 2003).
2.2.1.2 Large-Eddy Simulation (LES) Models
Large-eddy simulation (LES) of turbulent flows is a numerical alternative to the

RANS equations. It builds on the observation that most of the turbulence energy
is contained in the largest eddies which, as they break down, cascade their energy
into the smaller isotropic turbulent eddies where energy is dissipated. Accordingly,
the LES approach explicitly resolves the larger energy-containing eddies and uses
turbulence theory to parameterize the effects on the turbulent flow of the smallest
scale eddies in the inertial sub-range of atmospheric motions. Because energy-
containing eddies are much smaller in size in stable boundary layers than in
convective boundary layers, the first LES simulations for inclined slopes were made
for upslope flows (Schumann 1990). A more detailed account on these findings
will be given in Sect. 2.4 below. LES simulations of downslope flows in the
stable boundary layer, with its much smaller turbulence length scales, requires finer
grids and more extensive computational resources. Thus, numerical simulations of

downslope flows have only recently become possible on highly idealized slopes with
recent increases in computer power (Skyllingstad 2003; Axelsen and van Dop 2008,
2009). Research in this area looks very promising, but is still at an early stage. Some
applications of LES in complex terrain are provided in Chap. 10.
2.2.1.3 Analytical Models
Useful slope flow models can sometimes be developed by further simplifying the
reduced set of RANS equations (2.1) through (2.5), retaining only the key terms,
and solving the simplified set analytically. Such analytical models are of two types –
profile and hydraulic flow models.
Profile Models
Profile models consider local equilibria in the buoyancy and momentum equations
and obtain analytical solutions for equilibrium vertical profiles of velocity and
temperature in the slope flow layer by neglecting or simplifying some of the terms
in the equations. The classical Prandtl (1942) model is an equilibrium gravity
flow solution in which constant eddy diffusivities are used for the friction term
and a simple thermodynamic relationship balances the diffusion of heat with the
temperature advection associated with the basic state stratification. The Prandtl
model agreed qualitatively with observations of both upslope and downslope
flows, successfully producing the typical jet-like wind profiles and temperature
deficit profiles seen in slope flows (Fig. 2.3). Prandtl’s model was modified by
Grisogono and Oerlemans (2001a, b) to allow eddy diffusivities to vary with
height, improving its agreement with observations. Extension of Prandtl’s (1942)
approach to reproduce nonstationary flows was accomplished by Defant (1949) and
Grisogono (2003). Other investigators have developed alternative models of slope
flows by finding analytical solutions to reduced sets of equations (Zammett and
Fowler 2007; Shapiro and Fedorovich 2009; see also in Barry 2008).
Hydraulic Flow Models
Hydraulic models start from the RANS equations, average the terms over the
slope flow depth and treat the drainage flow as a single layer interacting with an
overlying stationary fluid. They focus on layer-averaged quantities such as mass
and momentum fluxes and can be used to determine how bulk quantities vary with
distance up or down the slope. Manins and Sawford (1979a, b), Brehm (1986), and
Horst and Doran (1986) were early developers of this approach for slope flows.
They, Papadopoulos et al. (1997), Papadopoulos and Helmis (1999), Haiden and
Whiteman (2005) and Martı́nez and Cuxart (2009) have used observations from
a b
Fig. 2.3 Observed (O) and theoretical (T) (a) downslope and (b) upslope wind profiles. Obser-
vations are from a slope on the Nordkette near Innsbruck, Austria. The theoretical curves come
from Prandtl’s (1942) model. Differences above the jet maximum between theory and observations
are indicated by the dotted lines (Adapted from Defant 1949, © Springer-Verlag Publishing Co.,
Vienna, Austria. Used with permission)
lines of instrumented towers or tethersondes running down a slope, or the output of

numerical model simulations, to estimate the magnitudes of the different terms in
the hydraulic flow equations.
2.2.1.4 Laboratory Tank Models
Laboratory water tanks have been used to investigate both up- and downslope
flows on sloping surfaces in basins containing stratified fluid. A bottom-heated,
salt-stratified water tank over a slope with adjacent plain and plateau was used by
Reuten et al. (2007) and Reuten (2008) to investigate layering, venting and trapping
of pollutants that are carried in upslope flows. Conditions conducive to trapping
include weak large-scale flows, strong sensible heat flux, weak stratification, a
short or no plateau, symmetric geometry, a low ridge, and inhomogeneities in
surface heat flux. Reuten et al. (2007) speculate that inhomogeneities in slope
angle and surface roughness can also produce trapping. In a similar laboratory
experiment by Princevac and Fernando (2008), but this time with a V-shaped tank
containing thermally stratified water that was heated uniformly on the sloping
surfaces, upslope flows were found to leave the sidewalls under certain conditions,
intruding horizontally into the stratified fluid over the valley center.
Upslope winds flow up the slope as a bent-over plume rather than rising
vertically, even in a neutral environment. These flows become turbulent when the
heat flux becomes large enough. Laboratory simulations by Princevac and Fernando
(2007) show that, for a certain range of Prandtl numbers,1 there is a minimum
slope angle above which upslope flows can be sustained. For average atmospheric
conditions the critical angle is only 0.1ı . Princevac and Fernando (2007) used this
result to explain Hunt et al.’s (2003) observations of a dominant upslope flow on a
0.18ı slope in the Phoenix valley.
Downslope flows into a stratified environment have been studied in laboratory
tanks, as well, by introducing a continuous source of negatively buoyant fluid at
the top of a constant-angle slope (Baines 2001). In this case, the flow maintains
a uniform thickness, with a distinct boundary at its top, until it approaches its
level of neutral buoyancy, where it leaves the slope. Turbulent transfers of mass
and momentum occur across the interface, causing a continuous loss of fluid
(detrainment) from the downslope flow. Flows of negatively buoyant fluid over steep
slopes, on the other hand, are in the form of entraining plumes (Baines 2005). It
should be mentioned that realistic downslope flows over valley sidewalls are not
produced by a continuous source of negatively buoyant fluid at the top of the slope.
Rather, they must lose heat continuously to the underlying surface to maintain their
negative buoyancy. Nonetheless, this laboratory tank analog provides important
information on interactions that occur at the top of the downslope flow layer.
2.2.2 Surface Radiative and Energy Budgets: The Driving

Force for Slope Flows
Thermally driven winds in complex terrain are produced by the formation of

inclined layers of temperature excess or deficit (relative to the ambient environment)
along terrain slopes of different scale. It is thus of paramount importance to
understand the physical processes that produce temperature changes above slopes.
The key principle is that of energy conservation, which relates local changes in
potential air temperature at any point to the budget of heat fluxes at that point. The
potential temperature tendency equation, neglecting thermal conduction and latent
heat release, can be written as
@
cp D r A r H r R (2.6)
@t T
1
The Prandtl number Pr is a nondimensional number defined, for any fluid, as the ratio of the
kinematic viscosity to the thermal diffusivity ›, i.e. Pr D /›. As an extension, for turbulent
flows the turbulent Prandtl number Prt is defined as the ratio of the eddy viscosity Km to the eddy
heat diffusivity Kh : Prt D Km /Kh . Both of these numbers provide, for laminar and turbulent flows
respectively, an estimate of the relative importance of convection versus diffusion in heat transfer
processes involved with the flow.
where A D cp u is the sensible heat flux associated with local advection of

warmer or colder air by the mean three-dimensional wind velocity u, H D cp u0 0
is the turbulent sensible heat flux produced by the coupling between turbulent
fluctuations of wind velocity and potential temperature, R is the radiative heat flux
(including both the shortwave and longwave spectra). In the absence of mean wind,
as for instance over flat terrain under anticyclonic synoptic-scale conditions, the
mean advection term A is negligible. In contrast, over complex terrain identical
weather conditions produce thermally driven flows, and thus the advection term is
usually a relevant one. While latent heat release is not included in (2.6), it may be
very important over vegetated areas and in mountainous regions adjacent to large
water bodies, and even in some arid environments where condensation of moisture
advected from elsewhere produces dew or fog (cf. Khodayar et al. 2008). Also
not considered explicitly in (2.6) is local heat storage in forest canopies and other
materials. During the course of a day all of the terms in (2.6) undergo changes with
time, and display strongly varying distributions in space.
Slope flows, the smallest-scale diurnal wind system, respond rapidly to temporal
and spatial variations in sensible heat fluxes caused by variations in the surface
energy budget. The surface energy budget is affected by changes in net radiation,
ground heat flux, soil moisture and its influence on the partitioning of energy
between sensible and latent heat fluxes, by changes in surface cover or vegetation,
cloud cover (Ye et al. 1989), break-in of background winds or turbulent episodes,
advection, etc. Surface radiation and energy budget principles, measurements in
mountain valleys, and examples of spatial variations across sample topographies
have been discussed by Whiteman (1990), Matzinger et al. (2003) and Oncley
et al. (2007).
Radiative and turbulent sensible heat flux divergences (convergences) cool (heat)
a layer of air above a slope (see [2.4]), creating a buoyancy force that has an along-
slope gravitational component (2.2) that is the basic driving force for downslope
(upslope) flows. In this section we will consider topographic effects on radiative
and sensible heat flux divergences, a topic that has received increasing attention in
the last 20 years.
2.2.2.1 Radiative Flux Divergence, r R
Radiative flux divergence contributes significantly to the cooling of clear air in

the nocturnal boundary layer over flat terrain (Garratt and Brost 1981; André and
Mahrt 1982; Ha and Mahrt 2003) and can be expected to play an important role in
producing the cooling over sloping surfaces that drives downslope flows. Despite
this expectation, slope flow models are generally driven by assumed or measured
rates of surface sensible heat flux or prescribed surface temperatures, with radiative
flux divergence in the atmosphere above the slope assumed negligible. While no
direct measurements of radiative flux divergence have yet been made over slopes,
simple radiative transfer (RT) models (e.g., Manins 1992) suggest that radiative
flux divergence cannot be neglected as a source of cooling or heating of the slope
boundary layer.
Because of recent improvements in RT models and increases in computer speeds,

it is now possible to treat both long- and shortwave RT in a realistic way for
complex terrain areas for clear sky conditions using Monte Carlo simulations
(Mayer 2009). Monte Carlo RT models track the paths of individual photons,
accounting in a stochastic way for emission, scattering, reflection and absorption
interactions between photons and air molecules, aerosols and ground surfaces. The
simulations, which often involve the tracking of millions of individual photons, can
be accomplished for situations where required input data are available, including a
digital terrain model, vertical atmospheric profiles of temperature, radiatively active
gases (e.g. water vapor and carbon dioxide) and aerosols, and spatially resolved
albedos and surface radiating temperatures. Improvements in complex terrain RT
modeling will come from the further development of these models to handle key
three-dimensional aspects of complex terrain atmospheres including the existence
of shallow cold and warm air layers on the slopes. Monte Carlo models can, in
principle, be used to simulate not only radiative fluxes (see, e.g., Chen et al. 2006)
but also radiative flux divergences, even over shallow slope layers. In the future,
such simulations can perhaps be tested against measurements and used to develop
better parameterizations for simpler models.
The MYSTIC Monte Carlo RT model (Mayer and Kylling 2005) has recently
been applied to the complex terrain of Arizona’s Meteor Crater and successfully
evaluated against longwave and shortwave irradiance measurements made on slope-
parallel broadband radiometers on the floor, sidewalls and rim of the crater on one
October day (Mayer et al. 2010). Figure 2.4 shows Monte Carlo simulations of
direct and diffuse irradiance at noon over the entire crater domain for the same day.
Following this successful evaluation of the model performance, further simulations
were performed to determine the cooling rates caused by radiative flux divergence
in idealized valley and basin atmospheres and in a realistic simulation of the Meteor
Crater (Hoch et al. 2011). In this parametric study the radiative flux divergence-
induced cooling rates for different temperature profiles representative of different
times of day were compared to those over flat terrain for valleys and basins of
different widths. Longwave radiative cooling in topographic depressions is generally
weaker than over flat terrain because of the counter-radiation from surrounding
terrain, but strong temperature gradients near the surface associated with nighttime
inversions and temperature deficit layers over slopes significantly increase longwave
cooling rates. The effects of the near-surface temperature gradients extend tens of
meters into the overlying atmosphere and can produce cooling rates on the order
of 30 K day1 that play an important role in the in-situ cooling which produces
and maintains drainage flows. Hoch et al. (2011) attributed nearly 30% of the total
nighttime cooling observed in the Meteor Crater on a calm October night to radiative
flux divergence. They approximated near-slope temperature gradients by specifying
the ground radiating temperature and the air temperature at the first model level at
5 m. Since model results for flat terrain show that radiative flux divergence depends
strongly on the detailed air temperature profile in the lowest meters above a surface
(Räisänen 1996), it will be important in future RT work to gain better resolution of
slope temperature structure profiles.
Fig. 2.4 Monte Carlo radiative transfer model simulations of direct (left) and diffuse (right)
shortwave radiation fluxes under clear sky conditions over a 4 4 km domain encompassing
Arizona’s Meteor Crater at noon, 15 October 2006 (From Sebastian Hoch)
2.2.2.2 Turbulent Sensible Heat Fluxes and Their Divergence, r H
Several field experiments (e.g., Manins and Sawford 1979a; Horst and Doran
1988; Doran et al. 1989, 1990) have measured turbulent sensible heat fluxes at
multiple heights on slope towers, but vertical gradients of these fluxes (i.e., the
vertical divergence of sensible heat fluxes) have generally not been reported because
of concerns regarding accuracy, with radiative flux divergence being the small
difference between two large quantities. An accurate measurement of the total
divergence through the slope flow layer would require a measurement very close to
the surface and a measurement at the top of the slope flow layer. Both measurements
are problematic: the upper measurement would have to be at the top of the slope
flow – which depends on background weather conditions, varies with time during
the day, and is not easily detected – and the lower measurement would suffer from
underreporting of fluxes caused by instrumental errors associated with near-ground
eddies being smaller than the sonic anemometer sampling volume (Kaimal and
Finnigan 1994). Moreover, many slopes are inhomogeneous, covered with patches
of forest, bushes, rocks, or uneven ground making it difficult to find representative
measurement sites (Van Gorsel et al. 2003a). Further requirements regarding the
appropriate coordinate system for the measurements, the representativeness of
the measurement location in terms of the homogeneity of the surface flux in the
upwind direction (i.e., flux footprint), and the post-processing of turbulence data
collected over sloping surfaces make turbulence measurements in slope flows prone
to errors. The reader will find further detailed discussions on these issues in papers
by Andretta et al. (2002), de Franceschi and Zardi (2003), Hiller et al. (2008) and
de Franceschi et al. (2009).
A surprising finding from evidence that has accumulated since the mid 1990s is
that significant inaccuracies are occurring with the measurement of surface heat
fluxes with existing research equipment. When individual terms of the surface
energy budget (see (2.7) below) are measured independently, an imbalance is often
found. This finding comes primarily from field experiments over flat, homogeneous
terrain in a variety of climate settings (see e.g., Oncley et al. 2007). The surface
energy budget imbalance during daytime is characterized by a sum of ground,
sensible and latent heat fluxes that is too small to balance the measured net radiation,
undershooting this value by 10–40%. Imbalances have also been found during
nighttime, again with the sum of the ground, sensible and latent heat flux terms
being smaller than the nighttime longwave loss. Since the Law of Conservation
of Energy must be satisfied, this suggests that experimental errors are the culprit.
In complex terrain, additional sources of error may arise because of the neglect
of budget terms under assumptions of shallow flows and horizontal homogeneity.
Ground heat flux is often poorly measured, but is expected to be relatively small
and thus incapable of explaining the error magnitude. Net radiation, on the other
hand, is relatively well measured, especially when individual components of net
radiation are measured with high-quality broadband radiometers. This suggests that
the turbulent heat fluxes are being underestimated, and attention has been focused
on advective effects, the differing flux footprints at different measurement heights,
and unmeasured flux convergences/divergences below the typical 5- to 10-m eddy
correlation measurement heights. Recent complex terrain research programs have
also encountered these measurement problems (e.g., Rotach et al. 2008).
Because sensible heat flux divergence measurements are unavailable over slopes,
most slope flow models have been driven by an assumed value of surface sensible
heat flux. Most models additionally assume that sensible heat flux is constant along
a slope, a useful approximation for a uniform slope on an isolated mountainside,
but questionable on valley sidewalls. For simple surfaces with uniform surface
cover, the surface heat flux QH can be expressed in terms of the surface energy
and radiation budgets as follows
QH D cp w0 0 s D .S C D C K " CL # CL " CQG C QE / (2.7)
where S is the incoming direct radiation, D is the incoming diffuse radiation, K " is
the reflected shortwave radiation, L # is the incoming longwave radiation, L "
is the outgoing longwave radiation, QG is the ground heat flux, and QE is the
latent heat flux. The individual terms represent the transfer of energy per unit time
through a sloping unit surface area, as measured in W m2 . In this equation, the sign
convention is that fluxes toward the surface are considered positive and fluxes away
from the surface are considered negative. Equation 2.7 illustrates the large number
of processes on which sensible heat flux depends.
During daytime, in complex, three-dimensional valley topography, the time
varying sensible heat flux is driven primarily by the input of direct radiation on a
slope (S in (2.7)). This input depends on the path of the sun’s movement through the
sky, on the azimuth and inclination angles of the slopes, the surface albedo (Pielke
et al. 1993), the physical and vegetative properties of the surface, and especially
near sunrise and sunset, on whether shadows are cast on the slope from surrounding
topography (Oliver 1992; Sun et al. 2003; Zoumakis et al. 2006). The complicated
Fig. 2.5 Modeled propagation of shadows and extraterrestrial insolation across the Meteor Crater
on October 15 at different times of day (MST). Shades of gray are insolation, with black indicating
0 and white indicating 1,364 W m2 . The horizontal scale is as in Fig. 2.4
time evolution of shadows and insolation on sloping surfaces in Arizona’s Meteor

Crater are shown as an example of this factor in Fig. 2.5. Matzinger et al. (2003) and
Rotach and Zardi (2007) suggest that the daytime spatial variation of sensible heat
fluxes over complex terrain can be reasonably well estimated by relating sensible
heat fluxes to global insolation on slopes. This approach is promising, but it is
clear from (2.7) that many other factors affect the sensible heat flux, including both
radiative and surface energy budget variables.
During nighttime, solar radiative fluxes become zero, but additional factors lead
to significant spatial variations in sensible heat flux on the sidewalls. Strong vertical
variations in temperature are superimposed on the slopes once an inversion forms
and grows in the valley, affecting outgoing longwave radiation. At the same time,
incoming longwave fluxes vary along the slope due to their varying views of the sky
and surrounding terrain, with significant amounts of back radiation received at low
elevation sites from the surrounding terrain.
A review paper by Duguay (1993) summarized some of the interactions between
the radiation field and topography, focusing on the challenge of modeling radiative
fluxes in complex topography. Key questions were the parameterization of diffuse
sky radiation and of the terrain-reflected shortwave radiation, and the difficulty
of dealing with anisotropic radiation fields. Other investigators have addressed
individual aspects of the radiative transfer problem. Olyphant (1986) investigated
the influence of longwave radiation from surrounding topography on the energy
balance of snowfields in the Colorado Front Range. Longwave loss is typically
reduced by about 50% when compared to the ridge tops. Plüss and Ohmura (1997)
addressed the influence of longwave counter-radiation from elevated snow-covered

terrain in mountainous terrain in a modeling study. They stress the importance of the
surface temperature of the sky-obstructing terrain, and also of the air temperature in
between the point of interest and the surrounding terrain, especially for inclined
surfaces. In snow-covered environments, neglecting the effects of the usually
warmer air temperatures leads to an underestimation of the radiation emitted from
the obstructed parts of the hemisphere. Height dependence of fluxes and of other
influences such as seasonal dependencies, cloud radiative forcing, the Greenhouse
Effect, etc. were investigated by Marty et al. (2002), Philipona et al. (2004) and
Iziomon et al. (2001).
The effects of the smaller scale topography on the radiation field arising
from variations in exposure, shading and sky view effects as well as surface
properties like albedo were investigated by Whiteman et al. (1989a) and Matzinger
et al. (2003) by measuring radiative fluxes on surfaces parallel to the underlying
topography. Oliphant et al. (2003) addressed these terrain effects by combining
simple radiometric measurements with modeling efforts. Whiteman et al. (1989a)
reported instantaneous values and daily totals of radiation budget components for
five sites in different physiographic regions of Colorado’s semiarid, northwest-
southeast-oriented Brush Creek Valley for a clear-sky day near the fall equinox,
finding significant differences in the radiation budgets between sites. A higher
daytime net radiation gain and a higher nighttime net radiation loss at a ridge top
site were attributed to its unobstructed view of the sky. Oliphant et al. (2003) found
slope aspect and slope angle to be the dominant features in the radiation budget of
a complex terrain area in New Zealand, followed by elevation, albedo, shading,
sky view factor and leaf area index. Matzinger et al. (2003), in a cross-section
through Switzerland’s Riviera Valley, found a decrease in downward longwave
radiation and an increase of net radiation with height. Müller and Scherer (2005)
recognized the importance of including topographic influences on the radiation
balance in mesoscale weather forecast models. They introduced a subgrid scale
parameterization of the topographic effects and managed to improve 2-m temper-
ature forecasts for areas where the interaction between topography and radiation
was most apparent. These include areas dominated by wintertime shading, areas
with increased daytime sun exposure, and deeper valleys where nighttime counter-
radiation is important. Others have addressed more specific topographic effects and
their impacts on slope boundary layer evolution. Colette et al. (2003) focused on the
effect of terrain shading on the modeled morning break-up of the nocturnal stable
boundary layer in an idealized valley, finding that shading can have a significant
impact on the timing of inversion break-up.
Soil moisture and spatial variations of soil moisture also affect slope flows
(Banta and Gannon 1995). Using a numerical model, they found that higher soil
moisture values during daytime partition more energy into latent heat flux (QE in
(2.7)), reducing sensible heat flux and decreasing upslope flow strength. The main
effect of high soil moisture during nighttime, in contrast, is an increase in soil heat
conductivity so that heat diffuses upward from a deeper layer of soil (QG in (2.7))
to replace the energy lost by longwave radiation at the surface.
2.2.3 Role of Turbulence in Slope Flows
The bulk of the research on slope flows to date has been concerned with the mean
characteristics of the slope flows, such as their depths, strengths, wind profiles,
relationships to temperature structure, etc. The important role of turbulence in
producing these characteristics has received much less attention. In contrast to flat
terrain, where calm or very weak winds often accompany fair weather conditions,
marked up- and downslope flows usually occur over slopes on fair weather days.
Turbulent kinetic energy (TKE) in slope flows is always produced, as in neutrally
stratified turbulent shear layers, by the coupling between vertical wind shear that
develops in the mean slope flow profile and the momentum flux. However buoyancy
contributes too, either to produce turbulent energy, when the sensible heat flux is
upwards, or to suppress it, when the flux is downward. This turbulence is critical
to the vertical mixing of heat and momentum that governs the height of the jet,
the depth of the flow and the ultimate equilibrium temperature deficit (Cuxart
et al. 2011).
Turbulence in upslope flows is generally considered to be a more tractable
problem than turbulence in downslope flows. This comes from parallels with
the study of turbulence over flat, homogeneous terrain, where understanding of
turbulence has progressed much faster for convective boundary layers than for stable
boundary layers. Towers on slopes are generally too short to fully encompass the
rapidly growing upslope flow layer, however. Thus, unless additional observational
tools can be developed for measuring turbulence over deeper layers above slopes,
future improvements in understanding of upslope flows may come primarily from
modeling studies.
The continuous turbulence production by shear in the wind profile in fully
developed shallow downslope flows is expected to be a much more manageable
problem than turbulence in the deeper stable boundary layers that form over both
flat and complex terrain areas, where the turbulence is intermittent and discontin-
uous (Mahrt 1999; Van de Wiel et al. 2003). The shallow downslope boundary
layers are more accessible to tower measurements, but appropriate measurement
techniques and post-processing methods (e.g. filtering, wind component rotations,
and alignment to an appropriate reference frame) are still problematical. Turbulence
data are generally obtained from high-rate sampling of the three components of
the wind. Additional sampling of scalar variables (temperature, humidity, CO2
concentrations, etc.) on the same time scale can be used to correlate fluctuations of
the wind components and scalars to determine fluxes. Representative measurements
require that the upwind fetch be homogeneous with respect to sources and sinks of
scalars and surface properties. Following the collection of wind data on the slope,
the wind measurements are usually rotated to align the coordinate system to the
mean wind, which is generally parallel to the underlying ground (at least, near the
surface). Flaws in the typical double- or triple-rotation method used over flat terrain
are encountered in sloping terrain (see Finnigan et al. 2003), so that the “planar-
fit” approach developed by Wilczak et al. (2001) is considered preferable. But an
appropriate post-processing also requires carefully designed filtering and averaging

procedures to extract turbulent quantities from the often non-stationary mean flows
encountered over slopes (de Franceschi and Zardi 2003; Weigel et al. 2007; de
Franceschi et al. 2009).
Few model simulations of turbulent quantities are yet available for downslope
flows since most of the RANS numerical models to date have been developed
for mesoscale simulations on much larger scales than the slope flows. The lack
of vertical resolution of the slope flows in these models, and uncertainty in the
applicability of the turbulence parameterizations have limited their utility. Recent
improvements in LES model formulations (also see Chap. 10), and corresponding
increases in computer power, are now leading to continuously improved simulations
as to grid resolution, numerical schemes and appropriateness of turbulence closures
(Skyllingstad 2003; Chow et al. 2006; Axelsen and Van Dop 2008, 2009; Serafin
and Zardi 2010a).
2.2.4 Upslope Flows
The radiative heating of the ground and the resulting upward turbulent sensible
heat fluxes, typically occurring in connection with upslope flows, increase rapidly
after sunrise, reaching magnitudes that are much higher than those associated with
downslope flows. As a consequence of increasing heat flux, upslope flows cannot
approach a steady state, but rather grow continuously with time in strength and
depth. Upslope flows are confined to a layer over the heated slope in the morning by
the highly stable air that formed above the slope overnight (the confining stability
can be especially strong over valley sidewalls, compared to slopes on isolated
mountainsides). Like the situation over flat ground, heating of the slopes causes
an unstable or convective boundary layer to grow upward into the remnants of
the nighttime stable layer. The daytime boundary layer over slopes, in contrast to
over flat ground, is a convective-advective boundary layer, with bent-over plumes
and convective elements moving up the slope. While the upslope flow grows in
depth and strengthens with time after sunrise, the volume flux of the flow increases
with upslope distance as air is entrained into the flow at its upper boundary. The
growing depth of the convective-advective upslope flow layer eventually allows the
convection to break through the stable layer. Following the breakup of the confining
stable layer, one might expect the convective currents to simply rise vertically from
the heated ground. In fact, the flow continues to rise up the slope. This is apparently
caused by a horizontal pressure gradient associated with the inclined superadiabatic
sub-layer that forms over the slope. The upslope flows are often disturbed, however,
by convective mixing that brings down the generally stronger background winds
from higher layers of the atmosphere.
Our understanding of upslope flows has suffered from the lack of observational
tools for making continuous measurements of the mean and turbulent profiles
through the full depth of the growing upslope flow layer. Towers are generally
not tall enough to penetrate this layer and the non-stationarity and inhomogeneity
of the flows has made observations particularly difficult. Some field observations
(Kuwagata and Kondo 1989; Van Gorsel et al. 2003a; Reuten et al. 2005; Geerts
et al. 2008; De Wekker 2008), laboratory simulations (Hunt et al. 2003; Princevac
and Fernando 2007; Reuten et al. 2007; Reuten 2008) and model simulations
(Kuwagata and Kondo 1989; Schumann 1990; Atkinson and Shahub 1994; Axelsen
and van Dop 2008) have improved this situation. Two interesting observations on
upslope flows came from a field experiment over an isolated slope northeast of
Vancouver, BC, in 2001. The observed depression of a mixed layer at the base of the
mountain slope is thought to have been caused by subsidence in the return branch of
the upslope circulation (De Wekker 2008; Serafin and Zardi 2010a). On a different
day, scanning Doppler lidar observations up the slope appeared to show a closed
slope flow circulation within the inclined convective boundary layer (Reuten et al.
2005), although these findings are controversial.
Schumann (1990) simulated upslope flows over an infinite, uniform, heated,
sloping surface for a range of slope angles and surface roughness. The simulations
were revolutionary, providing new insight into the role of turbulence in upslope
flows and the utility of LES modeling to improve understanding of geophysical
phenomena on small length scales. Schumann’s work, as a parametric study, was
limited to a constant ambient stability (3 K km1 ) and sensible heating rate (about
100 Wm2 ), with a quiescent atmosphere above the slope flow layer. The simulated
steady-state profiles of upslope wind speed and excess temperature (with respect to
the unperturbed thermal structure) for slopes of different inclination angle from 2ı to
30ı are shown in Fig. 2.6, where the axis variables are normalized by a scale height
H of 58 m, a wind speed scale v of 0.58 m s1 , and a temperature scale of 0.17 K.
A dry adiabatic lapse rate curve is provided on the temperature profile plot for
comparison with the simulated profiles. Several interesting features are seen in the
simulations. The steady state was attained by a balance between the buoyancy term
and surface friction in (2.2) and between the surface heating and upslope advection
of heat against the mean temperature gradient in (2.4), just like in Prandtl’s model.
As expected, the temperature excess present at the ground decayed with height.
Temperature profiles, except those over the steepest slopes, were characterized
by a near-adiabatic temperature decrease with altitude, indicating good mixing in
the turbulent flows. Strong local minima occurred in the velocity and temperature
profiles at the upper edge of the mixed layer. Upslope flow depths and volume fluxes
were greatest over low-angle slopes. Shallow and weak upslope flows occurred on
steeper slopes. Wind maxima occurred within about 50 m of the surface over slopes
of all inclination angles. The turbulence kinetic energy profile in the upslope flow
was generated by vertical shear, with high positive values of shear just above the
ground but below the height of the jet maximum and high negative values of shear
above the jet maximum. This resulted in a slope-normal transport of momentum
both upward and downward from the jet level (see Fig. 2.2).
Fig. 2.6 Normalized upslope mean wind component and temperature profiles over heated slopes,
with indicated slope angles from 2ı to 30ı , as simulated using an LES model. H, v* and
* are respectively the scaling values adopted to normalize the slope-normal coordinate n,
the wind component <u> and the temperature <T>. Error bars indicate standard deviations
(From Schumann 1990, © Quarterly Journal of the Royal Meteorological Society. Reprinted with
permission)
LES simulations such as this have the advantage of being able to produce profiles
of all turbulence quantities above the slope and to determine all contributions to
the turbulence kinetic energy budget. One caveat, however, is that subgrid scale
TKE contributions are parameterized. The contributions of these terms near the
surface, where eddies are small, have not been verified. It is important to note that
detailed observations are, so far, unavailable to properly test LES simulations of the
upslope boundary layer. The idealized nature of the LES simulations should also be
emphasized. In real, three-dimensional topography, slopes are very inhomogeneous
in terms of slope angle, roughness, vegetative cover, sensible heat flux, non-
planarity, etc. and the flows are distinctly non-stationary. In contrast, the LES
simulations were conducted on uniform slopes and the equations were integrated
until steady-state solutions were reached. The simulations certainly have interesting
implications for upslope flows on sidewalls with varying inclination angles. In these
cases, along-slope convergences or divergences of mass in the upslope layer will
cause detrainment and entrainment from the slope flow layer (see, also, Vergeiner
and Dreiseitl 1987), with implications for the air mass over the valley center. There
are ample opportunities to extend Schumann’s (1990) initial work.
2.2.5 Downslope Flows
Downslope flows on isolated mountainsides and on valley sidewalls have been stud-
ied extensively over the years. In this section we will summarize current knowledge
about the mean and turbulence structure of downslope flows using observations and
simulations, concentrating on research conducted since about 1990.
For downslope flows, the terms of importance in the governing equations
(2.2–2.4), have been evaluated mainly with hydraulic models (Manins and Sawford
1979b; Horst and Doran 1988; Papadopoulos et al. 1997; Haiden and Whiteman
2005). Steady-state flow on mid-latitude slopes is typically an “equilibrium flow”
balance between buoyancy and turbulent momentum flux divergence (i.e., friction),
in agreement with the local equilibrium assumption made in Prandtl’s (1942)
analytical model. The surface stress is thought to contribute more friction than
the entrainment of slower moving air from above the downslope layer. The
balance can sometimes include a significant along-slope advective contribution in
the momentum balance near the height of the wind maximum, where shear changes
sign (Skyllingstad 2003), in agreement with Horst and Doran’s (1988) results. The
slope-perpendicular momentum equation, since the RHS of (2.3) is near zero, is
a balance between the two advection terms. The thermodynamic energy equation,
(2.4), is a balance between the radiative and sensible heat flux divergences and
temperature advection.
Horst and Doran (1988) found that the turbulence structure over slopes was
consistent with local shear production of turbulence. Turbulence, which is generated
primarily by vertical wind shear and destroyed by viscous dissipation, is critical for
the vertical mixing of momentum and buoyancy in the mean flow. The jet-like wind
profile associated with downslope flows produces a turbulence structure (Fig. 2.2)
that is quite different from that over flat terrain. Vertical shear is positive below the
jet, disappears at the height of the jet and is negative above the jet. This vertical
distribution of wind shear produces TKE profiles with a maximum near the surface,
and a local minimum at the height of the wind maximum. Above that height, TKE
is highly dependent on the speed and direction of the ambient wind. Turbulence
above the wind maximum is decoupled from the surface so that normal surface
similarity theory is not expected to be applicable above the jet. Vertical velocity
variance profiles depend on cooling rates and external wind speeds (Coulter and
Martin 1996).
In the remainder of this section we will highlight the important influences of
slope angle, ambient stability, external flows and other disturbances on downslope
flows. We begin with slope angle.
2.2.5.1 Impact of Slope Angle on Downslope Flows
If a continuous source of cold air were introduced at the top of a slope (as
in Burkholder et al. 2009), the downslope flow would be stronger on steeper
slopes where the along-slope component of the gravity vector is maximized. But,
downslope flows in valleys and on mountainsides are not generally fed by a source
of cold air at the top of the slope. Rather, cold air forms in place on the slope through
a downward flux of sensible heat to the underlying radiatively cooled surface. The
air moves down the slope only as long as it is negatively buoyant with respect to
the ambient environment. The air in the flow is heated adiabatically at 9.8ı C km1
as it moves down the slope. To continue its descent it must be continuously cooled
by a downward sensible heat flux to the cold underlying surface. The faster the
flow, the lower the rate of cooling per unit mass of flow. On low angle slopes, the
air travels a greater horizontal distance along a slope to descend the same vertical
distance compared to a steeper slope. The air loses heat through downward sensible
heat flux along its entire trajectory, which is longer on a shallow slope than on a
steep slope.
Slope flows will not form on horizontal surfaces, where radiative loss to the
sky and downward sensible heat flux occur but the along-surface component of
the gravity vector is zero. At the other limit, a vertical surface will have a strong
component of the gravity vector along the slope, but a vertical surface will not
radiate well to the sky and, because of the small distance that the air would travel
along the slope to lose a unit of altitude, the rate of cooling of the flow to the
underlying surface would be small. The strongest downslope flows will thus develop
on slopes of intermediate angle where there is a combination of good outgoing
longwave radiative loss to the sky that drives a slope-normal sensible heat flux and
a component of the gravity vector acting along the slope.
Observations (Table 2.1) on simple slopes show that downslope flows on low-
angle slopes are deeper and stronger than those on steeper slopes. Slope flows are
often strongest in the early evening when the ambient stability is weak. For slopes
on valley sidewalls, this is the time before the overlying valley flows strengthen to
produce a cross-slope shear that disturbs the slope flows. The temperature deficit
layer over the slope (i.e., inversion depth) often has a depth corresponding to
about 5% of the vertical drop from the top of the slope and the height of the
wind maximum is usually within the temperature inversion layer, often at heights
that are 30–60% of the inversion depth. The downslope flow itself usually extends
much higher than the height of the wind maximum or the height of the inversion
top because of the entrainment of ambient air by the downslope flow (Princevac
et al. 2005). But, this height is much more variable than the inversion height and
is very dependent on the speed and direction of the ambient flow above the slope
flow layer. In general, the downslope flow extends to heights of several times the
inversion depth.
The speed and depth of downslope flows (and, thus mass or volume flux)
increase with downslope distance. This has been seen in experiments on Rattlesnake
Mountain, Washington (Horst and Doran 1986), on a low angle slope in Utah’s
Salt Lake Valley (Haiden and Whiteman 2005; Whiteman and Zhong 2008), and on
Greece’s Mt. Hymettos (Amanatidis et al. 1992; Papadopoulos and Helmis 1999).
An example illustrating this increase in depth and speed is shown in Fig. 2.7.
Downslope flows accelerate with downslope distance while increasing their depth
Table 2.1 Observations of downslope flow on slopes of different inclination

Angle Inversion Inversion Max speed Height of max
(ı ) depth (m) strength (K) (m s1 ) speed (m) References
21 4–8 5 1–2 0.6–2 Horst and Doran (1986)
9 16 4 1–2 6 Papadopoulos et al. (1997)
6.5 40–50 10 1.5 25 Horst and Doran (1986)
4 50 3 3.5–4 40 Monti et al. (2002)
1.6 25 7 4–6 15–20 Whiteman and Zhong (2008),
Haiden and Whiteman (2005)
1700
a b c
1650
Height (m MSL)
1600
1550
1500
1450
290 295 300 305 -3 -2 -1 0 1 2 3 4 5 -3 -2 -1 0 1 2 3 4 5
-1)
Potential temp (K) Along-slope wind (m s Cross-slope wind (m s-1)
Fig. 2.7 Vertical profiles of (a) potential temperature, (b) downslope wind speed component
(m s–1 ) and (c) cross-slope wind speed component (m s1 ) at 2217 MST 8 October 2000 from
four tethersondes running down a 1.6ı slope at the foot of the Oquirrh Mountains at various
distances from the top of the slope, namely 4,200 m (light gray), 5,100 m (dark gray), and 6,200 m
(black) (From Whiteman and Zhong 2008. © American Meteorological Society. Reprinted with
permission)
as additional mass is entrained into the flowing layer at its upper boundary. Some of
the simple steady-state analytical slope flow models (see, e.g., those reviewed
by Barry 2008) are in conflict with observations and with predictions from full
physics models. Because the models are often run using the same sensible heat
flux forcing on slopes of all angles, they have led to the mistaken concept that
downslope flows on upper steeper slopes are stronger and deeper than flows on
lower slopes, and will overrun the flows on lower slopes or will travel down the
slope as frontal disturbances. While such frontal propagation episodes can occur in
special circumstances (e.g., when the upper slopes go first into shadow while the
lower slopes are still in sunlight), they appear to be atypical.
RANS and LES models are in agreement with field data, predicting weaker
downslope flows (i.e. displaying both smaller maximum jet speed and smaller
mass flux) on steeper slopes (Skyllingstad 2003; Zhong and Whiteman 2008;
Axelsen and Van Dop 2009). The effect of stepwise changes in slope angle partway
down a mountainside were investigated using both RANS and LES models by Smith
and Skyllingstad (2005) for an 11.6ı slope that changed suddenly to a 1.6ı slope
at a lower elevation. The slope flow depth increased rapidly over a distance of
500 m at the slope angle juncture and the height of the maximum velocity moved
upward about 5–10 m. Below the juncture, a new equilibrium was established with
a decreased rate of growth in slope flow depth and more rapid cooling of the
near-surface temperatures. Near-surface downslope velocity decreased, producing
a distinct detached and elevated jet structure. Potential energy generated at the top
of the slope by continued cooling was transported downslope and converted into
kinetic energy at the slope base.
2.2.5.2 Impact of Ambient Stability on Downslope Flows
The effect of ambient stability on downslope flows is now well known from
both simulations on simple slopes and from field experiments. The downslope
flow must adjust to any changes in ambient stability of the atmosphere adjacent
to the slope. If the flow, for example, encounters a sudden increase in ambient
stability, it must be cooled at a higher rate to be advected along the surface at
the same speed. Downslope flow intensity is thus inversely proportional to ambient
stability (Rao and Snodgrass 1981; King et al. 1987; Ye et al. 1990; Helmis and
Papadopoulos 1996; Zhong and Whiteman 2008). The nighttime buildup in ambient
stability within valleys reduces downslope flow speed and may even suppress
downslope currents. In enclosed basins, downslope flows die in the late evening as
ambient stability increases and the atmosphere becomes increasingly horizontally
stratified and quiescent (Clements et al. 2003; Whiteman et al. 2004a, c; Steinacker
et al. 2007).
2.2.5.3 Impact of Background Flow on Downslope Flows
Downslope flows, which are usually rather weak and shallow air currents, are quite
sensitive to disturbance by background or ambient flows (Fitzjarrald 1984; Barr and
Orgill 1989; Doran 1991) that may oppose or follow the downslope flows or may
have a significant cross-slope component. Background flows can originate as valley
flows, larger mesoscale circulations, synoptic flows, sea or lake breezes, seiches,
gravity waves, wakes from flow over ridges, etc. The effects of these differing
phenomena on downslope flows are difficult to separate using field data, but it is
generally understood that background flows with strengths of only 2–3 m s1 can
have far reaching effects on the structure of the downslope flows and even their
existence.
Downslope flows on valley sidewalls are routinely subject to the development

of overlying valley flows. These valley flows, which have jet-like vertical and
horizontal profiles within the valley cross-section, produce horizontal shears above
the slope flows that vary with elevation. Because these flows are oriented along the
valley axis, while the slope flows are perpendicular to the valley axis, they produce a
strong directional shear through the shallow slope flow layer. When the valley winds
get strong enough, horizontal and vertical shears and the associated turbulence can
greatly modify the downslope flows or remove them entirely from valley sidewalls
(Doran et al. 1990).
As cross-slope winds increase, slope flows deepen and behave more like a
weakly stratified sheared boundary layer (Skyllingstad 2003). The cross-slope flow
produces an extra source of mixing from the additional vertical shear, which weak-
ens the vertical stratification. Because the velocity near the surface becomes lower
there is less loss of momentum to surface drag. But, as cross-slope flow increases,
the downslope jet becomes weaker and deeper and is eventually overpowered by the
cross-slope-generated turbulence.
2.2.5.4 The Effects of Slope Inhomogeneity on Downslope Flows
Vergeiner and Dreiseitl (1987) pointed out that real mountain slopes are much
more complicated than those assumed in existing models. Since, as emphasized
above, radiative and turbulent fluxes vary in time and space, upslope and downslope
flows must vary continuously in space and time in response to small-scale slope
inhomogeneities. This variation makes it difficult to find representative measure-
ment locations on slopes (e.g., sites with uniform slope angles and surface cover)
to compare with idealized models. Slopes on mountain sides vary not only in
topography and surface cover, but they are also subject to time- and space-varying
interactions with larger scale flows and to varying ambient stabilities.
Only a few observational and modeling studies have investigated the response
of slope flows to slope inhomogeneities. As mentioned, several studies have
investigated the effects of changes in slope angle. Others have looked into the
effects on downslope flows of forests (e.g., Lee and Hu 2002; van Gorsel 2003);
differentially cooled sloping surfaces (Shapiro and Fedorovich 2007; Burkholder
et al. 2009) and sunset time variations on complex slopes (Papadopoulos and
Helmis 1999). Haiden and Whiteman’s (2005) investigation of downslope flows
on a visually rather uniform slope at the foot of Utah’s Oquirrh Mountains found
imbalances in both the momentum and heat budget equations that were attributed to
small scale convexities and concavities on the slope that channeled the downslope
currents. Monti et al. (2002) observed multi-layered stability above a slope that
appeared to come from elevated downslope currents. They suggested that slope
flows can develop features similar to hydraulic flows, such as jump conditions or
turbulent transitions.
2.2.5.5 Drainage Flow Oscillations
Oscillations in drainage (i.e., downslope and down-valley) flow strengths with

periodicities between about 5 and 90 min have been frequently observed. They can
be produced by a variety of different mechanisms, including interactions between
valley flows and tributary flows (Porch et al. 1989), variation of cold air sources
from upstream (Allwine et al. 1992), interactions with mountain waves (Stone and
Hoard 1989; Poulos et al. 2000, 2007), modulation of slope and valley flows by
interactions of synoptic and mesoscale flows with surrounding topography (Mahrt
and Larsen 1982; Doran et al. 1990; De Wekker 2002), interactions between slope
flows and sea breezes (Bastin and Drobinski 2005), meandering motions in stratified
flows and, in the case of downslope flows, by internal waves impinging normal to
the slope (Princevac et al. 2008) and by overshooting and oscillations caused by the
resulting buoyancy disequilibrium (Fleagle 1950; Doran and Horst 1981; McNider
1982; Helmis and Papadopoulos 1996; van Gorsel et al. 2003b; Chemel et al. 2009;
Viana et al. 2010). Additionally, cold air forming over plateaus or other elevated
terrain source areas sometimes cascades down the slopes at intervals as cold air
avalanches (Küttner 1949). Atmospheric analogs to the seiches that form in ocean
and lake basins have also been observed in basin temperature inversions, where it
is common to see smoke plumes from chimneys and open fires that exhibit weak
oscillatory motions in which the plume is carried back and forth quasi-horizontally
relative to its source. The back-and-forth sloshing, if produced by seiches, should
have a characteristic frequency that is dependent on basin topography, atmospheric
stability and wind speed. Seiche-like oscillations with a periodicity of about 15 min
were recently documented on the floor of Arizona’s Meteor Crater (Whiteman et al.
2008), and a spectral element model (Fritts et al. 2010) has produced seiches in
idealized topography similar to the crater.
2.2.5.6 Glacier Winds
Slope flows are generally diurnally varying flows, but over a snow surface where
the surface energy budget is consistently negative the resulting downward sensible
heat flux from the adjacent atmosphere produces a shallow cold air layer with a
continuous downslope flow. Over cold, high albedo glaciers or snow surfaces, this
downslope flow has been called a glacier wind. A special feature of downslope
winds over snow surfaces is the constrained upper limit value of surface tempera-
ture, 0ı C. In daytime during the melt season when ambient air temperatures undergo
their normal diurnal variation, this upper limit of surface temperature results in a
temperature inversion above the melting surface whose maximum strength occurs
in mid- to late-afternoon in the ablation zone, and at night at higher elevations where
radiative cooling of the surface is more important (Oerlemans 1998).
Glacier winds, which occur on a variety of scales, have received increased
attention in the last 20 years. Oerlemans et al. (1999) found temperature inversions
of about 20 m depth and 8ı C strength over a large maritime ice cap in Iceland.
Near-steady-state glacier winds were nearly always present, except for periods with
traveling intense storms, suggesting that glacier winds are the key factor shaping
the microclimate of glaciers, at least in summer. Maximum downslope wind speeds
of 3–10 m s1 were found at heights from several to a few tens of meters above
the surface, although the winds often extended to heights exceeding 100 m. The
momentum balance that produces the glacier wind appears to be a balance between
buoyancy force and friction, as expressed in the Prandtl (1942) and Oerlemans
and Grisogono (2002) models. In contrast, for a small, mid-latitude glacier in
the Alps the buoyancy force was balanced by both friction and the mesoscale
pressure gradient that drives the valley wind above the downslope layer (Van den
Broeke 1997a, b).
2.3 The Valley Wind System
Diurnal valley winds are thermally driven winds that blow along the axis of a
valley, with up-valley flows during daytime and down-valley flows during nighttime.
Valley winds are the lower branch of a closed circulation that arises when air in
a valley is colder or warmer than air that is farther down-valley or over the adjacent
plain at the same altitude. Unlike slope winds, valley winds are not primarily a
function of the slope of the underlying valley floor. In fact, they have been observed
even in valleys with horizontal floors (Egger 1990; Rampanelli et al. 2004).
Instead, they depend on other geometrical factors, such as the shape and aspect
of the valley cross-section and their along-valley variations, including tributaries
(Steinacker 1984), as discussed below. The diurnal reversal of the flow requires
a larger daily temperature range within the valley than over the adjacent plain
(Nickus and Vergeiner 1984; Vergeiner and Dreiseitl 1987). Horizontal pressure
gradients that develop as a function of height between air columns with different
vertical temperature structures over the valley and the adjacent plain drive the valley
wind (Fig. 2.8). These pressure gradients can be observed directly (Khodayar et al.
2008; Cogliati and Mazzeo 2005). Along the Inn Valley, for example, thermally
induced pressure differences of up to 5 hPa over a distance of 100 km have been
observed (Vergeiner and Dreiseitl 1987). As a consequence of the changing pressure
distribution in the vertical, the vertical profile of along valley winds undergoes a
cyclic evolution during the day (Gross 1990). Valley winds evolve gradually over
the daily cycle and produce weak to moderate wind speeds. Peak wind speeds over
the valley center are frequently in the range from 3 to 10 m s1 .
Ideally, the upper branch of the closed circulation consists of an elevated
horizontal flow running in the opposite direction from the winds in the valley below.
Because the upper branch is unconfined by topography and thus broader in horizon-
tal extent, it is generally quite weak and can be obscured by stronger synoptic-scale
flows. The existence of these return flows or compensation currents was debated at
the early stage of the scientific investigation of valley winds (cf. Wagner 1938) but
is today considered a well-documented feature (McGowan 2004; Zängl and Vogt
Fig. 2.8 Idealized picture of the development of daytime up-valley winds (upper panel) and
nighttime down-valley winds (lower panel) in a valley-plain system with a horizontal floor. The red
and blue curves are vertical profiles of the horizontal valley wind component at a location close to
the valley inlet. Two columns of air are shown – one over the valley floor and one above the plain.
Red and blue sections of the columns indicate layers where potential temperature is relatively warm
(W) or cold (C). The free atmosphere is assumed to be unperturbed by the daily cycle at the tops
of the columns (Adapted from Whiteman 2000)
2006). Figure 2.9, for example, compares modeled temperature profiles within a
valley and over the adjacent plain to illustrate the reversing temperature gradients
with height that drive the elevated return circulations (from Rampanelli et al. 2004).
Investigations by Buettner and Thyer (1966) in valleys that radiate out from the
isolated volcano of Mt. Rainier, Washington (USA), found that the upper branches
of the valley circulations there could even be observed within the terrain-confined
Fig. 2.9 (a) Comparison of different vertical profiles of potential temperature in an idealized
valley-plain system (sketched in the insert). The profiles are taken along the valley axis (solid
line) and over the plain (line of circles) along the dashed vertical lines indicated in the insert at
t D 6 h after simulated sunrise. (b) Difference between the two profiles at various heights. The
horizontal dashed line at 1,000 m indicates the sidewall ridge height (Adapted from Rampanelli
et al. 2004. © American Meteorological Society. Reprinted with permission)
upper altitudes of the valleys. They coined the term anti-winds for these confined
currents, but the deeper currents above the ridgetops are more frequently called
return or compensating flows.
The evening reversal of the daytime up-valley wind begins in late afternoon or
early evening when the daytime wind system begins to lose strength. The gradual
loss of strength, taking place over several hours, is caused by a decrease in the
along-valley pressure gradient as temperatures along the valley axis are equilibrated
by advection and as the sensible heat flux that maintained the temperature gradient
decreases. Once the sensible heat flux reverses on the sidewalls or in the shadowed
parts of the valley, downslope flows begin, transferring the cooling experienced in
a shallow layer above the slopes to the valley atmosphere through compensatory
rising and cooling motions. Radiative processes and turbulent sensible heat flux
divergences also play a role. As the valley atmosphere cools, the horizontal pressure
gradient between the valley and adjacent plain reverses and a down-valley flow
begins. The valley wind reversal lags that of the slope flows because the large
mass of the valley atmosphere must be cooled before the pressure gradient can
reverse (Vergeiner 1987). The reversal typically begins at the floor of the valley,
where cooling is most intense, with the formation of a temperature inversion. As
the cooling progresses upwards, the down-valley flow deepens and strengthens.
Within the nighttime inversion, the coldest temperatures are at the valley floor, and
temperatures increase with height. This results in relatively warmer temperatures on
the valley sidewalls than on the valley floor, i.e. a warm slope zone on the valley
sidewalls (Koch 1961).
The morning reversal of the down-valley wind, which occurs after sunrise,
similarly requires a reversal of the along-valley pressure gradient as the valley
atmosphere warms. The reversal occurs several hours after upslope flows are
initiated when the heat transfer processes become effective in warming the entire
valley atmosphere. Further details on the heat transfer mechanisms are provided
in the next section that focuses on interactions between the slope and valley wind
circulations.
The warmer temperatures during daytime and colder temperatures during night-
time in the valley atmosphere compared to the atmosphere over the adjacent plain
(or farther down the valley) are mainly caused by two factors. First, under low
background wind conditions the valley atmosphere is protected by the surrounding
topography from mass and heat exchange with the atmosphere above the valley so
that heating will be more effectively concentrated in the valley during daytime and
cooling will be enhanced in the valley during nighttime. A second, more important,
factor has been termed the “area-height relationship” or “topographic amplification
factor”, TAF. The concept (see Whiteman 1990 or 2000 for fuller discussions), is
that the daily temperature range in the valley will be amplified by the smaller mass
of air that is heated or cooled within the confined valley volume than within the
larger volume of the same depth and surface area at its top over the adjacent plain.
The higher mean temperature inside the valley during daytime causes a pressure
gradient to develop between valley and plain that drives an up-valley wind. The
lower mean temperature and higher pressure in the valley during nighttime drive a
down-valley wind.
The mechanisms by which valley geometry might affect cooling and warming
of the valley atmosphere have been explored by various authors. Wagner (1932)
was the first to express the TAF concept. Steinacker (1984) carried the concept
further, computing the distribution of drainage area as a function of height in the Inn
Valley and comparing it to that over the adjacent plain in southern Germany. McKee
and O’Neal (1989) then suggested that along-valley pressure gradient differences
produced by changes in terrain cross-section along a valley’s axis could explain
along-valley wind speed variations. While the intuitive TAF concept, which focuses
on valley geometry, helps to conceptualize the important thermodynamics leading
to up- and down-valley flows, the simple concept involves inherent assumptions that
have made it difficult to apply in practice.
There are inherent difficulties in evaluating precisely both the valley volume,
and the heat fluxes across its boundaries. As to the volume, its top boundary is
difficult to define in practice. An intuitive choice would be to locate it at crest
height, yet in real valleys the crest height varies with down-valley distance and
differs between the opposing sidewalls. Further, it appears that the volume of the
valley tributaries has to be taken into account (Steinacker 1984). As to the fluxes,
Gauss’s divergence theorem can be used with a volume integral of (2.7) to determine
the rate of change of the volume-average potential temperature in the valley. This
rate of change is driven by the surface integral of the advective, turbulent and
radiative heat fluxes across the surfaces bounding the volume, including the valley
floor, sidewalls and the top boundary. However the flux distributions along the
volume boundaries are generally not well known (e.g. on the valley sidewall slopes:
see Sect. 2.2). Yet boundary conditions allow some simplifications: for instance,
advection through the ground surface vanishes, since wind speed vanishes there.
However this is not the case on the top surface, where heat advection through
slope flows on the sidewalls and compensating subsidence at the valley core will
play a crucial role in the daytime heating of the valley atmosphere (Rampanelli
et al. 2004; Weigel et al. 2007; Schmidli and Rotunno 2010; Serafin and Zardi
2010b, c). Because the top boundary is not a solid surface, heat gain or loss in the
volume is not normally confined to the volume. This is especially the case during
daytime, when convective boundary layers can grow above ridge height. Similarly,
the down-and up-valley ends of the volume are open: once an along-valley flow
begins, advection acts to reduce temperature differences between valley and plain,
or between segments of a valley. Further, the TAF concept in its simplest form does
not consider that the rate of heat gain or loss may vary with height in the valley or
plains volumes, and assumes that rates of heat gain and loss are equal in the two
volumes. In fact, the surface energy budgets and heat transfer processes may be
quite different in the two volumes. Heat transfer processes in the valley volume, for
example, involve heat transfer over the slopes associated with up- and downslope
flows, compensatory rising and sinking motions over the valley center, and radiative
transfer processes that include back-radiation from the sidewalls (cf. Haiden 1998).
Further information on these and other factors that lead to variations in the thermally
driven pressure gradient along a valley including the partitioning between sensible
and latent heat fluxes in the surface energy budget and the confinement of heated
or cooled air within a valley can be found in publications by Zängl et al. (2001),
Rampanelli (2004), Zängl (2004), Zängl and Vogt (2006), Serafin (2006), Rucker
et al. (2008), Schmidli and Rotunno (2010), and Serafin and Zardi (2010a, b, c).
In the early evening, down-valley flows often converge into sub-basins or build
up cold-air pools or lakes behind terrain constrictions (cf. Bodine et al. 2009). In
general, though, if the valley does not have major constrictions along its length or
major changes in surface microclimates, the nighttime down-valley flows accelerate
and increase in depth with down-valley distance, resulting in an along-valley mass
or volume flux divergence. Through the law of mass conservation, the existence
of along-valley divergence requires the import of potentially warmer air into the
valley from aloft or from tributaries. Interestingly, recent field measurements and
Fig. 2.10 Average up- and down-valley wind profiles from four valleys using all daytime and
nighttime data from the locations indicated (Ekhart 1944) (© E. Schweizerbart Science Publishers
www.borntraeger.cramer.de. Used with permission)
model simulations have also shown a divergence of along-valley mass flux in

the daytime up-valley flows for the Inn Valley (Vergeiner 1983; Freytag 1988;
Zängl 2004), the Kali Gandaki Valley (Egger et al. 2000; Zängl et al. 2001) and
in the Wipp Valley (Rucker et al. 2008). This along-valley mass flux divergence
is somewhat counterintuitive, since one might expect an up-valley flow in the
main valley to dissipate with up-valley distance as the flows turn up the slopes
and tributaries. Various mechanisms have been invoked to explain the mass flux
divergence, including compensating subsidence over the main valley, transition to
supercritical flow in the widening part of a valley (possibly favored by gravity waves
produced by protruding ridges), and intra-valley change in the horizontal pressure
gradient induced by differential heating rates.
In different valleys (Fig. 2.10), the vertical structure, strength and duration of
up- and down-valley flows differ, depending on climatic and/or local factors (Ekhart
1948; Löffler-Mang et al. 1997). One cannot assume that valleys with strong up-
valley flows will also have strong down-valley flows. In the Kali Gandaki valley
of Nepal, the daytime up-valley winds are much stronger than the nocturnal down-
valley winds (Egger et al. 2000, 2002). In California’s San Joaquin Valley, the up-
valley flows can persist all night (Zhong et al. 2004). In other cases, such as in
Chile’s Elqui Valley, nocturnal down-valley winds are not observed at all (Kalthoff
et al. 2002), while down-valley winds may persist all day in snow-covered valleys
(Whiteman 1990).
Down-valley winds are often enhanced where valleys issue onto the adjacent
plains or into wider valleys. Drainage flows in the valley decrease in depth,
accelerate, and fan out onto the plain at the valley exit, producing a valley exit jet.
The valley exit jet at the mouth of Austria’s Inn Valley reaches speeds of 13 m s1
at heights only 200 m above ground (Pamperin and Stilke 1985; Zängl 2004). Valley
exit jets have also been observed at the mouths of tributary canyons that issue from
the Wasatch Mountains into the wide Salt Lake Valley (Banta et al. 1995, 2004;
Fast and Darby 2004; Darby and Banta 2006; Darby et al. 2006) and at canyons that
flow onto the plains south of Boulder, Colorado (Coulter and Gudiksen 1995; Doran
1996; Varvayanni et al. 1997). They have also been observed at the exit of valleys
issuing onto the Snake River Plain in eastern Idaho (Stewart et al. 2002) and at the
exit of the Lech Valley into the Bavarian Plain (Spengler et al. 2009). Valley exit jets,
which are often a source of non-polluted or relatively clean air, can have important
air pollution consequences (Darby et al. 2006) and can bring nighttime relief from
high temperatures at canyon exits. The outflows of valley exit jets can be modulated
by larger-scale influences including the thermally driven flows beyond the valley
exit (Darby and Banta 2006).
2.3.1 The Effects of Boundary Layer Processes

and Turbulence on Valley Flows
Only recently have field research programs investigated the impact of turbulence
on valley winds (Rao and Schaub 1990; Poulos et al. 2002; Rotach et al. 2004;
Rotach and Zardi 2007). Turbulence measurements have been made primarily from
instrumented towers using eddy correlation techniques, although some airborne
measurements were made in the Riviera Valley during the Mesoscale Alpine
Programme (Weigel et al. 2006, 2007; Rotach and Zardi 2007). Many good methods
and techniques have been developed for processing and analyzing turbulence
measurements from towers (see, e.g., Kaimal and Finnigan 1994), but airborne
turbulence measurements are more difficult to make and analyze. Wind velocity
components are retrieved as a difference between absolute airplane motion with
respect to a fixed frame of reference and airspeed as indicated by the aircraft
(Crawford and Dobosy 1992). These two terms are at least an order of magnitude
larger than the difference between them. Thus, high accuracy in the wind velocity
components requires quite high accuracy in the original measurements. Airborne
measurements are, nonetheless, the most promising tool for analysis of turbulence
properties far from the surface in the middle of valley volumes (Rotach et al. 2004;
Weigel et al. 2007).
The dynamics of diurnal valley winds results from the combination of slope
flows on the sidewalls and boundary layer processes on and above the valley floor,
including such factors as free convection, shear, and compensatory subsidence
in the valley core atmosphere. As a consequence, turbulent processes associated
with valley winds occur on different space scales (height above the slope, slope
length, boundary layer depth over the valley floor, valley width) and time scales
(hour, morning, afternoon, daytime, nighttime). As a consequence it is difficult to
ascertain their individual roles and interactions. A high site-to-site variability must
be considered in designing measurement strategies (Doran et al. 1989). Even the
application of standard turbulence analysis procedures, such as eddy covariance,
requires consideration of site-specific characteristics. (Kaimal and Finnigan 1994;
de Franceschi and Zardi 2003; Hiller et al. 2008; de Franceschi et al. 2009).
Over a homogeneous plain under fair weather conditions with weak synoptic
winds, shear will be minimized and convection will be the dominant process
producing turbulence. Under the same synoptic conditions, thermally driven flows
will develop in valleys, and the dominant process leading to turbulence will be shear
production, as turbulence in valleys is rarely caused by convection alone (Weigel
et al. 2007; de Franceschi et al. 2009).
Some observations suggest that the classical Monin-Obukhov surface layer
scaling for velocity and temperature variances holds in the middle of a wide valley
with a flat floor (Moraes et al. 2004; de Franceschi 2004; de Franceschi et al. 2009).
However, at other locations (e.g. sloping sidewalls) scaling approaches that are
valid for flat horizontally homogeneous terrain need to be extended and/or modified
(Andretta et al. 2001, 2002; van Gorsel et al. 2003a; Grisogono et al. 2007; Catalano
and Moeng 2010).
In Switzerland’s Riviera Valley during daytime in summer, the vertical TKE
distribution throughout the valley core atmosphere scales with the convective
velocity scale
1=3
g 0 0
w D w s zi (2.8)
0
defined, as usual, with gravity acceleration g, a temperature reference value 0 ,

the convective boundary layer (CBL) height zi , and the surface temperature flux
w0 0 s , provided the latter is determined at the slope rather than at the valley center
(Weigel and Rotach 2004; Weigel 2005; Weigel et al. 2006, 2007; Rotach et al.
2008). Despite highly convective conditions, it is mainly shear (as opposed to
buoyancy) that is responsible for TKE production. However the maximum of shear
production occurs not at the valley surface, but rather at higher levels above the
valley. This suggests that wind shear is mainly produced by the interaction of the
up-valley wind with the background winds aloft, rather than by surface friction. As a
consequence shear can be related to the wind strength in the core of the valley wind
and investigators have found that the scaling velocity determined from the surface
heat flux on the slope is closely related to the valley wind strength.
Surface fluxes of momentum, heat and water vapor are strongly affected by
topography and land surface cover (the reader will find various examples of such
variability in Whiteman et al. 1989b; Grimmond et al. 1996; Wotawa and Kromp-
Kolb 2000; Grossi and Falappi 2003; Antonacci and Tubino 2005; Bertoldi et al.
2006; Kalthoff et al. 2006; Serafin 2006; Martı́nez et al. 2010). In particular, vege-
tation can strongly affect turbulence production and surface layer structure, e.g. by
imposing displacement heights and roughness lengths (cf. de Franceschi et al. 2009),
and affecting fluxes and budgets of momentum, energy and mass. Additional
complications arise when boundary layer processes occur in urban areas located
in mountain valleys (Kuttler et al. 1996; Piringer and Baumann 1999; Savijärvi and
Liya 2001; Kolev et al. 2007; Giovannini et al. 2011; Fernando 2010). Transport
and mixing of pollutants in complex terrain have been addressed by many authors
(see, e.g., Salerno 1992; Prévôt et al. 1993, 2000; Wong and Hage 1995; Prabha and
Mursch-Radlgruber 1999; Grell et al. 2000; Doran et al. 2002; Gohm et al. 2009;
De Wekker et al. 2009; de Franceschi and Zardi 2009; Aryal et al. 2009; Kitada and
Regmi 2003; Regmi et al. 2003; Lehner and Gohm 2010; Panday 2006; Panday and
Prinn 2009; Panday et al. 2009; Szintai et al. 2010). Boundary layer processes are
also addressed in Chap. 5.
2.3.2 The Effects of Other Winds on Valley Flows
Valley wind systems are often modified, or even overpowered, by larger-scale

thermally driven flows or by dynamically driven flows above a valley produced by
mesoscale or synoptic-scale processes. Dynamically driven winds, including those
in valleys, are discussed in Chap. 3, so we give here only a short summary.
Background flows above mountainous terrain are driven by large-scale pressure
gradients that are superimposed on the along-valley pressure gradients produced
locally by intra-valley temperature differences. When the valley atmosphere has a
high static stability, the background flows will often be channeled along the valley
axis (Whiteman and Doran 1993). These flows, are not, however, thermally driven
flows. They occur under more active synoptic conditions and do not reverse twice
per day at the normal morning and evening transition times. They are also generally
stronger than the diurnal valley flows and, since they are imposed from aloft, do
not necessarily exhibit the typical jet-like vertical profile usually seen with valley
flows. Nonetheless their interaction with the valley atmosphere may be significantly
affected by changes in the surface heat flux during the diurnal cycle (see e.g. Jiang
and Doyle 2008).
The direction of the flow in a valley (either up- or down-valley) may be driven by
the pressure gradient aloft, rather than by locally generated temperature differences
along the length of the valley. As valleys are often curved or bent, the along-valley
component of the synoptic-scale pressure gradient may differ from one segment of
the valley to another. Accordingly, changes in wind speed and direction will occur
along the axis of a bent valley, depending on factors such as valley orientation, width
and depth (Kossmann and Sturman 2003). This phenomenon in which winds within
the valley are driven by pressure differences between the two ends of the valley is
called pressure-driven channeling. The flow direction is from the high pressure end
toward the low pressure end of the valley. This type of channeling is in contrast to
forced channeling in which momentum from the winds aloft is carried down into
the valley, producing winds along the valley axis whose direction is determined by
the component along the valley axis of the wind velocity aloft. Forced channeling
is often seen near mountain passes. In addition to these major types of channeling,
smaller scale airflow separations, eddies, and dynamic instabilities can be produced
by background flow/terrain interactions that will affect thermally driven flows within
a valley (Orgill et al. 1992).
When static stability is low, downward momentum transport, associated with
convection, can bring winds aloft into a valley with wind directions not necessarily
aligned along the valley axis. These cross-valley flows are more likely to be present
in the upper altitudes of a valley, and are unlikely to be confused with thermally
driven diurnal valley winds. Flows over valley ridges can have smaller-scale
dynamic effects on valley winds. In addition to the channeling processes mentioned
above, other effects on the structure and evolution of diurnal wind circulations
within a valley come from large-scale background flows (see, e.g., Barr and Orgill
1989; Doran 1991; Orgill et al. 1992; Mursch-Radlgruber 1995; Zängl 2008; and
Schmidli et al. 2009a). During daytime, ridge-level winds may significantly modify
the subsidence over the valley center and produce asymmetries in the cross-valley
circulation. A weak ridgetop wind may flush out the upper part of the valley and
modify the valley circulation, whereas a stronger one may intrude into the valley and
even reach the valley bottom. During nighttime, upper winds can penetrate into the
valley and “erode” the existing ground based inversion. However, valley drainage,
once established, is rather resistant to erosion from above, because of atmospheric
stability, especially in valleys having large drainage areas. As an example, Orgill
et al. (1992) investigated the mesoscale external factors affecting nocturnal drainage
winds in four different valleys in western Colorado during the ASCOT project. They
found that in such valleys wind erosion can reduce drainage flows to less than half of
the valley depth, the principal contributing factors to wind erosion processes being
an along-valley wind component above the valley that opposes the drainage, a weak
stability in the valley, and winds aloft with speeds exceeding a threshold around
5 m s1 and accelerating at rates greater than 0.0004 m s2 (Orgill et al. 1992). In
the Rosalian mountain range in eastern Austria, Mursch-Radlgruber (1995) found
that an opposing ambient flow decreased the down-valley flow depth in the shallow
valleys studied, but did not reduce the jet speed maximum. In contrast, an ambient
flow in the same direction as the drainage flow suppressed the drainage flow and
led to stagnation or pooling on one night. This effect was explained by observing
that the ambient wind produced a pressure low in the upper valley in the lee of
the Rosalian range, thus counteracting the normally existing hydrostatic pressure
gradient responsible for classical drainage flow.
The contrast in temperature between air over a water surface and air over the
adjacent land typically drives diurnal sea/lake breezes that can interact with valley
winds, significantly modifying both phenomena. The sea/lake breeze and the up-
valley wind in a nearby valley may merge in a connected wind field called the
Extended Sea Breeze (Kondo 1990) or Extended Lake Breeze (McGowan et al.
1995), producing long-range moisture transport (Sturman and McGowan 1995).
This transport may provide significant moisture through dew deposition to arid
valleys, as seen at the Elqui Valley in the Andes (Khodayar et al. 2008). Small
bodies of water on valley floors do not develop appreciable lake breezes, although
they can significantly affect surface energy budgets, valley wind strength and extent,
and moisture distribution (Sturman et al. 2003a, b). Relatively large lakes on valley
floors can develop lake breezes that interact with slope and valley wind systems,
with internal boundary layers forming at the lake-land boundary (McGowan and
Sturman 1996; Zardi et al. 1999; de Franceschi et al. 2002). Bergström and Juuso
(2006) used a numerical simulation to compare up-valley winds occurring with and
without a water body at the bottom of the valley, finding that the lower lake surface
temperatures lead to higher daily average wind speeds.
2.3.3 Topographic Effects on Valley Flows
2.3.3.1 Tributary Valleys
In valleys with tributaries, upstream flows divide at each confluence of a side valley
with the main valley, and downstream flows merge into a single down-valley wind.
Bifurcating flows and merging flows have not been extensively researched, but
one would expect that the topography in the vicinity of the merging valleys and
wind system characteristics in the main and tributary valleys would be determining
factors for the structure of the valley wind system. Because of the many different
forms of topography and the range of wind system characteristics, it is not clear if
investigation of a specific instance would provide information general enough to be
applicable to other valleys. However detailed studies of the structure and dynamics
of flow mergings or bifurcations at single situations are crucial to provide a better
understanding of local pollutant concentrations, as shown by Banta et al. (1997),
Fiedler et al. (2000) and Zängl and Vogt (2006). The requirement of mass flux
balance may result in a closed circulation in which return flows produce enhanced
subsidence or rising motions over the valley juncture, affecting temperature and
wind structures there (Freytag 1988; Zängl 2004).
The bulk effect on nighttime along-valley mass flux in a main valley by
inflow from a tributary box canyon was investigated in the Atmospheric Studies
in Complex Terrain (ASCOT) program in the 1980s (Coulter et al. 1989, 1991;
Porch et al. 1991). The tributary provided more mass flux to the main canyon
than would have come from a simple valley sidewall. Simulations of ideal valley-
tributary systems by O’Steen (2000) supported this conclusion. They also showed
that the increase in down-valley mass flux in the main valley begins well before
the tributary is encountered, suggesting that the mechanism by which a tributary
provides additional mass flow to the main valley is more complicated than suggested
by the intuitive notion of a simple side-stream contribution at the confluence with a
concomitant increase in valley flux downstream of the tributary. In some cases the
flow down the main valley actually decreases in a layer corresponding to the outflow
height of the tributary flow. This leads to a negative correlation of flow down the
main valley and flow out of the tributary. Cold air flowing out of the tributary is
apparently blocked by cold air flow coming down the main valley. Cold air thus
tends to accumulate in the tributary and then enter the main valley through surges.
Alternatively, the tributary contribution may glide over the main valley’s core flow.
Conservation of momentum requires that the increased mass of cold air from the
tributary slows the main valley flow, allowing additional flow to exit the tributary.
The latter may assume the form of a regularly pulsating outflow, and modulate low-
frequency oscillations in the main valley wind, as shown by Porch et al. (1991).
Drainage flow from tributaries depends not only on cooling rates, but also
on ambient winds. Unexpectedly, an external wind blowing in opposition to the
drainage wind can strengthen the drainage wind. The mechanism is not fully
understood: perhaps circulation cells introduce mass from the ambient wind into the
tributary system (Coulter et al. 1989). In any case the orientation of the tributaries
to the main valley appears to be a controlling factor in the interactions between
tributary and valley flows, and periodic flow out of the tributaries can modulate
oscillations in the main valley drainage flow (Coulter et al. 1991).
2.3.3.2 Canyons
Canyons, valleys having relatively high depth-to-width ratios and steep sidewalls,
are subject to the same up- and down-valley flow regimes as valleys. A canyon
that is a segment of a longer valley may have one of two effects on drainage
flows. If the canyon is narrow or tortuous, it may restrict the valley flow. During
nighttime this may cause a cold-air pool or cold air lake to develop above the
canyon constriction (Whiteman 2000). If, on the other hand, the canyon is not
too narrow, the air may accelerate through the constricted canyon to conserve the
total momentum of the downvalley flow. The flow and temperature structure in a
canyon are affected by several physical processes. Deep narrow canyons have weak
downslope flows on the steep slopes (see Sect. 2.5.1) because of a reduced longwave
radiation loss caused by the restricted view of the sky and by enhanced radiative
interactions between the sidewalls that tend to drive the canyon atmosphere toward
isothermalcy. Turbulence generated at the ridgeline by terrain interactions with
above-canyon flows, or turbulence generated in valley flows by terrain elements,
tend to mix the canyon atmosphere (Start et al. 1975). Simulations of nocturnal
drainage flows in small rugged canyons (Lee et al. 1995) have shown that coupling
with upper winds may produce multi-vortex structures that are very sensitive to
the radiation budget. Canyons sometimes act as conduits between segments of a
valley having quite different climates (e.g., the Columbia Gorge, a canyon between
Oregon and Washington, USA). Strong temperature contrasts between the segments
can accelerate the normal canyon flows (e.g., Sharp and Mass 2004).
Arizona’s Grand Canyon of the Colorado River, one of only a few deep canyons
that have been investigated, maintains only weak stability during both day and night
in winter (Whiteman et al. 1999c). This may be due to terrain-enhanced interactions
between above-canyon flows and air within the canyon or to overturning of the
canyon atmosphere by cold air generated on snow-covered mesas to the north and
south of the canyon. Temperature jumps form at the top of the canyon when warm
air advection occurs across the canyon rim. Because stability in the canyon is near-
neutral, wintertime valley flows are weak and the flow along the canyon is driven by
pressure gradients that are superimposed on the canyon by traveling synoptic-scale
weather disturbances, with flow directed from the high pressure to the low pressure
end of the canyon. Nonetheless, weak diurnal valley flows are sometimes present
at the north end of the canyon in the vicinity of the Little Colorado River (Banta
et al. 1999).
2.3.4 Modeling of the Valley Flows
The state of the art on modeling of thermally driven flows was reviewed by Egger
(1990). Contributions to the understanding of diurnal valley flows have come from
a range of modeling approaches including conceptual, laboratory tank, Navier–
Stokes, analytical and large-eddy simulation models. By means of a simplified
mixed layer model over idealized topography, Kimura and Kuwagata (1995) showed
how the sensible and latent heat that accumulate in an atmospheric column are
dependent on the scale of the topography. Whiteman and McKee (1982) used
a simple thermodynamic model to explain the basic mechanisms governing the
morning breakup of nocturnal inversions in valleys. Zoumakis and Efstathiou
(2006a, b) extended this approach to obtain a criterion for estimating the time
required for the complete breakup and development of a CBL. Conceptual and
analytical models relating the energy budget and valley geometry to the rate of
atmospheric cooling in a valley were developed by McKee and O’Neal (1989).
Vergeiner (1987) used an analytical model to study basic relationships between
the strength and response time of a valley flow as a function of topographic
characteristics such as valley length, width, or depth. A hydraulic model reproducing
the development of diurnal up-valley wind, based on concepts of mixed layer
schemes, was proposed by Zardi (2000).
Similarly to what has been discussed above for slope winds, a more complete
understanding of the atmospheric processes leading to diurnal valley flows has been
provided by numerical models. In the last two decades the achievements in this field
have benefitted from the increasing availability of suitable computers and from the
impressive progress in the development of numerical weather prediction models.
Further information on numerical modeling is provided in Chaps. 9, 10 and 11.
Diurnal mesoscale circulations such as slope and valleys flows are not resolved
by the coarse grids in general circulation and climate models, and must accordingly
be parameterized as turbulent motions. However, only smaller-scale turbulence and
convection are presently represented in boundary-layer parameterizations, based on
similarity arguments and data from flat terrain experiments. Even though vertical
heat fluxes induced by the diurnal slope and valley flows are as large as or even
larger than turbulent fluxes produced by smaller-scale turbulence and convection,
they are not yet considered in these parameterizations (Noppel and Fiedler 2002;
Rotach and Zardi 2007).
Simulations of idealized valleys, with simplified initial and boundary conditions,

have led to a deeper insight into individual mechanisms affecting the dynamics of
diurnal valley winds. These simulations are particularly valuable in identifying the
relative roles of processes that are difficult to isolate in more realistic topography and
weather situations. While many idealized simulations have already been mentioned
in this chapter, space is unavailable here to provide a detailed account of all of
them. Representative articles include those of Enger et al. (1993); Atkinson (1995);
Anquetin et al. (1998); Li and Atkinson (1999); O’Steen (2000); Rampanelli et al.
(2004); Bergström and Juuso (2006); Bitencourt and Acevedo (2008); Schmidli
et al. (2010); Schmidli and Rotunno (2010); and Serafin and Zardi (2010a, b, c).
The complexity of real mountain valleys provides a stimulating challenge for
numerical simulations. The spatial resolution of present models is not yet fine
enough to reproduce all the small-scale features of valley winds. High resolution
digital elevation databases have to be significantly smoothed before they are used
for model runs, and information on land cover, land usage, and soil moisture is
often approximate or incomplete. Nevertheless, many simulations of real cases have
proven useful, as the reader may appreciate by inspecting the reference papers
by Leone and Lee 1989; Gross 1990; Koračin and Enger 1994; Fast 1995, 2003;
Ramanathan and Srinivasan 1998; Colette et al. 2003; Zhong and Fast 2003;
De Wekker et al. 2005; Chow et al. 2006; Lee et al. 2006; Weigel et al. 2006;
and Bischoff-Gauß et al. 2006, 2008. Model outputs must be carefully compared
with available observations to make sure that the simulations capture the relevant
mechanisms. Understanding of the model outputs benefits from good visualization
tools and from physical insight and conceptual models. Additional confidence is
gained when comprehensive field observations are available for comparison. Thus,
simulations are usually done in conjunction with large observational programs.
2.3.5 Exceptions and Anomalies
The occurrence of anomalous valley winds, i.e. developing in apparent contradiction

to established concepts and theories, has stimulated broad debates and deeper
analyses since earlier investigations in the field. Wagner’s (1938) seminal paper, for
example, mentioned two Alpine wind systems that were known to contradict valley
wind theory. The Maloja wind in Switzerland’s Upper Engadine Valley northeast of
the Maloja Pass and the Ora del Garda in the Adige (in German, Etsch) Valley near
Trento, Italy, are both known to blow down-valley during daytime.
The Maloja wind was investigated in a modeling study by Gross (1984, 1985),
who also focused on the development of a narrow, elongated cloud (the Maloja
Schlange, English: Maloja snake) that sometimes forms in conjunction with the
wind and stretches northeastward from the pass. The anomalous Maloja wind is
an up-valley wind from the Bergell Valley to the southwest of the pass that intrudes
into the Upper Engadine Valley as a down-valley wind. The explanation for the
anomalous wind is the peculiar topography in the Maloja Pass region, where the
Bergell ridgelines extend beyond the pass into the Upper Engadine Valley.
Fig. 2.11 A well known example of confluence between valleys (Wagner 1938): the area north
of the city of Trento, where the “Ora del Garda” wind, blowing from the Valle dei Laghi (a)
through an elevated saddle (), interacts with the up-valley wind blowing in the Adige Valley
from South (b) (From de Franceschi et al. 2002. © American Meteorological Society. Reprinted
with permission)
The outflow of the Ora del Garda in the Adige Valley during the early afternoon
appears as an “abnormal flow development” (Wagner 1938) over a whole segment
of the valley including the city of Trento. There, the wind near the surface flows
down-valley for the whole afternoon, rather than blowing up-valley as expected.
The down-valley flow is caused by the spillover into the Adige Valley of relatively
cold air through an elevated saddle from the upper Valle dei Laghi, which joins
the Adige Valley just up-valley from Trento (Fig. 2.11). In the late morning a
lake breeze develops at Lake Garda at the southern inlet of the Valle dei Laghi.
The cold lake air is carried up-valley and over the elevated saddle. The airflow
blowing down from this saddle onto the Adige valley floor in the early afternoon
on fair weather days brings potentially cooler air, which then flows underneath the
up-valley wind blowing at that time in the Adige Valley. The flow splits into an up-
valley ground-level flow north of the junction and a down-valley flow south of the
junction. Earlier researchers, relying only on surface observations, could not explain
the phenomenon. Subsequent field campaigns (reported in Wagner 1938) and recent
ones (de Franceschi et al. 2002) identified the vertical structure of the valley
atmosphere and demonstrated that the flows are compatible with known dynamics.
In recent years, other cases of anomalous winds have been observed and
explained. In most cases the aberrant winds are produced by a large source of
cold or warm air at one end of the valley or the other. For example, nighttime
down-valley winds often fail to reverse during daytime in winter when the valleys
are covered with snow (Emeis et al. 2007; Schicker and Seibert 2009). Austria’s
Inn Valley (Whiteman 1990) and Germany’s Loisach Valley (Whiteman 2000) are
examples. Similarly, valleys with large sources of cold air at their lower end (e.g.,
an ocean or a long lake) can have up-valley winds that fail to reverse at night. The
strong and continuous cold air advection can cause the valley atmosphere to remain
stably stratified during daytime with only weak development of a shallow convective
boundary layer at the valley floor. Examples of valleys exhibiting this behavior
include Chile’s Elqui Valley (Kalthoff et al. 2002; Khodayar et al. 2008), which
ends at the Pacific Ocean, and California’s San Joaquin Valley (Zhong et al. 2004;
Bianco et al. 2011). Up-valley flows in the San Joaquin Valley are also promoted by
a large source of warm air in a desert area at its upper end.
Anomalous nighttime up-valley and daytime down-valley winds occur at several
sites along the Colorado River upstream of the Grand Canyon (Whiteman et al.
1999a). The valley floor in this area is a series of basins connected by narrow
canyons. The anomalous winds represent flows that converge at the lower ends of
the basins during nighttime and diverge out at the upper ends of the basins during
daytime. Another anomaly can be observed in the Marble Canyon area north of the
Grand Canyon where nighttime winds flow up-valley. This has been attributed to
the presence of a major geological formation that tilts downward in the up-valley
direction (Whiteman et al. 1999a).
2.4 Phases of and Interactions Between the Slope

and Valley Winds
The evolution of the coupled slope and valley systems during a typical day
progresses through four distinct phases, described below.
2.4.1 Daytime Phase
This phase typically begins in mid to late morning, after the nighttime valley
temperature inversion has been removed by daytime heating. The air in the valley is
warmer than over the adjacent plain, producing a lower pressure in the valley than
over the plain, and the resulting horizontal pressure gradient causes air to flow up
the valley axis (Fig. 2.8, upper panel).
Once the nocturnal temperature inversion is destroyed, convection from the
heated valley surfaces extends above the valley ridges. Vertical mixing associated
with the convection couples the flow in the valley to the background or synoptic-
scale flows aloft. The pressure gradients producing these flows are superimposed on
the local, thermally produced pressure gradient within the valley. Winds within the
valley can be affected by this coupling, although the background flows are generally
channeled along the valley axis. It is typical, in fact, that wind speeds, both in a
valley and over a plain, increase in the afternoon due to the coupling with the usually
stronger winds aloft (Whiteman 2000).
Maintenance of the daytime up-valley flow requires that temperatures in the
valley cross-section be greater than at the same level over the adjacent plain (or
at the same altitude farther down the valley). The TAF contributes to additional
heating in the valley, as solar radiation coming across the top of the valley at
ridgetop level, once converted to sensible heat flux at the valley floor and sidewalls,
overheats the air in the valley relative to the adjacent plain because of the smaller
volume in the valley. Radiative processes are also more effective within the valley
than over the surrounding plain because the valley air mass is partially enclosed
by radiating surfaces. In addition, advective heat transfer to the valley atmosphere
is more effective than the convective heat transfer to the same elevation over the
lower-altitude plain, again due to the smaller volume of air to be heated. The extra
heating in the valley invigorates the convection from the valley after the inversion is
destroyed, but this effect decreases as the mixed layer grows deeper above the ridges.
2.4.2 Evening Transition Phase
The surface energy budget reverses above valley surfaces when outgoing longwave
radiation first exceeds incoming shortwave radiation. This usually occurs initially in
late afternoon in areas of the valley that are shadowed by terrain. With this energy
budget reversal, the ground begins to cool and a downward turbulent sensible heat
flux begins to remove heat from a layer of air above the surface. A shallow layer
of cooled air thus forms over the sidewalls and floor of the valley. Over the slopes,
the air in this layer becomes negatively buoyant compared to the air adjacent to the
slope, and begins to move down the slope. The downslope flows converge onto
the valley floor or, after a temperature inversion forms on the valley floor, into
the elevated portion of the inversion. Rising motions over the valley center that
compensate for downslope flows on the sidewalls produce upward cold air advection
that cools the entire valley cross-section (Whiteman 2000; Fast and Darby 2004;
Brazel et al. 2005). Other processes play a role, including radiative transfer and
the continuing cold air advection associated with the daytime up-valley flow. Up-
valley winds continue to blow for several hours after the energy budget reversal
because of inertial effects and because it takes some time for the cooling over the
valley slopes to be transferred to the entire valley volume through compensatory
rising motions and other processes. The temporary post-sunset continuance of the
up-valley flow generates shear that increases turbulent sensible heat flux in the
surface layer (Kuwagata and Kimura 1995, 1997; de Franceschi et al. 2002, 2009;
de Franceschi 2004). Once the air in the valley cross-section becomes colder
than the air over the adjacent plain, the pressure gradient reverses and the winds
reverse from up-valley to down-valley. The reversal occurs first in the ground-based
stable layer (temperature inversion) that forms on the valley floor. The stable layer
containing down-valley winds grows rapidly in depth in the early evening, causing
the down-valley winds to increase in strength and depth. The rate of growth of the
inversion gradually decreases as it fills the valley (Whiteman 1986) and the cooling
rate decreases exponentially for the remainder of the night. The bulk of the total
nighttime cooling occurs in the first few hours following sunset (De Wekker and
Whiteman 2006). The residual air layer, previously flowing up-valley, is then lifted
above the growing stable layer. This up-valley flow gradually decreases in strength
until it is overcome by background flows above the valley, which are often channeled
along the valley axis.
During the evening transition period the downslope flows that defined the
beginning of the period gradually decay as a strong stable layer builds up within the
valley and as down-valley flows in the stable layer expand to fill the valley volume.
The downslope flows on the valley sidewalls within the growing stable layer respond
to the increasing ambient stability by decreasing their strength. They continue on the
upper sidewalls of the valley above the growing stable layer, but leave the sidewall
and converge into the main valley volume once encountering the growing stable
layer (cf. Catalano and Cenedese 2010).
During the period of weakening up-valley flow and reversal to down-valley
flow, strong radiative cooling along the uneven valley floor and slopes results in
the formation of shallow cold air layers, that preferentially fill low spots on the
valley floor and extend up the sidewalls. These near-surface stable layers decouple
the surface from the flow aloft. During this period of light and variable winds
the surface-based temperature inversion often strengthens (Fast and Darby 2004).
The formation of the shallow cold air layer on the floor and sidewalls tends to
decouple the valley winds (whether up- or down-valley) from the friction of the
earth’s surface. Sometimes, an acceleration in valley wind caused by this decoupling
increases the shear across the shallow layers and briefly mixes them out, producing
transient evening temperature rises at the valley floor or on the slopes (Whiteman
et al. 2009). Oscillations may be generated in the valley and slope winds if the cold
air layer repeatedly reforms and decays.
2.4.3 Nighttime Phase
The beginning of the nighttime phase is hard to define because it is tied to the
gradual buildup of a stable layer containing down-valley winds rather than to a
discrete event. Typically, in clear undisturbed conditions the nighttime phase starts
sometime during the period from several hours after sunset until about midnight.
The main characteristic of the nighttime phase is the predominance of down-
valley flows throughout the valley volume. The maintenance of the down-valley
wind requires that the atmospheric cross-section continues to cool relative to the
adjacent plain. Because the downslope flows become weaker and shallower during
the latter part of the evening transition phase and are more and more affected by
(i.e., turned into the direction of) the down-valley flows (Whiteman and Zhong
2008) the importance of compensatory rising motions over the valley center to
this cooling is reduced. Instead, horizontal and vertical shears at the valley surfaces
produced by the increasing strength of the down-valley flow throughout the cross-
section become increasingly responsible for turbulent transport of heat toward the
radiatively cooling surfaces and the continued cooling of the valley cross-section.
The pressure gradient that drives the valley flow is at a maximum at the valley
floor. The vertical wind profile, however, takes a jet-like shape with the wind
maximum displaced 10–100 s of meters above the floor. The wind speed is reduced
at the surface by friction (Khodayar et al. 2008), and the winds above the jet
gradually give way to the gradient flow above (Zhong et al. 2004). Wind profiles
on a horizontal valley cross-section also show a jet-like shape with the peak wind
over the valley center where the frictional influence of the sidewalls is minimized.
The nighttime down-valley wind current is thus sometimes called a down-valley
jet (DVJ), and vertical and horizontal cross-sections through the current in a deep
valley can be described in terms of Prandtl’s well-known mathematical expression
and a parabolic profile, respectively (Clements et al. 1989; Coulter and Gudiksen
1995; Fast and Darby 2004). Because the peak winds generally occur over the
valley center, measurements taken at one location on a valley cross-section may
be not representative of the overall wind structure (King 1989). The establishment
of the DVJ is sometimes associated with abrupt warming at low levels as a result
of downward mixing and vertical transport of warm air from the inversion layer
above (Pinto et al. 2006). Time-periodic phenomena, such as pulses in the strength
of the DVJ (possibly contributing to vertical transport by creating localized areas
of low-level convergence) as well as gravity waves and Kelvin–Helmholtz waves,
which facilitate vertical mixing near the surface and atop the DVJ, are commonly
observed. These are reflected in the spectra of velocity and temperature fluctuations
(Stone and Hoard 1989). Typical periods of these oscillations are around 20 min
(Porch et al. 1991).
Nocturnal drainage flow along the valley floor displays very irregular behavior,
its structure being strongly determined by the local orography (Trachte et al. 2010).
Experimental evidence confirms the intuitive idea that drainage winds “slide” along
the ground and strongly adhere to its shape (Clements et al. 1989; Eckman 1998;
Fiedler et al. 2000; Haiden and Whiteman 2005). The down-valley flow sloshes
toward the outside of valley bends due to inertial forces (Sakiyama 1990).
2.4.4 Morning Transition Phase
The morning transition begins shortly after sunrise, when incoming solar radiation
exceeds longwave radiation loss and the surface energy budget reverses on the
sidewalls. The end of the morning transition is marked by the break-up of the
ground-based inversion that fills the valley at night. Inversions break up occurs in
a fundamentally different way in valleys (Whiteman 2000) than over plains (Stull
1988). Whereas the plains atmosphere is heated primarily by the upward growth of
an unstable or convective boundary layer, heating of the valley atmosphere results
not only from the (limited) growth of an unstable boundary layer (cf. Weigel et al.
2007), but also from subsiding motions in the valley core that compensate for
upslope flows on the sidewalls.
After sunrise, as the slopes and valley floor are illuminated by the sun, the
surface energy budget reverses, and heat is transferred upward from the surface to
an initially shallow but progressively growing layer of air above the heated surfaces.
The layer of warming air over the sidewalls is on an inclined surface. Air at any point
within this layer is warmer than the air outside the layer at the same height, resulting
in an upslope flow. Differences in insolation on slopes – due to slope exposure,
shadows, ground cover, etc. – cause the incipient upslope flows to be somewhat
intermittent and inhomogeneous. Convection above the heated slopes and valley
floor entrains air from the elevated remnant of the nocturnal inversion (“stable core”)
into the upslope flow, which carries it to the valley ridgelines and vents it into the
free atmosphere. During this process, some horizontal transport of mass and heat can
contribute to the heating of the valley atmosphere, especially when mid-level stable
layers are encountered, or when topography projects into the valley atmosphere.
The transport of heated air up the sidewalls is compensated by sinking of the stable
core over the valley center. Because the stable core is statically stable, subsidence
causes downward advection of potentially warmer air, heating the atmosphere over
the valley center, and eventually causing the valley atmosphere to become warmer
than the air over the plain. The resulting horizontal pressure difference between the
valley atmosphere and the atmosphere over the adjacent plain initiates an up-valley
flow.
The break-up of nocturnal inversions in valleys was investigated in western
Colorado valleys by Whiteman (1982), resulting in a conceptual model and a simple
thermodynamic model (Whiteman and McKee 1982), and by Brehm and Freytag
(1982) in the Alps. Break-up requires that a valley be warmed from its initial
inversion temperature profile to a neutral profile having the potential temperature
of the inversion top at sunrise. The two main processes accomplishing the breakup
are the warming associated with convective boundary layer growth upward from
the heated ground and the warming caused by subsidence that compensates for
mass removal from the valley atmosphere by upslope flows. The warming can be
amplified by valley geometry as explained by the TAF principle. The time required
to destroy the Colorado inversions depended on inversion strength at sunrise and
the rate of input of sensible heat from the valley surfaces after sunrise, which
varies with season and snow cover. The Colorado valley inversions typically broke
up within 3–5 h after sunrise. Bader and McKee (1983), in a numerical model
study, found that a third mechanism, horizontal heat transfer from the sidewalls,
could also heat the valley atmosphere and contribute to inversion breakup. This
process is particularly effective in the short period of time between sunrise and
the time when upslope flows extend continuously up the sidewalls. Upslope flows
may detach from the slopes and intrude horizontally into the valley atmosphere at
elevations where sub-layers of high stability occur within the stable core (Harnisch
et al. 2009). Eckman (1998) and Rampanelli et al. (2004) performed simulations
in idealized valleys to investigate along-valley aspects of the inversion breakup.
They found that if the plain and the valley floor are perfectly horizontal, effects
of valley-plain contrasts concentrate in a region close to the mouth of the valley
and weaken with distance up the valley. Accordingly, the valley winds develop only
in a limited region upstream and downstream of the mouth of the valley. However,
when the valley floor is even weakly inclined, buoyancy effects can drive valley
winds farther up the valley (Rampanelli et al. 2004). By modifying Whiteman
and McKee’s thermodynamic model of valley inversion breakup, Zoumakis and
Efstathiou (2006a, b) focused on the relative partitioning of energy into convective
boundary layer growth and subsidence warming, but without considering horizontal
heat transfer. The occurrence of warm or cold air advection over the valley during
the inversion breakup period can change the heating requirement (Whiteman 1982;
Whiteman and McKee 1982). Warm air advection increases the time required to
destroy the inversion, while cold air advection has the opposite effect. Further, cold
air advection within the valley atmosphere increases the heating requirement, and
thus the time required to break the inversion.
The warming effect of compensatory sinking motions decreases in wider valleys
(Serafin and Zardi 2010b) and simulations show that compensatory sinking plays
only a small role in inversion breakup if the valley’s width to depth ratio exceeds
about 24 (Bader and McKee 1985). Inversions take longer to break up in wintertime
and in deep valleys (Whiteman 1982; Kelly 1988; Colette et al. 2003). The
asymmetric distribution of surface sensible heat flux between the opposing sidewalls
of a valley may result in a nonuniform inversion breakup (Urfer-Henneberger 1970;
Kelly 1988; Ramanathan and Srinivasan 1998; Anquetin et al. 1998; Matzinger
et al. 2003; Whiteman et al. 2004b). Gravity waves and horizontal mixing within
the stable layer effectively redistribute warm air and reduce temperature differences
due to differential sidewall heating (Bader and McKee 1983). Cross-valley flows
may develop and be superimposed on the valley circulations when sensible heat
fluxes differ greatly on the opposing sidewalls of a valley (Gleeson 1951; Bader and
Whiteman 1989; Atkinson 1995).
Especially in narrow valleys, the stable core inhibits the growth of a convective
boundary layer above the valley floor and sidewalls and delays the development
of deep convection (Furger et al. 2000; Bischoff-Gauß et al. 2006; Chemel and
Chollet 2006). Indeed, convective boundary layers over a valley floor are typically
not as deep as over an adjacent plain (Dosio et al. 2001; de Franceschi et al.
2003; Rampanelli and Zardi 2004; Rampanelli et al. 2004; Weigel and Rotach
2004; Bergström and Juuso 2006; Chemel and Chollet 2006; Weigel et al. 2006;
Bischoff-Gauß et al. 2008). Convective eddies in the relatively shallow convective
boundary layer over the valley floor efficiently mix and horizontally homogenize the
convective boundary layer. It thus exhibits weak cross- and along-valley variability
over the valley floor (Zängl et al. 2001; De Wekker et al. 2005; Chemel and Chollet
2006) unless strong terrain heterogeneities exist (Ramanathan and Srinivasan 1998).
Switzerland’s Riviera Valley (Weigel and Rotach 2004; De Wekker et al. 2005;
Weigel et al. 2006) exhibits some hitherto unknown meteorological characteristics
that are produced by a bend in the valley. During the morning transition period
inertia causes the up-valley flow that enters the valley from the Magadino plain to
drift toward the outside of a bend at the lower end of the valley and develop a cross-
valley wind component. This produces a cross-valley tilt of the inversion layer and
anomalous downslope flows on the sunlit western sidewall that transport warm air
downward toward the valley floor, stabilizing the valley atmosphere. Cold-air ad-
vection in the up-valley flow, subsidence of warm air from the free atmosphere aloft
(compensating for up-slope flows), and curvature-induced secondary circulations
in the southern half of the valley produce a tendency for the valley atmosphere to
remain rather stable all day, although a shallow convective boundary layer does
grows slowly upward from the valley floor during daytime. The valley circulation
thus has a prolonged morning transition phase, rather than a daytime phase. Other
investigators (Cox 2005; Bergström and Juuso 2006; Bischoff-Gauß et al. 2006)
have also reported valleys that maintain persistent stable layers during daytime up-
valley flows. A common characteristic of several such valleys is a large source of
cold air at the valley mouth, such as a long lake or an ocean. During daytime, a stable
stratification and negative sensible heat fluxes can be maintained over lake and ocean
surfaces while the nearby land surfaces maintain strong positive fluxes (McGowan
and Sturman 1996; Bischoff-Gauß et al. 2006). These cases provide examples of
another mechanism by which cold air can be advected continuously up the valley,
reducing temperatures, maintaining stability and suppressing CBL growth.
2.5 The Mountain-Plain Wind System
The daytime heating and nighttime cooling contrast between the mountain atmo-
sphere over the outer slopes of the mountain massif and the free atmosphere over the
surrounding plain produces the horizontal pressure differences that drive this wind
system. The mountain-plain wind system brings low level air into the mountain
massif during daytime (the plain-to-mountain wind), with a weak return circulation
aloft. During nighttime, the circulation reverses, bringing air out of the mountain
massif at lower levels (the mountain-to-plain wind), with a weak return circulation
aloft. The daytime circulations are naturally somewhat deeper and more energetic
than the nighttime circulations because of the larger daytime heat fluxes and the
correspondingly deeper daytime boundary layer. The reversal of this large-scale
wind system is somewhat delayed relative to the slope and valley wind systems
because of its larger mass.
The mountain-plain circulation is associated with a diurnal pressure oscillation
between the mountainous region and the adjacent plains (for the Alps, cf. Frei and
Davies 1993; for the Rocky Mountains, cf. Li and Smith 2010). For the largest
mountain, this oscillation can be detected in one or more modes of the atmospheric
tides observed at global scale (Dai and Wang 1999).
According to Ekhart (1948), the mountain-plain wind system occupies a separate
outer layer from the valley and slope wind systems (see Fig. 2.1). An alternative way
of thinking of the mountain-plain wind would be to consider it as encompassing both
the slope and valley wind systems, as all three wind systems interact to transport
mass to and from a mountain massif from its surroundings.
Mountain-plain circulations play a key role in moisture transport and initiation of
moist convection (see also Chaps. 6 and 7) as well as in long-range transport of air
pollutants and their precursors (see also Chap. 5). Convergence of the daytime flows
over the mountains produces afternoon clouds and air mass thunderstorms, and the
divergent nighttime sinking motions produce late afternoon and evening clearing.
Pollutants associated with population centers on the plains adjacent to mountains are
carried toward the mountains during daytime, where they are transported vertically
to higher altitudes (Henne et al. 2005).
Due to the large spatial extent of mountain-plain circulations, their basic features
can be monitored by conventional networks of surface and upper air observing
stations (Kalthoff et al. 2002; Lugauer and Winkler 2005; Henne et al. 2005, 2008;
Taylor et al. 2005; Steinacker et al. 2006; Bica et al. 2007; Li et al. 2009). Mountain-
plain flows are generally weak, usually below 2 m s1 , with much lower speeds
in the broader and often deeper return flows aloft, and can easily be overpowered
by prevailing large scale flows. The mountain-plain circulation is diurnal, with its
evolution occurring in phases (Bossert and Cotton 1994a, b) and may be influenced
by large-scale conditions (Lugauer and Winkler 2005). The characteristic weather
associated with the plain-to-mountain circulation depends on the moisture levels
in the surroundings. Where dry air masses are present, such as in the Colorado
Plateau (Bossert et al. 1989; Bossert and Cotton 1994a, b), the circulations are
marked by strong winds, low cloudiness, and isolated thunderstorms. Where the
mountains are surrounded by oceans or other sources of low-level moisture, as in
Western Sumatra (Sasaki et al. 2004; Wu et al. 2009) or the Alps (Ekhart 1948;
Weissmann et al. 2005), the daytime rising motions transport moisture to higher
levels of the atmosphere and produce widespread clouds and precipitation (Kalthoff
et al. 2009). In some cases mountain-plain circulations also interact with sea breezes
(Taylor et al. 2005; Hughes et al. 2007).
Persistent temperature differences between the mountain atmosphere and the
surroundings occur seasonally in some mountain ranges. Seasonal thermal contrasts
are the primary driver of monsoonal circulations, that bring cold air out the
mountain massif during winter and into the mountain massif during summer. These
large-scale seasonal circulations are a type of mountain-plain circulation, often with
diurnal oscillations of wind strength superimposed on them. The initiation of the
summer monsoon caused by the heating of the mountain atmosphere in spring
is responsible for the initiation, for example, of the Himalayan Monsoon (Barros
and Lang 2003) and the variously named North American, Southwest, Mexican or
Arizona Monsoon that affects the southwestern United States (Higgins et al. 2006).
The remainder of this section will discuss mountain-plain circulations that

develop around the east–west oriented Alps and around the north–south oriented
Rocky Mountains of North America. The mountain-plain system develops regard-
less of the orientation of a mountain range. The mountain-plain circulations that
develop between plains and elevated basins or plateaus are discussed in Sect. 2.6.
2.5.1 Mountain-Plain Winds in the Alps
The theory of diurnally reversing mountain wind circulations published by Prof.

A. Wagner (1932, 1938) included a postulated “equalizing flow” between the
plains and mountains. The evidence for these flows was collected by Wagner and
his students, most notably E. Ekhart and A. Burger. Burger and Ekhart (1937)
analyzed summer wind data obtained by the optical tracking of pilot balloons
released twice-daily from 20 stations located in and around the Alps to find that
a daytime circulation brought air into the Alps from the surrounding plains, with
a weak compensation current (“Ausgleichströmung”), or return circulation, aloft.
The thermal origin of this phenomenon and the associated horizontal pressure
differences that drive the mountain-plain circulation were investigated further by
Ekhart (1948). In recent years new analysis tools such as the Vienna Enhanced
Resolution Analysis (VERA; Steinacker et al. 2006; Bica et al. 2007) have been
used with the dense network of surface stations in the Alps to provide additional
information on the thermal forcing of mountain-plain circulations.
New information on the Alpine mountain-plain circulation has come from a
research program called Vertikaler Austausch und Orographie (VERTIKATOR,
or Vertical Exchange and Orography), which supported a major field experiment
in July 2002 on the north side of the Alps in the Bavarian foreland (Lugauer
et al. 2003). On summer days radiative heating of the Alps produces a horizontal
transport of air from the foreland into the Alps, and a vertical transport from the
boundary layer into the free troposphere above the mountains. The daily vertical
transport of air and moisture over the Alps by the mountain-plain circulation has
been termed “Alpine pumping” (Lugauer et al. 2003). Climatological investigations,
using the extensive network of surface stations in the Alpine foreland, have provided
additional information on the spatial and temporal extent of the circulation (Lugauer
and Winkler 2005; Winkler et al. 2006). The circulation is strongest on days when
background upper-air flows are weak and when daily incoming solar radiation
exceeds about 20 MJ m2 . Because incoming radiation is strongest in summer,
the flows are most common and best developed between April and September.
The diurnally reversing mountain-plain circulation extends approximately 100 km
north of the Alps to the Danube River. The low-level flow coming into the Alps
during daytime, with speeds generally less than 2 m s1 , extends gradually farther
into the plains during daytime and extends vertically to reach about 1–2 km
depths at the edge of the Alps by mid-afternoon. Rising motions over the Alps
produce cloudiness and precipitation, while the descending return branch of the
circulation suppresses convective cloud formation north of the Alps. An intensive

VERTIKATOR field study on 12 July 2002 used an airborne Doppler lidar, a wind-
temperature radar, dropsondes, and rawinsondes to gain additional information on
the daytime circulation (Weissmann et al. 2005). Vertical velocities in the rising
branch of the circulation were on the order of 0.05–0.10 m s1 . Despite its large
scale, the plain-to-mountain circulation responded readily to local geographical and
dynamic features. High-resolution simulations with numerical models were able
to reproduce the general flow structure and, when combined with vertical motion
data from the lidar, were able to estimate the daytime mass fluxes into the Alps
(Weissmann et al. 2005).
2.5.2 Mountain-Plain Winds in the Rocky Mountains
A robust mountain-plain circulation, forced by the diurnal cycle of solar heating and
longwave cooling, is a key feature of flows around the Rocky Mountains (Reiter and
Tang 1984). The circulation is strongest on clear days and nights, when background
synoptic-scale flows are weak. These conditions occur most frequently in summer
and fall. The diurnal geopotential height differences between the Rocky Mountains
and the surrounding plains, and the resulting near-surface diurnal inflows and
outflows, cause moisture convergence and divergence (Reiter and Tang 1984). The
daytime convergence over the mountain crests produces afternoon thunderstorms
and affects precipitation patterns. The nighttime divergence causes sinking motions
over the mountains that tend to reduce cloudiness and precipitation. The inflow
and outflow cycle can be detected in wind observations at mountaintop locations,
with inflows and outflows at speeds of several meters per second (Bossert et al.
1989). Transition periods tend to be long. The transition from the daytime inflow to
nighttime outflow occurs from 1900 to 0200 MST, the outflow to inflow transition
takes place from 0600 to 1100 MST. The transition from inflow to outflow is much
faster on highly convective days when evaporation of precipitation cools the heated
boundary layer over the mountains.
Harmonic analysis of summer wind data from the network of radar wind profilers
in the central United States shows that diurnal circulations extend up to 1,200 km
eastward from the Rocky Mountains over the gently sloping Great Plains to the
Mississippi River (Whiteman and Bian 1998). The strongest mountain-plain flows
occur in the lowest kilometer of the atmosphere where speeds, as averaged over the
100-m range gates of the radars, reach up to 3.5 m s1 just east of the mountains.
The return flow aloft is distributed through a layer extending above the highest
mountains (4,000 m) at speeds of a few tenths of a meter per second. The
mountain-plain circulation decays with distance from the mountains.
Much of the research on the mountain-plain circulation has focused on the
extensive plains to the east of the Rocky Mountains, where observational networks
are fairly dense, but the wind system also is present on the western slopes of the
Rockies and on the periphery of the mountain massif (Reiter and Tang 1984).
2.6 Diurnal Wind Systems in Basins and Over Plateaus
Basins and plateaus, common landforms in many parts of the world, influence
atmospheric motions on a variety of scales, from tens of meters to thousands of
kilometers. Basins are hollows or depressions in the earth’s surface, wholly or partly
surrounded by higher land. An example of a basin that is only partly surrounded
by mountains is the Los Angeles basin (Lu and Turco 1995), which is open to
the Pacific Ocean on its western side. Closed basins with level unbroken ridgetops
are rarely encountered. Plateaus are elevated, comparatively level expanses of land.
Large plateaus include the Tibetan Plateau and the Bolivian Altiplano. A plateau
partially or completely surrounded by mountains can also be referred to as an
elevated basin.
Diurnal wind systems in basins and over plateaus are driven by horizontal pres-
sure gradients that develop between the atmosphere within or over the topography
and the free atmosphere of its surroundings. Basins and plateaus experience diurnal
slope and mountain-plain wind systems but, if there is no well-defined channel,
they lack a valley wind system. The large-scale mountain-plain wind system that
carries air between an elevated basin and the surrounding plains is called a basin-
plain wind. The daytime branch is the plain-to-basin wind and the nighttime branch
is the basin-to-plain wind (Kimura and Kuwagata 1993). Similarly, the mountain-
plain wind system over a plateau is called a plateau-plain wind. The daytime branch
is the plain-to-plateau wind and the nighttime branch is the plateau-to-plain wind
(Mannouji 1982).
2.6.1 Diurnal Wind Systems in Basins
Meteorological experiments have been conducted in basins all around the world,
including the Aizu (Kondo et al. 1989) and Ina Basins (Kimura and Kuwagata
1993) and a small frost hollow (Iijima and Shinoda 2000) in Japan, the McKenzie
Lake basin (Kossman et al. 2002; Zawar-Reza et al. 2004; Zawar-Reza and Sturman
2006a, b) in New Zealand, the Grünloch (Pospichal 2004; Whiteman et al. 2004a,
b, c; Steinacker et al. 2007) and Danube Valley (Zängl 2005) basins in Austria,
other small Alpine basins or dolines (Vertacnik et al. 2007), the Duero Basin in
Spain (Cuxart 2008); a basin on the island of Majorca (Cuxart et al. 2007; Martı́nez
2011), in the Columbia (Whiteman et al. 2001; Zhong et al. 2001) and Colorado
Plateaus basins (Whiteman et al. 1999b) of the western United States, and in
Colorado’s Sinbad Basin (Whiteman et al. 1996; Fast et al. 1996), Utah’s Peter
Sinks basin (Clements et al. 2003), and Arizona’s Meteor Crater (Savage et al. 2008;
Whiteman et al. 2008; Yao and Zhong 2009; Fritts et al. 2010; Lehner et al. 2010;
Whiteman et al. 2010; Haiden et al. 2011).
Slope winds are a feature of basin meteorology, just as they are for valley
and plateau meteorology. Basins, however, if surrounded by mountains, have no
Table 2.2 Extreme minimum temperatures in selected basins (Data from most U.S. locations
come from the U. S. National Climatic Data Center [NCDC])
Location Temperature minimum Date Reference
ı ı
Peter Sink, Utah 56.3 C (69.3 F) Feb 1, 1985 Pope and Brough (1996)
West Yellowstone, Montana 54ı C (66ı F) Feb 1933 NCDC
Taylor Park, Colorado 51ı C (60ı F) Feb 1951 NCDC
Fraser, Colorado 47ı C (53ı F) Jan 1962 NCDC
Stanley, Idaho 48ı C (54ı F) Dec 1983 NCDC
Gruenloch Basin, Austria 52.6ı C (63ı F) 19 Feb– Aigner (1952)
4 Mar 1932
equivalent to the valley wind. The absence of this wind system is responsible for the
unusually cold nighttime minimum temperatures often experienced in basins. The
low nighttime temperature minima (see examples in Table 2.2) can be explained by
contrasting the situation in a closed basin to that in a valley. Air in a valley at night,
once it becomes colder than air at the same level over the plains, begins to flow
down the valley axis and is replaced by air that sinks into the valley from above.
The air that sinks into the valley is statically stable and also warms adiabatically
as it sinks, so that it is warmer than the air it is replacing. The down-valley flow
thus causes a continuous import of warm air into a valley from aloft, reducing the
potential for cold nighttime minima. Basins, in contrast, trap air in place and cool it
continuously throughout the night. Extremes of minimum temperature are attained
by radiative cooling on clear, dry, winter nights following synoptic incursions of
cold air masses into a region. Cooling is increased on long winter nights and can be
especially strong if a fresh cover of new snow insulates the basin air from ground
heat fluxes. When winds aloft are weak, buoyancy forces tend to cause isentropes
to become horizontal within the inversion in a confined basin. Thus, temperature
measurements on the sidewalls become good proxies for free air temperatures over
the basin center, allowing temperature “soundings” to be made from surface-based
instruments (Whiteman et al. 2004a).
Nocturnal cooling is often confined within a basin during the night, although in
shallow basins the cooling may extend above the ridgeline and produce basin-to-
plain outflows over the lower passes (Pospichal et al. 2003). Because TAF causes
additional cooling in basins compared to the surrounding plains, a temperature jump
may develop at night above the basin between the air cooled within the basin and
the still-warm free atmosphere above (see e.g. Clements et al. 2003).
2.6.1.1 The Diurnal Cycle in Basins
In fair weather conditions, undisturbed by clouds or background flows, the diurnal

temperature structure in basins evolves similarly from day to day. During daytime,
the basin atmosphere becomes well mixed, with a neutral temperature profile. In
late afternoon, as shadows are cast on the sidewalls, the net longwave loss exceeds
the net solar input, and the local energy budget reverses, with downward turbulent
sensible heat flux removing heat from a shallow air layer above the slopes. As the
shadows progress, the energy budget reverses over more and more of the surface
area of the basin and downslope flows begin as cold air layers form over the slopes.
A common misperception, which arises by making an analogy between air and
water, is that the nighttime cooling of the basin air mass is produced solely by the
convergence of downslope flows onto the basin floor and the cooling induced by
compensatory rising motions over the basin center. If this were the case, a series of
soundings of the basin atmosphere would show the continuous rising of the top of a
temperature inversion as the cold pool became deeper and deeper. In fact, as a layer
of cold air builds up on the floor of a basin, the downslope flows no longer have
the temperature deficit to reach all the way to the basin floor, and they converge
near the top of the growing inversion layer. Observations show that much of the
cooling in a basin takes place continuously throughout the basin’s full depth, and
this cooling continues throughout the night, even after the downslope flows slow
and die as the ambient stability builds up within the basin. The radiative and sensible
heat flux divergences that produce the cooling responsible for inversion buildup over
flat plains, where no downslope flows are involved, also play a major role in basins.
De Wekker and Whiteman (2006) have compared the cooling rates of valley, plain
and basin atmospheres and have found that basin atmospheres cool more rapidly
than those over plains or valleys. This is, no doubt, due to the TAF, the sheltering
of the basin from background winds by surrounding terrain (Vosper and Brown
2008), the divergence of heat from the confined volume, and the exclusion of diurnal
valley flows (which advect warm air into valleys from above and thus reduce their
cooling rates). The mechanisms by which the basin atmosphere cools have been dis-
cussed by Magono et al. (1982), Maki et al. (1986), Maki and Harimaya (1988), Neff
and King (1989), Kondo and Okusa (1990), and Vrhovec (1991). The cooling rates
of basin (and valley) atmospheres usually decrease during the night, but the relative
roles of turbulent and radiative flux divergences and advection are still unresolved.
Cooling rates in basins and valleys can be reduced by cloudiness or when heat is
released at the ground with the formation of dew or frost (Whiteman et al. 2007).
Nighttime temperature inversions in basins break up after sunrise (Triantafyllou
et al. 1995; Whiteman et al. 2004b) following the same patterns that have been
observed in valleys. The warming of the basin atmosphere and the ultimate
destruction of the nocturnal inversion is accomplished by convective currents from
the heated floor and sidewalls, the transport of mass from the basin by advection in
upslope flows over the sidewalls, and warming due to subsiding motions over the
basin center that compensate for mass carried up the sidewalls. Subsiding motions
are weaker in wider basins. Cross-basin asymmetries in the warming and in the
convective boundary layer and wind structure can occur because of shadows cast
into the basin by the surrounding terrain and by spatial variations in the receipt of
solar radiation on sloping surfaces of different aspect (Hoch and Whiteman 2010).
Cross-basin differences in insolation produce cross-valley temperature and pressure
differences that produce cross-basin flows toward the warmer sidewall (Lehner et al.
2010). Inversions in small basins in Austria and the Western US, despite their
great strength, break up somewhat earlier than inversions in western U.S. valleys
(Whiteman et al. 2004b).
10
0
Temperature (°C)
-5
-10
-15
-20
-25
-30
12 00 12 00 12 00 12 00 12
30 Nov 1 Dec 2 Dec 3 Dec 4 Dec
Time (MST)
Fig. 2.12 Temperature time series for surface-based temperature sensors at different heights on
the sidewall of the Gruenloch Basin, Austria. Because the coldest air settles into the bottom of the
basin, the lowest elevation sensors report the lowest temperatures and the highest elevation sensors
report the highest temperatures
The daytime cycle of warming often extends high above a basin, because of
strong daytime sensible heat fluxes and convection. The horizontal pressure gra-
dients between the air over the basin and the air over the surrounding plain produce
a plain-to-basin pressure gradient during daytime. In large basins the inflows are
delayed by upslope circulation cells from inside and outside the basin that become
anchored to the ridgeline surrounding the basin. This mechanism is explained
further in the plateaus section below. A modeling study by De Wekker et al. (1998)
shows that flows into an idealized basin (one with consistent ridge height) come
over the basin ridgetop when horizontal pressure differences are present above the
ridgeline, but the pressure gradient between air confined inside the basin and air
outside the basin at the same height plays no role in initiating or maintaining the
plain-basin flow unless there are gaps in the ridgeline.
Temperature inversions in basins often develop quite regularly on undisturbed
nights, but there are a variety of perturbations that can occur in their development
under disturbed conditions. A typical time series of temperature measurements at
different heights on the sidewalls of the 150-m deep Gruenloch Basin are shown in
Fig. 2.12 for a period of four consecutive winter days following new snow on 30
November. The night of 1–2 December was a typical undisturbed night with clear
skies and weak background winds. On disturbed nights, temperature inversions in
this basin exhibit a variety of disturbances to the normal evolution of temperature
structure, as shown in Fig. 2.13. These schematic illustrations of selected cold-pool
disturbances were observed using the same sensors as for Fig. 2.12, but on disturbed
nights in the 9-month period from October 2001 to early June 2002. Inversion
structure is sensitive to cloudiness, to background flows (especially Föhn or strong
Fig. 2.13 Schematic depiction of disturbances to nocturnal temperature inversions in Austria’s

Gruenloch Basin as revealed in temperature traces from instruments located at different heights
on the sidewalls (compare Fig. 2.12). The different forms of disturbances in the sub-figures can
be compared to the first sub-figure in which an undisturbed temperature evolution is shown and a
rough indication of the time and temperature scales for this and subsequent sub-figures is given.
The disturbances are produced by strong background winds, shear at the top of the inversion, and
variations in cloudiness and surface energy budget (From Dorninger et al. 2011)
winds above the basin), turbulent erosion at the top of the cold pool (Petkovšek
1992; Rakovec et al. 2002; Zhong et al. 2003), to cold-air intrusions coming over
the basin ridgeline (Whiteman et al. 2010; Haiden et al. 2011) and to variations in
the surface energy budget. A numerical modeling study by Zängl (2003) determined
the effects of upstream blocking, drainage flows and geostrophic pressure gradient
on cold-air pool persistence for a basin similar in shape to the Gruenloch Basin.
2.6.1.2 Persistent Wintertime Inversions in Basins
Persistent or multi-day temperature inversions or cold-air pools have generally

adverse effects on valley and basin populations (Smith et al. 1997). Temperatures
remain cold during such events, suppressing the normal diurnal temperature range.
The strong stability in the inversion causes air pollution and moisture to build up
from sources within the pool. Poor air quality may persist for days (cf. Cuxart and
Jiménez 2011). Fog may form and lift to produce stratus. Light rain, snow or drizzle
may be produced if clouds persist or become thicker, or if they are seeded by higher
clouds. Rain or drizzle may freeze when reaching the cold ground. Low cloud bases,
visibility restrictions in fog, rain, snow or drizzle, and icy runway conditions may
close airports and cause transportation problems.
Multi-day temperature inversions are primarily a wintertime phenomenon. They
are produced when incoming energy during daytime is insufficient to remove cold
air that has settled into the basin or valley. The cold air may form initially through
the usual nocturnal inversion formation processes or may be left in low-lying terrain
following the passage of a cold front. Inversions, whether at the incipient or fully
formed stages, can be strengthened by differential advection when relatively warm
air is advected over the basin or can be weakened when cold air is advected over the
basin (Whiteman et al. 1999b). The shorter days of winter, lower sun angles, and the
presence of snow cover are usually responsible for the changes in the surface energy
budget at the basin floor that reduce the daytime turbulent sensible heat flux and
convection below that required to break the inversion. Additionally, fog or stratus
clouds that form in the cold-air pool may reduce insolation and sensible heat flux at
the cold pool base.
Both the onset and cessation times of persistent cold-air pools are difficult to
forecast. Forecast “busts” lead to especially large errors in maximum and minimum
temperature forecasts. Forecasting the cessation of cold-air pools is particularly
problematic because the many different processes that can affect breakup occur on
different scales. Local processes such as evaporation or melting at the basin surface
or changes in cloud cover within the pool have an effect, as do synoptic-scale warm
or cold advection above the pool, turbulent erosion at the top of the cold pool, or
drag or pressure forces that may be strong enough to push the cold pool to one side
of a basin, to cause seiches within the pool, or to produce vertical oscillations at
the top of the cold-air pool. Forecasts of cold pool breakup usually are based on
the forecast passage of troughs or low-pressure weather systems (cf. Reeves and
Stensrud 2009). These passages are accompanied by advection of colder air above
the pool, destabilizing the cold-air pool, and by stronger winds that may assist in
breakup. Cold air remnants are likely to persist at the lowest altitudes of a valley
or basin even when the cold pool is thought to be fully removed. Studies of cold-
air pools in the Colorado Plateaus Basin show that the final breakup, including any
remnants at the lowest elevations, usually occurs in the afternoon with assistance
from a final burst of convection (Whiteman et al. 1999b).
While most of the research on persistent cold-air pools to date has been on non-
cloudy pools, cloudy cold-air pools sometimes form as moisture builds up within the
inversion layer due to evaporation from the underlying ground. These pools have
a weaker, near-moist-adiabatic stability. This is apparently caused by the sinking
of cold air produced at cloud top by longwave radiation loss and the resultant
overturning and vertical mixing between the cloud top and the ground. Because
of the weaker stability, air pollutants generated inside cloudy cold-air pools will be
better mixed in the vertical than for non-cloudy pools.
2.6.2 Diurnal Wind Systems over Plateaus
Diurnal wind systems over plateaus have received much less research attention than
diurnal wind systems in valleys and basins. Nonetheless, simulations of idealized
plateaus have appeared (Egger 1987; De Wekker et al. 1998; Zängl and Chico 2006)
and field studies have been conducted on the Mexican Plateau (Bossert 1997; Doran
et al. 1998; Fast and Zhong 1998; Doran and Zhong 2000; Whiteman et al. 2000),
especially in connection with air pollution problems (de Foy et al. 2008), in a valley
feeding the Himalayan Plateau (Egger et al. 2000, 2002; Zängl et al. 2001) and
on the Altiplano of Bolivia (Zängl and Egger 2005; Egger et al. 2005; Reuder and
Egger 2006).
Plateaus provide elevated surfaces on which nighttime cooling and daytime
heating can take place. During nighttime, air is cooled above the plateau. When
the plateau is a tabletop, much of the cold air that forms (especially if the tabletop is
narrow or has a small area) drains off the plateau sides, leaving only a shallow and
relatively weak inversion on the plateau. If the plateau is surrounded by mountains
the cooling can be confined in the elevated basin and produce stronger and deeper
inversions than over plateaus with no surrounding mountains. During daytime, a
convective boundary layer grows above the elevated heating surface. The daytime
destruction of the nocturnal inversions often involves interactions with flows aloft
(Banta and Cotton 1981; Banta 1984), which are generally stronger at higher
elevations. If there is a weak nocturnal inversion on the plateau, the convection
is able to grow quickly upwards through the inversion and can often reach quite
high levels of the atmosphere. The warm air above a plateau can be advected off
the plateau into the surrounding atmosphere if background winds are present. For
example, the easterly advection of warm air from the South Park plateau in the
central Rocky Mountains of Colorado can sometimes be detected east of the plateau
where it affects the development of afternoon thunderstorms over the adjacent plains
(Arritt et al. 1992).
An interesting latency occurs in the development of the daytime wind over
plateau regions. There is often a very late break-in of plain-to-plateau or plain-to-
basin winds that occurs in the late afternoon or early evening rather than in the
late morning when the temperature difference between the air over the plateau
and its surroundings starts to build up, as illustrated in a conceptual model in
Fig. 2.14 based on observations and modeling of the Mexican Plateau (Whiteman
et al. 2000). During daytime, convection builds up a deep boundary layer above
an elevated plateau that is warmer than the air surrounding the plateau. Upslope
flows form on the outer periphery of the plateau and on the plateau itself, with
return circulations aloft. The circulation cells converge at the plateau rim or edge,
and the convergence anchors the circulation cells there during daytime (Bossert and
Cotton 1994a, b). The normal tendency of the atmosphere to produce circulations
to equilibrate temperature differences that develop between the air over the plateau
and its surroundings is therefore restrained. Cold air can intrude onto the plateau,
Fig. 2.14 Schematic diagram of daytime circulations and temperature profiles above and along-
side an elevated circular plateau, illustrating the daytime buildup of a baroclinic zone at the edges
of the plateau that leads to a break-in of cold winds onto the plateau in the evening. The dry
adiabatic lapse rate d is provided as a reference
however, through passes and gaps at the plateau edge and through valleys that cut
into the plateau (Egger 1987; Doran and Zhong 2000; Egger et al. 2000, 2002, 2005;
Zängl et al. 2001; Zängl and Egger 2005). Because the circulations are anchored to
the plateau edge, a baroclinic zone forms during daytime around the periphery of
the plateau, where the temperature gradient becomes strong. As the sun starts to
set, the slope flows weaken, the circulation cell boundaries are no longer anchored
to the plateau edge, and a strong flow of cold air comes across the baroclinic
zone onto the plateau. This sudden break-in of cold air onto the plateau from all
sides often occurs in early evening, and its progression across the plateau depends
on plateau width. Model simulations show that gravity waves are an important
mechanism that allows the warm air in the convective column above the plateau
to equilibrate with the surrounding air during the nighttime (Whiteman et al. 2000;
Zängl and Chico 2006).
As an example, the afternoon or early evening break-in of flows from outside a
plateau or elevated basin is depicted in the conceptual model of Fig. 2.15 for the
South, Middle and North Park basins of Colorado, as based on observations and
numerical modeling by Bossert and Cotton (1994a, b).
Fig. 2.15 Diurnal flow patterns in the Central Colorado Rocky Mountains, looking northward
from southern Colorado to the elevated basins of South Park, Middle Park and North Park. (a)
During daytime, upslope flows from inside and outside the basins converge over the basin rims. (b)
During late afternoon, cold air from the surroundings breaks into the basins across the rims (From
Bossert and Cotton 1994a. © American Meteorological Society. Reprinted with permission)
2.7 Forecasting
The many practical applications of knowledge concerning diurnal mountain winds

were enumerated in the introduction to this chapter. Because diurnal mountain winds
affect so many aspects of human life and the environment, there is a need to make
accurate forecasts of diurnal mountain winds.
Forecasting of diurnal mountain winds is, in some respects, relatively simple.
Climatological investigations have told us much about the regular development and
characteristics of diurnal mountain winds. They form regularly and reliably under
high-pressure synoptic conditions when skies are clear and background winds are
weak. The underlying topography provides the channels for these flows. Airflows
are down the terrain (slopes, valleys and mountains) during nighttime and up
the terrain during daytime, with lags that correspond to the mass of the wind
system and the corresponding rate of heat transfer required to reverse it. The larger
the temperature contrasts that form inside the terrain or between the terrain and
the surrounding plain, the stronger the flows. Stronger temperature contrasts and
stronger flows occur when the diurnal cycle of sensible heat flux is strongest,
in the summer half-year. Increasing cloudiness and interactions with strong flows
aloft decrease the temperature and pressure contrasts that drive the wind systems.
A decreased amplitude of the sensible heat flux or failure of the sensible heat
flux to change sign diurnally decreases the intensity of the diurnal wind systems
or eliminates the diurnal reversal that is their chief characteristic. The decreased
amplitude could be caused by rain, high soil moisture, increasing cloudiness, or
an increased albedo due to snow. Changes in climate along a valley can affect
the amplitude of temperature contrast along a valley’s axis, affecting wind system
strength. The diurnal wind system is often present, but weaker and less regular in its
time of reversal, on cloudy and windy days.
The more difficult aspect of forecasting comes with the need to predict wind
system strength and characteristics for specific locations and at specific times.
Present-day operational weather models are presently able to provide accurate
simulations of general synoptic weather conditions and are now reaching the
horizontal resolution (1–2 km) necessary to resolve large valleys such as the
Inn, Rhine and Rhone Valleys of the Alps. These models, however, need further
improvements to provide accurate forecasts for individual valleys. Horizontal and
vertical grid and terrain resolutions should be improved, as well as boundary layer
and surface property parameterizations. We know that some valleys in a region
have stronger wind systems than others and that wind systems in some valleys are
more susceptible to interference from external conditions. The increasingly high
resolution attained by numerical research models is expected to provide more and
more precise indications of the relative strengths of wind systems in different valleys
under synoptic conditions, or even of different segments of an individual valley.
Such simulations, however, will need to be evaluated with field data, to verify that
the models are performing adequately. Some simpler models, such as the German
Weather Service’s proprietary KLAM 21 drainage flow model (Sievers 2005) have
proven useful in simulations on a range of scales from an individual vineyard to
a domain of several tens of kilometers. A simpler approach (McKee and O’Neal
1989) using topographic maps investigates the along-valley variation of TAF (i.e.,
terrain cross-sections) and may have some utility, although the results to date have
been mixed.
A number of pattern recognition approaches have been used to determine
connections between synoptic-scale patterns and observed features of local winds
in multi-valley target areas where wind and supporting data are available (Guardans
and Palomino 1995; Weber and Kaufmann 1995, 1998; Kaufmann and Weber
1996, 1998; Kastendeuch and Kaufmann 1997; Kaufmann and Whiteman 1999;
Kastendeuch et al. 2000; Ludwig et al. 2004). In the Basel, Switzerland area, for
example, smaller valleys in the mountain complex have a high frequency of drainage
flows, while larger valleys more often experience interactions with synoptic-scale
winds that produce channeled flows. Up-valley flows are generally stronger in this
valley complex than down-valley flows, and the frequency of days with thermally
driven flows varies little with season (Weber and Kaufmann 1998). Although the
theoretical number of possible wind patterns is very large in a complicated terrain
area, complex topography tends to constrain the observed fields to a small number
of typical patterns. Recurrent wind patterns can also be determined by means of
conditional sampling (Zaremba and Carroll 1999). Such information should prove
useful for forecasting.
To overcome the inherent complexity and highly resource-demanding dynamical
modeling, an alternative approach is offered by the increasing performance and
flexibility of Geographic Information Systems (GIS). The latter can easily represent
small-scale local features strongly affecting diurnal flows such as topography,
land cover, surface budgets, etc. Supported by statistical analysis of long term
observations, and with an input of ambient conditions from either station networks
or large-scale numerical models, they can provide a first guess of wind and
temperature fields at local scale, as well as a basis for the estimate of related
quantities such as surface fluxes of heat, water vapor and passive scalars (Carrega
1995; Fast 1995; Ciolli et al. 2004; Deng and Stull 2005, 2007; Bertoldi et al. 2006;
Schicker and Seibert 2009).
When forecasts are needed for a mountainous region, both real time and clima-
tological data from weather stations can be analyzed to get a basic understanding
of the strength and timing of the diurnal wind systems. Where quantitative data
are available, general conceptual models of the phenomena can be used along
with advice from individuals with local experience. Major federal projects, such as
fighting a wildland fire, can sometimes justify the on-site assignment of a weather
observer/forecaster, who can assist with interpretation and provide useful daily
feedback on forecast accuracy.
2.8 Past Progress, Future Objectives
Significant progress has been made in the understanding of diurnal mountain winds
since the last AMS book on mountain meteorology was published (Blumen 1990).
Field experiments, both large international efforts and single investigator projects,
have been conducted taking advantage of new instrumentation and data platforms.
Modeling of diurnal mountain winds has improved as advances have been made in
hardware and software and as more detailed field data have become available.
Over the past two decades, field studies and modeling of diurnal mountain
winds have worked in tandem. This collaboration should continue: field studies
provide the data required to evaluate and refine models, and modeling work is
invaluable in conceptualizing the processes observed in the field. Improvements can
be expected in both areas. The introduction of higher horizontal and vertical grid
and terrain resolution, better numerical approaches, more adequate boundary layer
parameterizations, better modeling of turbulent processes and better handling of
non-uniform surface cover and irregular terrain into models will bring significant
advances to our understanding of mountain and valley atmospheres. The use
of remote sensing instruments capable of scanning larger target volumes (e.g.,
De Wekker and Mayor 2009) is expected to produce more comprehensive and
highly resolved observational data sets, which will allow the proper evaluation of
increasingly sophisticated models.
In recent years, international cooperation has become a hallmark of research
of diurnal mountain winds. Yearly international conferences alternating between
the AMS Mountain Meteorology Conference and the International Conference
on Alpine Meteorology, periodic short courses and large, international research
programs will continue to strengthen international ties.
Future research into diurnal mountain winds will benefit not only from broad
cooperation within the global mountain meteorology community but also from
interdisciplinary cooperation. Researchers in related fields – including mathematics,
numerical analysis, physics, fluid mechanics, turbulence dynamics, chemistry, and
hydrology – will contribute to a clearer understanding of the mountain and valley
atmosphere.
In the coming decades, an effort must be made to make research findings
available to the applications community. Weather forecasters, environmental and
civil engineers, land use planners, agricultural experts and wildland fire managers,
among others, would benefit from the knowledge gained and the tools developed by
basic research.
To date, research on diurnal mountain winds has mostly focused on individual
phases of the wind systems and individual processes, including the morning
transition, the evening transition, slope winds, surface energy budgets and so on.
This research will continue to be valuable. A number of interesting questions invite
further investigation.
• With respect to slope flows, additional research is needed on upslope flows, on
turbulent and radiative processes, on the external influences of ambient stability
and ambient flows, and on the effects on slope flows of inhomogeneities of

surface properties and heat budget components along a slope. Models should
be developed to apply to realistic slopes and should be tested against new, higher
resolution data sets. Large eddy simulation models should be applied to gain
knowledge of mean and turbulent processes in slope flows.
• Understanding of valley flows would benefit from a comprehensive investigation
of heat transport processes within the valley atmosphere, of the factors affecting
inversion development and characteristics, and of the interactions between flows
from multiple valleys.
• Research programs on mountain-plain circulations occurring in the major moun-
tain ranges in the world could build on the VERTIKATOR results from Europe,
with many scientific and practical benefits. For instance, concerning the Rocky
Mountains, such a program could improve understanding of the role of the
mountain-plain circulations on the transport of moisture and pollutants to the
Rocky Mountains in diurnal and monsoonal wind systems. These circulations
are closely connected to the Great Plains Low-Level Jet and the North American
Monsoon, and play a role in the timing and movement of thunderstorm and
mesoscale convective systems over the Great Plains.
• Research is needed on persistent wintertime cold-air pools to address increasing
air pollution problems being experienced in urban basins and to improve the
forecasting of their buildup and breakup. Because wintertime cold-air pools are
experienced in many countries, an international research program on this topic
would be warranted.
Additionally, and perhaps more importantly, there is a clear need to take
a comprehensive look at thermally driven circulations in complex terrain, the
interactions of the component parts, the interactions of these winds with winds
aloft and the role of turbulence in wind characteristics. The initiatives investigating
these processes should be multi-national, interdisciplinary and include weather
forecasters and applications scientists.
Acknowledgments We thank the book editors for the opportunity to contribute to this book and
the associated 2008 Mountain Weather Workshop. Special thanks go to authors and organizations
which provided figures for this chapter. The University of Trento and Prof. Dino Zardi are thanked
for providing the financial support and warm hospitality for Dr. Whiteman’s 2-month visit to
the University of Trento in 2008 to begin the collaborative work on this chapter. Moreover
D. Zardi is grateful to the University of Trento for granting him a sabbatical leave during the
academic year 2009/2010. We greatly appreciate the editorial assistance of Johanna Whiteman,
the initial reviews of portions of this chapter by Thomas Haiden, Thomas W. Horst, Stephan
De Wekker, Tina Katopodes Chow, and Stefano Serafin and the valuable comments provided by
anonymous reviewers. Dr. Whiteman wishes to acknowledge partial funding support from National
Science Foundation grants ATM-0837870 and ATM-0444205 and from the National Oceanic and
Atmospheric Administration’s National Weather Service CSTAR grant NA07NWS4680003.
References
Aigner, S., 1952: Die Temperaturminima im Gstettnerboden bei Lunz am See, Niederösterreich
[The minimum temperatures in the Gstettner basin near Lunz, Lower Austria]. Wetter und
Leben, Sonderheft, 34–37.
Allwine, K. J., B. K. Lamb, and R. Eskridge, 1992: Wintertime dispersion in a mountainous basin
at Roanoke, Virginia: Tracer study. J. Appl. Meteor., 31, 1295–1311.
Amanatidis, G. T., K. H. Papadopoulos, J. G. Bartzis, and C. G. Helmis, 1992: Evidence of
katabatic flows deduced from a 84 m meteorological tower in Athens, Greece. Bound.-Layer
Metor., 58, 117–132.
André, J. C., and L. Mahrt, 1982: The nocturnal surface inversion and influence of clear-air
radiative cooling. J. Atmos. Sci., 39, 864–878.
Andretta M., A. Weiss, N. Kljun, and M. W. Rotach, 2001: Near-surface turbulent momentum flux
in an Alpine valley: Observational results. MAP newsletter, 15, 122–125.
Andretta, M., A. H. Weigel, and M. W. Rotach, 2002: Eddy correlation flux measurements in
an Alpine valley under different mesoscale circulations. Preprints 10th AMS Conf. Mountain
Meteor., 17–21 June 2002, Park City, UT, 109–111.
Anquetin, S., C. Guilbaud, and J.-P. Chollet, 1998: The formation and destruction of inversion
layers within a deep valley. J. Appl. Meteor., 37, 1547–1560.
Antonacci, A., and M. Tubino, 2005: An estimate of day-time turbulent diffusivity over complex
terrain from standard weather data. Theor. Appl. Climatol., 80, 205–212.
Arritt, R. W., J. M. Wilczak, and G. S. Young, 1992: Observations and numerical modeling of an
elevated mixed layer. Mon. Wea. Rev., 120, 2869–2880.
Aryal, R. K., B.-K. Leeb, R. Karki, A. Gurung, B. Baral, and S.-H. Byeonb, 2009: Dynamics of
PM2.5 concentrations in Kathmandu Valley, Nepal. J. Haz. Mat., 168, 732–738.
Atkinson, B. W., 1981: Meso-scale Atmospheric Circulations. Academic Press, 495pp.
Atkinson, B. W., 1995: Orographic and stability effects on valley-side drainage flows. Bound.-
Layer Meteor., 75, 403–428.
Atkinson, B. W., and A. N. Shahub, 1994: Orographic and stability effects on day-time, valley-side
slope flows. Bound.-Layer Meteor., 68, 275–300.
Axelsen, S. L., and H. van Dop, 2008: Large-eddy simulation and observations of slope flow. Acta
Geophysica. 57, 803–836.
Axelsen, S. L., and H. van Dop, 2009: Large-eddy simulation of katabatic flow over an infinite
slope. Acta Geophysica. 57, 837–856.
Bader, D. C., and T. B. McKee, 1983: Dynamical model simulation of the morning boundary layer
development in deep mountain valleys. J. Climate Appl. Meteor., 22, 341–351.
Bader, D. C., and T. B. McKee, 1985: Effects of shear, stability and valley characteristics on the
destruction of temperature inversions. J. Climate Appl. Meteor., 24, 822–832.
Bader, D. C., and C. D. Whiteman, 1989: Numerical simulation of cross-valley plume dispersion
during the morning transition period. J. Appl. Meteor., 28, 652–664.
Baines, P. G., 2001: Mixing in flows down gentle slopes into stratified environments. J. Fluid
Mech., 443, 237–270.
Baines, P. G., 2005: Mixing regimes for the flow of dense fluid down slopes into stratified
environments. J. Fluid Mech., 538, 245–267.
Banta, R. M., 1984: Daytime boundary-layer evolution over mountainous terrain. Part I: Observa-
tions of the dry circulations. Mon. Wea. Rev., 112, 340–356.
Banta, R., and W. R. Cotton, 1981: An analysis of the structure of local wind systems in a broad
mountain basin. J. Appl. Meteor., 20, 1255–1266.
Banta, R. M., and P. T. Gannon, 1995: Influence of soil moisture on simulations of katabatic flow.
Theor. Appl. Climatol., 52, 85–94.
Banta, R. M., L. D. Olivier, W. D. Neff, D. H. Levinson, and D. Ruffieux, 1995: Influence of
canyon-induced flows on flow and dispersion over adjacent plains. Theor. Appl. Climatol., 52,
27–42.
Banta, R. M., and Coauthors, 1997: Nocturnal cleansing flows in a tributary valley. Atmos.
Environ., 31, 2147–2162.
Banta, R. M., L. S. Darby, P. Kaufmann, D. H. Levinson, and C.-J. Zhu, 1999: Wind-flow patterns
in the Grand Canyon as revealed by Doppler lidar. J. Appl. Meteor., 38, 1069–1083.
Banta, R. M., L. S. Darby, J. D. Fast, J. Pinto, C. D. Whiteman, W. J. Shaw, and B. D. Orr, 2004:
Nocturnal low-level jet in a mountain basin complex. I: Evolution and effects on local flows.
J. Appl. Meteor., 43, 1348–1365.
Barr, S., and M. M. Orgill, 1989: Influence of external meteorology on nocturnal valley drainage
winds. J. Appl. Meteor., 28, 497–517.
Barros, A. P., and T. J. Lang, 2003: Monitoring the monsoon in the Himalayas: Observations in
Central Nepal, June 2001. Mon. Wea. Rev., 131, 1408–1427.
Barry, R. G., 2008: Mountain Weather And Climate, 3rd Edition. Cambridge University Press,
506pp.
Bastin, S., and P. Drobinski, 2005: Temperature and wind velocity oscillations along a gentle slope
during sea-breeze events. Bound.-Layer Meteor., 114, 573–594.
Bergström, H., and N. Juuso, 2006: A study of valley winds using the MIUU mesoscale model.
Wind Energy, 9, 109–129.
Bertoldi, G., R. Rigon, and T. M. Over, 2006: Impact of watershed geomorphic characteristics on
the energy and water budgets. J. Hydrometeor., 7, 389–403.
Bhatt, B. C., and K. Nakamura, 2006: A climatological-dynamical analysis associated with
precipitation around the southern part of the Himalayas. J. Geophys. Res., 111, D02115.
Bianco, L., I. V. Djalalova, C. W. King, and J. M. Wilczak, 2011: Diurnal evolution and annual
variability of boundary-layer height and its correlation to other meteorological variables in
California’s Central Valley. Bound.-Layer Meteor., 140, 491–511.
Bica, B., and Coauthors, 2007: Thermally and dynamically induced pressure features over complex
terrain from high-resolution analyses. J. Appl. Meteor. Climatol., 46, 50–65.
Bischoff-Gauß, I., N. Kalthoff, and M. Fiebig-Wittmaack, 2006: The influence of a storage lake in
the Arid Elqui valley in Chile on local climate. Theor. Appl. Climatol., 85, 227–241.
Bischoff-Gauß, I., N. Kalthoff, S. Khodayar, M. Fiebig-Wittmaack, and S. Montecinos, 2008:
Model simulations of the boundary-layer evolution over an arid Andes valley. Bound.-Layer
Meteor., 128, 357–379.
Bitencourt, D. P., and O. C. Acevedo, 2008: Modelling the interaction between a river surface and
the atmosphere at the bottom of a valley. Bound.-Layer Meteor., 129, 309–321.
Blumen, W. (Ed.), 1990: Atmospheric Processes over Complex Terrain. Meteor. Monogr., 23
(no. 45), Amer. Meteor. Soc., Boston, Massachusetts.
Bodine, D., P. M. Klein, S. C. Arms, and A. Shapiro, 2009: Variability of surface air temperature
over gently sloped terrain. J. Appl. Meteor. Climatol., 48, 1117–1141.
Bossert, J. E., 1997: An investigation of flow regimes affecting the Mexico City region. J. Appl.
Meteor., 36, 119–140.
Bossert, J. E., and W. R. Cotton, 1994a: Regional-scale flows in mountainous terrain. Part I: A
numerical and observational comparison. Mon. Wea. Rev., 122, 1449–1471.
Bossert, J. E., and W. R. Cotton, 1994b: Regional-scale flows in mountainous terrain. Part II:
Simplified numerical experiments. Mon. Wea. Rev., 122, 1472–1489.
Bossert, J. E., J. D. Sheaffer, and E. R. Reiter, 1989: Aspects of regional-scale flows in mountainous
terrain. J. Appl. Meteorol., 28, 590–601.
Brazel, A.J., H. J. S. Fernando, J. C. R. Hunt, N. Selover, B. C. Hedquist, and E. Pardyjak, 2005:
Evening transition observations in Phoenix, Arizona. J. Applied Meteor., 44, 99–112.
Brehm, M. 1986: Experimentelle und numerische Untersuchungen der Hangwindschicht und ihrer
Rolle bei der Erwärmung von Tälern. [Experimental and numerical investigations of the slope
wind layer and its role in the warming of valleys]. Wiss. Mitt. Nr. 54, dissertation, Universität
München – Meteorologisches Institut, München, FRG.
Brehm, M., and C. Freytag, 1982: Erosion of the night-time thermal circulation in an Alpine valley.
Arch. Meteor. Geophys. Bioclimatol., Ser. B, 31, 331–352.
Buettner, K. J. K., and N. Thyer, 1966: Valley winds in the Mount Rainier area. Arch. Meteor.
Geophys. Bioclimatol., Ser. B, 14, 125–147.
Burger, A., and E. Ekhart, 1937: Über die tägliche Zirkulation der Atmosphäre im Bereiche der
Alpen [The daily atmospheric circulation in the Alpine region]. Gerl. Beitr., 49, 341–367.
Burkholder, B. A., A. Shapiro and E. Fedorovich, 2009: Katabatic flow induced by a cross-slope
band of surface cooling. Acta Geophysica, 57, 923–949.
Carrega, P., 1995: A method for the reconstruction of mountain air temperatures with cartographic
applications. Theor. Appl. Climatol., 52, 69–84.
Catalano, F., and A. Cenedese, 2010: High-resolution numerical modeling of thermally driven
slope winds in a valley with strong capping. J. Appl. Meteor. Climatol., 49, 1859–1880.
Catalano, F., and C.-H. Moeng, 2010: Large-eddy simulation of the daytime boundary layer in an
idealized valley using the Weather Research and Forecasting numerical model. Bound.-Layer
Meteor., 137, 49–75.
Chemel, C., and J. P. Chollet, 2006: Observations of the daytime boundary layer in deep Alpine
valleys. Bound.-Layer Meteor., 119, 239–262.
Chemel, C., C. Staquet, and Y. Largeron, 2009: Generation of internal gravity waves by a katabatic
wind in an idealized alpine valley. Meteor. Atmos Phys., 103, 187–194.
Chen, Y., A. Hall, and K. N. Liou, 2006: Application of 3D solar radiative transfer to mountains.
J. Geophys. Res., 111, D21111.
Chow, F. K., A. P. Weigel, R. L. Street, M. W. Rotach, and M. Xue. 2006. High-resolution
large-eddy simulations of flow in a steep Alpine valley. Part I: Methodology, verification, and
sensitivity experiments. J. Appl. Meteor. Climatol., 45, 63–86.
Ciolli, M., M. de Franceschi, R. Rea, A. Vitti, D. Zardi, and P. Zatelli, 2004: Development and
application of 2D and 3D GRASS modules for simulation of thermally driven slope winds.
Transactions in GIS, 8, 191–209.
Clements, W. E., J. A. Archuleta, and D. E. Hoard, 1989: Mean structure of the nocturnal drainage
flow in a deep valley. J. Appl. Meteor., 28, 457–462.
Clements, C. B., C. D. Whiteman, and J. D. Horel, 2003: Cold-air-pool structure and evolution in
a mountain basin: Peter Sinks, Utah. J. Appl. Meteor., 42, 752–768.
Cogliati, M. G., and N. A. Mazzeo, 2005: Air flow analysis in the upper Rio Negro Valley
(Argentina). Atmos. Res., 80, 263–279.
Colette, A., F. K. Chow, and R. L. Street, 2003: A numerical study of inversion-layer breakup and
the effects of topographic shading in idealized valleys. J. Appl. Meteor., 42, 1255–1272.
Coulter, R. L., and P. Gudiksen, 1995: The dependence of canyon winds on surface cooling and
external forcing in Colorado’s Front Range. J. Appl. Meteor., 34, 1419–1429.
Coulter, R. L., and T. J. Martin, 1996: Effects of stability on the profiles of vertical velocity and its
variance in katabatic flow. Bound.-Layer Meteor., 81, 23–33.
Coulter, R. L., M. Orgill, and W. Porch, 1989: Tributary fluxes into Brush Creek Valley. J. Appl.
Meteor., 28, 555–568.
Coulter, R. L., T. J. Martin, and W. M. Porch, 1991: A comparison of nocturnal drainage flow in
three tributaries. J. Appl. Meteor., 30, 157–169.
Cox, J. A. W., 2005: The sensitivity of thermally driven mountain flows to land-cover change.
Dissertation, Univ. Utah, Meteorology Department. 109pp.
Crawford, T. L., and R. J. Dobosy, 1992: A sensitive fast-response probe to measure turbulence
and heat flux from any airplane. Bound.-Layer Meteor., 59, 257–278.
Cuxart, J., 2008: Nocturnal basin low-level jets: An integrated study. Acta Geophysica, 56,
100–113.
Cuxart, J., and M. A. Jiménez, 2011: Deep radiation fog in a wide closed valley:
Study by numerical modeling and remote sensing. Pure Appl. Geophys. In press. DOI:
10.1007/s00024-011-0365-4.
Cuxart, J., M. A. Jiménez, and D. Martı́nez, 2007: Nocturnal meso-beta basin and katabatic flows
on a midlatitude island. Mon. Wea. Rev., 135, 918–932.
Cuxart, J., J. Cunillera, M. A. Jiménez, D. Martı́nez, F. Molinos, and J. L. Palau, 2011:

Study of mesobeta basin flows by remote sensing. Bound.-Layer Meteor. In press. DOI:
10.1007/s10546-011-9655-8.
Dai, A., and J. Wang, 1999: Diurnal and semidiurnal tides in global surface pressure fields.
J. Atmos. Sci., 56, 3874–3891.
Darby, L. S., and R. M. Banta, 2006: The modulation of canyon flows by larger-scale influences.
Ext. Abstr., 12th Conf. Mountain Meteor., Santa Fe, NM. Amer. Meteor. Soc., Boston, MA.
Darby, L. S., K. J. Allwine, and R. M. Banta, 2006: Nocturnal low-level jet in a mountain basin
complex. Part II: Transport and diffusion of tracer under stable conditions. J. Appl. Meteor.
Climatol., 45, 740–753.
Defant, F., 1949: Zur Theorie der Hangwinde, nebst Bemerkungen zur Theorie der Berg- und
Talwinde. [A theory of slope winds, along with remarks on the theory of mountain winds
and valley winds]. Arch. Meteor. Geophys. Bioclimatol., Ser. A, 1, 421–450 (Theoretical and
Applied Climatology). [English translation: Whiteman, C.D., and E. Dreiseitl, 1984: Alpine
meteorology: Translations of classic contributions by A. Wagner, E. Ekhart and F. Defant. PNL-
5141/ASCOT-84-3. Pacific Northwest Laboratory, Richland, Washington, 121 pp].
Defant, F., 1951: Local Winds. In: Compendium of Meteorology, T. M. Malone (Ed.), Boston,
Amer. Meteor. Soc., 655–672.
de Foy, B., and Coauthors, 2008: Basin-scale wind transport during the MILAGRO field campaign
and comparison to climatology using cluster analysis. Atmos. Chem. Phys., 8, 1209–1224.
de Franceschi, M., 2004: Investigation of atmospheric boundary layer dynamics in Alpine valleys.
Ph.D. thesis in Environmental Engineering, University of Trento, Italy, 136pp. http://www.ing.
unitn.it/dica/eng/monographs/index.php.
de Franceschi, M., and D. Zardi, 2003: Evaluation of cut-off frequency and correction of filter-
induced phase lag and attenuation in eddy covariance analysis of turbulence data. Bound.-Layer
Meteor., 108, 289–303.
de Franceschi, M., and D. Zardi, 2009: Study of wintertime high pollution episodes during the
Brenner-South ALPNAP measurement campaign. Meteor. Atmos. Phys., 103, 237–250.
de Franceschi, M., G. Rampanelli, and D. Zardi, 2002: Further investigations of the “Ora del
Garda” valley wind. 10th Conf. Mount. Meteor. and MAP Meeting, 13–21 June 2002, Park
City, UT, Amer. Meteor. Soc., Boston, MA. Available at http://ams.confex.com/ams/pdfpapers/
39949.pdf.
de Franceschi, M., G. Rampanelli, D. Sguerso, D. Zardi, and P. Zatelli, 2003: Development of
a measurement platform on a light airplane and analysis of airborne measurements in the
atmospheric boundary layer. Ann. Geophys., 46, 269–283.
de Franceschi, M., D. Zardi, M. Tagliazucca, and F. Tampieri, 2009: Analysis of second order
moments in the surface layer turbulence in an Alpine valley. Quart. J. Roy. Meteor. Soc., 135,
1750–1765.
De Wekker, S. F. J., 2002: Structure and morphology of the convective boundary layer in
mountainous terrain. Dissertation, University of British Columbia, Department of Earth and
Ocean Sciences, Vancouver, Canada, 191pp.
De Wekker, S. F. J., 2008: Observational and numerical evidence of depressed convective boundary
layer heights near a mountain base. J. Appl. Meteor., 47, 1017–1026.
De Wekker, S. F. J., and S. D. Mayor, 2009: Observations of atmospheric structure and dynamics in
the Owens Valley of California with a ground-based, eye-safe, scanning aerosol lidar. J. Appl.
Meteor. Climatol., 48, 1483–1499.
De Wekker, S. F. J., and C. D. Whiteman, 2006: On the time scale of nocturnal boundary layer
cooling in valleys and basins and over plains. J. Appl. Meteor., 45, 813–820.
De Wekker, S. F. J., S. Zhong, J. D. Fast, and C. D. Whiteman, 1998: A numerical study of the
thermally driven plain-to-basin wind over idealized basin topographies. J. Appl. Meteor., 37,
606–622.
De Wekker, S. F. J., D. G. Steyn, J. D. Fast, M. W. Rotach, and S. Zhong, 2005: The performance of
RAMS in representing the convective boundary layer structure in a very steep valley. Environ.
Fluid Mech., 5, 35–62.
De Wekker S. F. J., A. Ameen, G. Song, B. B. Stephens, A. G. Hallar and I. B. McCubbin,

2009: A preliminary investigation of boundary layer effects on daytime atmospheric CO2
concentrations at a mountaintop location in the Rocky Mountains Acta Geophysica, 57,
904–922.
Denby, B., 1999: Second-order modelling of turbulence in katabatic flows. Bound.-Layer Meteor.,
92, 67–100.
Deng, X., and R. Stull, 2005: A mesoscale analysis method for surface potential temperature in
mountainous and coastal terrain. Mon. Wea. Rev., 133, 389–408.
Deng, X., and R. Stull, 2007: Assimilating surface weather observations from complex terrain into
a high-resolution numerical weather prediction model. Mon. Wea. Rev., 135, 1037–1054.
Dixit, P. N., and D. Chen, 2011: Effect of topography on farm-scale spatial variation in extreme
temperatures in the Southern Mallee of Victoria, Australia. Theor. Appl. Climatol., 103,
533–542.
Doran, J. C., 1991: The effect of ambient winds on valley drainage flows. Bound.-Layer Meteor.,
55, 177–189.
Doran, J. C., 1996: The influence of canyon winds on flow fields near Colorado’s Front Range.
J. Appl. Meteor., 35, 587–600.
Doran, J. C., and T. W. Horst, 1981: Velocity and temperature oscillations in drainage winds.
J. Appl. Meteor., 20, 361–364.
Doran, J. C., and S. Zhong, 2000: Thermally driven gap winds into the Mexico City basin. J. Appl.
Meteor., 39, 1330–1340.
Doran, J. C., M. L. Wesley, R. T. McMillen, and W. D. Neff, 1989: Measurements of turbulent heat
and momentum fluxes in a mountain valley. J. Appl. Meteor., 28, 438–444.
Doran, J. C., T. W. Horst, and C. D. Whiteman, 1990: The development and structure of nocturnal
slope winds in a simple valley. Bound.-Layer Meteor., 52, 41–68.
Doran, J. C., and Coauthors, 1998: The IMADA-AVER boundary-layer experiment in the Mexico
City area. Bull. Amer. Meteor. Soc., 79, 2497–2508.
Doran, J. C., J. D. Fast, and J. Horel, 2002: The VTMX 2000 campaign. Bull. Amer. Meteor. Soc.,
83, 537–551.
Dorninger, M., Whiteman, C. D., B. Bica, S. Eisenbach, B. Pospichal, and R. Steinacker, 2011:
Meteorological events affecting cold-air pools in a small basin. J. Appl. Meteor. Climatol.
Accepted.
Dosio, A., S. Emeis, G. Graziani, W. Junkermann, and A. Levy, 2001: Assessing the meteorological
conditions of a deep Italian Alpine valley system by means of a measuring campaign and
simulations with two models during a summer smog episode. Atmos. Environ., 35, 5441–5454.
Duguay, C. R., 1993: Radiation modeling in mountainous terrain: Review and status. Mount. Res.
Dev., 13, 339–357.
Eckman, R. M., 1998: Observations and numerical simulations of winds in a broad forested valley.
J. Appl. Meteor., 37, 206–219.
Egger, J., 1987: Valley winds and the diurnal circulation over plateaus. Mon. Wea. Rev., 115,
2177–2186.
Egger, J., 1990: Thermally forced flows: Theory. In W. Blumen (Ed.), Atmospheric Processes over
Complex Terrain. Amer. Meteor. Soc., Boston, 43–57.
Egger, J., S. Bajrachaya, U. Egger, R. Heinrich, J. Reuder, P. Shayka, H. Wendt, and V. Wirth,
2000: Diurnal winds in the Himalayan Kali Gandaki Valley. Part I: Observations. Mon. Wea.
Rev., 128, 1106–1122.
Egger, J., and Coauthors, 2002: Diurnal winds in the Himalayan Kali Gandaki valley. Part III:
Remotely piloted aircraft soundings. Mon. Wea. Rev., 130, 2042–2058.
Egger, J., and Coauthors, 2005: Diurnal circulation of the Bolivian Altiplano. Part I: Observations.
Mon. Wea. Rev., 133, 911–924.
Ekhart, E., 1944: Beiträge zur alpinen Meteorologie. [Contributions to Alpine Meteorology].
Meteor. Z., 61, 217–231. [English translation: Whiteman, C.D., and E. Dreiseitl, 1984: Alpine
Meteorology: Translations of Classic Contributions by A. Wagner, E. Ekhart and F. Defant.
PNL-5141/ASCOT-84-3. Pacific Northwest Laboratory, Richland, Washington, 121 pp].
Ekhart, E., 1948: De la structure thermique de l’atmosphere dans la montagne [On the thermal
structure of the mountain atmosphere]. La Meteorologie, 4, 3–26. [English translation: White-
man, C.D., and E. Dreiseitl: 1984. Alpine Meteorology: Translations of Classic Contributions
by A. Wagner, E. Ekhart and F. Defant. PNL-5141/ASCOT-84-3. Pacific Northwest Laboratory,
Richland, Washington, 121 pp].
Emeis, S., C. Jahn, C. Münkel, C. Münsterer, and K. Schäfer, 2007: Multiple atmospheric layering
and mixing-layer height in the Inn valley observed by remote sensing. Meteor. Z., 16, 415–424.
Enger, L., D. Koračin, and X. Yang, 1993: A numerical study of boundary-layer dynamics in a
mountain valley. Part 1: Model validation and sensitivity experiment. Bound.-Layer Meteor.,
66, 357–394.
Fast, J. D., 1995: Mesoscale modeling and four-dimensional data assimilation in areas of highly
complex terrain. J. Appl. Meteor., 34, 2762–2782.
Fast, J. D., 2003: Forecasts of valley circulations using the terrain-following and step-mountain
vertical coordinates in the Meso-Eta model. Wea. Forecasting, 18, 1192–1206.
Fast, J. D., and S. Zhong, 1998: Meteorological factors associated with inhomogeneous ozone
concentrations within the Mexico City basin. J. Geophys. Res., 103 (D15), 18,927-18,946.
Fast, J. D., and L. S. Darby, 2004: An evaluation of mesoscale model predictions of down-valley
and canyon flows and their consequences using Doppler lidar measurements during VTMX
2000. J. Appl. Meteor., 43, 420–436.
Fast, J., S. Zhong, and C. D. Whiteman, 1996: Boundary layer evolution within a canyonland
basin. Part II. Numerical simulations of nocturnal flows and heat budgets. J. Appl. Meteor., 35,
2162–2178.
Fernando, H. J. S., 2010: Fluid dynamics of urban atmospheres in complex terrain. Ann. Rev. Fluid
Mech., 42, 365–89.
Fiedler, F., I. Bischoff-Gauß, N. Kalthoff, and G. Adrian, 2000: Modeling of the transport and
diffusion of a tracer in the Freiburg-Schauinsland area. J. Geophys. Res., 105, D1, 1599–1610.
Finnigan, J. J., R. Clement, Y. Malhi, R. Leuning, and H. A. Cleugh, 2003: A re-evaluation of long-
term flux measurement techniques. Part I. Averaging and coordinate rotation. Bound.-Layer
Meteor., 107, 1–48.
Fitzjarrald, D. R., 1984: Katabatic winds in opposing flow. J. Atmos. Sci., 41, 1143–1158.
Fleagle, R. G., 1950: A theory of air drainage. J. Meteor., 7, 227–232.
Frei, C., and H. C. Davies, 1993: Anomaly in the Alpine diurnal pressure signal: Observations and
theory. Quart. J. Roy. Meteor. Soc., 119, 1269–1289.
Freytag, C., 1988: Atmosphärische Grenzschicht in einem Gebirgstal bei Berg- und Talwind.
[Atmospheric boundary layer in a mountain valley during mountain and valley winds]. Wiss.
Mitteil. Nr. 60, Münchener Universitätsschriften, Facultät für Physik, Universität München –
Meteorologisches Institut, 197pp.
Fritts, D. C., D. Goldstein, and T. Lund, 2010: High-resolution numerical studies of stable
boundary layer flows in a closed basin: Evolution of steady and oscillatory flows in an
axisymmetric Arizona Meteor Crater. J. Geophys. Res., 115, D18109.
Furger, M., and Coauthors, 2000: The VOTALP Mesolcina Valley campaign 1996 – Concept,
background and some highlights. Atmos. Environ., 34, 1395–1412.
Garratt, J. R., and R. A. Brost, 1981: Radiative cooling effects within and above the nocturnal
boundary layer. J. Atmos. Sci., 38, 2730–2746.
Geerts, B., Q. Miao, and J. C. Demko, 2008: Pressure perturbations and upslope flow over a heated,
isolated mountain. Mon. Wea. Rev., 136, 4272–4288.
Geiger, R., R. H. Aron, and P. Todhunter, 2009: The Climate Near the Ground, 7th Edition. Roman
and Littlefield Publishers, Maryland, 523pp.
Giovannini, L., D. Zardi, and M. de Franceschi, 2011: Analysis of the urban thermal fingerprint of
the city of Trento in the Alps. J. Appl. Meteor. Climatol., 50, 1145–1162.
Gleeson, T. A., 1951: On the theory of cross-valley winds arising from differential heating of the
slopes. J. Meteor., 8, 398–405.
Gohm, A., F. Harnisch, J. Vergeiner, F. Obleitner, R. Schnitzhofer, A. Hansel, A. Fix, B. Neininger,

S. Emeis, and K. Schäfer, 2009: Air pollution transport in an Alpine valley: Results from
airborne and ground-based observations. Bound.-Layer Meteor., 131, 441–463.
Grell, G. A., S. Emeis, W. R. Stockwell, T. Schoenemey, R. Forkel, J. Michalakes, R. Knoche, and
W. Seidl, 2000: Application of a multiscale, coupled MM5/chemistry model to the complex
terrain of the VOTALP valley campaign. Atmos. Environ., 34, 1435–1453.
Grimmond, C. S. B., C. Souch, and M. D. Hubble, 1996: Influence of tree cover on summertime
surface energy balance fluxes, San Gabriel Valley, Los Angeles. Climate Research, 6, 45–57.
Grisogono, B., 2003: Post-onset behaviour of the pure katabatic flow. Bound.-Layer Meteor., 107,
157–175.
Grisogono, B., and J. Oerlemans, 2001a: A theory for the estimation of surface fluxes in simple
katabatic flow. Quart. J. Roy. Meteor. Soc., 127, 2125–2139.
Grisogono, B., and J. Oerlemans, 2001b: Katabatic flow: Analytic solution for gradually varying
eddy diffusivities. J. Atmos. Sci., 58, 3349–3354.
Grisogono, B., L. Kraljevič, and A. Jeričevič, 2007: The low-level katabatic jet height versus
Monin-Obukhov height. Quart. J. Roy. Meteor. Soc., 133, 2133–2136.
Gross, G., 1984: Eine Erklärung des Phänomens Maloja-Schlange Mittels Numerischer Simula-
tion. [An Explanation of the Maloja Snake (Cloud) Phenomenon by Means of a Numerical
Simulation]. Dissertation, Technischen Hochschule Darmstadt, Darmstadt, FRG.
Gross, G., 1985: An explanation of the Maloja-serpent by numerical simulation, Beitr. Phys.
Atmos., 58, 441–457.
Gross, G., 1990: On the wind field in the Loisach Valley–Numerical simulation and comparison
with the LOWEX III data. Meteor. Atmos. Phys., 42, 231–247.
Grossi, G., and L. Falappi, 2003: Comparison of energy fluxes at the land surface-atmosphere
interface in an Alpine valley as simulated with different models. Hydrology and Earth System
Sciences, 7, 920–936.
Guardans, R., and I. Palomino, 1995: Description of wind field dynamic patterns in a valley and
their relation to mesoscale and synoptic-scale situations. J. Appl. Meteor., 34, 49–67.
Gustavsson, T., M. Karlsson, J. Bogren, and S. Lindqvist, 1998: Development of temperature
patterns during clear nights. J. Appl. Meteor., 37, 559–571.
Ha, K.-J., and L. Mahrt, 2003: Radiative and turbulent fluxes in the nocturnal boundary layer.
Tellus, 55A, 317–327.
Haiden, T., 1998: Analytical aspects of mixed-layer growth in complex terrain. Preprints, Eighth
Conf. Mountain Meteorology, Amer. Meteor. Soc., Flagstaff, Arizona, 368–372.
Haiden, T., 2003: On the pressure field in the slope wind layer. J. Atmos. Sci., 60, 1632–1635.
Haiden, T., and C. D. Whiteman, 2005: Katabatic flow mechanisms on a low-angle slope. J. Appl.
Meteor., 44, 113–126.
Haiden, T., C. D. Whiteman, S. W. Hoch, and M. Lehner, 2011: A mass-flux model of nocturnal
cold air intrusions into a closed basin. J. Appl. Meteor. Climatol., 50, 933–943.
Harnisch, F., A. Gohm, A. Fix, R. Schnitzhofer, A. Hansel, and B. Neininger, 2009: Spatial
distribution of aerosols in the Inn Valley atmosphere during wintertime. Meteor. Atmos. Phys.,
103, 215–228.
Hauge, G., and L. R. Hole, 2003: Implementation of slope irradiance in Mesoscale Model version
5 and its effect on temperature and wind fields during the breakup of a temperature inversion.
J. Geophys. Res., 108, D2, 4058.
Hawkes, H. B., 1947: Mountain and valley winds with special reference to the diurnal mountain
winds of the Great Salt Lake region. Dissertation, Ohio State University, 312pp.
Helmis, C. G., and K. H. Papadopoulos, 1996: Some aspects of the variation with time of katabatic
flows over a simple slope. Quart. J. Roy. Meteor. Soc., 122, 595–610.
Henne, S., M. Furger, and A. S. H. Prévôt, 2005: Climatology of mountain venting-induced
elevated moisture layers in the lee of the Alps. J. Appl. Meteor., 44, 620–633.
Henne, S., W. Junkermann, J. M. Kariuki, J. Aseyo, and J. Klausen, 2008: The establishment of
the Mt. Kenya GAW station: Installation and meteorological characterization. J. Appl. Meteor.
Climatol., 47, 2946–2962.
Higgins, W., and Coauthors, 2006: The NAME 2004 field campaign and modeling strategy. Bull.
Amer. Meteor. Soc., 87, 79–94.
Hiller, R., M. Zeeman, and W. Eugster, 2008: Eddy-covariance flux measurements in the complex
terrain of an Alpine valley in Switzerland. Bound.-Layer Meteor., 127, 449–467.
Hoch, S. W., and C. D. Whiteman, 2010: Topographic effects on the surface radiation balance in
and around Arizona’s Meteor Crater. J. Appl. Meteor. Climat., 49, 1894–1905.
Hoch, S. W., C. D. Whiteman, and B. Mayer, 2011: A systematic study of longwave radiative
heating and cooling within valleys and basins using a three-dimensional radiative transfer
model. J. Appl. Meteor. Climatol. In press.
Holton, J. R., 1966: The diurnal boundary layer wind oscillation above sloping terrain. Tellus, 19,
199–205.
Horst, T. W., and J. C. Doran, 1986: Nocturnal drainage flow on simple slopes. Bound.-Layer
Meteor., 34, 263–286.
Horst, T. W., and J. C. Doran, 1988: The turbulence structure of nocturnal slope flows. J. Atmos.
Sci., 45, 605–616.
Hughes, M., A. Hall, and G. Fovell, 2007: Dynamical controls on the diurnal cycle of temperature
in complex topography, Clim. Dyn., 29, 277–292.
Hunt, J. C. R., H. J. S. Fernando, and M. Princevac, 2003: Unsteady thermally driven flows on
gentle slopes. J. Atmos. Sci., 60, 2169–2182.
IIjima, Y., and M. Shinoda, 2000: Seasonal changes in the cold-air pool formation in a subalpine
hollow, central Japan. Int. J. Climatol., 20, 1471–1483.
Iziomon, M. G., H. Mayer, W. Wicke, and A. Matzarakis, 2001: Radiation balance over low-lying
and mountainous areas in southwest Germany. Theor. Appl. Climatol., 68, 219–231.
Jiang, Q., and J. D. Doyle, 2008: Diurnal variation of downslope winds in Owens Valley during
the Sierra Rotor experiment. Mon. Wea. Rev., 136, 3760–3780.
Kaimal, J. C., and J. J. Finnigan, 1994: Atmospheric boundary-layer flows: Their structure and
measurement. Oxford University Press, New York, 289pp.
Kalthoff, N., and Coauthors, 2002: Mesoscale wind regimes in Chile at 30ı S. J. Appl. Meteor., 41,
953–970.
Kalthoff, N., M. Fiebig-Wittmaack, C. Meißner, M. Kohler, M. Uriarte, I. Bischoff-Gauß, and E.
Gonzales, 2006: The energy balance, evapo-transpiration and nocturnal dew deposition of an
arid valley in the Andes. J. Arid Environments, 65, 420–443.
Kalthoff, N., and Coauthors, 2009: The impact of convergence zones on the initiation of deep
convection: A case study from COPS. Atmos. Res., 93, 680 – 694.
Kastendeuch, P. P., and P. Kaufmann, 1997: Classification of summer wind fields over complex
terrain. Int. J. Climatol., 17, 521–534.
Kastendeuch, P. P., P. Lacarrere, G. Najjar, J. Noilhan, F. Gassmann, and P. Paul, 2000: Mesoscale
simulations of thermodynamic fluxes over complex terrain. Int. J. Climatol., 20, 1249–1264.
Kaufmann, P., and R. O. Weber, 1996: Classification of mesoscale wind fields in the MISTRAL
field experiment. J. Appl. Meteor., 35, 1963–1979.
Kaufmann, P., and R. O. Weber, 1998: Directional correlation coefficient for channeled flow and
application to wind data over complex terrain. J. Atmos. Oceanic. Technol., 15, 89–97.
Kaufmann, P., and C. D. Whiteman, 1999: Cluster-analysis classification of wintertime wind
patterns in the Grand Canyon region. J. Appl. Meteor., 38, 1131–1147.
Kavčič, I., and B. Grisogono, 2007: Katabatic flow with Coriolis effect and gradually varying eddy
diffusivity. Bound.-Layer Meteor., 125,377–387.
Kelly, R. D., 1988: Asymmetric removal of temperature inversions in a high mountain valley.
J. Appl. Meteor., 27, 664–673.
Khodayar, S., N. Kalthoff, M. Fiebig-Wittmaack, and M. Kohler, 2008: Evolution of the atmo-
spheric boundary-layer structure of an arid Andes valley. Meteor. Atmos. Phys., 99, 181–198.
Kimura, F., and T. Kuwagata, 1993: Thermally induced wind passing from plain to basin over a
mountain range. J. Appl. Meteor., 32, 1538–1547.
Kimura, F., and T. Kuwagata, 1995: Horizontal heat fluxes over complex terrain computed using a
simple mixed-layer model and a numerical model. J. Appl. Meteor., 34, 549–558.
King, C. W., 1989: Representativeness of single vertical wind profiles for determining volume
fluxes in valleys. J. Appl. Meteor., 28, 463–466.
King, J. A., F. H. Shair, and D. D. Reible, 1987: The influence of atmospheric stability on pollutant
transport by slope winds. Atmos. Environ., 21, 53–59.
Kitada, T., and R. P. Regmi, 2003: Dynamics of air pollution transport in late wintertime over
Kathmandu Valley, Nepal: As revealed with numerical simulation. J. Appl. Meteor., 42,
1770–1798.
Koch, H. G., 1961: Die warme Hangzone. Neue Anschauungen zur nächtlichen Kaltluftschichtung
in Tälern und an Hängen. [The warm slope zone. New views of the nocturnal cold air layer in
valleys and on slopes]. Z. Meteor., 15, 151–171.
Kolev, N., I. Grigorov, I. Kolev, P. C. S. Devara, P. E. Raj, and K. K. Dan, 2007: Lidar and sun
photometer observations of atmospheric boundary-layer characteristics over an urban area in a
mountain valley. Bound.-Layer Meteor., 124, 99–115.
Kondo, H., 1990: A numerical experiment on the interaction between sea breeze and valley wind
to generate the so-called “extended sea breeze.” J. Meteor. Soc. Japan, 68, 435–446.
Kondo, J., and N. Okusa, 1990: A simple numerical prediction model of nocturnal cooling in a
basin with various topographic parameters. J. Appl. Meteor., 29, 604–619.
Kondo, J., T. Kuwagata, and S. Haginoya, 1989: Heat budget analysis of nocturnal cooling and
daytime heating in a basin. J. Atmos. Sci., 19, 2917–2933.
Koračin, D., and L. Enger, 1994: A numerical study of boundary-layer dyamics in a mountain
valley. Part 2: Observed and simulated characteristics of atmospheric stability and local flows.
Bound.-Layer Meteor., 69, 249–283.
Kossmann, M., and A. P. Sturman, 2003: Pressure-driven channeling effects in bent valleys. J. Appl.
Meteor., 42, 151–158.
Kossman, M., A. P. Sturman, P. Zawar-Reza, H. A. McGowan, A. J. Oliphant, I. F. Owens, and R.
A. Spronken-Smith, 2002: Analysis of the wind field and heat budget in an alpine lake basin
during summertime fair weather conditions. Meteor. Atmos. Phys., 81, 27–52.
Küttner, J., 1949: Periodische Luftlawinen. [Periodic air avalanches]. Meteor. Rundsch., 2,
183–184.
Kuttler, W., A.-B. Barlag, and F. Roßmann, 1996: Study of the thermal structure of a town in a
narrow valley. Atmos. Environ., 30, 365–378.
Kuwagata, T., and J. Kondo, 1989: Observation and modeling of thermally induced upslope flow.
Kuwagata, T., and F. Kimura, 1995: Daytime boundary layer evolution in a deep valley. Part I:
Observations in the Ina Valley. J. Appl. Meteor., 34, 1082–1091.
Kuwagata, T., and F. Kimura, 1997: Daytime boundary layer evolution in a deep valley. Part II:
Numerical simulation of the cross-valley circulation. J. Appl. Meteor., 36, 883–895.
Lee, I. Y., R. L. Coulter, H. M. Park, and J.-H. Oh, 1995: Numerical simulation of nocturnal
drainage flow properties in a rugged canyon. Bound.-Layer Meteor., 72, 305–321.
Lee, S.-M., W. Giori, M. Princevac, and H. J. S. Fernando, 2006: Implementation of a stable PBL
turbulence parameterization for the mesoscale model MM5: Nocturnal flow in complex terrain.
Lee, X., and X. Hu, 2002: Forest-air fluxes of carbon, water and energy over non-flat terrain.
Bound.-Layer Meteor, 103, 277–301.
Lehner, M., and A. Gohm, 2010: Idealised simulations of daytime pollution transport in a steep
valley and its sensitivity to thermal stratification and surface albedo. Bound.-Layer Meteor.,
134, 327–351.
Lehner, M., C. D. Whiteman, and S. W. Hoch, 2010: Diurnal cycle of thermally driven cross-basin
winds in Arizona’s Meteor Crater. J. Appl. Meteor. Climatol., 50, 729–744.
Leone, J. M., Jr., and R. L. Lee, 1989: Numerical simulation of drainage flow in Brush Creek,
Colorado. J. Appl. Meteor., 28, 530–542.
Li, J.-G., and B. W. Atkinson, 1999: Transition regimes in valley airflows. Bound.-Layer Meteor.,
91, 385–411.
Li, Y., and R. B. Smith, 2010: The detection and significance of diurnal pressure and potential
vorticity anomalies east of the Rockies. J. Atmos. Sci., 67, 2734–2751.
Li, Y., R. B. Smith, and V. Grubišić, 2009: Using surface pressure variations to categorize the
diurnal valley circulations: Experiments in Owens Valley. Mon. Wea. Rev., 137, 1753–1769.
Lindkvist, L., and S. Lindqvist, 1997: Spatial and temporal variability of nocturnal summer frost
in elevated complex terrain. Agric. Forest Meteor., 87, 139–153.
Lindkvist, L., T. Gustavsson, and J. Bogren, 2000: A frost assessment method for mountainous
areas. Agric. Forest Meteor., 102, 51–67.
Liu, X., A. Bai, and C. Liu, 2009: Diurnal variations of summertime precipitation over the
Tibetan Plateau in relation to orographically-induced regional circulations. Environ. Res. Lett.,
4, 045203.
Löffler-Mang, M., M. Kossmann, R. Vögtlin, and F. Fiedler, 1997: Valley wind systems and their
influences on nocturnal ozone concentrations. Beitr. Phys. Atmosph., 70, 1–14.
Lu, R., and R. P. Turco, 1995: Air pollutant transport in a coastal environment – II. Three-
dimensional simulations over Los Angeles basin. Atmos. Environ., 29, 1499–1518.
Ludwig, F. L., J. Horel, and C. D. Whiteman, 2004: Using EOF analysis to identify important
surface wind patterns in mountain valleys. J. Appl. Meteor., 43, 969–983.
Lugauer, M., and P. Winkler, 2005: Thermal circulation in South Bavaria – climatology and
synoptic aspects. Meteor. Z., 14, 15–30.
Lugauer, M., and Coauthors, 2003: An overview of the VERTIKATOR project and results of
Alpine pumping. Extended Abstracts, Vol. A, International Conf. on Alpine Meteor. and MAP-
Meeting 2003, May 19–23, 2003, Brig, Switzerland. Publications of MeteoSwiss, No. 66,
129–132.
Luhar, A. K., and S. K. Rao, 1993: Random-walk model studies of the transport and diffusion of
pollutants in katabatic flows. Bound.-Layer Meteor., 66, 395–412.
Magono, C., C. Nakamura, and Y. Yoshida, 1982: Nocturnal cooling of the Moshiri Basin,
Hokkaido in midwinter. J. Meteor. Soc. Japan, 60, 1106–1116.
Mahrt, L., 1982: Momentum balance of gravity flows. J. Atmos. Sci., 39, 2701–2711.
Mahrt, L., 1999: Stratified atmospheric boundary layers. Bound.-Layer Meteor., 90, 375–396.
Mahrt, L., 2006: Variation of surface air temperature in complex terrain. J. Appl. Meteor. Climatol.,
45, 1481–1493.
Mahrt, L. J., and S. Larsen, 1982: Small scale drainage flow. Tellus, 34, 579–587.
Mahrt, L., D. Vickers, R. Nakamura, M. R. Soler, J. Sun, S. Burns, and D. H. Lenschow, 2001:
Shallow drainage flows. Bound.-Layer Meteor., 101, 243–260.
Maki, M., and T. Harimaya, 1988: The effect of advection and accumulation of downslope cold air
on nocturnal cooling in basins. J. Meteor. Soc. Japan, 66, 581–597.
Maki, M., T. Harimaya, and K. Kikuchi, 1986: Heat budget studies on nocturnal cooling in a basin.
J. Meteor. Soc. Japan, 64, 727–740.
Manins, P. C., 1992: Vertical fluxes in katabatic flows. Bound.-Layer Meteor., 60, 169–178.
Manins, P. C., and B. L. Sawford, 1979a: Katabatic winds: A field case study. Quart. J. Roy. Meteor.
Soc., 105, 1011–1025.
Manins, P. C., and B. L. Sawford, 1979b: A model of katabatic winds. J. Atmos. Sci., 36, 619–630.
Mannouji, N., 1982: A numerical experiment on the mountain and valley winds. J. Meteorol. Soc.
Jap., 60, 1085–1105.
Martı́nez, D., 2011: Topographically induced flows and nocturnal cooling in the atmospheric
boundary layer. Doctoral dissertation. Department of Physics, University of the Balearic
Islands.
Martı́nez, D., and J. Cuxart, 2009: Assessment of the hydraulic slope flow approach using a
mesoscale model. Acta Geophys., 57, 882–903.
Martı́nez, D., J. Cuxart, and J. Cunillera, 2008: Conditioned climatology for stably stratified nights
in the Lleida area. Tethys, 5, 13–24.
Martı́nez, D., M. A. Jiménez, J. Cuxart, and L. Mahrt, 2010: Heterogeneous nocturnal cooling in a
large basin under very stable conditions. Bound.-Layer Meteor., 137, 97–113.
Marty, C., R. Philipona, C. Fröhlich, and A. Ohmura, 2002: Altitude dependence of surface
radiation fluxes and cloud forcing in the alps: results from the alpine surface radiation budget
network. Theor. Appl. Climatol., 72, 137–155.
Matzinger, N., M. Andretta, E. van Gorsel, R. Vogt, A. Ohmura, and M. W. Rotach, 2003: Surface
radiation budget of an Alpine valley. Quart. J. Roy. Meteor. Soc., 129, 877–895.
Mayer, B., 2009: Radiative transfer in the cloudy atmosphere. Eur. Phys. J. Conf., 1, 75–99.
Mayer, B., and A. Kylling, 2005: Technical note: The libRadtran software package for radiative
transfer calculations – Description and examples of use. Atmos. Chem. Phys., 5, 1855–1877.
Mayer, B., S. W. Hoch, and C. D. Whiteman, 2010: Validating the MYSTIC three-dimensional
radiative transfer model with observations from the complex topography of Arizona’s Meteor
Crater. Atmos. Chem. Phys. Discuss., 10, 13373–13405.
McGowan, H. A., 2004: Observations of anti-winds in a deep Alpine valley, Lake Tekapo, New
Zealand. Arctic, Antarctic, and Alpine Res., 36, 495–501.
McGowan, H. A., and A. P. Sturman, 1996: Interacting multi-scale wind systems within an alpine
basin, Lake Tekapo. New Zealand. Meteor. Atmos. Phys., 58, 165–177.
McGowan, H. A., I. F. Owens, and A. P. Sturman, 1995: Thermal and dynamic characteristics of
alpine lake breezes, Lake Tekapo, New Zealand. Bound.-Layer Meteor., 76, 3–24.
McKee, T. B., and R. D. O’Neal, 1989: The role of valley geometry and energy budget in the
formation of nocturnal valley winds. J. Appl. Meteor., 28, 445–456.
McNider, R. T., 1982: A note on velocity fluctuations in drainage flows. J. Atmos. Sci., 39,
1658–1660.
Monti, P., H. J. S. Fernando, M. Princevac, W. C. Chan, T. A. Kowalewski, and E. R. Pardyjak,
2002: Observations of flow and turbulence in the nocturnal boundary layer over a slope.
J. Atmos. Sci., 59, 2513–2534.
Moraes, O. L. L., O. C. Acevedo, R. Da Silva, R. Magnago, and A. C. Siqueira, 2004: Nocturnal
surface-layer characteristics at the bottom of a valley. Bound.-Layer Meteor., 112, 159–177.
Müller, M. D., and D. Scherer, 2005: A grid- and subgrid-scale radiation parameterization of
topographic effects for mesoscale weather forecast models. Mon. Wea. Rev., 133, 1431–1442.
Mursch-Radlgruber, E., 1995: Observations of flow structure in a small forested valley system.
Theor. Appl. Climatol., 52, 3–17.
Neff, W. D., and C. W. King, 1989: The accumulation and pooling of drainage flows in a large
basin. J. Appl. Meteor., 28, 518–529.
Nickus, U., and I. Vergeiner, 1984: The thermal structure of the Inn Valley atmosphere. Arch.
Meteor. Geophys. Bioclimatol., Ser. A, 33, 199–215.
Noppel, H., and F. Fiedler, 2002: Mesoscale heat transport over complex terrain by slope winds –
A conceptual model and numerical simulations. Bound.-Layer Meteor., 104, 73–97.
O’Steen, L. B., 2000: Numerical simulation of nocturnal drainage flows in idealized valley-
tributary systems. J. Appl. Meteor., 39, 1845–1860.
Oerlemans, J., 1998: The atmospheric boundary layer over melting glaciers. Chapter 6 in: Clear
and Cloudy Boundary Layers (Eds: A. A. M. Holtslag and P. G. Duynkerke), Royal Netherlands
Academy of Arts and Sciences,129–153.
Oerlemans, J., and B. Grisogono, 2002: Glacier winds and parameterisation of the related surface
heat fluxes. Tellus A, 54, 440–452.
Oerlemans, J., H. Björnsson, M. Kuhn, F. Obleitner, F. Palsson, C. J. P. P. Smeets, H. F. Vugts,
and J. De Wolde, 1999: Glacio-meteorological investigations on Vatnajökull, Iceland, summer
1996: An overview. Bound.-Layer Meteor., 92, 3–26.
Oliphant, A. J., R. A. Spronken-Smith, A. P. Sturman, and I. F. Owens, 2003: Spatial variability of
surface radiation fluxes in mountainous terrain. J. Appl. Meteor., 42, 113–128.
Oliver, H. R., 1992: Studies of the energy budget of sloping terrain. Int. J. Climatol., 12, 55–68.
Olyphant, G. A., 1986: Longwave radiation in mountainous areas and its influence on the energy
balance of alpine snowfields. Water Resources Res., 22, 62–66.
Oncley, S. P., and Coauthors, 2007: The Energy Balance Experiment EBEX-2000. Part I: Overview
and energy balance. Bound.-Layer Meteor., 123, 1–28.
Orgill, M. M., J. D. Kincheloe, and R. A. Sutherland, 1992: Influences on nocturnal valley drainage
winds in Western Colorado valleys. J. Appl. Meteor., 31, 121–141.
Pamperin, H., and G. Stilke, 1985: Nächtliche Grenzschicht und LLJ im Alpenvorland nahe dem
Inntalausgang. [Nocturnal boundary layer and low level jet near the Inn Valley exit]. Meteor.
Rundsch., 38, 145–156.
Panday, A. K., 2006: The diurnal cycle of air pollution in the Kathmandu Valley, Nepal.
Dissertation No. 37361. Massachusetts Institute of Technology. [http://hdl.handle.net/1721.1/
37361]
Panday, A. K., and R. G. Prinn, 2009: Diurnal cycle of air pollution in the Kathmandu Valley,
Nepal: Observations, J. Geophys. Res., 114, D09305.
Panday, A. K., R. G. Prinn, and C. Schär, 2009: Diurnal cycle of air pollution in the Kathmandu
Valley, Nepal: 2. Modeling results. J. Geophys. Res., 114, D21308.
Papadopoulos, K. H., and C. G. Helmis, 1999: Evening and morning transition of katabatic flows.
Papadopoulos, K. H., C. G. Helmis, A. T. Soilemes, J. Kalogiros, P. G. Papageorgas, and
D. N. Asimakopoulos, 1997: The structure of katabatic flows down a simple slope. Quart.
J. Roy. Meteor. Soc., 123, 1581–1601.
Parish, T. R., and L. D. Oolman, 2010: On the role of sloping terrain in the forcing of the Great
Plains low-level jet. J. Atmos. Sci., 67, 2690–2699.
Petkovšek, Z., 1992: Turbulent dissipation of cold air lake in a basin. Meteor. Atmos. Phys., 47,
237–245.
Philipona, R., B. Dürr, and C. Marty, 2004: Greenhouse effect and altitude gradients over the
Alps – by surface longwave radiation measurements and model calculated LOR. Theor. Appl.
Climatol., 77, 1–7.
Pielke, R. A., J. H. Rodriguez, J. L. Eastman, R. L. Walko, and R. A. Stocker, 1993: Influence of
albedo variability in complex terrain on mesoscale systems. J. Climate, 6, 1798–1806.
Pinto, J. O., D. B. Parsons, W. O. J. Brown, S. Cohn, N. Chamberlain, and B. Morley, 2006:
Coevolution of down-valley flow and the nocturnal boundary layer in complex terrain. J. Appl.
Piringer, M., and K. Baumann, 1999: Modifications of a valley wind system by an urban area –
Experimental results. Meteor. Atmos. Phys., 71, 117–125.
Plüss, C., and A. Ohmura, 1997: Longwave radiation on snow-covered mountainous surfaces.
J. Appl. Meteor., 36, 818–824.
Porch, W. M., R. B. Fritz, R. L. Coulter, and P. H. Gudiksen, 1989: Tributary, valley and sidewall
air flow interactions in a deep valley. J. Appl. Meteor., 28, 578–589.
Porch, W. M., W. E. Clements, and R. L. Coulter, 1991: Nighttime valley waves. J. Appl. Meteor.,
30, 145–156.
Pope, D., and C. Brough, 1996: Utah’s Weather and Climate. Publishers Press, 245pp.
Pospichal, B., 2004: Struktur und Auflösung von Temperaturinversionen in Dolinen am Beispiel
Grünloch [Structure and breakup of temperature inversions in Dolines, with an example from
the Grünloch]. Diplomarbeit, Instit. Meteor. u. Geophysik, Univ. Wien, 69pp.
Pospichal, B., S. Eisenbach, C. D. Whiteman, R. Steinacker, and M. Dorninger, 2003: Observations
of the cold air outflow from a basin cold pool through a low pass. Ext. Abstr., Vol. A, Intl Conf.
Alpine Meteor. and the MAP-Meeting 2003, Brig, Switzerland, 19–23 May 2003, 153–156.
Available as MeteoSwiss Publication 66, MeteoSwiss, Zurich, Switzerland.
Poulos, G. S., and S. Zhong, 2008. The observational history of small-scale katabatic winds in
mid-latitudes. Geography Compass, 2, 1–24.
Poulos, G. S., J. E. Bossert, T. B. McKee, and R. A. Pielke, 2000: The interaction of katabatic
flow and mountain waves. Part I: Observations and idealized simulations. J. Atmos. Sci., 57,
1919–1936.
Poulos, G. S., and Coauthors, 2002: CASES-99: A comprehensive investigation of the stable
nocturnal boundary layer. Bull. Amer. Meteor. Soc., 83, 555–581.
Poulos, G. S., J. E. Bossert, T. B. McKee, and R. A. Pielke, Sr., 2007: The interaction of katabatic
flow and mountain waves. Part II: Case study analysis and conceptual model. J. Atmos. Sci.,
64, 1857–1879.
Prabha, T. V., and E. Mursch-Radlgruber, 1999: Modeling of diffusion in a wide Alpine valley.
Theor. Appl. Climatol. 64, 93–103.
Prandtl, L., 1942: Strömungslehre [Flow Studies]. Vieweg und Sohn, Braunschweig, 382pp.
Prévôt, A. S. H., J. Staehelin, H. Richner, and T. Griesser, 1993: A thermally driven wind system
influencing concentrations of ozone precursors and photo-oxidants at a receptor site in the
Alpine foothills. Meteor. Z., N.F., 2, 167–177.
Prévôt, A. S. H., J. Dommen, M. Bäumle, and M. Furger, 2000: Diurnal variations of volatile
organic compounds and local circulation systems in an Alpine valley. Atmos. Environ., 34,
1413–1423.
Princevac, M., and H. J. S. Fernando, 2007: A criterion for the generation of turbulent anabatic
flows. Phys. Fluids, 19, 105102.
Princevac, M., and H. J. S. Fernando, 2008: Morning breakup of cold pools in complex terrain.
J. Fluid Mech., 616, 99–109.
Princevac, M., H. J. S. Fernando, and C. D. Whiteman, 2005: Turbulent entrainment into natural
gravity-driven flows. J. Fluid Mech., 533, 259–268.
Princevac, M., J. C. R. Hunt, and H. J. S. Fernando, 2008: Quasi-steady katabatic winds on slopes
in wide valleys: Hydraulic theory and observations. J. Atmos. Sci., 65, 627–643.
Räisänen, P., 1996: The effect of vertical resolution on clear-sky radiation calculations: tests with
two schemes. Tellus, 48A, 403–423.
Rakovec, J., J. Merše, S. Jernej, and B. Paradiž, 2002: Turbulent dissipation of the cold-air pool
in a basin: Comparison of observed and simulated development. Meteor. Atmos. Phys., 79,
195–213.
Ramanathan, N., and K. Srinivasan, 1998: Simulation of airflow in Kashmir valley for a summer
day. J. Appl. Meteor., 37, 497–508.
Rampanelli, G., 2004: Diurnal thermally-induced flows in Alpine valleys. Ph.D. thesis in Environ-
mental Engineering, University of Trento, Italy.
Rampanelli, G., and D. Zardi, 2004: A method to determine the capping inversion of the convective
boundary layer. J. Appl. Meteor., 43, 925–933.
Rampanelli, G., D. Zardi, and R. Rotunno, 2004: Mechanisms of up-valley winds. J. Atmos. Sci.,
61, 3097–3111.
Rao, S. K., and M. A. Schaub, 1990: Observed variations of ¢ ™ and ¢ ¥ in the nocturnal drainage
flow in a deep valley. Bound.-Layer Meteor., 51, 31–48.
Rao, K. S., and H. F. Snodgrass, 1981: A nonstationary nocturnal drainage flow model. Bound.-
Layer Meteor., 20, 309–320.
Reeves, H. D., and D. J. Stensrud, 2009: Synoptic-scale flow and valley cold pool evolution in the
Western United States. Wea. Forecasting, 24, 1625–1643.
Regmi, R. P., T. Kitada, and G. Kurata, 2003: Numerical simulation of late wintertime local flows in
Kathmandu Valley, Nepal: Implication for air pollution transport. J. Appl. Meteor., 42, 389–403.
Reiter, E. R., and M. Tang, 1984: Plateau effects on diurnal circulation patterns. Mon. Wea. Rev.,
112, 638–651.
Reuder, J., and J. Egger, 2006: Diurnal circulation of the South American Altiplano: Observations
in a valley and at a pass. Tellus A, 58, 254–262.
Reuten, C., 2008: Uplsope flow systems: Scaling, structure, and kinematics in tank and atmosphere.
VDM Verlag Dr. Mueller e.K., Saarbrücken, Germany, 200pp.
Reuten, C., D. G. Steyn, K. B. Strawbridge, and P. Bovis, 2005: Observations of the relation
between upslope flows and the convective boundary layer in steep terrain. Bound.-Layer
Meteor., 116, 37–61.
Reuten, C., D. G. Steyn, and S. E. Allen, 2007: Water tank studies of atmospheric boundary layer
structure and air pollution transport in upslope flow systems. J. Geophys. Res., 112, D11114.
Rotach, M. W., and D. Zardi, 2007: On the boundary layer structure over complex terrain: Key
findings from MAP. Quart. J. Roy. Meteor. Soc., 133, 937–948.
Rotach M. W., and Coauthors, 2004: Turbulence structure and exchange processes in an Alpine
valley: The Riviera project. Bull. Amer. Meteor. Soc., 85, 1367–1385.
Rotach, M. W., M. Andretta, P. Calanca, A. P. Weigel, and A. Weiss, 2008: Turbulence character-
istics and exchange mechanisms in highly complex terrain. Acta Geophysicae, 56, 194–219.
Rucker, M., R. M. Banta, and D. G. Steyn, 2008: Along-valley structure of daytime thermally
driven flows in the Wipp Valley. J. Appl. Meteor. Climatol., 47, 733–751.
Sakiyama, S. K., 1990: Drainage flow characteristics and inversion breakup in two Alberta
mountain valleys. J. Appl. Meteor., 29, 1015–1030.
Salerno, R., 1992: Analysis of flow and pollutant dispersion by tracer experiments in south Alpine
valleys. Theor. Appl. Climatol. 45, 19–35.
Sasaki, T., and Coauthors, 2004: Vertical moisture transport above the mixed layer around the
mountains in western Sumatra. Geophys. Res. Lett., 31, L08106.
Savage, L. C., III, S. Zhong, W. Yao, W. J. O. Brown, T. W. Horst, and C. D. Whiteman, 2008:
An observational and numerical study of a regional-scale downslope flow in northern Arizona.
J. Geophys. Res., 113, D14114.
Savijärvi H., and J. Liya, 2001: Local winds in a valley city. Bound.-Layer Meteor., 100, 301–319.
Schicker, I., and P. Seibert, 2009. Simulation of the meteorological conditions during a winter smog
episode in the Inn Valley. Meteor. Atmos. Phys., 103, 203–214.
Schmidli, J., and R. Rotunno, 2010: Mechanisms of along-valley winds and heat exchange over
mountainous terrain. J. Atmos. Sci., 67, 3033–3047.
Schmidli, J., G. S. Poulos, M. H. Daniels, and F. K. Chow, 2009a: External Influences on nocturnal
thermally driven flows in a deep valley. J. Appl. Meteor. Climatol., 48, 3–23.
Schmidli, J., B. Billings, F. K. Chow, J. Doyle, V. Grubišić, T. Holt, Q. Jiang, K. A. Lundquist,
P. Sheridan, S. Vosper, S. F. J. de Wekker, C. D. Whiteman, A. A. Wyszogrodzki, and G. Zängl,
2010: Intercomparison of mesoscale model simulations of the daytime valley wind system.
Mon. Wea. Rev., 139, 1389–1409.
Schumann, U., 1990: Large-eddy simulation of the up-slope boundary layer. Quart. J. Roy. Meteor.
Soc., 116, 637–670.
Serafin, S., 2006: Boundary-layer processes and thermally driven flows over complex terrain. Ph.D.
thesis in Environmental Engineering, Univ. of Trento, Italy, 200pp. http://www.ing.unitn.it/
dica/eng/monographs/index.php.
Serafin, S., and D. Zardi, 2010a: Structure of the atmospheric boundary layer in the vicinity of a
developing upslope flow system: A numerical model study. J Atmos. Sci., 67, 1171–1185.
Serafin, S., and D. Zardi, 2010b: Daytime heat transfer processes related to slope flows and
turbulent convection in an idealized mountain valley. J Atmos. Sci., 67, 3739–3756.
Serafin, S., and D. Zardi, 2010c: Daytime development of the boundary layer over a plain and in a
valley under fair weather conditions: a comparison by means of idealized numerical simulations
J Atmos. Sci., 68, 2128–2141.
Shapiro, A., and E. Fedorovich, 2007: Katabatic flow along a differentially cooled sloping surface.
J. Fluid Mech., 571, 149–175.
Shapiro, A., and E. Fedorovich, 2008: Coriolis effects in homogeneous and inhomogeneous
katabatic flows. Quart. J. Roy. Meteor. Soc., 134, 353–370.
Shapiro, A., and E. Fedorovich, 2009: Nocturnal low-level jet over a shallow slope. Acta
Geophysica, 57, 950–980.
Sharp, J., and C. F. Mass, 2004: Columbia Gorge gap winds: Their climatological influence and
synoptic evolution. Wea. Forecasting, 19, 970–992.
Sievers, U., 2005: Das Kaltluftabflussmodell KLAM 21 – Theoretische Grundlagen, Anwendung
und Handhabung des PC-Modells [The cold air drainage model KLAM 21 – Theoretical basis,
application, and operation of the PC model]. Berichte des Deutschen Wetterdienstes no. 227,
Offenbach am Main, Germany, 101pp.
Simpson, J. E., 1999: Gravity currents, 2nd Edition. Cambridge Univ. Press, 259pp.
Skyllingstad, E. D., 2003: Large-eddy simulation of katabatic flows. Bound.-Layer Meteor., 106,
217–243.
Smith, C. M., and E. D. Skyllingstad, 2005: Numerical simulation of katabatic flow with changing
slope angle. Mon. Wea. Rev., 133, 3065–3080.
Smith, R., and Coauthors, 1997: Local and remote effects of mountains on weather: research needs
and opportunities. Bull. Amer. Meteor. Soc., 78, 877–892.
Soler, M. R., C. Infante, P. Buenestado, and L. Mahrt, 2002: Observations of nocturnal drainage
flow in a shallow gully. Bound.-Layer Meteor., 105, 253–273.
Spengler, T., J. H. Schween, M. Ablinger, G. Zängl, and J. Egger, 2009: Thermally driven flows at
an asymmetric valley exit: Observations and model studies at the Lech Valley exit. Mon. Wea.
Rev., 137, 3437–3455.
Start, G. E., C. R. Dickson, and L. L. Wendell, 1975: Diffusion in a canyon within rough
mountainous terrain. J. Appl. Meteor., 14, 333–346.
Steinacker, R., 1984: Area-height distribution of a valley and its relation to the valley wind.
Contrib. Atmos. Phys., 57, 64–71.
Steinacker, R., and Coauthors, 2006: A mesoscale data analysis and downscaling method over
complex terrain. Mon. Wea. Rev., 134, 2758–2771.
Steinacker, R., and Coauthors, 2007: A sinkhole field experiment in the eastern Alps. Bull. Amer.
Meteor. Soc., 88, 701–716.
Stewart, J. Q., C. D. Whiteman, W. J. Steenburgh, and X. Bian, 2002: A climatological study of
thermally driven wind systems of the US Intermountain West. Bull. Amer. Meteor. Soc., 83,
699–708.
Stone, G. L., and D. E. Hoard, 1989: Low frequency velocity and temperature fluctuations in
katabatic valley flows. J. Appl. Meteor., 28, 477–488.
Stull, R. B., 1988: An Introduction to Boundary Layer Meteorology. Kluwer Academic, 670pp.
Sturman, A. P., and H. A. McGowan, 1995: An assessment of boundary-layer air mass character-
istics associated with topographically-induced local wind systems. Bound.-Layer Meteor., 74,
181–193.
Sturman, A. P., H. A. McGowan, and R. A. Spronken-Smith, 1999: Mesoscale and local climates.
Progress Physical Geography, 23, 611–635.
Sturman, A. P., and Coauthors, 2003a: The Lake Tekapo Experiment (LTEX): An investigation
of atmospheric boundary layer processes in complex terrain. Bull. Amer. Meteor. Soc., 84,
371–380.
Sturman, A. P., and Coauthors, 2003b: Supplement to the Lake Tekapo Experiment (LTEX): An
investigation of atmospheric boundary layer processes in complex terrain. Bull. Amer. Meteor.
Soc., 84, 381–383.
Sun, J., S. P. Burns, A. C. Delany, S. P. Oncley, T. W. Horst, and D. H. Lenschow, 2003: Heat
balance in the nocturnal boundary layer during CASES-99. J. Appl. Meteor., 42, 1649–1666.
Szintai, B., P. Kaufmann, and M. Rotach, 2010: Simulation of pollutant transport in complex
terrain with a numerical weather prediction–particle dispersion model combination. Bound.-
Layer Meteor., 137, 373–396.
Taylor, J. R., M. Kossmann, D. J. Low, and P. Zawar-Reza, 2005: Summertime easterly surges in
southeastern Australia: A case study of thermally forced flow. Austr. Met. Mag., 54, 213–223.
Thompson, B. W., 1986: Small-scale katabatics and cold hollows. Weather, 41, 146–153.
Trachte, K., T. Nauss, and J. Bendix, 2010: The impact of different terrain configurations on the
formation and dynamics of katabatic flows: Idealised case studies. Bound.-Layer Meteor., 134,
307–325.
Triantafyllou, A. G., C. G. Helmis, D. N. Asimakopoulos, and A. T. Soilemes, 1995: Boundary
layer evolution over a large and broad mountain basin. Theor. Appl. Climatol., 52, 19–25.
Urfer-Henneberger, C., 1970: Neure Beobachtungen über die Entwicklung des
Schönwetterwindsystems in einem V-förmigen Alpental (Dischmatal bei Davos). [New
observations of the development of a clear weather wind system in a v-shaped mountain valley
(Dischma Valley near Davos)]. Arch. Meteor. Geophys. Bioclimatol., Ser. B, 18, 21–42.
Van de Wiel, B. J. H., A. F. Moene, O. K. Hartogensis, H. A. R. De Bruin, and A. A. M. Holtslag,
2003: Intermittent turbulence and oscillations in the stable boundary layer over land. Part III.
A classification for observations during CASES-99. J. Atmos. Sci., 60, 2509–2522.
Van den Broeke, M. R., 1997a: Momentum, heat and moisture budgets of the katabatic wind layer
over a midlatitude glacier in summer. J. Appl. Meteor., 36, 763–774.
Van den Broeke, M. R., 1997b: Structure and diurnal variation of the atmospheric boundary layer
over a mid-latitude glacier in summer. Bound.-Layer Meteor., 83, 183–205.
Van Gorsel, E., 2003: Aspects of flow characteristics and turbulence in complex terrain. Results
from the MAP-Riviera project. Dissertation. University of Basel, Geography Department.
ISBN 3-85977-247-1. 58pp.
Van Gorsel, E., A. Christen, C. Feigenwinter, E. Parlow, and R. Vogt, 2003a: Daytime turbulence
statistics above a steep forested slope. Bound.-Layer Meteor., 109, 311–329.
Van Gorsel, E., R. Vogt, A. Christen, and M. W. Rotach, 2003b: Low frequency temperature and
velocity oscillations in katabatic winds. Ext. Abstr., Vol A, Int. Conf. Alpine Meteor. and MAP
Meeting 2003. Publ. MeteoSwiss, No. 66, 251–254.
Varvayanni, M., J. G. Bartzis, N. Catsaros, P. Deligianni, and C. E. Elderkin, 1997: Simulation of
nocturnal drainage flows enhanced by deep canyons: The Rocky Flats case. J. Appl. Meteor.,
36, 775–791.
Vergeiner, I., 1983: Dynamik alpiner Windsysteme. [Dynamics of Alpine wind systems]. Bericht
zum forschungsvorhaben “3556” des Fonds zur Forderung der wissenschaftlichen Forschung,
Wien, Austria, 129pp.
Vergeiner, I., 1987: An elementary valley wind model. Meteor. Atmos. Phys., 36, 255–263.
Vergeiner, I., and E. Dreiseitl, 1987: Valley winds and slope winds – observations and elementary
thoughts. Meteor. Atmos. Phys., 36, 264–286.
Vertacnik, G, I. Sinjur, and M. Ogrin, 2007: Temperature comparison between some Alpine do-
lines in winter time. ICAM 2007. http://www.cnrm.meteo.fr/icam2007/html/PROCEEDINGS/
ICAM2007/extended/manuscript 95.pdf
Viana, S., E. Terradellas, and C. Yagüe, 2010: Analysis of gravity waves generated at the top of a
drainage flow. J. Atmos. Sci., 67, 3949–3966.
Vosper, S. B., and A. R. Brown, 2008: Numerical simulations of sheltering in valleys: The
formation of nighttime cold-air pools. Bound.-Layer Meteor., 127, 429–448.
Vrhovec, T., 1991: A cold air lake formation in a basin – a simulation with a mesoscale numerical
model. Meteor. Atmos. Phys., 46, 91–99.
Wagner, A., 1932: Neue Theorie der Berg- und Talwinde [New theory of mountain and valley
winds]. Meteor. Z., 49, 329–341.
Wagner, A., 1938: Theorie und Beobachtung der periodischen Gebirgswinde. [Theory and
observation of periodic mountain winds]. Gerlands Beitr. Geophys. (Leipzig), 52, 408–449.
[English translation: Whiteman, C.D., and E. Dreiseitl, 1984: Alpine Meteorology: Translations
of Classic Contributions by A. Wagner, E. Ekhart and F. Defant. PNL-5141/ASCOT-84-3.
Pacific Northwest Laboratory, Richland, Washington, 121pp].
Weber, R. O., and P. Kauffmann, 1995: Automated classification scheme for wind fields. J. Appl.
Meteor., 34, 1133–1141.
Weber, R. O., and P. Kaufmann, 1998: Relationship of synoptic winds and complex terrain flows
during the MISTRAL field experiment. J. Appl. Meteor., 37, 1486–1496.
Weigel, A. P., 2005: On the atmospheric boundary layer over highly complex topography.
Dissertation No. 15972, ETH, Zurich. [http://e-collection.ethbib.ethz.ch/view/eth:27927].
Weigel, A. P., and M. W. Rotach, 2004: Flow structure and turbulence characteristics of the daytime
atmosphere in a steep and narrow Alpine valley. Quart. J. Roy. Meteor. Soc., 130, 2605–2627.
Weigel, A. P., F. K. Chow, M. W. Rotach, R. L. Street, and M. Xue, 2006: High-resolution large-
eddy simulations of flow in a steep Alpine valley. Part II. Flow structure and heat budgets.
J. Appl. Meteor. Climatol., 45, 87–107.
Weigel, A. P., F. K. Chow, and M. W. Rotach, 2007: On the nature of turbulent kinetic energy in a
steep and narrow Alpine valley. Bound.-Layer Meteor., 123, 177–199.
Weissmann, M., F. J. Braun, L. Gantner, G. J. Mayr, S. Rahm, and O. Reitebuch, 2005: The Alpine
mountain-plain circulation: Airborne Doppler lidar measurements and numerical simulations.
Mon. Wea. Rev., 33, 3095–3109.
Whiteman, C. D., 1982: Breakup of temperature inversions in deep mountain valleys: Part I.
Observations. J. Appl. Meteor., 21, 270–289.
Whiteman, C. D., 1986: Temperature inversion buildup in Colorado’s Eagle valley. Meteor. Atmos.
Physics, 35, 220–226.
Whiteman, C. D., 1990: Observations of thermally developed wind systems in mountainous terrain.
In: W. Blumen, Editor, AMS Meteorological Monographs, 45 (23), 5–42.
Whiteman, C. D., 2000: Mountain Meteorology: Fundamentals and Applications. Oxford Univer-
sity Press, New York, 355pp.
Whiteman, C. D., and T. B. McKee, 1982: Breakup of temperature inversions in deep mountain
valleys: Part II. Thermodynamic model. J. Appl. Meteor., 21, 290–302.
Whiteman, C. D., and E. Dreiseitl, 1984: Alpine Meteorology. Translations of Classic Contri-
butions by A. Wagner, E. Ekhart and F. Defant. PNL-5141/ASCOT-84-3. Pacific Northwest
Laboratory, Richland, Washington, 121 pp.
Whiteman, C. D., and J. C. Doran, 1993: The relationship between overlying synoptic-scale flows
and winds within a valley. J. Appl. Meteor., 32, 1669–1682.
Whiteman, C. D., and X. Bian, 1998: Use of radar profiler data to investigate large-scale thermally
driven flows into the Rocky Mountains. Proc. Fourth Int. Symp. on Tropospheric Profiling:
Needs and Technologies, Snowmass, CO.
Whiteman, C. D., and S. Zhong, 2008: Downslope flows on a low-angle slope and their interactions
with valley inversions. I. Observations. J. Appl. Meteor. Climatol., 47, 2023–2038.
Whiteman, C. D., K. J. Allwine, L. J. Fritschen, M. M. Orgill, and J. R. Simpson, 1989a: Deep
valley radiation and surface energy budget microclimates. I. Radiation. J. Appl. Meteor., 28,
414–426.
Whiteman, C. D., K. J. Allwine, M. M. Orgill, L. J. Fritschen, and J. R. Simpson, 1989b: Deep
valley radiation and surface energy budget microclimates. II. Energy budget. J. Appl. Meteor.,
28, 427–437.
Whiteman, C. D., T. B. McKee, and J. C. Doran, 1996: Boundary layer evolution within a
canyonland basin. Part I. Mass, heat, and moisture budgets from observations. J. Appl. Meteor.,
35, 2145–2161.
Whiteman, C. D., X. Bian, and J. L. Sutherland, 1999a: Wintertime surface wind patterns in the
Colorado River valley. J. Appl. Meteor., 38, 1118–1130.
Whiteman, C. D., X. Bian, and S. Zhong, 1999b: Wintertime evolution of the temperature inversion
in the Colorado Plateau Basin. J. Appl. Meteor., 38, 1103–1117.
Whiteman, C. D., S. Zhong, and X. Bian, 1999c: Wintertime boundary-layer structure in the Grand
Canyon. J. Appl. Meteor., 38, 1084–1102.
Whiteman, C. D., S. Zhong, X. Bian, J. D. Fast, and J. C. Doran, 2000: Boundary layer evolution
and regional-scale diurnal circulations over the Mexico Basin and Mexican Plateau. J. Geophys.
Res., 105, 10081–10102.
Whiteman, C. D., S. Zhong, W. J. Shaw, J. M. Hubbe, X. Bian, and J. Mittelstadt, 2001: Cold pools
in the Columbia Basin. Wea. Forecasting, 16, 432–447.
Whiteman, C. D., S. Eisenbach, B. Pospichal, and R. Steinacker, 2004a: Comparison of vertical
soundings and sidewall air temperature measurements in a small Alpine basin. J. Appl. Meteor.,
43, 1635–1647.
Whiteman, C. D., B. Pospichal, S. Eisenbach P. Weihs, C. B. Clements, R. Steinacker, E. Mursch-
Radlgruber, and M. Dorninger, 2004b: Inversion breakup in small Rocky Mountain and Alpine
basins. J. Appl. Meteor., 43, 1069–1082.
Whiteman, C. D., T. Haiden, B. Pospichal, S. Eisenbach, and R. Steinacker, 2004c: Minimum
temperatures, diurnal temperature ranges and temperature inversions in limestone sinkholes of
different size and shape. J. Appl. Meteor., 43, 1224–1236.
Whiteman, C. D., S. F. J. De Wekker, and T. Haiden, 2007: Effect of dewfall and frostfall on
nighttime cooling in a small, closed basin. J. Appl. Meteor., 46, 3–13.
Whiteman, C. D., and Coauthors, 2008: METCRAX 2006 – Meteorological experiments in
Arizona’s Meteor Crater. Bull. Amer. Meteor. Soc., 89, 1665–1680.
Whiteman, C.D., S.W. Hoch, and G.S. Poulos, 2009: Evening temperature rises on valley floors
and slopes: Their causes and their relationship to the thermally driven wind system. J. Appl.
Whiteman, C. D., S. W. Hoch, M. Lehner, and T. Haiden, 2010: Nocturnal cold air intrusions
into Arizona’s Meteor Crater: Observational evidence and conceptual model. J. Appl. Meteor.
Climatol., 49, 1894–1905.
Wilczak, J. M., S. P. Oncley, and S. A. Stage, 2001: Sonic anemometer tilt correction algorithms.
Winkler, P., M. Lugauer, and O. Reitebuch, 2006: Alpines Pumpen [Alpine pumping]. Promet
(Deutscher Wetterdienst), 32, 34–42. [In German].
Wong, R. K. W., and K. D. Hage, 1995: Numerical simulation of pollutant dispersion in a small
valley under conditions with supercitical Richardson number. Bound.-Layer Meteor., 73, 15–33.
Wotawa, G., and H. Kromp-Kolb, 2000: The research project VOTALP – general objectives and
main results. Atmos. Environ., 34, 1319–1322.
Wu, P., M. Hara, J.I. Hamada, M.D. Yamanaka, and F. Kimura, 2009: Why a large amount of rain
falls over the sea in the vicinity of western Sumatra island during nighttime. J. Appl. Meteor.
Climatol., 48, 1345–1361.
Yang, K., T. Koike, H. Fujii, T. Tamura, X. Xu, L. Bian, and M Zhou, 2004: The daytime evolution
of the atmospheric boundary layer and convection over the Tibetan Plateau: Observations and
simulations. J. Meteor. Soc. Japan, 82, 1777–1792.
Yao, W. Q., and S. Zhong, 2009: Nocturnal temperature inversions in a small enclosed basin and
their relationship to ambient atmospheric conditions. Meteor. Atmos. Phys., 103, 195–210.
Ye, Z., M. Segal, J. R. Garratt, and R. A. Pielke, 1989: On the impact of cloudiness on the
characteristics of nocturnal downslope flows. Bound.-Layer Meteor., 49, 23–51.
Ye, Z. J., J. R. Garratt, M. Segal, and R. A. Pielke, 1990: On the impact of atmospheric thermal
stability on the characteristics of nocturnal downslope flows. Bound.-Layer Meteor., 51, 77–97.
Yoshino, M. M., 1975: Climate in a Small Area. Tokyo, University of Tokyo Press, 549pp.
Zammett, R. J., and A. C. Fowler, 2007: Katabatic winds on ice sheets: A refinement of the Prandtl
model. J. Atmos. Sci., 64, 2707–2716.
Zängl, G., 2003: The impact of upstream blocking, drainage flow and the geostrophic pressure
gradient on the persistence of cold-air pools. Quart. J. Roy. Meteor. Soc., 129, 117–137.
Zängl, G., 2004: A reexamination of the valley wind system in the Alpine Inn Valley with numerical
simulations. Meteor. Atmos. Phys., 87, 241–256.
Zängl, G., 2005: Wintertime cold-air pools in the Bavarian Danube Valley Basin: Data analysis
and idealized numerical simulations. J. Appl. Meteor., 44, 1950–1971.
Zängl, G., 2008: The impact of weak synoptic forcing on the valley-wind circulation in the Alpine
Inn Valley. Meteor. Atmos. Phys., 105, 37–53.
Zängl, G., and J. Egger, 2005: Diurnal circulation of the Bolivian Altiplano. Part II: Theoretical
and model investigations. Mon. Wea. Rev., 133, 3624–3643.
Zängl, G., and S. G. Chico, 2006: The thermal circulation of a grand plateau: Sensitivity to the
height, width, and shape of the plateau. Mon. Wea. Rev., 134, 2581–2600.
Zängl, G., and S. Vogt, 2006: Valley-wind characteristics in the Alpine Rhine Valley: Measure-
ments with a wind-temperature profiler in comparison with numerical simulations. Meteor. Z.,
15, 179–186.
Zängl, G., J. Egger, and V. Wirth, 2001: Diurnal winds in the Himalayan Kali Gandaki Valley. Part
II. Modeling. Mon. Wea. Rev., 129, 1062–1080.
Zardi, D., 2000: A model for the convective boundary layer development in an Alpine valley,
Proceedings, 26th Int. Conf. Alpine Meteor. ICAM2000, Innsbruck, Austria, 11–15 September
2000. Österreichische Beiträge zur Meteorologie und Geophysik. Heft Nr. 23/Publ. Nr. 362.
ISSN 1016–6254.
Zardi, D., R. Gerola, F. Tampieri, and M. Tubino, 1999: Measurement and modelling of a valley
wind system in the Alps. Proceedings of the 13th AMS Symposium on Boundary Layers and
Turbulence. 10–15 January 1999, Dallas, TX, 28–31.
Zaremba, L. L., and J. J. Carroll, 1999: Summer wind flow regimes over the Sacramento Valley.
J. Appl. Meteor., 38, 1463–1473.
Zawar-Reza, P., and A. P. Sturman, 2006a: Numerical analysis of a thermotopographically-induced
mesoscale circulation in a mountain basin using a non-hydrostatic model. Meteor. Atmos. Phys.,
93, 221–233.
Zawar-Reza, P., and A. P. Sturman, 2006b: Two-dimensional numerical analysis of a thermally
generated mesoscale wind system observed in the MacKenzie Basin, New Zealand. Aust.
Meteor. Mag., 55, 19–34.
Zawar-Reza, P., H. McGowan, A. Sturman, and M. Kossman, 2004: Numerical simulations of wind
and temperature structure within an Alpine lake basin, Lake Tekapo, New Zealand. Meteor.
Atmos. Phys., 86, 245–260.
Zhong, S., and J. Fast, 2003: An evaluation of the MM5, RAMS, and Meso-Eta models at
subkilometer resolution using VTMX field campaign data in the Salt Lake Valley. Mon. Wea.
Rev., 131, 1301–1322.
Zhong, S., and C. D. Whiteman, 2008: Downslope flows on a low-angle slope and their interactions
with valley inversions. II. Numerical modeling. J. Appl. Meteor. Climatol., 47, 2039–2057.
Zhong, S., C. D. Whiteman, X. Bian, W. J. Shaw, and J. M. Hubbe, 2001: Meteorological processes
affecting evolution of a wintertime cold air pool in the Columbia Basin. Mon. Wea. Rev., 129,
2600–2613.
Zhong, S., X. Bian, and C. D. Whiteman, 2003: Time scale for cold-air pool breakup by turbulent
erosion. Meteor. Z., 12, 229–233.
Zhong, S., C. D. Whiteman, and X. Bian, 2004: Diurnal evolution of three-dimensional wind and
temperature structure in California’s Central Valley. J. Appl. Meteor., 43, 1679–1699.
Zoumakis, N. M., and G. A. Efstathiou, 2006a: Parameterization of inversion breakup in idealized
valleys. Part I: The adjustable model parameters. J. Appl. Meteor. Climatol., 45, 600–608.
Zoumakis, N. M., and G. A. Efstathiou, 2006b: Parameterization of inversion breakup in idealized
valleys. Part II: Thermodynamic model. J. Appl. Meteor. Climatol., 45, 609–623.
Zoumakis, N. M., G. A. Efstathiou, A. G. Kelessis, J. Triandafyllis, D. Papas, M. Chasapis, M.
Petrakakis, P. Karavelis, 2006: A simple scheme for daytime estimates of surface energy budget
in complex terrain. Fresenius Environmental Bulletin, 15, 923–927.
Chapter 3
Dynamically-Driven Winds
Peter L. Jackson, Georg Mayr, and Simon Vosper
Abstract This chapter is concerned with dynamically-forced atmospheric flow

phenomena which occur when the wind encounters mountains. The range of effects
is wide and therefore attention is restricted to arguably the most important phenom-
ena in terms of weather forecasting. These are mountain waves, rotors, downslope
windstorms, gap winds and barrier jets. The essence of many of these phenomena is
described by mountain wave theory. Recent advances in observation technologies
and their application in field programs, as well as in numerical modeling, have
led to new understanding, including the incorporation of complicating factors
like boundary-layer processes. This chapter describes current theory for each of
these phenomena, along with recent observational studies and the latest forecast
techniques and models.
P.L. Jackson ()

Environmental Science & Engineering, Natural Resources and Environmental Studies Institute,
University of Northern British Columbia, 3333 University Way, Prince George, British Columbia,
Canada V2N 4Z9
e-mail: peterj@unbc.ca
G. Mayr
Institute of Meteorology and Geophysics, University of Innsbruck, Innrain 52,
6020 Innsbruck, Austria
e-mail: georg.mayr@uibk.ac.at
S. Vosper
Atmospheric Processes and Parametrizations, Met Office, FitzRoy Road, Exeter,
United Kingdom EX1 3PB
e-mail: simon.vosper@metoffice.gov.uk

122 P.L. Jackson et al.
3.1 Introduction: Flow Over, Around and Through

Mountain Ranges
This chapter deals with flow phenomena that are dynamically forced when wind
interacts with orography. As air approaches a mountain barrier, it can pass over
it, flow through gaps or valleys that dissect it, or be blocked by the mountain
and diverted horizontally around it. Each of these scenarios results in different
dynamically-driven wind phenomena. When air is able to pass over a mountain
barrier under specific conditions, mountain waves are generated that can result in lee
waves, rotors and downslope windstorms. Air flowing through channels that dissect
a mountain barrier can result in gap flow arising from along-channel acceleration.
If the air is blocked by a mountain barrier and diverted around it, then a barrier jet
may form. This chapter describes these various phenomena.
For air to be able to pass over a mountain, it must either have low stability
or large inertia (wind speed). Air that approaches a mountain barrier will be
forced to lift as it encounters higher terrain. As the air lifts, if the atmosphere is
stable, it will encounter negative buoyancy that will slow the wind in the across-
barrier direction. Unless the approaching air has sufficient inertia to cross the
barrier, it may be blocked entirely by the barrier and forced horizontally around
it. The ratio of energy needed to overcome the negative buoyancy imposed by
the mountain, to the horizontal inertia that can provide this energy, is given by a
fundamental dimensionless parameter, called the non-dimensional mountain height,
HO D NH=U , where N is the Brunt Väisälä, or buoyancy frequency, H is the
mountain barrier height, and U is the upstream wind speed component perpendicular
to the mountain barrier. The square of the Brunt Väisälä frequency is defined as
g d
N2 D ;
dz
where z is height, g is the acceleration due to gravity and is the potential

temperature. This will be discussed further in Sect. 3.2.2. If the air does not have
sufficient inertia to go over the mountain, then it will be diverted around the
mountain and may form an elevated barrier jet, as described in Sect. 3.6.
If the approaching air does have sufficient inertia to overcome stability and cross
a mountain barrier, the displacement caused by the barrier can have either horizontal
or vertical consequences depending on how long it takes for air to cross the barrier.
If the mountain barrier is so wide that it takes longer than f1 (f is the Coriolis
parameter) for air to cross it, then Coriolis effects and conservation of potential
vorticity will result in predominantly horizontal perturbations to the flow. This will
result in the generation of large-scale horizontal topographic Rossby waves with
ridges over the mountain barrier and troughs to the lee of the mountains. These
conditions depend on values of another important dimensionless number, the across-
barrier Rossby number, Ro D U/fL 1, where L is the mountain barrier width.
3 Dynamically-Driven Winds 123
On the other hand, when Ro 1, which would occur for mountain barriers less
than about 100 km in width for typical values of U (10 m s1 ) and f (104 s1 ),
the horizontal displacements are reduced, Coriolis effects can be largely ignored to
a linear approximation, and air parcels are displaced vertically. If the air is stably
stratified, but still has sufficient inertia to cross the barrier, the vertically displaced
air parcels will form gravity waves that oscillate about their equilibrium level. These
gravity waves, called mountain waves, may, depending on conditions, propagate
vertically or horizontally. Horizontally propagating gravity waves that are vertically
trapped can form lee waves and rotors (Sect. 3.3), and vertically propagating
gravity waves in some circumstances may amplify and cause downslope windstorms
(Sect. 3.4).
Dynamically-driven winds both affect, and are affected by, many other aspects
of mountain weather discussed in this book. For example, while early work
on dynamically-driven winds ignored “complicating” factors such as thermal,
boundary-layer, and moist processes (see discussions in Chaps. 2, 5, 6 and 7), it
is now recognized that considering these factors is often essential to go beyond a
first-order understanding of forced flows. In turn, strongly-forced dynamic flows
can overwhelm thermal and boundary-layer processes. The foehn, a well-known
and studied type of forced mountain flow is discussed in much more detail in
Chap. 4. Early progress in understanding dynamically-driven flows, was mainly
due to theoretical developments. While theory continues to be important, new
observations of the mountain atmosphere made possible by advances in technology
(Chap. 8) and high-resolution numerical simulation made possible by advances in
computing and models (Chaps. 9 and 10) have opened new avenues of inquiry
and understanding. As numerical weather prediction models (Chap. 11) begin to
achieve the resolution and model development necessary to adequately represent
dynamically-forced flows, they improve in their ability to forecast these flows
directly.
The chapter starts in Sect. 3.2 by introducing basic concepts: the equations used,
mountain waves, and hydraulic theory. The sections that follow this are organized
by flow phenomena. Sections 3.3 and 3.4 describe phenomena that are forced by
flow over mountains: Lee waves and rotors, and downslope windstorms. Section 3.5
describes gap flow that occurs when air is forced through valleys that dissect
mountain barriers. Section 3.6 describes barrier jets that arise when flow impinging
on a mountain barrier is blocked by it.
3.2 Fundamentals of Forced Flows in Complex Terrain
In order to understand the complex and diverse behavior of atmospheric flow in

mountainous terrain, knowledge of the most fundamental processes is first required.
Perhaps the two most important phenomena to consider are internal gravity waves
and the hydraulic behavior of shallow-layer flow across mountains and through
passes. Many aspects of dynamically driven flows in complex terrain are influenced
by these processes. In this section we shall briefly review the relevant equations
of motion and then separately consider internal gravity waves and hydraulic flow
theory. A number of simplifications and approximations can be applied to the
equations in order to bring out the most important aspects of the dynamics.
3.2.1 Flow Equations
The main aspects of the flows considered here are represented by the Euler equations
which govern the motion of inviscid, frictionless adiabatic flow of dry air. For flow
in a rotating frame of reference (the Earth), these equations can be written as
Du 1
C f k u D rp gk (3.1)
Dt
D
C r .u/ D 0 (3.2)
Dt
D
D0 (3.3)
Dt
where u is velocity, ¡ is density, p is pressure and k is the unit vector in the vertical
(z) direction. Dissipative processes, such as the effects of viscosity or turbulent
mixing, and moist processes are not represented by these equations. Such processes
are clearly important for certain phenomena but our aim here is merely to illustrate
the primary aspects of dynamically driven winds. This can be accomplished without
the need to consider these additional effects.
Equations 3.1–3.3 describe the motion of a compressible gas and therefore permit
sound waves. Since these have limited dynamical significance and complicate
further analysis, it is useful to remove them from the system. This can be achieved
through the anelastic (e.g. Ogura and Phillips 1962) or Boussinesq approximations.
Subtly different versions of each exist and a comprehensive description is given
by Durran (1999) and references therein. It is useful to split the thermodynamic
variables, pressure, density and potential temperature into steady hydrostatically
balanced reference profiles (that are a function of height, z, only) and perturbations
which vary in space and time. In other words we define p.x; t/ D p.z/ N C p 0 .x; t/,
.x; t/ D .z/ 0 N 0
N C .x; t/ and .x; t/ D .z/ C .x; t/. It is also helpful to introduce
the Exner pressure, D .p=p0 /R=cp and let .x; t/ D .z/ N C 0 .x; t/ where p0
is a reference pressure and R and cp are the gas constant and specific heat capacity
for dry air, respectively. The anelastic approximation consists of neglecting time
derivatives of the density term in the continuity equation and replacing (3.2) by
r .u/
N D0 (3.4)
The momentum equation, (3.1), can also be written in a more useful form as:
Du g 0
C f k u C cp r 0 D k (3.5)
Dt N
Equations 3.5, 3.4 and 3.3 are the anelastic equations and their solutions do not
contain sound waves. The term on the right-hand side of (3.5) is the buoyancy term
and is of fundamental importance to stratified flows over mountains.
Perhaps the simplest example of a stratified flow is that of a shallow homoge-
neous layer of fluid with a free surface. The flow of such a fluid over an obstacle is
analogous to certain types of dynamically driven flows in mountainous terrain. For
example, the stationary patterns of ripples and the turbulent hydraulic jumps which
form in a stream as it flows over undulations in the river bed are analogous to some
of the features observed in atmospheric flows over mountains. The acceleration of
a channel flow through a narrow constriction also resembles that of an atmospheric
flow through a mountain pass. Consequently much progress can be made toward
understanding atmospheric flows by considering the equations which govern the
motion of these simpler single layers, the shallow water equations. For a free-surface
height, (x,y), the shallow water equations can be written as:
@u @u @u @
Cu Cv f v D g (3.6)
@t @x @y @x
@v @v @v @
Cu Cv C fu D g (3.7)
@t @x @y @y
@ @.u/ @.v/
C C D0 (3.8)
@t @x @y
In the atmospheric context, (x,y) might describe the height of a discontinuity

in density or temperature, such as an inversion. The analogy between this simple
shallow water system, or its more complicated counterpart where the flow consists
of multiple homogeneous layers, provides a useful basis for interpretation of more
complex atmospheric flows.
3.2.2 Mountain Waves
Internal gravity waves, or buoyancy waves, are a fundamental feature of stably

stratified fluids and occur as a result of the buoyancy forces which act on fluid
parcels when they are vertically displaced from their equilibrium position. In the
atmosphere, internal gravity waves occur on the mesoscale, with typical horizontal
wavelengths lying in the range of a few hundred meters to several hundred
kilometers. Any process which involves lifting of air parcels or perturbations to
buoyancy through heating can result in gravity wave motion. Such processes include
convection, shear instability, convergence at frontal zones, drainage flows in stable

boundary layers (see Chap. 2) and, of course, the vertical displacements associated
with air flow over orography. The general term “mountain wave” is used to describe
this latter kind of internal gravity-wave motion.
Mountain waves are ubiquitous to flows over orography and are known to
have an important, if not controlling influence on the flow dynamics of a wide
range of physical processes. For example, the windstorms which form on the lee-
side of mountain ranges are intimately linked to the presence of mountain waves.
Mountain waves are also sometimes cited as the cause of clear-air turbulence, a
particular cause of concern for aviation safety. From the point of view of global
atmospheric circulation, mountain waves also play an important role. The waves
carry momentum and energy into the stratosphere, creating turbulence where they
break and imparting a net drag on the flow, influencing the planetary circulation and
thus the passage of synoptic-scale weather systems. Recently satellites have begun
to detect evidence for mountain waves in the middle and upper atmosphere (e.g.
Jiang et al. 2002). The use of microwave limb sounders to detect fluctuations in
temperature enables the distribution of gravity-wave activity to be examined on a
global scale. The most active areas appear to be the mountainous regions of the high
latitudes in wintertime, where the winds are strongest. It is likely that the observed
gravity waves are mountain waves which have propagated high into the stratosphere.
There is, of course, a vast body of literature on the subject of mountain waves
and there are several excellent and extensive reviews. Notably those by Smith
(1979, 1989), Baines (1995) and Nappo (2002) serve as very informative and useful
reviews of the subject and provide a thorough background to the theory of mountain
waves. In this section we aim only to present sufficient background to explain later
aspects of this chapter on downslope windstorms and lee waves and rotors. Over
the last two decades there have been several major field programs in which various
aspects of orographic flow, including mountain wave related phenomena, have been
examined in detail. Notable examples are the Pyrenees Experiment (PYREX, see
Bougeault et al. 1993), the Mesoscale Alpine Programme (MAP, see Bougeault
et al. 2001), and most recently the Terrain induced Rotor Experiment (T-REX, see
Grubišić et al. 2008). The observations made during these experiments have led
to advances in our understanding of various aspects of mountain-wave behaviour.
These include the generation, propagation and dissipation of mountain waves, the
drag associated with the waves, wave induced turbulence and the impacts of the
waves on the near-surface flow.
Two important non-dimensional parameters which determine the response of a
stratified fluid to orography are the parameters FL D U/(NL) and HO D NH=U ,
where L is the mountain width scale. As we shall see below, the quantity FL provides
information about the propagation characteristics of mountain waves. As described
in Sect. 3.1, the non-dimensional mountain height, HO , predicts the extent to which
the approaching flow will traverse over the mountain summit, or will be forced
around the mountain sides. In general when HO 1 the upstream flow contains
sufficient kinetic energy to overcome the potential energy barrier represented by the
mountain and the flow will therefore be largely over the mountain. When HO 1,
however, the buoyancy forces associated with the stable stratification act to restrict
vertical motion and the flow is constrained to be largely horizontal. The flow will
then be blocked upwind and is forced to travel around the flanks of the mountain.
The parameter HO provides a measure of the importance of nonlinearity. For HO 1
the flow response to the mountains is generally well described by linear theory. For
HO > 1, however, nonlinear processes are important. Sometimes Ĥ1 D U/NH is
referred to as a Froude number, although this is different from the Froude number
used in hydraulic theory that will be discussed later in which the length scale is the
depth of the fluid, not the height of the mountain barrier.
In the interests of simplicity and because here our focus will be on waves
generated by individual peaks rather than very broad continental-scale mountain
ranges such as the Rockies or Andes, we will now neglect the effects of rotation
in our analysis. This amounts to neglecting the Coriolis term in (3.5). Neglecting
rotation is valid when the Rossby number, Ro D U/(fL) 1 and this is generally true
for typical mountain widths and values of U and f. For example, with a mountain
width of L D 10 km, a wind speed of U D 10 m s1 and a mid-latitude value
of f D 104 s1 , Ro D 10. Note, however, that while rotation may be relatively
unimportant for the dynamics of the mountain waves on these scales, it does affect
the dispersion properties of gravity waves with very long horizontal wavelengths
(e.g. several hundred kilometers). It is also known that in the nonlinear flow regime
(HO > 1), for moderate values of Ro (roughly 1 Ro 10), rotation can have
significant effects on the flow. These range from asymmetry in the flow as it splits
around the flanks of the mountain (e.g. Hunt et al. 2001), acceleration of the flow
through mountain passes (e.g. Sprenger and Schär 2001), asymmetry in mountain
wave breaking (e.g. Grisogono and Enger 2004) and modification of the drag
exerted by the mountain on the atmosphere (e.g. Ólafsson and Bougeault 1997).
Further progress can be made if we now assume that disturbances to the mean
flow are small. By splitting the velocity vector, u, into a mean height dependent
profile and a perturbation and then linearizing the equations (by neglecting products
of primed quantities) we obtain the linearized anelastic equations for a dry non-
rotating inviscid atmosphere. These can be written as
@u0 N u0 C u0 r U N
N C cp r 0 g 0
C Ur D k (3.9)
@t N
N 0/ D 0
r .u (3.10)
@ 0 N
N r 0 C w0 d D 0
CU (3.11)
@t dz
where UN D .UN ; VN / is the horizontal mean flow vector and u0 .x; y; z; t/ D .u0 ; v0 ; w0 /
is the perturbation velocity vector. For mountain waves, this linearization of the
equations of motion is strictly only valid when both the mountain slope and the non-
dimensional mountain height are small (H/L 1 and HO 1). However the linear
equations prove to be useful both from the point of view of explaining many
of the dynamical properties of the flow, as well as providing useful quantitative
predictions well beyond their formal limits of applicability. We shall discuss the
role of nonlinearity, when HO is O(1) or larger, later in this chapter.
For the sake of simplicity we shall now further assume that the mean flow
contains no vertical wind shear (d U=d N z D 0) and that N is independent of
height. For now, we shall also make the Boussinesq approximation, which assumes
that the effect of changes in density can be neglected in the continuity equation
(3.10), allowing N to be replaced by a reference value 0 . Note that the Boussinesq
approximation is also commonly applied in studies of the atmospheric boundary
layer (see Chap. 2). Strictly, the Boussinesq approximation requires that the vertical
scale of motions is much less than the density scale height, H, defined as

1 d N 1
H D (3.12)
N dz
This approximation simplifies the following analysis, but it also removes an im-
portant feature of mountain-wave dynamics, namely the growth of wave amplitude
with height which occurs as a result of the decrease in density, which can ultimately
lead to wave breaking in the stratosphere.
Without loss of generality we shall further assume that the flow direction is
aligned with the x-direction (this simplifies the problem to a two-dimensional one)
and assume wave-like solutions for all perturbation variables with wavevector
D .k; m/ and frequency ! of the form
n o
O i.kxCmz!t / ;
0 D Re e (3.13)
where k and m are wavenumbers in the x and z directions, respectively, O is a

complex wave amplitude and Re denotes the real part. Substituting such expressions
into the Boussinesq versions of (3.9)–(3.11) gives the linear dispersion relation for
internal gravity wave motion, namely:
2 N 2k2
! D .! UN k/ D 2
_ 2
(3.14)
k C m2
The quantity !O is the intrinsic frequency of the gravity wave motion and
describes the frequency of motion experienced in a frame of reference travelling
with the mean flow speed, UN . Equation 3.14 states that the intrinsic frequency of
gravity wave motion is always limited by the Brunt Väisälä frequency and there is
therefore a natural limit to the rapidity of oscillations. Note that !O D N can in
fact only be achieved when the wave motion has no phase variation with height (i.e.
when m D 0), the case where the lines of constant phase are vertical. For phase lines
more inclined towards the horizontal the intrinsic frequency decreases.
Rearranging (3.14) we obtain

N2
m2 D k 2 1 (3.15)
!O 2
from which we can deduce that if !O 2 < N 2 then the vertical wavenumber is real
and the gravity wave motion is able to propagate in the vertical. If !O 2 > N 2 then
m is imaginary and the physically realistic solution is for the disturbances to decay
exponentially with height away from their source, in this case the orography. For
mountain waves, the source is of course stationary, implying that the frequency of
the waves (in a fixed frame of reference), ¨, is zero. Equation 3.15 then reduces to a
relationship between the vertical and horizontal wavenumbers and the background
flow properties, namely
N2
m2 D k2: (3.16)
UN 2
According to (3.16) we should therefore expect mountain waves with long
horizontal wavelengths to have vertical wavelengths which are relatively short.
The vertical wavelength will increase with decreasing horizontal wavelength until,
for disturbances which are sufficiently short, m becomes imaginary and the wave
motion decays in the vertical. If we consider the case of monochromatic waves gen-
erated by a simple sinusoidal mountain range with wavelength,
, the significance
of the non-dimensional parameter FL becomes clear. Mountain waves will only exist
when N=UN > k, or alternatively when FL D UN =.NL/ < 1, where the scale L is
related to the wavelength by L D
/(2 ). For values of FL > 1 the waves generated
have a frequency greater than N, so that the flow cannot support mountain waves
and disturbances will decay exponentially with height. Such flows are illustrated
schematically by Fig. 3.1.
It is now useful to define the group velocity, cg D .@!=@k; @!=@m/. This
describes the direction and rate of propagation of information or energy by the
waves. The group velocity is the velocity with which a “packet” of waves travels
through the atmosphere. From (3.14) we can show that
@! Nm2
D UN ˙ 3 (3.17)
@k
and
@! Nmk
D
3 ; (3.18)
@m
where D jj. If the background wind was to vary sufficiently slowly with height
that the dispersion relations still hold locally, then from (3.17) we can see that as
UN ! 0, m ! 1 and from (3.18), @!=@m ! 0. This implies that as a wave
a b
5000 5000
4000 4000
3000 3000
z (m)
z (m)
2000 2000
1000 1000
0 0
-10000 -5000 0 5000 10000 -10000 -5000 0 5000 10000
x (m) x (m)
Fig. 3.1 Streamlines over sinusoidal mountains with wavelength 10 km. The wind speed and
Brunt Väisälä frequency are independent of height with N D 0.01 s1 and (a) UN D 10 ms1
and (b) UN D 20 ms1 . In (a) the value of FL is approximately 0.628 and therefore propagating
mountain waves occur. In (b) FL 1.26 and the wave motion decays exponentially with height
approaches a level where the wind (resolved in the direction of the horizontal
wavevector) falls to zero, the vertical wavelength tends to zero and the rate at which
wave energy propagates in the vertical also tends to zero. Therefore wave energy
does not pass through such a level. The wave energy is dissipated at such “critical
levels” and reflection may occur as a result of nonlinearity. The latter is thought to
be an important mechanism in the dynamics of downslope windstorms, where the
breaking of waves in the troposphere may result in a mountain wave induced critical
level above the mountain.
The angle at which a packet of waves will propagate away from the mountain
is determined by the ratio of expressions (3.17) and (3.18). Choosing the sign of
vertical propagation to be positive (so that wave energy radiates vertically away
from the mountain) it can be shown that the angle of wave energy propagation, ˛,
away from the mountain is given by

@! @! 1
tan ˛ D D m=k: (3.19)
@m @k
Equations 3.16 and 3.19 tell us that when the horizontal wavelength of mountain
waves is sufficiently long that k N=UN , the vertical wavenumber m N=UN
and the wave motion is non-dispersive; the angle of propagation is near vertical.
The motion associated with these long wavelengths is, to a good approximation,
in hydrostatic balance. For flows across broad smooth mountains (with FL 1)
for which the majority of the mountain spectrum consists of long wavelengths, the
wave motion is therefore contained in a narrow beam above the mountain. Wave
energy radiates vertically and little wave motion is encountered downwind. For
shorter horizontal wavelengths whose k is less than, but not significantly less than
N=UN , the waves are dispersive. In this case the angle ˛is reduced (and depends on
the wavenumber, k) and the wave motion is non-hydrostatic. Therefore for narrow
Fig. 3.2 The mountain waves generated by flow over (a) broad and (b) narrow ridges when
the upwind profile consists of constant stability and wind speed (N D 0.02 s1 and U D 20 m
s1 ). Color shading denotes vertical velocity (m s1 ) and isentropes are denoted by line contours
(interval 5 K). In (a) the width-scale L is 10 km and FL D 0.1, resulting in a hydrostatic mountain
wave response in which the waves propagate vertically and the wave motion is restricted to a
vertical beam above the mountain. In (b) L is 1 km and FL D 1. In this case the wave motion is
non-hydrostatic. The waves propagate horizontally as well as vertically and wave motion is present
in a dispersive tail to the lee of the ridge
mountains, where FL < 1 and a significant portion of the mountain spectrum is made
up of short wavelengths, wave energy radiates in the downstream direction as well
as in the vertical, resulting in the appearance of “lee wave” motion downwind.
Examples of the hydrostatic and non-hydrostatic wave fields associated with
broad and narrow mountain ridges, respectively, are illustrated in Fig. 3.2. The
wave fields shown are those which occur in flow over two-dimensional ridges of the
1
commonly used bell shaped form h.x/ D H 1 C x 2 =L2 . The upwind profiles
have constant stability and wind speed (N D 0.02 s1 and UN D 20 ms1 ) and the
wave fields shown are linear solutions, computed using (3.21) (see below). For the
broad mountain, where L D 10 km and FL D 0.1, wave motion is only observed
immediately above the mountain. These hydrostatic mountain waves propagate
vertically into the upper atmosphere. In contrast, for the narrower mountain, when
L D 1 km and FL D 1, the dispersive nature of the energy propagation (implied by
the k dependence in (3.19)) means that short wavelength waves have a horizontal
component to their propagation and hence wave motion appears downwind of the
ridge as well as above it.
In practice, the complexity of real mountain ranges means that they are made
up of a wide spectrum of horizontal scales and thus we should expect a range
of responses from the mountain wave field. The very shortest scales, perhaps
associated with individual peaks or narrow valleys, may in general be too short
to force wave motion and their associated disturbances may decay with height. For
longer scales, we can expect both a vertically propagating hydrostatic wave response
as well as non-hydrostatic waves which trail downwind of the mountain range. It is
worth bearing in mind that the increase of wind strength with height through the
troposphere, typical of mid-latitudes, implies that wave motion which is hydrostatic
(and hence vertically propagating) at low levels, may in fact respond in a more
dispersive non-hydrostatic manner aloft. The shorter, non-hydrostatic, wavelengths
may themselves be greatly affected by changes in wind and stability with height. As
we shall see in Sect. 3.3, this will often result in a trapped lee-wave response, where
the wave energy is contained within the troposphere.
In order to understand how mountain waves are affected by height variations in
wind and stability we shall return to (3.9–3.11) but this time drop the assumptions
of zero shear and constant stability. Note that we have dropped the Boussinesq
approximation for the time being, since density is a function of height in (3.10).
We seek solutions to the steady versions of the equations with the form
O
0 D .z/e i.kxCly/
: (3.20)
Substituting such expressions into (3.9–3.11) results, after some algebra, in the
vertical structure, or Taylor-Goldstein equation for mountain waves in a sheared
flow with non-uniform stability, namely

d 1 d _
_
Nw C
2 h 2 w D 0; (3.21)
dz N dz
where h D .k; l/ is the horizontal wavevector, h D jh j, wO (h , z) is a complex

vertical velocity amplitude and the function
2 , often called the Scorer parameter,
is given by
N 2 h 2 1 d2 N

2 .z/ D .U:h /: (3.22a)
N h/
.U:
2 N h dz2
U:
For flow over a two-dimensional ridge, the wave field is made up of horizontal
wavevectors with a single direction, perpendicular to the ridge. In this case the
expression for
2 simplifies further to
N2 1 d 2 Ur

2 .z/ D ; (3.22b)
Ur 2 Ur d z2
where Ur is the wind speed resolved in the direction perpendicular to the ridge.
The presence of density in the first term of (3.21) results in the general growth
of mountain wave amplitude with height which can lead to wave breaking in the
upper atmosphere. This breakdown of wave motion can lead to the generation
of turbulence and a net drag force on the mean flow. The latter is commonly
parameterized in global weather prediction and climate models. Note that this wave
growth effect would not have been accounted for if we had retained the Boussinesq
approximation in the derivation of (3.21); the first term on the left-hand side would
O z2 .
then have been simply d 2 w=d
The propagation of mountain waves is affected by the variations in

2 .z/ which
arise from changes in the background wind profile or static stability. As a rule,
the increase of wind speed with height through the troposphere means that
2 will
decrease with altitude, implying that short wavelength wave motion (for which
2
h 2 > 0 at low levels) may decay with height at higher altitudes. In general the
first term on the right-hand sides of (3.22a and 3.22b) tend to dominate over the
second, although the latter term can be important across jets where there may be
rapid changes in vertical wind shear. The importance of variations in
2 will become
more apparent in Sect. 3.3 when we consider the phenomenon of wave reflection,
which can result in the channeling of wave energy downwind and the long trains of
waves, commonly known as trapped lee waves.
Traditionally most studies of mountain waves have assumed that the atmospheric
boundary layer is sufficiently shallow that it will have only a small impact on
their generation. Its role has been largely ignored and frictionless, free-slip lower
boundary conditions have been assumed. Recently, however, the importance of
boundary-layer processes has come to light in a number of studies. From the point
of view of wave generation, modeling studies have demonstrated how the boundary
layer will generally have a damping effect, reducing the amplitude of the waves, the
strength of downslope winds and the tendency for wave breaking (e.g. Richard et al.
1989; Peng and Thompson 2003; Vosper and Brown 2007). Notably Smith et al.
(2002) concluded that the presence of a stagnant boundary layer beneath the Alpine
mountain peaks had a significant attenuating effect on mountain waves observed
during MAP. Recently the influence of the boundary layer on wave generation has
been illuminated using linear theory by Smith et al. (2006), Smith (2007) and Jiang
et al. (2006) who showed that absorption of wave energy in the boundary layer can
result in a decay of trapped lee-wave motion downstream. We shall return to this
point in Sect. 3.3.
For situations where the combination of mountain height, upstream wind and
stability are such that the non-dimensional mountain height, HO , is of order unity
or greater, the linear equations of motion no longer provide an accurate description
of the flow. In such situations the flows are nonlinear and the deceleration of the
wind on the upwind side of the mountain can be sufficient to cause stagnation and
the appearance of a region of blocked flow upstream (e.g. Smith 1980). The flow
below mountain crest may then be diverted around the flanks rather than over the
summit, sometimes separating from sides and forming a wake region downwind.
In general the mountain waves will steepen as HO increases above unity, eventually
leading to wave breaking immediately above and downwind of the mountain (e.g.
Smith 1980; Miranda and James 1992). This low-level wave breaking enhances the
lee-slope acceleration and is thought to be an important process in the development
of downslope wind storms. The turbulent dissipation in the wave breaking region
also contributes to the development of the downstream wake (e.g. Schär and Durran
1997). The wake may take the form of a long uniform region of decelerated flow
extending from the mountain (e.g. Smith et al. 1997) or consist of an attached
vortex pair (Smolarkiewicz and Rotunno 1989) which under some circumstances
b(x)
F<1 F<1 F<1
(x) F=1
F<1
hydraulic
F>1
F>1
jump
F<1
h(x) F>1 F>1
Fig. 3.3 Schematics of hydraulic flow configuration through a channel (top, left) with a sill.
Flow is from left to right. Asymmetric flow configuration common for downslope windstorms
and gap flows with subcritical flow upstream and supercritical flow downstream adjusting to the
downstream conditions in a hydraulic jump (bottom, left). Subcritical flow everywhere (top, right);
supercritical flow everywhere (bottom, right)
will be unstable, resulting in the shedding of discrete and long-lived vertically

orientated vortices (Etling 1989; Schär and Durran 1997). Damping processes such
as boundary layer friction have been shown to limit the instability of the wake, and
thus prevent such vortex shedding (Grubišić et al. 1995).
3.2.3 Hydraulic Theory
Stratified air flowing over a mountain or through a valley is analogous to the flow
of water in a channel. This analogy can be exploited for a physical explanation of
dynamically forced flow in complex terrain as well as forecasting. Here, a short
introduction to the basic concepts of hydraulic theory will be given. The theory will
be developed further in the sections on downslope windstorms (Sect. 3.4) and gap
flow (Sect. 3.5). Baines (1995) describes hydraulic theory in even greater detail.
Starting with the one-dimensional version of the shallow water equations (3.6–
3.8), we consider flow in a channel of width b(x) over some topography h(x) as
shown in Fig. 3.3. The height of the free surface of the fluid is (x). As a further
simplification of (3.6), we are only interested in well-adjusted, steady-state flow
without friction, external pressure gradient, and Coriolis force. The advection in its
fully non-linear form, however, is kept. Then the momentum equation becomes
@u @ @h
u Cg D g (3.23)
@x @x @x
where the pressure gradient term has been split into a part resulting from the change
of the height of the free surface and a part due to changes in terrain elevation. The
continuity equation requires the volume flux q through the cross-sectional area A to
be constant: @(A u)/@x D 0. If we take a channel with vertical sidewalls, the volume
flux q D u b. The continuity equation becomes
@u @ u @b
Cu D (3.24)
@x @x b @x
Note that when the free surface is an interface between two layers of comparable
densities, reduced gravity g0 D g (1 2 )/1 has to be used instead of g. The two
equations can be written as a quasi-linear differential equation relating derivatives of
the dependent flow variables (u, ) to the derivatives of the independent topographic
variables (h, b) (Armi 1986)

ug g 0 u h
C 'x D D‰x with C D ; DD ; 'D ; D
u 0 q b 1
(3.25)
Solutions exist if det(C) ¤ 0 or if certain regularity conditions are fulfilled in the

neighborhood of locations where det(C) D 0. These locations are called “controls”
since they establish unique flow solutions. Regularity conditions ensure that the
magnitude of the slope of the free surface remains finite.
Characteristic velocities are given by det(C
I) D 0 with I being the identity
matrix, and
p

Du˙ g (3.26)
The square-root term is the speed of (long) gravity waves at the fluid interface.
Its relation to the advective velocity u yields the Froude number
u2
F2 D (3.27)
g
The Froude number determines the characteristics of the flow: when F < 1 is
“subcritical” waves can travel up and down stream. When F > 1 is “supercritical”
waves can only travel downstream. The special situation of F D 1 is “critical” will
be discussed in more detail. In this case, det(C) D 0 and the characteristic velocity
in (3.26) vanishes.
The differential solutions to (3.25) are

1 @u 1 1 @b 1 1 @h
D C
u @x 1F 2 b @x 1F 2 @x
2
1 @ F 1 @b 1 1 @h
D (3.28)
@x 1F 2 b @x 1F 2 @x
The terms in square brackets are termed “influence coefficients”. Two special
cases occur for @b/@x D 0 and @h/@x D 0. For the special case of @b/@x D 0 (channel
of constant width or at the narrowest part), the change of height of the free surface
is directly but with opposite sign related to the change of elevation:

@ 1 @h
D (3.29)
@x 1F 2 @x
In the limit of F D 0, as the terrain rises the top of the fluid has to come down
equally. When the flow is subcritical (F < 1) everywhere, the top of the fluid still
changes in opposite direction to the terrain. However, it descends further than the
terrain rises on the upstream side and rises further than the terrain drops on the
downstream side (Fig. 3.3, top, right). The fluid depth is shallowest at the crest.
When the flow is supercritical (F > 1) everywhere, terrain and the height of the
interface move in tandem but the excursion of the interface height is less than the
excursion of the terrain (Fig. 3.3, bottom, right).
The most common situation for downslope windstorms, however, is an asym-
metric descent of the interface from upstream across the crest to some distance
downstream (Fig. 3.3, bottom, left). For that to happen, the influence coefficient
(i.e. the term in square parentheses) has to change sign as the flow crosses from
the upstream to the downstream side. A location has to exist where F2 D 1. For a
solution to exist, this location has to be at the crest where @h/@x D 0. Conditions of
asymmetry and regularity render the crest a so-called “control”. We are no longer
free to set an arbitrary combination of layer depth and volume flux rate q. If one is
set, e.g. the layer depth, the flow rate follows. And, since the interface has to descend
on both sides in the asymmetric case, the flow has to be subcritical (slow and thick)
upstream and supercritical (fast and shallow) downstream. The height descends with
subcritical flow upstream and supercritical flow downstream. At the control, i.e. the
crest, the Bernoulli equation (conservation of energy) is ½ u2 C g c D gH0 , where
H0 is the far-upstream depth of the fluid. Combining the Bernoulli equation with
(3.27) for F2 D 1, results in c D 2/3 H0 . At the control, the fluid depth has already
decreased by one third of its value far upstream.
Note that for the more general case when @b/@x ¤ 0 at @h/@x D 0, the control need
not be at the crest but where
1 @ 1 @h
C D0 (3.30)
b @x @x
This regularity condition includes one of the dependent variables, the height of
the interface .
Further downstream the flow will eventually return from a supercritical to a
subcritical state, slow down and thicken again. The turbulent transition zone is a
hydraulic jump, in which momentum is conserved but kinetic energy is dissipated.
For the asymmetric case of interest for gap flow and downslope windstorms, the
depth of the reservoirs on either side of the topographical constriction/crest will
differ with the larger depth on the upstream side.
Using a single layer of constant density is the most frequent application of

hydraulic theory to the atmosphere. However, hydraulic theory can be formulated
for two layers (Armi 1986; Armi and Riemenschneider 2008) or more including
continuous stratification. For continuous stratification, an analytical solution exists,
which is described in Sect. 3.5 on gap flow.
3.3 Lee Waves and Rotors
A special kind of non-hydrostatic wave motion occurs when the variation of the
Scorer parameter,
2 , with height is such that vertically propagating wave motion
is partially reflected back down towards the ground. In the typical case where
2 > h 2 in the lower and mid-troposphere (due to relatively low wind speeds or
high stability, for example) and
2 < h 2 in the upper troposphere, the wave energy
will become trapped within the lower layer and ducted horizontally downstream.
Such trapping is generally only partial, due to the fact that
2 h 2 will often
be positive in the lower stratosphere and wave motion is able to “tunnel” through
layers of
2 < h 2 . In cases where the trapping is strong the waves may be ducted
horizontally for hundreds of kilometers downstream. Such waves are commonly
referred to as trapped lee waves and they are responsible for the long regular trains
of clouds which form in the wave crests downwind of mountain ranges and which
are readily seen in satellite imagery. A schematic diagram, illustrating the structure
of the trapped lee wave is provided in Fig. 3.4. Examples of satellite images of
lee-wave cloud formations are shown later in Figs. 3.6 and 3.9.
Typical horizontal wavelengths of trapped lee waves observed in the troposphere
lie in the range 5–30 km. Increased stability at low levels (for example due to
an inversion layer at the boundary layer top) or reduced low-level wind speeds
will favor the existence of shorter horizontal wavelengths, while reduced stability
and increased winds will increase the wavelength. As we shall see in Sect. 3.3.2,
according to linear theory at least, the horizontal wavelength of trapped lee waves
is a function of the upwind profile of wind and stability rather than the mountain
shape.
Mountain waves are often associated with turbulence, and one particularly severe
phenomenon which causes turbulence at low-levels is the mountain-wave rotor.
This term is used to refer to a region of recirculating airflow with horizontal
vorticity, orientated parallel to the ridge crest. Rotors are typically accompanied by
strong near-surface downslope winds which decelerate rapidly at the rotor’s leading
upwind edge, where boundary-layer flow separation takes place and air parcels
are lifted away from the surface. Downwind of the separation point, the mean
near-surface winds are generally slack or weakly reversed below the rotor, but are
usually highly turbulent, exhibiting a high degree of temporal and spatial variability.
Rotors are recognized as posing a significant hazard to aviation and have been
cited as contributing to numerous accidents and damage to aircraft. Additionally
2
kH
Height
Trapping layer
rgy
Wa
ene
ve
ve
ene
Wa
gyr
l2
Separation point Lee−wave (Type I) rotor
Fig. 3.4 Schematic diagram illustrating the structure of the trapped lee wave (flow is from left to
right). The Scorer parameter,
2 , profile is such that a trapping layer exists in which
2 h 2 > 0.
Upward propagating wave energy is reflected back down towards the ground at the top of this
layer, resulting in net downwind horizontal propagation of wave energy. In the situation shown
here the waves are of sufficient amplitude that flow separation occurs beneath the wave crests and
a turbulent (type I) rotor forms
rotor circulations may also have an important impact on the transport of pollutants
and aerosols in mountainous terrain, serving as a mechanism for transport from the
boundary layer into the free troposphere.
Based on observations and recent modeling (Hertenstein and Kuettner 2005),
rotors have been classified into two types: the rotors which form under the crests
of trapped lee waves (type I) and the jump-like rotor (type II), in which the wave
motion is replaced by a deep turbulent layer downwind of the mountain. The latter
is thought to contain the more severe turbulence, but is probably relatively rare. The
distinction between the type I and II rotor is clear, although as discussed later the
definition of the type II rotor itself is rather loose, and covers a variety of different
flow types including low-level wave breaking and hydraulic jumps.
Remarkable examples of rotor clouds and dust transport are provided by Fig. 3.5
which shows instances of rotor formation in the Owens Valley, to the east of the
Sierra Nevada mountain range in California, USA. The deep turbulent region and
lack of smooth wave clouds seen in Fig. 3.5a and b seems to fit the description of the
type II rotor. Figure 3.5c provides a clear example of rotor clouds associated with
a type I rotor beneath a train of lee waves which formed downwind of the Rocky
Mountains near Boulder, Colorado.
Fig. 3.5 Rotor clouds and

blowing dust over the Owens
Valley during mountain wave
events in the lee of the Sierra
Nevada on (a) 5 March 1950
during the Sierra Wave
Program (SWP),
(photographed by Robert
Symons) and (b) 25 March
2006 during IOP-6 of the
Terrain induced Rotor
Experiment (T-REX)
(photographed by Barbara
Brooks from the FAAM
BAe-146 aircraft). In both
cases the flow is westerly
(from right to left) and the
Sierra Nevada Mountains are
on the right-hand side of the
photographs. (c) Rotor clouds
on 2 December 2003 in a
train of trapped lee waves in
the lee of the Rocky
Mountains, near Boulder,
Colorado (photographed by
Rolf Hertenstein) (From
Hertenstein and Kuettner
2005, reproduced by
permission of Tellus)
While mountain waves and trapped lee waves have been the subject of much
attention from the atmospheric science community over the last few decades, rotors
have received relatively little attention. However, improvements in observational
techniques and increases in computing power have prompted a surge in interest in
the rotor problem and there have been several recent field campaigns and numerical
modeling studies aimed at understanding the formation and characterizing the
properties of rotors.
a b
8
5
z/km
0
−4 −3 −2 −1 0 1 2 3 4
Vertical velocity /ms−1
Fig. 3.6 Measurements of lee-wave activity over Scotland on 11 March 2000 made by (a)
radiosonde measurements at 1012 UTC and (b) high-resolution visible satellite imagery at 1507
UTC. Plot (a) shows fluctuations in the ascent rate of a radiosonde launched from the Isle of
Arran after the mean ascent rate has been removed. Image (b) shows lee-wave cloud over much of
Scotland. The coastline and Isle of Arran are also shown (Figures are adapted from Vosper 2003.
Reproduced by permission of the Royal Meteorological Society)
3.3.1 Observations
3.3.1.1 Lee Waves
As discussed previously, lee-wave cloud patterns are frequently observed in high-

resolution satellite imagery. Their vertical motion is also often readily seen in
fluctuations in the ascent rates of radiosondes. Figure 3.6a shows an example of
a moderate amplitude lee wave as observed by a radiosonde launched on the Isle of
Arran, in south-west Scotland, UK. Fluctuations of amplitude around 2 m s1 are
evident in the ascent rate of the balloon between the ground and 6 km. As shown
by visible satellite imagery in Fig. 3.6b, the waves observed by the radiosonde form
part of an extensive train of lee waves in a north-westerly airstream. The origin of the
waves is clearly not the mountains and hills of Arran itself (though they undoubtedly
contribute towards the wave activity) since the wave motion extends over much of
Scotland and far upwind.
Research aircraft are commonly used to make measurements of lee waves
during field observation programs. Over the past few decades there have been
numerous observational research campaigns in which lee waves have been measured
by research aircraft above and downwind of mountain ranges. Recent examples
include the measurements made during PYREX, MAP and, most recently, T-REX.
Aircraft observations are particularly powerful when conducted with multiple
Fig. 3.7 Vertical velocity (m s1 ) measured by three research aircraft over the Sierra Nevada
Mountains during a coordinated T-REX mission on 25 March 2006 (IOP-6). Measurements made
by the NCAR HIAPER are shown in blue, UK FAAM BAe146 measurements are shown in red
and University of Wyoming King Air (UWKA) measurements are shown in black. The UWKA
data above 7 km and just below 6 km share vertical velocity scales with the neighboring BAe146
legs. Data from the three aircraft were obtained over a period of just under 5 h. The HIAPER and
BAe146 tracks, which were almost perfectly aligned, were offset 3–4 km to the north from the
shown UWKA tracks (From Grubišić et al. 2008. © American Meteorological Society. Reprinted
with permission)
aircraft, coordinated in such a way to provide a comprehensive picture of the wave

field. Figure 3.7 provides such an example, with lee-wave measurements made
during T-REX over the Sierra Nevada Mountains. The measurements were made
using three research aircraft flying simultaneously at different altitudes. The utility
of aircraft measurements is enhanced further through the use of remote sensing
instruments, such as backscatter lidar, which can detect the presence of clouds. This
technique was used to good effect during MAP to obtain a detailed picture of the
wave and cloud field over the Alps (see Fig. 3.8).
Fig. 3.8 Vertical displacements (solid curves) inferred from aircraft vertical velocity measure-
ments and airborne lidar backscatter (colour, dBZ) on a north–south cross section over the Alps
on 20 September 1999 during MAP. The airflow is from the south (left-hand side) (From Smith
et al. 2007 reproduced from Doyle and Smith (2003). Reproduced by permission of the Royal
Meteorological Society)
3.3.1.2 Rotors
Compared with the waves themselves, observations of rotors are relatively scarce.
One of the earliest observations of rotors was that by Mohorovičić (1889), who
observed rotor clouds during bora events along the northern Adriatic coast in
Croatia. Grubišić and Orlić (2007) provide an interesting account of Mohorovičić’s
(largely forgotten) work, along with observations of rotor clouds elsewhere which
were published around the same time. Some of the best observations have come
from the pioneering work of the Sierra Wave Project (SWP) and the Jet Stream
Project (JSP) in the 1950s (Holmboe and Klieforth 1957; Grubišić and Lewis 2004)
of mountain waves and rotors over the Sierra Nevada. Rotors were also observed
during the Colorado Lee Wave Program in the late 1960s and early 1970s (Lester
and Fingerhut 1974). Following this, there have been few organized campaigns
targeted at observing rotors. Recently, however, rotors have again become the focus
of observational programs. Mobbs et al. (2005) have obtained measurements of
the near-surface flow downwind of the Wickham range on the Falkland Islands
during episodes of rotor activity. Similar ground based measurements of lee-wave
induced flow separation were made by Sheridan et al. (2007) downwind of the
Pennines in the UK and Darby and Poulos (2006) observed rotor evolution in
the Rocky Mountains using several measurement techniques, including research
aircraft, Doppler lidar and wind profiler systems. Very recently, attention has
returned to the Sierra Nevada Mountains with a two-part observational campaign
focused on the rotor formation in the Owens Valley. The first, pilot-phase of
the campaign, the Sierra Rotors Project (SRP), took place during the spring
of 2004 and consisted primarily of ground based measurements in the Owens
Valley (Grubišić and Billings 2007). The second phase, the Terrain induced Rotor
Experiment (T-REX) took place during spring 2006 and involved a comprehensive
set of observing systems, including research aircraft and a wide range of ground
based and remote sensing instrumentation. T-REX has provided detailed systematic
measurements of mountain-wave rotors and the datasets will shed new light on the
dynamics of rotors.
For obvious safety and logistical reasons, in-situ aircraft measurements of rotors
are relatively rare, the exceptions being those made during the early SRP and JSP
and very recent T-REX campaigns. Most observations, by necessity, have been made
by ground-based instrumentation or use remote sensing instruments on airborne
platforms. One technique deployed in recent field campaigns (e.g. Mobbs et al.
2005; Sheridan et al. 2007; Grubišić and Billings 2007) has involved the deployment
of large numbers of automatic weather stations (AWSs) over a small area to provide
a well resolved picture of the surface “footprint” of rotor motion aloft. Figure 3.9
shows an example of such measurements obtained in the lee of the Pennine hills in
northern England, UK. The measurements show a region of relatively low surface
pressure, within which the near-surface winds undergo considerable acceleration.
Downwind of this region the flow is decelerated by an adverse (positive) pressure
gradient. A region of high surface pressure sits downwind of the low pressure and
the flow within this region is very slack, and at times is reversed. As we shall see
in Sect. 3.2, this type of pattern is consistent with the theoretical ideas of how lee
waves influence the near-surface flow.
Clearly AWS measurements offer some interesting insights into the impacts of
lee waves and rotors on the surface flow, but remote sensing techniques provide
the best chance of observing the vertical and three-dimensional flow structure.
Figure 3.10 shows Doppler lidar observations of the velocity field made during the
T-REX campaign. The lidar was situated in the Owens Valley and the measurements
shown are for IOP-6 which included periods of large amplitude mountain wave
Fig. 3.9 Lee-wave cloud over the UK and near-surface flow observations beneath the lee waves
in the Vale of York, northern England. Frame (a) shows high resolution visible satellite imagery
on 17 March 2005 at 1000 UTC. Frame (b) shows a snapshot of the 10 min average 2 m wind
vectors measured by AWSs in the Vale of York (in the lee of the Pennines hills) on the same day
at 2010 UTC. Colour contours indicate the surface pressure perturbations (units hPa) at each site
in the AWS array. Terrain contours are depicted by solid lines with an interval of 50 m. The inset
in (b) shows the flow measurements upwind of the Pennines. The boxed area in (a) indicates the
location of the northern Pennines and the Vale of York (Figures are adapted from Sheridan et al.
2007. Reproduced by permission of the Royal Meteorological Society)
and rotor activity. For the times shown in Fig. 3.10 strong westerly and south-
westerly winds were present on the west side of the valley (west of the lidar site)
with relatively weak northerly along-valley winds to the east. The scans show clear
evidence for a rapid reversal of the westerly flow approximately 1–2 km to the
west of the lidar, indicating the existence of a recirculating rotor within the valley.
The lidar scans showed a high degree of temporal variability at this time, indicating
that the flow within the rotor was highly turbulent and that the structure of the rotor
was very transient.
The few existing in-situ measurements made by aircraft in rotors suggest that the
presence of extreme turbulence is most likely connected with vortices whose scale is
perhaps only a few hundred meters, and therefore much smaller than that of the main
rotor. For example, several vertical gusts exceeding 10 m s1 (and as large as 20 m
s1 ) were experienced in a 50 s period during a JSP research flight when an aircraft
penetrated a rotor (Holmboe and Klieforth 1957; Doyle and Durran 2007). Lidar
measurements obtained near Boulder, Colorado (Banta et al. 1990) also suggest the
presence of small scale vortices within the main rotor recirculation. Similarly, the
recent T-REX lidar observations support the existence of these small-scale intense
subrotors.
Fig. 3.10 Lidar scans of the low-level flow in the Owens Valley, east of the Sierra Nevada
mountains, during an episode of large amplitude mountain wave and rotor activity. The data
shown are from IOP-6 of the T-REX campaign. Frame (a) shows a PPI scan of Doppler lidar
radial velocity measurements obtained in the late afternoon on 25 March 2006. Green and blue
colors indicate flow toward the lidar while yellow and red indicate flow away from the lidar. The
dashed line indicates the orientation of the RHI scan shown in (b). Gray lines in (a) indicate terrain
contours, which are plotted every 200 m. The RHI scan was obtained about 4 min after the PPI
scan (From Grubišić et al. 2008. © American Meteorological Society. Reprinted with permission)
3.3.2 Theory
3.3.2.1 Lee Waves
As discussed earlier, we can expect trapped lee waves to form when the Scorer
parameter decreases with height in a suitable manner. The form of
2 .z/ determines
both the wavelength and the wave amplitude variation with height. At this stage it
is helpful to consider a very simple problem in order to illustrate the theory for lee
waves in more complex and realistic situations. Consider the case of a two-layer
stratified flow in which the stability takes constant (but different) values, Nl and Nu
in lower and upper layers, respectively. We shall assume that the wind speed, UN , is
constant across both layers and for convenience define the height of the interface
between the two layers to be at z D 0, with the ground located at z D d. We shall
also make the Boussinesq approximation, thus eliminating the density terms in
(3.21). We seek wave-like solutions to (3.21) of the form
O l D Al e iml z C Bl e iml z
w (3.31)
and
wO u D Au e imu z ; (3.32)
where wO l and wO u are the complex vertical velocity amplitudes in the lower and
upper layers, respectively, ml 2 D Nl 2 =UN 2 k 2 and mu 2 D Nu 2 =UN 2 k 2 are
the corresponding vertical wavenumbers (squared) and Al , Bl and Au are complex
constants. Note that (3.31) expresses the fact that we expect both upward and
downward propagation of wave energy in the lower layer, the latter being caused
by partial reflection at the layer interface, whereas (3.32) assumes that wave energy
in the upper layer will radiate freely in the vertical. We now impose matching
conditions at the interface. Firstly we require that an air parcel on the interface must
move with the vertical velocity of the interface. If the interface has shape z D &.x/
then this is described by
D&
D w on z D &: (3.33)
Dt
Letting & D &e O ikx and linearizing (3.33) then gives the matching condition
_
ikUN D w on z D 0 and this must be satisfied by air parcels both immediately above
_
and below the interface. Since the mean wind, UN , is the same in both layers this
becomes wO l D wO u at z D 0. The second matching condition applied at the interface
states simply that the pressure on both sides is the same, namely, POl D POu on z D 0.
This implies that
d wO l d wO u
D on z D 0: (3.34)
dz dz
Substituting these conditions into (3.31) and (3.32) and manipulating the result-
ing expressions for Al , Bl and Au then gives
2Al
wO l D Œml cos.ml z/ C i mu sin.ml z/ (3.35)
ml C mu
and
2Al ml imu z
wO u D e : (3.36)
ml C mu
Fig. 3.11 The theoretical 12

resonant lee-wave horizontal
wavelength plotted as a d=3 km
function of wind speed U for d=4 km
the two-layer stability case 10 d=5 km
with Nu D 0.01 and
Nl D 0.02 s1 . Data are
plotted for lower layer depths,
Wavelength (km)
d, of 3, 4 and 5 km 8
2
10 12 14 16 18 20
U (ms-1)
The constant Al is obtained from a free-slip lower boundary condition which

simply states that air parcels at ground level follow the mountain profile, z D h(x).
Thus, on the mountain surface we have w D Dh/ Dt. Linearizing and writing h.x/ D
_
O ikx then gives w
he
_
N
l D i U kh on z D d. Substituting this into (3.35) then gives
_ m cos.m z/ C im sin.m z/
O l D UN ikh
l l u l
w : (3.37)
ml cos.ml d / imu sin.ml d /
as the solution in the lower layer. Equation 3.37 contains singularities whenever ml
and mu are such as that the denominator is zero. This occurs when
tan.ml d / D i ml =mu (3.38)
and corresponds to the existence of resonant wavemodes. These resonant waves are
the naturally occurring trapped lee waves and as such require only an infinitesimal
amount of forcing by the orography. Their horizontal wavelength is determined
by the implicit relationship described by (3.38). Note that the occurrence of the
imaginary number, i, on the right-hand side of (3.38) means that solutions are only
possible when mu itself is imaginary, and hence the wave amplitude will decay
with height in the upper layer, consistent with the fact that the wave response is
trapped in the lower layer. Examples of solutions to (3.38) are presented in Fig. 3.11
which shows how the resonant horizontal wavelength varies with wind speed for a
case where the lower and upper-layer stabilities are Nl D 0.02 and Nu D 0.01 s1 ,
respectively.
According to (3.38) the theoretical horizontal wavelength of the resonant trapped

waves is purely a function of the background profile rather than the shape of the
orography. This is also true more generally, for any profile of wind and stability.
In general theoretical predictions for the resonant wavelength are only available
for a small class of idealized profiles such as the two-layer case considered above.
Vosper (2004) performed a similar analysis to that presented here and showed how
linear theory can be used to describe the behaviour of trapped lee waves which
form in a temperature inversion layer. More generally, however, for real profiles of
wind and stability, such resonant solutions cannot be determined analytically and
numerical approaches must be used. By discretizing (3.21) the resonant wavemodes
can be sought and the linear trapped lee wave solutions can then be computed
(e.g. Sawyer 1960; Vergeiner 1971). For three-dimensional flows the resonant
wavemodes will in general lie along lines in (k,l) space (Sawyer 1962) and the wave
field will be made up of contributions from wavemodes along such lines. Note that
although it is generally accepted that linear theory can provide a reliable prediction
of the horizontal wavelength of trapped lee waves (at least in cases where the non-
dimensional mountain height, HO 1) the same is not always true of the wave
amplitude. Although linear theory has been used widely for lee-wave prediction (and
as discussed in Sect. 3.3 certainly can prove useful from a forecasting perspective)
several studies (e.g. Smith 1976; Durran and Klemp 1982; Durran 1992; Vosper
2004) have shown how broad mountains, whose horizontal scale exceeds the lee-
wave wavelength, will provide little forcing at the resonant wavenumbers. The
amplitude of the waves will then be determined by nonlinear processes. For HO > 1
nonlinearity can also have an important influence on the horizontal wavelength,
when for instance, the mountain range consists of two (or more) ridges, separated by
a valley. Grubišić and Stiperski (2009) have shown how, in this case, when the two
ridges are of similar height, the ridge separation has a strong controlling influence
on the horizontal wavelength.
In common with most previous studies, the above discussion of trapped lee
wave behaviour makes no attempt to account for the influence of the atmospheric
boundary layer. As well as exhibiting a rich variety of phenomena itself in complex
terrain (see Chaps. 2 and 5), recent work has shown that the boundary layer can have
a significant impact on both the generation and then the attenuation of lee waves.
Notably Smith et al. (2002), motivated by observations of lee waves during MAP,
developed a simple theory for the absorption of wave energy by a stagnant layer
which filled the valleys between individual mountain peaks. Whereas a solid lower
boundary reflects downward propagating rays back upwards, the theory allows
for partial absorption of wave energy by the stagnant layer, which results in the
downwind decay of the lee wave motion. The process is illustrated by the schematic
diagram in Fig. 3.12. More recently, studies (Smith et al. 2006; Jiang et al. 2006)
have examined how the boundary layer acts to attenuate the trapped lee wave
motion and suggest a faster downwind decay in shallow nocturnal boundary layer
conditions, with slower decay when the boundary layer is deep and convective.
The boundary layer also plays an important role in mountain wave induced flow
separation, and thus has a crucial part to play in the dynamics of rotors.
Fig. 3.12 A schematic of the propagation of wave energy for trapped lee waves when a stagnant
layer of air fills the valleys between mountain peaks. The triangles represent the lower mountains,
with the lee waves generated by the higher peak which penetrates into the ambient flow. The
dashed line labelled zref marks the top of the stagnant layer. The upper dashed line represents
the level of wave reflection due to a decreasing Scorer parameter. The slanting wave rays are
parallel to the local group velocity and indicate the direction of energy propagation. The down-
going wave amplitude, B, is reduced by partial absorption at zref (Figure taken from Smith et al.
2007. Reproduced by permission of the Royal Meteorological Society)
3.3.2.2 Rotors
The theoretical and numerical challenges posed by mountain wave rotors have
meant that, until recently, relatively little progress has been made toward under-
standing their dynamics or the conditions in which they form. Nevertheless, the
recent surge in interest in this area has resulted in some significant progress. By
necessity, most of the recent advances in understanding the dynamics of rotors
have come from numerical simulation rather than analytical theory, although some
limited progress towards understanding the process of flow separation, with which
rotors are intimately linked, has emerged from analytical treatments. The focus, to
date, has tended to be on the type I, lee-wave rotor.
The formation of rotors is clearly connected to the occurrence of boundary-layer
separation and the relationship between separation and the mountain-wave motion
aloft. Although theories have existed for several decades as to how the pressure
gradient due to lee waves might induce flow separation (e.g. Lyra 1943) these ideas
have only recently been tested using numerical simulations. Doyle and Durran
(2002) have shown, through numerical simulations of two-dimensional trapped
lee waves, how the adverse (positive) pressure gradient which occurs beneath the
leading upwind edge of a wave crest will act to decelerate the flow. Once this
gradient is sufficiently large the flow will separate away from the surface and a
(type I) rotor will form. Doyle and Durran (2002) and later Vosper (2004) have
shown how surface friction is vitally important to obtaining realistic simulations of
rotors. Its effect is to both reduce the wind speed towards zero at the surface, thereby
enhancing the influence of the pressure gradient, thus promoting separation, and to
generate a sheet of high horizontal vorticity along the lee slope of the mountain,
Fig. 3.13 Streamlines and horizontal vorticity (units s1 ) for a simulation of lee-wave (type 1)
rotors in flow over a two-dimensional ridge. Horizontal vorticities greater than 0.03 s1 are shaded
in color. Horizontal wind speeds less than or equal to zero are shown using blue isotachs (every
2 m s1 ) (From Doyle and Durran 2002. © American Meteorological Society. Reprinted with
permission)
which is lifted aloft by the waves when the flow separates. This is illustrated by
Fig. 3.13 which shows a two-dimensional simulation of a type I rotor (Doyle and
Durran 2002).
The occurrence of flow separation and its relationship with the trapped lee-wave
amplitude has been studied in further detail by Vosper et al. (2006) and Jiang et al.
(2007). In the former study a theory was developed which suggests that, for two-
dimensional waves, a critical lee-wave pressure amplitude exists, beyond which
flow separation will occur under the wave crests. This critical amplitude can be
expressed non-dimensionally as p=.u2 /, where p is the amplitude of the
pressure fluctuations exerted by the waves on the ground (assumed constant through
the boundary layer) and u is the surface friction velocity of the background flow.
Vosper et al. have shown how this critical non-dimensional amplitude increases as
a function of surface roughness and this can be explained by the fact that, for a
smoother surface, the surface stress will be reduced and the near-surface winds will
be stronger. A larger pressure gradient is therefore required to reverse the flow. Note
that this result is contrary to what occurs for a laminar flow where separation occurs
more readily for a smooth surface. Jiang et al. (2007) have also demonstrated the
link between boundary-layer separation and trapped lee wave amplitude and found
that separation depended largely on the ratio of the vertical velocity maximum to
the surface wind speed. Separation was far less sensitive to other factors such as the
Fig. 3.14 The along ridge component of horizontal vorticity (colour shading, interval 0.02 s1 )
and flow vectors in a cross section through a type I rotor. The results shown are for a three-
dimensional simulation of flow over an infinitely long two-dimensional ridge. The vortex sheet is
lifted away from the surface and then broken up by shear instability, resulting in intense subrotors
which move along the interface between the main rotor and the wave field (From Fig. 7(c) of Doyle
and Durran 2007. © American Meteorological Society. Reprinted with permission)
mountain height and slope. Jiang et al. also found that the onset of separation was
sensitive to the surface sensible heat flux. A positive heat flux tends to inhibit the
occurrence of separation by increasing vertical turbulent mixing and thus enhancing
the surface wind speed. A negative (downward) heat flux appears to have the
opposite effect.
By performing very fine-resolution numerical simulations, Doyle and Durran
(2007) have begun to illuminate the internal dynamics of trapped lee-wave (type I)
rotors. Their simulations have shown how, once lifted away from the surface by flow
separation, shear instability along the upstream leading edge of the rotor can break
up the sheet of high horizontal vorticity into small intense vortices, or subrotors.
Doyle and Durran’s simulations illustrate the importance of three-dimensional
effects in the dynamics of these subrotors. Even for waves generated by a strictly
two-dimensional ridge, the tilting and in particular the stretching of vorticity which
is present in a three-dimensional turbulent flow, contributes to the vorticity of the
subrotors, meaning they can contain horizontal vorticity which far exceeds that of
the main rotor. The subrotors are relatively long lived, being swept downstream
along the interface between the main rotor and the lee wave and it seems possible
that their existence is connected to pulsations in the wind that have been observed
during downslope wind events (see Sect. 3.4.1). An example of this behaviour is
shown in Fig. 3.14 (Doyle and Durran 2007). In the less realistic situation where
the turbulence is restricted to be purely two-dimensional the subrotors are much
less intense and are short-lived, being eventually entrained into the main rotor
recirculation. Recently, new studies using eddy-resolving simulations have been
used to explore the role of turbulence in mountain-wave breaking (Epifanio and
Qian 2008) and interactions between the turbulent boundary layer and lee-wave
rotors (Smith and Skyllingstad 2009). This approach is clearly a powerful one and
the ability to explicitly resolve turbulent eddy motion will perhaps allow a full
understanding of the internal dynamics of rotors.
Compared with the type I rotors, the more severe jump-like rotors have received
much less attention. Vosper (2004), in a study of temperature inversion effects on
lee waves, found that a hydraulic-jump-like response occurred in place of trapped
lee waves if the potential temperature difference, , across the inversion was
sufficiently large. The results from a set of two-dimensional simulations were
used to construct a flow regime diagram which contained, in order of increasing
, trapped lee waves on the inversion layer, lee waves which were sufficiently
strong to induce flow separation and form (type I) rotors and, finally, regions of
overturning and severe turbulence in the lee of the mountain. These latter conditions
were somewhat loosely referred to as a “hydraulic jump” state. Hertenstein and
Kuettner (2005) also performed two-dimensional simulations for flow with a strong
temperature inversion and found that the solutions were sensitive to vertical wind
shear across the inversion in the upstream flow. For sufficiently strong forward
shear trapped lee waves formed with attendant type I rotors. Reduction in the shear,
however, resulted in a downstream response more akin to a hydraulic jump (see
Fig. 3.15). The relatively smooth lee-wave field is replaced by a deep turbulent layer
which extends far downwind. Hertenstein and Kuettner explained the sensitivity to
the wind shear in terms of the horizontal vorticity budget. The type of rotor that
forms depends on the dominant sign of the horizontal vorticity as the flow separates
from the surface. This balance is modified by the presence of vertical wind shear
across the upstream inversion.
Recently Jiang et al. (2007) have also examined the rotor types which can occur
in the presence of an elevated temperature inversion. They found that a variety of
different flow regimes could occur: (1) a steady undular jump regime, in which
a hydraulic jump forms in the lee of the ridge; (2) a regime in which trapped
lee-waves ride along the inversion layer behind the jump; (3) a propagating jump
regime in which the jump propagates away downwind and (4) a mixing jump regime
in which low-level wave breaking merges with a hydraulic jump below. These
flow regimes all involve the generation of turbulence and are capable of causing
boundary-layer separation, resulting in reversal in the lee of the mountain. In the
sense that these flows are all clearly different to that of the type I, trapped lee-wave
rotor, it is tempting to label the rotors associated with them all as type II, despite their
differences.
3.3.3 Models and Forecasting
The implications for aviation safety and risks to ground transport and property mean
that there is a clear need to forecast lee-wave motion and rotor activity. Glider pilots
can also benefit from wave forecasts since they can use the associated updrafts to
stay aloft for long periods of time and to achieve high altitudes and great distances.
Fig. 3.15 Potential temperature (contour interval 0.5 K) downwind of a two-dimensional ridge in
a type II rotor simulation. The C signs mark the positions of three propagating eddies along the
top of the rotor (From Fig. 7(b) of Hertenstein and Kuettner 2005. Reproduced by permission of
Tellus)
The typical horizontal wavelengths of trapped lee waves mean that they are
generally poorly resolved by numerical weather prediction (NWP) models. Global
forecast models, whose horizontal resolutions are typically 25–50 km cannot hope
to explicitly represent lee wave motion at all. The same is true for most current
operational limited area models, whose resolutions are typically around 10 km. It is
only with grid spacings of 1–2 km that we can really expect to properly resolve
the waves. As for the prospects of explicitly resolving rotors, whose scales are
considerably smaller than those of the lee waves themselves, this is likely to remain
beyond the reach of operational models for some time.
The above limitations of NWP models mean that other techniques must be
used for forecasting lee-waves. Many of the methods used are based on obtaining
approximate solutions to (3.21) with profiles of wind and stability taken from
either NWP model or radiosonde data. Although it is possible to obtain numerical
solutions to (3.21) which utilize the full available resolution of the forecast or
radiosonde profiles, in the past the lack of computing resources available to bench
forecasters has necessitated far simpler approaches. For example, the widely used
Casswell (1966) calculation involves the assumption that
2 decays exponentially
with height and an estimate for this vertical decay scale is obtained using profile
data at only three heights. Because it utilizes such a small amount of information
about the atmospheric state, and incorporates no detail of the actual orography, the
Casswell method can at best provide only a very crude forecast of lee-wave activity.
Nowadays numerical solutions to (3.21) can be obtained (e.g. Lane et al. 2000;
Vosper and Mobbs 1996) with, by modern standards, only very modest computing
resources. Predictions can be made of the wavelength and vertical structure of
the resonant lee waves with very fine vertical resolution (e.g. tens of metres). An
example of such an approach, adopted for use in opertational forecasting is given by
Shutts (1997). A set of solutions to (3.21) is obtained over a range of horizontal
wavevector directions (centered on the low-level wind direction) and then one
particular solution is chosen. The actual vertical velocity associated with the chosen
wavemode can only be calculated if the spectral composition of the underlying
orography is known. Since the Fourier spectrum of real orography is generally
very irregular and for practical purposes can only be computed for a discrete set
of wavelengths, Shutts instead suggests assuming a power-law dependence (on the
horizontal wavenumber) for the vertical velocity at some height near the ground
(the wave “launching height”). A vertical profile of horizontally averaged vertical
velocity can then be estimated by applying suitable scaling to the solutions to (3.21).
Shutts (1997) has demonstrated the success of the technique by comparing forecasts
with observations.
Although the Shutts method can provide the forecaster with a prediction of wave
amplitude and horizontal wavelength, it does not provide a detailed picture of the
lee-wave motion. In terms of assessing the likelihood of rotors, since (3.21) is
based on inviscid theory which takes no account of boundary-layer processes, it
cannot be used to predict how the lee waves will affect the near-surface winds.
Detailed predictions such as these can only be made using a more complex three-
dimensional numerical model. While the computational requirements of using a
complex NWP model are, in practice, too restrictive, it is possible to obtain detailed
three-dimensional solutions using simplified models. One such example is provided
by Vosper (2003) who constructed a numerical finite-difference model to solve the
linearized equations of motion, the Boussinesq version of (3.9)–(3.11). The model
provides detailed predictions of the wave field over the real complex orography
during steady upwind conditions and the simplifications to the equations of motion
permit these solutions to be obtained with only a relatively modest computer, such
as a desktop PC. The neglect of nonlinear terms and moist and turbulent physical
processes undoubtedly compromises the accuracy of the predictions, however
Vosper (2003) demonstrated that for the Isle of Arran, which contains mountain
peaks as high as almost 900 m, the model provides accurate predictions of lee waves.
Comparisons with radiosonde observations show that, with sufficiently accurate
upwind profile data, both the amplitude and phase of the waves can be accurately
predicted.
Recent developments to the above linearized numerical model include the
incorporation of a linear turbulent boundary layer scheme (King et al. 2004) which
allows the representation of the influence of waves on the near-surface flow. The
model, 3DVOM (3-D Velocities Over Mountains), is currently used operationally
at the UK Met Office to produce 1 km-horizontal resolution predictions of the
lee-wave field every 6-h for separate hilly regions of the UK and Ireland, as well
as the Falkland Islands. An example of such a forecast for the Pennines in northern
England, validated against observations made by a research aircraft, is shown in
Fig. 3.16. The model provides a good quantitative description of the wave field,
both in terms of wave amplitude and phase. Also shown are the predictions made
by a full NWP model (the Met Office Unified Model) run in a nested configuration
at 1 km resolution. The results are similar to those of the far simpler and cheaper
3DVOM calculation.
For larger mountains, where the linearized equations are no longer valid, we
would not expect a linear model to provide a quantitatively accurate forecast,
and a fully nonlinear model is required. Such an approach has been adopted as
a research activity in support of various recent field campaigns. For instance, the
NRL COAMPS model was used routinely to provide guidance for mission planning
during T-REX. Wave forecasts for the Sierra Nevada and Owens Valley area were
provided throughout the campaign on a series of nested grids, the finest of which
had a 2 km resolution (Grubišić et al. 2008). These forecasts proved to be a highly
useful and valuable part of the experiment planning process. An example of such
a forecast is shown in Fig. 3.17. The requirements of such models are discussed in
detail in Chap. 9, along with the issue of predictability of nonlinear mountain waves.
Compared with the techniques available for forecasting lee waves, the knowledge
available for forecasting rotors amounts to very little. Field campaigns have
provided some insight into the conditions under which rotors occur, but this tends to
be location-specific and the scientific knowledge is not sufficiently well developed
to be applied more generally. For example, the occurrence of a strong upwind
temperature inversion at, or around, the mountain height, along with unidirectional
flow across the mountain ridge, is known to be conducive to lee-wave motion
and it is generally thought that backward wind shear above the mountain top
will promote rotor formation. While these ideas are not inconsistent with current
scientific understanding, the theoretical and numerical studies which have addressed
the problem are so far too limited to provide any firm guidance as to the kind of rotor
motion which might occur, let alone the severity of the turbulence which might be
encountered. The ability of the very high-resolution numerical models to simulate
the dynamics of rotors (e.g. Doyle and Durran 2007) provides some hope that in
the future operational NWP models will have the ability to explicitly represent rotor
effects. Note, however, that much further research in this area is required, both in
terms of increasing our understanding, and also of the techniques needed to make
the best use of the high-resolution models.
3.4 Downslope Windstorms
Downslope windstorms have a long and rich history of scientific investigation and
have formed an important part of the meteorological literature for over 70 years.
They are a type of large-amplitude mountain wave that can occur downwind of
mountain barriers resulting in strong, often gusty low-level flow that accelerates
Fig. 3.16 3DVOM forecasts of lee-wave vertical velocity (blue line) over the Pennines, northern
England, UK, valid at 12 UTC on 17 November 2004. The forecasts are compared with
observations made by the UK FAAM research aircraft along east–west flight legs at 1,350 and
1,080 m above sea level (black line). Also shown are the predictions made by the Met Office
Unified Model run at 1 km resolution (red line) and the underlying terrain (shaded)
Fig. 3.17 Vertical velocity

(colour, interval 0.5 m s1 )
and potential temperature
(contour interval 6 K) from a
real-time 30-h COAMPS
forecast for the innermost
2 km resolution grid valid at
2100 UTC 25 March 2006
during T-REX IOP 6. The
cross section illustrates the
type of product that was
available for mission
planning purposes (From
Grubišić et al. 2008, Fig. 5.
© American Meteorological
Society. Reprinted with
permission)
down the lee slopes of the mountain. Under the right conditions, wind speeds in
downslope windstorms can be two to three times the wind speed at the mountain top
level. Necessary ingredients for downslope windstorms include a sufficiently large
mountain barrier, as well as strong across-barrier winds and a stable atmosphere,
both near the mountaintop level. They are observed in many places downwind of
mountains and are often given local names.
Several mechanisms have been proposed to explain the strong winds associated
with these events. These include an analog to hydraulic flow (Long 1954; Smith
1985; Durran 1986; Klemp and Durran 1987), the amplification of vertically
propagating gravity waves by constructive interference as they reflect off a pre-
existing critical layer, usually near the tropopause (Klemp and Lilly 1975), and
a similar reflection of upwardly propagating waves off a self-induced critical
layer created by wave breaking (Peltier and Clark 1979). The theoretical and
practical understanding of downslope windstorms has also been informed by rich
datasets from several critically important field programs, most notably the Pyrenees
Experiment (PYREX Bougeault et al. 1993), the Alpine Experiment (ALPEX Smith
1987), the Mesoscale Alpine Programme (MAP Bougeault et al. 2001; Smith et al.
2007) and the Terrain induced Rotor Experiment (T-REX Grubišić et al. 2008).
3.4.1 Observations
Downslope windstorms are observed in many places downwind of mountains and

are often given names. Locally named downslope windstorms in specific areas in-
clude: Rocky Mountains (Boulder windstorm, chinook), Western Washington (East
winds), Alps (deep foehn winds – refer to Chap. 4 in this volume), Eastern Adriatic
(bora winds), Southern California (Santa Ana winds), Argentina (Zonda), Utah
(Wasatch winds), southeast Alaska (Taku wind) etc. These winds are sometimes
categorized depending upon whether they are a warm wind or a cold wind. Warm
downslope windstorms are often referred to as foehn or chinook-type winds, while
cold downslope windstorms are often called bora-type winds. Chinook or foehn
events normally occur when an existing cold pool is displaced by an air mass
descending a lee slope that is warmed adiabatically as it descends (see Chap. 4
for a detailed discussion). It is thought that latent heat release on the windward
side of the mountain barrier may contribute to the relative warmth of foehn winds.
Bora-type downslope windstorms typically occur (in Western Washington State, and
on the Adriatic coast) when cold continental air displaces a warmer maritime air
mass, so is manifested as a relatively cold wind. The circumstances forming a bora
may instead form a gap wind if there is a channel that cuts through the mountain
barrier. The classification of downslope windstorms as bora- (cold) type or as foehn-
(warm) type is not always helpful, as it is sometimes difficult to make a definitive
categorization and some phenomena like the Boulder downslope windstorm can
occur as both warm and cold events. Also, a foehn is a general warm downslope
wind that may or may not be classified as a windstorm.
One of the best known downslope windstorms is the storm that occurs period-
ically, usually during winter, in Boulder Colorado, that is characterized by very
strong winds with peak gusts as high as 60 m s1 . This downslope windstorm has
been studied extensively due to its intensity, but also due to the high concentration
of meteorologists who live nearby and directly experience it. Brinkmann (1974)
analysed 20 Boulder windstorms finding that they are characterized by strong
gusts, are non-stationary, are associated with a surface pressure minima under the
disturbance, followed by a pressure jump downstream, that results from a hydraulic
jump. Brinkmann found that common upstream conditions include an inversion or
stable layer just above the mountain top, in combination with strong winds at that
level. The most studied Boulder downslope windstorm event is that of 11 January
1972 (e.g. Klemp and Lilly 1975, 1978; Clark and Peltier 1977; Peltier and Clark
1979, 1983; Clark and Farley 1984; Durran 1986; Doyle et al. 2000), since it was
the subject of special aircraft observations that provided rare in-situ observations
of the mountain wave structure (Lilly and Zipser 1972; Lilly 1978) (Figs. 3.18
and 3.19). This event was characterized by a very large amplitude mountain wave
extending through the troposphere into the lower stratosphere. A stable layer below
the mountain wave is behaving like a hydraulic supercritical flow layer as the
isentropes lower in the lee of the mountain (Fig. 3.18) with associated strong winds
in excess of 60 m s1 (Fig. 3.19) and terminating in a hydraulic jump that is
indicated by sharply rising isentropes and decelerating wind speeds. The surface
winds during this event were very strong and gusty (Fig. 3.20). Observations of
severe gustiness are quite common in downslope windstorms and may be caused by
subrotors embedded in the flow. For the 11 January 1972 Boulder windstorm, gusts
appear to have a period of slightly more than a minute, however longer period wind
pulsations have also been observed for less severe events (Durran 1990). The gust
structure of a moderate Boulder windstorm was documented by Neiman et al. (1988)
using Doppler lidar. They found the gusts had a period of approximately 4 min,
extended several hundred meters above the surface and appeared as a coherent
structure that propagates downwind at approximately the mean wind speed.
The foehn-type wind (see Chap. 4) can take the form of a downslope windstorm.
An important similarity between the 11 January 1972 Boulder windstorm in
Figs. 3.18–3.20 and the deep Alpine foehn (look ahead at Fig. 3.30 (bottom)) is
the presence of a weakly stable, nearly mixed region above the descending flow.
The Alpine deep foehn case additionally had a nearly mixed region below the
flowing layer on the upstream side. An example of an Alpine foehn downslope
windstorm event from MAP is shown in Fig. 3.21 (Jiang and Doyle 2004). It has
many similarities with the 11 January 1972 Boulder downslope windstorm shown
in Figs. 3.18–3.20: specifically the lower stable layer that descends and accelerates
that is surmounted by a turbulent unstable region with sharply rising isentropes and
across-barrier flow decreasing or reversing.
The bora is a downslope windstorm that occurs along the northeastern Adriatic
coast from Trieste, Italy to southern Croatia (Dalmatia) and has been frequently
documented in the European literature since the mid-nineteenth century (Yoshino
1976; Jurčec 1981; Grisogono and Belušić 2009). Other bora-like winds that have
Fig. 3.18 Cross section of potential temperature field observed in the 11 January 1972 downslope
wind storm illustrating the strong mountain wave extending through the troposphere and the lower-
level zone of supercritical flow and hydraulic jump between Boulder and Denver (From Lilly 1978.
© American Meteorological Society. Reprinted with permission)
Fig. 3.19 Cross section of u-component of the wind (m s1 ) from aircraft flight data and sondes
observed in the 11 January 1972 downslope wind storm (From Klemp and Lilly 1975. © American
Fig. 3.20 Anemometer trace from Southern Hills Junior High School in Boulder Colorado, on 11
January 1972 from 2,000 to 2,330 MST. Speed is in miles per hour (1 mph D 0.45 m s1 ) (From
Klemp and Lilly 1975. © American Meteorological Society. Reprinted with permission)
been studied occur near Novorossiysk on the Black Sea’s northeastern coast, at
Novaya Zemlya in the Russian arctic, and near Lake Baykal (Yoshino 1976). The
bora occurs under similar synoptic conditions as the gap flow through the fjords
and windstorms along the western coast of North America, described in Sect. 3.5.
A strong anticyclone over eastern or/and central Europe often in conjunction with
cyclonic conditions in the Mediterranean results in a strong pressure gradient
directed across the Dinaric Alps that separate the interior of the continent from
the Adriatic coast (Tutiš and Ivančan-Picek 1991; Ivančan-Picek and Tutiš 1995,
1996). Arctic air associated with the anticyclone deepens and eventually spills
through passes onto the coast (Jurčec 1981; Yoshino 1976; Smith 1987), usually
resulting in a decrease in air temperature. The strongest bora winds occur during
winter and at night, with generally ten “bora days” per month each winter season
(Yoshino 1972; Bajić 1989). A stable layer usually exists at mountain crest level
(Yoshino 1976) indicating the importance of hydraulic dynamics in describing
the flow.
Before ALPEX it was thought that the bora was a type of katabatic flow –
accelerating downslope due to its density. It is now understood that the bora is a
type of downslope windstorm, with hydraulic/mountain wave characteristics (Smith
1987). Klemp and Durran (1987) conducted two dimensional numerical simulations
of the bora and found that the flow resembled hydraulic flow over an obstacle,
and that the pressure gradients created by thermal differences on either side of the
Dinaric Alps were sufficient to generate the bora. The bora winds are strongest
downwind of gaps in the Dinaric Alps and can maintain their speed as localized jets
across the Adriatic sea for some distance (Zecchetto and Cappa 2001), presumably
decreasing speed when a transition from supercritical shooting flow to subcritical
flow occurs in a hydraulic jump. Gohm and Mayr (2005) simulated an Adriatic
deep bora case with a high resolution mesoscale model and found the strongest bora
Fig. 3.21 Cross section through the central Alps on 21 October 1999 showing (a) isentropes
(every K), backscatter from SABL lidar (in grayscale) depicting clouds and regions of strong
turbulence (indicated by symbol ƒ), (b) has the along-flight wind component (every 2 m s1 )
added in black contours with the region of reversed flow hatched. The flow pattern resembles
classical downslope wind event with plunging isentropes indicating the shallow hydraulic flow
separated from the air aloft by a turbulent region with overturning isentropes and a wind reversal
(From Jiang and Doyle 2004. © American Meteorological Society. Reprinted with permission)
flow downwind of gaps. They found the simulated bora was diurnally modulated:
during the night the stable surface layer upstream increased the bora strength, while
the developing convective boundary layer during the day reduced the gravity wave
amplitude and strength of the downslope bora flow. Gohm et al. (2008) simulated a
bora downwind of a gap that formed a jet with hydraulic jump features in the middle
of the gap and lee wave/rotor features on the edges of the jet.
In western Washington, downslope windstorms have been documented to occur
when cold arctic air and associated high surface pressure moves southward over the
interior of the continent. The coast-parallel Cascade mountain chain separates the
interior arctic air from the maritime air on the coast resulting in a large pressure
gradient across the mountain barrier as the continental high moves south. In this
situation, strong downslope winds can form, especially downwind of passes or gaps
in the mountain barrier such as the Stampede gap (Mass and Albright 1985; Reed
1981) where winds during these events typically reach 15–25 m s1 (Mass and
Albright 1985). The air flows down the slope and over the inland waters of Puget
Sound where a readjustment in boundary layer depth takes place and wind speeds
diminish with the flow splitting around the Olympic Mountains resulting in strong
easterly flow through Juan de Fuca Strait.
A very similar downslope windstorm has been documented near Juneau in
Alaska, the so-called Taku windstorm. This windstorm also involves downslope
acceleration in the lee of a gap (Colman and Dierking 1992; Dierking 1998; Bond
et al. 2006) under very similar synoptic conditions as the Cascade windstorms. In the
case of the Taku wind, there is an interplay between downslope windstorm dynamics
as discussed below, and gap wind dynamics discussed in Sect. 3.5, since this area is
also comprised of many gaps dissecting the coastal mountain barrier, and both gap
winds through channels and the stronger downslope winds tend to occur at the same
time with the lee trough induced by the downslope windstorm enhancing gap flows
through local channels.
The Santa Ana is a type of downslope windstorm occurring along the coast of
southern California when a large area of high pressure forms over the Great Basin
between the Rocky Mountains and Sierra Nevada Mountains. When a thermal low
pressure area over coastal California is advected offshore, the pressure gradient and
wind speeds can be enhanced. Santa Anas tend to occur most frequently between
October and February with December being the month of peak occurrence (Raphael
2003). About 20 events occur per year with the average duration being 1.5 days
(Raphael 2003). Santa Anas are often enhanced during the night by a land breeze
with typical observed wind speeds of 15–20 m s1 , gusting to over 25 m s1 . The air
flowing out of the high turns anticyclonically, is channelled through the canyons and
warms adiabatically due to compression as it flows downslope towards the coast.
The hot, dry, strong wind is associated with increased wildfire risk in California
and can modify temperature and currents in the nearby coastal upper ocean. The
winds can persist offshore for a hundred or so kilometers and are associated with
high wave conditions on the northeast side of the Southern California Bight Channel
Islands.
3.4.2 Theory
Downslope windstorms are large-amplitude mountain waves that undergo a hy-

draulic transition giving strong, supercritical flow on the lee slope. There is a
rich history of theoretical work on this phenomenon that has suggested several
mechanisms to account for downslope windstorms. These theories have been well
described and developed in a number of excellent reviews of this subject (e.g.
Smith 1979, 1989; Durran 1990; Baines 1995), and so will not be repeated in
detail here. The primary theories include an analog to hydraulic flow (e.g. Long
1954; Smith 1985; Durran 1986; Klemp and Durran 1987), the amplification of
vertically propagating gravity waves by constructive interference as they reflect off
a pre-existing critical layer, usually the tropopause (Klemp and Lilly 1975), and
a similar reflection and amplification of upwardly propagating waves off a self-
induced critical layer created by wave breaking (Peltier and Clark 1979). While
there has been debate over the relative importance and frequency of downslope
windstorms caused by these different mechanisms, it seems that the fundamental
dynamics are qualitatively explained as an analog to the hydraulic flow of water
over an obstacle resulting in rapid, supercritical flow along the obstacle’s lee slope,
which terminates in a turbulent hydraulic jump (see subsection 3.2.3). In the lower
left panel of Fig. 3.3, the supercritical flow of hydraulic theory over the lee slope,
corresponds to the zone of strong downslope winds near the surface.
The first mechanism, based on two-layer hydraulic theory with a deep upper
layer, was originally proposed by Long (1954), and used by Houghton and Kasahara
(1968), Houghton and Isaacson (1968) and Arakawa (1968, 1969). In the classical
hydraulic approach the flow of air over a mountain is modeled by fluid flowing over
a barrier using the shallow water equations (Eqs. 3.6–3.8, development described
in Sect. 3.2.3). Strong winds occur on the lee slope when the fluid (air) goes
from subcritical flow upstream, to supercritical flow over the mountain, becoming
subcritical again downstream as kinetic energy dissipates in a turbulent hydraulic
jump. In a classical hydraulic model, if the incoming subcritical flow is too rapid
to become critical at the mountain crest which acts as a control point, then a
wave can propagate upstream to “correct” the incoming flow. When applied to
the atmosphere, the main weaknesses of this theory are that it does not allow
continuous stratification, or vertical propagation of energy above either the free
surface or rigid lid upper boundary condition. However this theory is attractive
because of its simplicity and its apparent inclusion of important features observed in
downslope windstorms. Also, it is not uncommon for the atmosphere to have a sharp
change in stability, via elevated inversions and stable layers, which encourages the
atmosphere to behave as a simple hydraulic system. When a stable layer is present
near mountaintop level, hydraulic effects can occur.
A mechanism for downslope windstorms which would allow for the possibility
of vertical energy transport (Klemp and Lilly 1975) is the amplification of linear
vertically propagating gravity waves, by waves reflected off the tropopause (or some
other layer where the Scorer parameter, (3.22a), changes rapidly). The presence of
Fig. 3.22 Isentropes from a two-dimensional numerical simulation of a two-layer atmosphere over
a 600 m high mountain. Flow is from left to right. (a) Top of the lower stable layer is at 1,571 m
(one quarter of a vertical wavelength). (b) Top of the lower stable layer is at 3,142 m (one half of a
vertical wavelength) which corresponds to nonlinear resonance resulting in shooting supercritical
flow on the lee slope terminating in a hydraulic jump (Figure from Durran 1986. © American
vertically propagating mountain waves can be diagnosed from the isentropes. If

there is an upstream tilt of the isentrope phase, this indicates vertically propagating
wave energy that can be reflected downward by a critical layer (see Fig. 3.2 for
examples of this). The wave amplitude is determined by the superposition of these
upward and downward propagating modes and will depend upon whether or not
the atmosphere is “tuned” to constructive interference. Klemp and Lilly found that
the strongest downslope wind response occurred when the tropopause was located
half a vertical wavelength above the ground. With this linearized model of the
flow equations the vertically propagating and amplifying gravity waves cause lower
pressure and accelerated winds on the lee slopes with the wind strength proportional
to the mountain height.
As the theory is linear, its general applicability to large amplitude waves has
not been determined (Durran 1986). It seems that while the mechanism involving
linear theory proposed by Klemp and Lilly in which the predicted downslope
wind response scales with H2 can account for the strongest Boulder downslope
windstorms in the case where the profiles of U and N are constant (i.e. a single
layer structure), it is unsuccessful with nonlinear effects becoming important when
there are vertical variations in these profiles and the flow more closely resembles
classical hydraulic flow (Durran 1986, 1990) as shown in Fig. 3.22. Durran (1986)
examined the importance of nonlinear amplification for an atmosphere with a two-
layer stability structure, finding that in this case, nonlinear effects become important,
even with non-dimensional mountain heights as low as 0.3 – values which would
have insignificant nonlinear effects in an atmosphere with a single layer stability
structure.
The third mechanism proposed (Clark and Peltier 1977, 1984; Peltier and Clark
1979, 1983; Clark and Farley 1984), suggests large amplitude waves and downslope
winds occur after a developing vertically propagating mountain wave breaks. Wave
breaking occurs when the isentropes become vertical and the air becomes unstable
and overturns. The area of wave breaking creates a self-induced critical layer with
strong mixing and wind reversal. In this layer, where the wind component in the
direction of the horizontal wavevector goes to zero, the vertical wavelength and the
rate of vertical energy propagation also tends to zero as discussed in Sect. 3.2.2.
In this situation, if the distance between the self-induced critical layer and the
mountain is suitably tuned, it may trap and reflect upward propagating waves
resulting in resonance and strong surface winds. Peltier and Clark’s (1983) theory of
wave amplification for the numerically simulated downslope windstorm uses linear
theory and the assumption that the wave breaking region acts like a critical layer
where the Richardson number, Ri D N2 /(dU/dz), is less than ¼. According to Peltier
and Clark’s results, the critical layer will produce amplification only when it is
located about ¼ C n/2 (n D 0,1,2 : : : ) vertical wavelengths over the terrain. The self-
induced critical layer produced by wave breaking is similar to a turbulent mean-state
critical layer that would be produced when the stability becomes too low to provide
a sufficient restoring force for gravity waves, which occurs when the Richardson
number is less than ¼. It is believed that nonlinear waves that encounter such a
critical layer are reflected by it without losing much amplitude (Durran 1990).
The importance of wave breaking for the 11 January 1972 downslope windstorm
as well as for other cases with constant upstream wind and stability (constant U
and N) has been verified by many numerical simulations (e.g. Doyle et al. 2000).
Whether or not wave breaking occurs will depend on the non-dimensional mountain
height exceeding a critical value. For flow across a two dimensional ridge, Huppert
and Miles (1969) found the critical non-dimensional mountain height for gravity
wave breaking was 0.85. Smith and Grønås (1993) found a critical non-dimensional
mountain height of 1.1 ˙ 0.1 for an isolated Gaussian hill in three dimensional
flow. In each simulation when the critical non-dimensional mountain height was
exceeded, the wind stagnated aloft over the lee slope, isentropes overturned in wave
breaking and there was a strong downslope wind response. A critical level will
explicitly exist when there is a reverse shear in the approaching wind – U decreases
with height and perhaps even reverses. If this occurs at the correct altitude for a
resonant response, then strong downslope winds can result.
The theory proposed by Smith (1985) and Smith and Sun (1987) extends classical
hydraulic theory, by using a concept of local hydraulics in which the atmosphere
can have constant stratification rather than layers separated by a density interface,
and the hydraulic nature of the flow only manifests itself when a transition to
supercritical flow occurs on the lee slope below a layer of stagnant fluid. In local
hydraulics, the flow approaching the mountain may “select” a critical streamline
which splits becoming the upper boundary of the accelerated flow (Fig. 3.23),
decoupling it from a turbulent layer of well-mixed air aloft. It is the presence of
this stagnant region that provides the decoupling allowing the flow to produce a
shallow layer-like flow that can spill over terrain and resemble classical hydraulic
Fig. 3.23 (a) Schematic diagram of a downslope windstorm based on aircraft observations and
numerical simulations. In this case a certain critical dividing streamline encompasses a turbulent
region of uniform density ¡c indicated by the inverted “v” shapes. In this local hydraulic approach,
continuously stratified air approaches the mountain barrier and acquires hydraulic characteristics
of supercritical flow after the critical streamline divides. (b) Flow over a mountain from the Smith
(1985) local hydraulics model. U0 D 20 m s1 , N0 D 0.01 s1 (From Smith 1985. © American
supercritical flow and even produce a hydraulic jump. In this theory, when the non-
dimensional mountain height, Ĥ D NH/U is small, or if a wind reversal exists at the
wrong altitude the fluid may not form a neutral decoupled region with hydraulic
flow, but instead form continuous vertically propagating gravity waves (Smith and
Sun 1987). According to this theory, amplification of a mountain wave is possible
across a range of critical layer heights from 1/4 C n to 3/4 C n, where (n D 0,1,2 : : : )
vertical wavelengths above the terrain, in contrast to the predictions of Peltier
and Clark.
Durran and Klemp (1987) and Bacmeister and Pierrehumbert (1988) used
numerical experiments to compare the theories of Smith (1985) and Peltier and
Clark (1979, 1983) and found that when there is a mean state critical layer, the
mountain waves interact with this layer to form a local region of turbulent flow, and
that the essential elements of hydraulic theory explain the acceleration on the lee
slopes of mountains. This occurred in Durran and Klemp’s experiments whenever
Ĥ exceeded 0.2 and there was a mean state critical layer between ¼ and ½ vertical
wavelengths above the terrain (Durran 1990). In their experiments a deep layer with
Ri < ¼ developed above the topography and a strong downslope wind response was
found with this configuration when Ĥ exceeded 0.4.
Reed (1981) and Mass and Albright (1985) were able to use a simple Bernoulli
equation and gap wind dynamics based on a balance between the pressure gradient
force and inertia to account for the observed wind speeds in the cases they studied
in US Pacific Northwest. Colle and Mass (1998) set out to examine the relative
importance of mountain wave/hydraulic dynamics versus simple gap flow dynamics
by analysing a number of past events in western Washington State, as well as
using idealized mesoscale model simulations. They found that a combination of the
across-barrier pressure gradient and across-barrier flow at crest level explained 82%
of the variance in windstorm severity there. The presence of a critical layer (region
of flow reversal needed for reflection of vertically propagating waves) appeared to
be of less importance in storm evolution. However in cases with strong cross-barrier
flow at mountain crest level, the lack of a critical layer below 400 hPa resulted in
weaker surface winds unless a region of strong stability existed near crest level, in
which case the flow behaved hydraulically. For cases with weak cross barrier flow
at crest level, the presence of a critical layer was less important, as they postulated
a self-induced critical layer could be created and strong surface winds could
ensue.
One remarkable feature of many severe downslope windstorm events is the severe
gustiness of the wind. The observed gustiness of the 11 January 1972 Boulder
downslope windstorm (Fig. 3.20) was investigated by Clark and Farley (1984) using
two and three-dimensional numerical models. They found gust structures were only
apparent in the three-dimensional simulations and hypothesized that the gustiness
was produced by an intrinsically three-dimensional competition between wave
enhancement by gravity wave dynamics and wave destruction by convective insta-
bility. The apparent requirement of three-dimensions for gustiness was contradicted
by Scinocca and Peltier (1989) however who were able to produce realistic gusts
(which they call pulsations) in their two-dimensional simulations. Peltier and
Scinocca (1990) demonstrate that the simulated pulsations are due to Kelvin-
Helmholtz instability at the wind shear interface between the low level shooting
flow and the wave-induced stagnant layer. The Kelvin-Helmholtz instability is a
secondary instability that occurs after the downslope windstorm has formed when
Ri < ¼ in the flow and appears to initiate just upstream of the mountain crest.
Afanasyev and Peltier (1998) demonstrate the importance of the Kelvin-Helmholtz
mechanism and further demonstrate how in a fully three-dimensional simulation,
the Kelvin-Helmholtz billows are eroded by a convective instability and dissolve
into a turbulent background flow as they move downstream. The Kelvin-Helmholtz
instability as a mechanism to explain the pulsations has been suggested and modeled
for the bora as well (Belušić et al. 2007).
Most of the earlier theoretical work using linear theory, as well as numerical mod-
eling studies in the past has assumed that boundary-layer effects are not important
for downslope windstorms. However numerical simulations of the 1972 Boulder
windstorm by Richard et al. (1989) as well as work by Georgelin et al. (1994),
Ólafsson and Bougeault (1997), Peng and Thompson (2003), Smith et al. (2002,
2007) and Smith et al. (2006) has illustrated the importance of the boundary layer
on absorbing mountain wave energy and both attenuating downslope windstorm
strength and limiting its extent downwind. Georgelin et al. (1994) studied the
effect of subgrid scale orographic roughness on orographic flows from PYREX.
They found that correctly accounting for subgrid scale orography improved model
predictions. It reduced mountain wave amplitude, but increased blocking as well as
leeside low-level turbulence. Ólafsson and Bougeault (1997) conducted idealized
simulations to investigate the effects of rotation (f ) and surface friction on mountain
wave drag. They found that the effect of friction was to suppress wave breaking,
especially for NH/U less than 3, and that the combined effects of rotation and drag
were to constrain the drag and flow patterns to resemble those predicted by linear
theory. Smith et al. (2002) discussed the effects of a low-level stagnant layer on
the mountain wave, finding that it reduced the wave amplitude and was able to
absorb downward reflected gravity waves, preventing lee wave generation. Jiang and
Doyle (2008) modeled a downslope wind event in Owen’s Valley during the Terrain-
Induced Rotor Experiment, and found in this narrow valley that the downslope
winds varied diurnally through an interaction with the thermal structure in the
valley. Overnight, when the valley filled with cool air, strong downslope winds
were restricted to the upper lee slope. After sunrise, surface heating in the valley
weakened the downslope winds, eventually decoupling the valley air from the strong
westerlies above the mountain top. As the valley air continued to warm during the
afternoon, the mixed layer extended beyond the top of the valley which caused a
transition as a shallow layer of cooler air from the westerlies above the mountain top
descended and flowed across the whole valley. These simulations serve to illustrate
the important connections with diurnal mountain circulations (Chap. 2).
Until recently, “dry” dynamics were used to understand downslope windstorm
formation. Doyle and Smith (2003) used a numerical model to isolate the influence
of latent heat in mountain wave generation by running the COAMPS model with and
without latent heating and examining the differences (Fig. 3.24). They found that if
the latent heating in a shallow layer upstream was strong, the lee downslope wind
response increased significantly. This occurred because the latent heating resulted in
a decrease in stability aloft creating a critical layer there that decoupled the low level
flow allowing it to behave as a hydraulic layer. Such a mechanism – the interaction
between low level latent heating and the establishment of critical layer aloft resulting
in a downslope windstorm – suggests a mechanism explaining how warm foehn
winds are able to descend on the lee side of the Alps.
The previous section has outlined the theoretical basis for understanding downslope
windstorm formation. This understanding has developed over time through a fusion
of analytical approaches involving careful analysis and judicious simplifications to
the governing equations, as well as through idealized and realistic experiments with
numerical models of increasing complexity and accuracy. This section will discuss
how this theoretical understanding can be applied to model and forecast downslope
windstorms using synoptic interpretation and NWP as well as statistical methods.
More information on the application of mesoscale models and NWP models is
provided in Chaps. 9, 10, and 11.
While NWP models are essential to understand and forecast most meteorological
phenomena, there are limitations in the application of NWP to downslope wind-
storms. These limitations, such as model resolution and the vertical coordinate
system, as well as predictability will be discussed later in this section as well as
Fig. 3.24 Cross section of potential temperature (contours every 3 K) and vertical velocity
(contours every 2 m s1 with upward velocities greater than 2 m s1 shaded) for a COAMPS
simulation of flow from south (left) to north (right) of a mountain wave event across the central
Alps on 20 September 1999, (a) includes latent heating and (b) has latent heating turned off.
Note the more strongly plunging flow and downslope wind response in (a). The shallow upslope
latent heating has tuned the atmosphere for nonlinear resonance (From Doyle and Smith 2003.
Reproduced by permission of the Royal Meteorological Society)
in Chaps. 9 and 10. However, even if NWP models do not accurately represent
mountain waves directly, they can still provide important information on upstream
vertical profiles of wind and stability. This information may be interpreted using
knowledge of downslope windstorm mechanisms, to infer whether or not a strong
downslope wind response is likely. Let us summarize some of the key findings from
observations and theoretical work that relate conditions in the upstream environment
to downslope windstorm occurrence. The type of mountain wave, and whether or
not any waves will occur depends on topography, wind and stability profiles. The
following conditions are necessary, although not necessarily sufficient for a strong
downslope wind response:
• A significant mountain barrier. Long mountain barriers that are asymmetrical
with a shallow windward slope and a steep lee slope are particularly conducive
to downslope windstorm development (Smith 1977; Lilly and Klemp 1979;
Hoinka 1985). This limits the occurrence of downslope windstorms to specific
geographic regions.
• The presence of a moderate to strong wind component in the across-barrier
direction (7–15 m s1 or more, depending on the mountain) at an elevation near
the mountaintop level.
• Stable stratification. This is necessary to provide a restoring force for a gravity
wave response.
• A non-dimensional mountain height, Ĥ D NH/U, somewhat close to 1 (i.e. Ĥ not
1 or 1). If Ĥ is too small, the flow will easily go over the mountain without
a mountain wave response. If Ĥ is too large, the flow will be completely blocked
by the mountain barrier and not able to flow over the mountain.
A strong downslope wind response closely resembles supercritical hydraulic

flow. The circumstances that encourage this type of flow can arise in several ways.
One way is if the background atmospheric state has a layered stratification. The
presence of an inversion or isothermal layer at or just above the mountain top
level enhances hydraulic effects and the possibility of a strong supercritical flow
response on the lee slope. The importance of such a stable layer is verified in
the Boulder downslope windstorm climatologies of Brinkmann (1974) and Colson
(1954). Reduced stability or cross barrier winds further aloft can also decouple
the low-level flow and allow it to behave like a shallow hydraulic layer (Smith
1985; Baines 1995). If the cross-barrier winds increase aloft in the presence of an
inversion near mountain top level (i.e. there is forward shear), there is a possibility
of either lee waves or enhanced downslope winds. A rule of thumb that can be used
to differentiate the two mountain wave forms, is to compare the cross-barrier wind
component at mountain top level with the winds 2,000 m above the mountain top. If
the winds aloft increase to over 1.6 times the winds at mountain top level, then lee
waves rather than a downslope windstorm are more likely (COMET 2008).
Since gravity waves require a restoring force (buoyancy) and therefore a stable
atmosphere – they will not propagate through a neutral or near neutral turbulent
layer, which can therefore act as a mean-state critical layer (Ri < ¼) resulting in the
reflection of vertically propagating mountain waves. Therefore the presence of such
a mean-state critical layer above a stable layer can be diagnostic of a downslope
windstorm event. Vertically propagating mountain waves can also be reflected by a
self-induced critical layer (a layer where the cross-barrier wind component goes to
zero) that can form when vertically propagating mountain waves break, forming
a turbulent “dead” zone of air that effectively decouples the lower level flow,
allowing it to behave hydraulically. Such a wave breaking situation can occur in an
atmosphere with constant U and N where the mountain is large enough (Clark and
Peltier 1977). In addition to wave breaking resulting in a self-induced critical layer
and the existence of a mean-state critical layer, a critical layer can also be formed
by wind shear, when the wind component in the cross-barrier direction goes to zero,
either by the wind speed decreasing or by the direction shifting to a barrier-parallel
direction.
It is thought that jet streak dynamics could also play a role in downslope
windstorm development. Specifically the right-exit region of a jet streak is a zone
of upper level convergence and downward vertical velocity. If this zone occurs near
a ridgeline of sufficient height in a jet that is oriented perpendicular to the ridge,
then it can contribute to a subsidence inversion or stable layer near the mountain
top, which can lead to a downslope windstorm. Many of these factors for practical
forecasting of downslope windstorms are well illustrated in a COMET MetEd
interactive module on forecasting downslope windstorms (register for MetEd and
then go to http://www.meted.ucar.edu/mesoprim/mtnwave/) (COMET 2008).
The previous section has discussed numerical modeling of downslope wind-
storms from a research and process perspective – high resolution models have
been used extensively in idealized and realistic simulations to shed light on the
mechanisms responsible for downslope windstorm events. There are also numerous
examples where numerical models have successfully simulated many aspects of
downslope windstorms, in hindcast mode (e.g. Doyle and Shapiro 2000; Colle and
Mass 1998). NWP models are of course a key tool for operational forecasting. Are
operational NWP models of use in downslope windstorm predictions? One major
issue is whether the models have sufficient vertical and horizontal resolution and
accuracy to capture mountain wave activity.
It is generally accepted that to properly resolve a wave-like feature in a
numerical model, there must be six to eight grid points per wave, so that the model
resolution must be finer than 1/6 of a wavelength. To properly resolve mountain
waves, model horizontal resolution should be on the order of 1 km, and certainly less
than 10 km. Current operational NWP models (as of 2011) are close to achieving
10 km or finer horizontal resolution, so in the future will be able to provide realistic
simulations of mountain wave activity and downslope windstorms, although there
are still questions concerning the predictability of these phenomena which will be
discussed below.
Other issues related to downslope wind forecasting by numerical models that
are summarized by Smith et al. (2007) include inaccuracies associated with the
vertical coordinate system over mountains. The commonly used terrain following
coordinate system has difficulty in accurately representing the horizontal pressure
gradient force. This led Schär et al. (2002) to propose a new coordinate system
that has smoother and more horizontal levels at mid and upper levels and thereby
reduce the truncation errors and high frequency noise in the model (see Chap. 9).
Klemp et al. 2003 found that errors and spurious gravity waves could be generated
in numerical models in the conversion between terrain-following and Cartesian
coordinates if the transformation of the advection terms and pressure gradient
terms was not done consistently. Another issue related to the sigma coordinate
system concerns the application of horizontal diffusion on sloping sigma surfaces.
Horizontal diffusion is commonly applied to control model errors. If it is applied
along the terrain following coordinates which slope with the terrain, this can conflate
vertical gradients with horizontal gradients leading to errors especially when there
are strong vertical gradients such as inversions (Smith et al. 2007). Zängl (2002b)
developed a true horizontal diffusion scheme for the MM5 model that was shown to
improve its ability to simulate topographic flows (see further discussion in Chap. 9).
Although increasing computer power and improving NWP models with ever-
increasing resolution holds promise for the prediction of downslope windstorms, the
predictability of these phenomena remains a major unresolved issue. Lorenz (1969)
hypothesized that the growth rate of perturbations increases as the scale of a phe-
nomenon decreases, implying that the predictability of mesoscale phenomena whose
scales are on the order of 10 km, may only be a few hours in advance. However, the
fact that downslope windstorms are forced by topography implies that they ought
to be inherently more predictable than mesoscale phenomena that do not have such
a strong organizing influence (Anthes et al. 1985). The enhanced predictability of
downslope windstorms was first demonstrated by Klemp and Lilly (1975), who used
a linear two-dimensional model to predict the occurrence of downslope windstorms
in Boulder, Colorado. They initialized their model using soundings from NWP
model forecasts, and made predictions valid approximately 5 h in advance of the
windstorm, successfully predicting events where observed gusts were in excess of
25 m s1 . More recent studies have not been as encouraging. Nance and Colman
(2000) used a high-resolution two-dimensional nonlinear mesoscale model that was
initialized with upstream soundings from 12 to 18 h NWP model forecasts in order
to support operational forecasting of downslope wind events in seven regions in the
United States between 1993 and 1997. They found that the model improved upon
existing forecasting methods which were based on a decision tree (Brown 1986), but
also had a high “false alarm” rate, implying difficulty in distinguishing windstorm
events from null events.
During MAP, many studies were able to successfully simulate mountain waves of
various types (e.g. Smith et al. 2002; Smith and Broad 2003; Doyle and Smith 2003;
Volkert et al. 2003). As pointed out by Smith et al. (2007) these successful cases
were characterized by stationary non-breaking waves, and results from other cases
that were characterized by non-stationary waves were not as successful (e.g. Jiang
and Doyle 2004; Smith 2004; Jiang et al. 2005; Doyle and Jiang 2006). Doyle et al.
(2000) compared two-dimensional simulations by 11 different numerical models
initialized with an identical sounding representing the 11 January 1972 Boulder
downslope windstorm event. While all models predicted downslope winds and wave
breaking in the lower stratosphere, the models had different temporal evolutions of
the wave breaking. Doyle and Reynolds (2008) tested the sensitivity of the same
downslope windstorm event to small variations in initial conditions by performing
ensemble simulations in a high-resolution two-dimensional numerical model. Each
ensemble member was constructed by perturbing the upstream sounding by small
amounts corresponding to typical radiosonde measurement errors. They found that
for this case there was a large spread in the ensembles with a 25 m s1 range in the
modeled downslope wind speed. Half the ensemble members simulated a trapped
lee wave response and the other half simulated large-amplitude wave breaking in the
lower stratosphere with a hydraulic jump and a strong downslope wind response.
The bimodal distribution of the ensemble members led them to conclude that when
there is such a transition across a regime boundary (between non-breaking and wave
breaking responses), the use of ensemble statistics such as the ensemble mean may
not be appropriate. Reinecke and Durran (2009a) used a three-dimensional NWP
model to conduct ensemble simulations of two downslope windstorms observed
during T-REX. One case was a wave breaking downslope windstorm, while the
second case was a layered case characterized by strong low-level stability in which
wave breaking was not important. For each case they conducted 70 ensembles
by perturbing the initial conditions across a range representing uncertainty in the
initial conditions. They found that the forecasts were very sensitive to the initial
conditions, especially for the wave breaking case where the difference in the
downslope wind speeds predicted by the strong ensemble members minus the weak
ensemble members grew to 28 m s1 after only 6 h of forecast. The layered case
was found to be somewhat more predictable with the difference in downslope wind
speeds growing to 22 m s1 after 12 forecast hours. The results of these recent
downslope wind predictability studies indicating sensitivity to initial conditions, do

not suggest much confidence should be placed in the accurate and precise strength,
timing, or location of downslope winds predicted beyond 12 h in advance. Despite
this, operational NWP models are capable of predicting the overall synoptic setting
for these storms up to several days in advance. This currently allows a forecast
of downslope windstorm occurrence without necessarily being able to precisely
predict the location, timing or strength. As computer power continues to increase,
high resolution NWP models could be implemented in operational forecast settings
as a nowcasting tool to make short-range precise and accurate forecasts. Further
discussion on predictability issues can be found in Chap. 9 (Sect. 9.9).
Žagar and Rakovec (1999) discuss a dynamic downscaling approach to obtain
a high resolution wind field that they call “dynamic adaptation”. The approach
interpolates a coarse resolution NWP forecast onto a fine-resolution mesoscale
model grid, and then runs a simplified version of the mesoscale model for between
30 and 45 min to obtain a high-resolution wind field. The approach was successfully
applied to a strong wind event: a foehn in the Ljubljana basin in Slovenia, in
which diurnal boundary layer variations which cannot be properly represented
by the approach, were not important. The advantage of the approach is its low
computational cost, and it seems to hold promise as a forecast tool in an operational
setting for strong dynamically forced flows.
As a less computationally and technically intensive forecast tool, it is possible
to develop statistical models to predict downslope windstorms at specific locations.
Mercer et al. (2008) used a 10 year Boulder windstorm dataset together with a set
of 18 predictors to train and test statistical models to predict the downslope wind
speed. They tested linear regression and neural network models but found that a
support vector regression model performed best and was able to predict downslope
wind speed with an RMSE of 4–6 m s1 .
Given the likelihood of limited predictability (i.e. short prediction lead times), at
least for strong, highly nonlinear downslope windstorms, the ability to observe their
occurrence is very important. Mountain waves, including downslope windstorms
can be observed in a variety of ways. Of course, direct observation of the downslope
wind response at surface observing stations is the best indication of occurrence;
however stations do not exist in every location. Satellite imagery can also be useful.
If there is sufficient moisture for clouds to be present, a zone of descending air is
often indicated in satellite imagery by a gap in the cloudiness that may indicate
mountain wave formation.
3.5 Gap Winds
Gaps criss-cross mountain ranges all over the world. Gaps also exist between
islands, or islands and the mainland. They can be level or elevated with an elevation
maximum in the gap whence they are commonly called passes or cols. Gaps provide
a means for an air mass to advance to the other side of the barrier formed by
Fig. 3.25 Cold continental air crossing the Coast Mountains through gaps and emanating as
accelerating flows onto the Pacific Ocean. Flow is from right to left. Wind speed (see color
bar) over water is derived from Synthetic Aperture Radar (SAR) data. Image date: 0224 UTC
29 January 2008 (Data courtesy of Alaska SAR Demonstration [AKDEMO] of NOAA’s STAR-
Center for Satellite Applications and Research)
mountain ranges. A distinctive feature of most gap flows is their asymmetry between
the region upstream of the gap entrance and downstream of the gap. Upstream, the
layer which flows through the gap is thick and relatively slow moving (subcritical
in hydraulic terminology). As it approaches the gap, it thins and accelerates and
becomes supercritical downstream of the gap.
3.5.1 Observations
Regions in the world where gaps join two persistently different air masses see
frequent gap flows. In this subsection, we will first look at observations from the
earth’s atmosphere and the laboratory of flows through level gaps. Next we will
look at elevated gaps, which have a vertical constriction added to the lateral one to
form a pass or a sill.
The northern part of the North American Pacific coast where long, coast-parallel
mountain ranges separate cold wintertime continental air masses from warmer
maritime ones is a preferred region for gap flows. Figure 3.25 shows a striking
visualization of the surface wind field with jets emanating from various gaps in
the Coast Mountains of British Columbia and Alaska. High wind speeds are only
visible downstream of gaps, which provide a connection to the continental air mass
east of the Coast Mountains. Highest wind speeds occur near the exit or even further
downwind as seen in the southern part of the figure close to the coast. There, several
gap jets merge to form one wide high wind speed streak. Even further away from
the gap exits, the flow slows down but has to pass through another gap formed by
the islands off the coast, where it reaches its overall maximum wind speeds near the
exit of that last gap.
Further visualizations of gap flow wind fields over water from synthetic aperture
radar (SAR) measurements can be found in Pan and Smith (1999), Winstead et al.
(2006), Liu et al. (2008), and Alpers et al. (2009), for example. Alpers et al. (2009)
provide highly resolved wind fields over the Adriatic Sea and the Black Sea during
six bora – gap flow events.
Shelikof Strait, formed by Kodiak Island and the Alaska Peninsula (Fig. 3.26),
is an example of a level gap. Colder air frequently lies northeastward on the other
side of the mountain range barely visible at the horizon of the figure. Research
aircraft flights (Lackmann and Overland 1989) provided the horizontal wind and
temperature field at 80 m AGL shown in Fig. 3.26. Wind speeds between the
entrance and exit regions nearly triple. Like in the previous example, the maximum
is close to the exit of the gap and not at its narrowest part as simple continuity
reasoning (“Venturi flow”) might suggest. Another important characteristic feature
is that the upstream air is colder than the dowstream air, here by about 4 K.
The Strait of Juan de Fuca (Fig. 3.27a) forms another level gap along the Pacific
North American coast. In wintertime it is fed by cold continental air from east of
the Cascade Range, which has passed through elevated gaps in the Cascade Range,
i.e. the Strait of Juan de Fuca is the second gap through which the air streams.
In the gap flow event shown in Fig. 3.27, the presence of hydrometeors provided
backscatter for a Doppler radar on board of the NOAA-P3 aircraft, from which the
three-dimensional wind field could be computed. The vertical structure of the flow
along the strait is shown in Fig. 3.27. Wind speeds between entrance and exit of
the gap double. It is remarkable that the abrupt thinning of the gap flow layer is
coincident with a further speed-up. The gap flow layer is isolated from the flow
aloft through a layer with southerly flow in almost perpendicular direction to the
gap flow.
Observations of gap flows in the laboratory (Armi and Williams 1993) repro-
duced in Fig. 3.28 show similar structures to the ones observed in the atmosphere.
Similar to its cold and stably stratified counterpart in the atmosphere, the fluid
in the laboratory experiment is continuously stratified, which is visualized by
dyeing selected isopycnals to give an equivalent to isentropic cross sections in the
atmosphere. An arrow indicates the narrowest part. When fluid is withdrawn at
the bottom of the channel, the flow is similar to the one observed in the Strait of
Juan de Fuca, shown in Fig. 3.27b. The flow accelerates through the channel as
the depth of the gap flow decreases. Near the exit, the flow accelerates even more
and its depth drops. A homogeneous stagnant layer forms as the streamlines split
at the gap exit. This layer isolates the gap flow from the fluid aloft. A similar,
nearly stagnant layer was observed in the Strait of Juan de Fuca example as well
Fig. 3.26 Horizontal wind vectors (see reference vector for scale), isotachs (m s1 , dotted), and
temperature (ı C, colored) of the gap flow through Shelikof Strait measured with the NOAA-P3
aircraft 80 m above the ocean on 28 March 1985. Inset shows a pseudo-3D view of Shelikof Strait
between the Alaskan Peninsula and Kodiak Island viewed from SSE (Adapted from Figs. 6 and 8
of Lackmann and Overland 1989. © American Meteorological Society. Reprinted with permission.
Inset figure © Google and NOAA)
as the location of the speed maximum closest to the ground (Fig. 3.27b). The
equivalent to withdrawing fluid near the bottom of the laboratory channel is having
the strongest pressure gradient in the flowing parts of the atmospheric gap flow close
to the surface. Since the temperature differences of the air masses on either side of
the gap are hydrostatically responsible for the bulk of the pressure difference, this
translates into having the strongest temperature differences between the upstream
and downstream side of the gap close to the surface.
Fig. 3.27 (a) A pseudo-3D

view from the W of the Strait
of Juan de Fuca between
Vancouver Island and the
Olympic Mountains. (b)
Wind vectors in the along-gap
section and isotachs of
along-section speed (contours
in 3 m s1 steps and shading)
computed from Doppler radar
measurements onboard of the
NOAA P3 aircraft. (c) Wind
vectors and isotachs of the
along-section speed (dashed
contours in 3 m s1 steps
andshading) and isentropes
(solid, 3 K intervals) from
numerical simulations. The
gap flow in (b) and (c) is in
the offshore direction from
right to left, and is capped by
southerly flow around 2 km
MSL and onshore flow
further aloft. (Part (a) ©
Google and NOAA. Parts (b)
and (c) are from Colle and
Mass 2000. © American
When fluid is withdrawn from mid-levels of the channel (Fig. 3.28b), the flow
response changes. The fluid accelerates through the gap and the flowing layer thins
and accelerates with a parabolic wind profile and the jet maximum at the center of
the gap flow layer at the exit of the gap. In addition to the isolating, stagnant and
Fig. 3.28 Steady continuously stratified flow through a contraction. The linear continuous stratifi-
cation is made visible by dyeing selected isopycnals in the upstream reservoir. The fluid accelerates
through the narrowest section of the contraction (marked by the arrow) as the flow moves from left
to right. (a) Level gap with fluid withdrawn through a slit at the bottom of the channel downstream
of the gap. (b) Level gap with fluid withdrawn at the center height of the channel downstream of the
gap. Vertical dye streaks are distorted by the flow and show the self-similar velocity profile of the
Wood (1968) solution (cf. subsection 3.5.2). (c) Elevated gap, i.e. a sill was added at the narrowest
section of the gap. Note the upstream flow in the contraction is the self-similar solution but that
this solution changes in the neighborhood of the sill and accelerates down the lee slope (Adapted
from Armi and Williams 1993. Reproduced by permission of Cambridge University Press)
homogeneous layer aloft, another such layer forms at the bottom. Just like for the
case of bottom withdrawal, the flow is self-similar in the sense that at any position,
the density and velocity profiles can be scaled with the height of the moving
stratified layer. This self-similar structure upstream is a necessary component of
any steady stratified flow from a reservoir and for it to exist, at least an isolating
Fig. 3.29 A pseudo-3D from the N of the Wipp Valley in the central Alps at the border between
Austria and Italy. The main gap with an altitude of about 2.1 km MSL with a small deep incision
of the Brenner Pass at only 1.4 km MSL is embedded in the main Alpine crest (approximately
3 km MSL). The Wipp Valley merges with the Inn Valley via a nearly 200 m terrain drop at
approximately right angles (© Google and NOAA)
stagnant intermediate layer needs to be formed above (cf. Armi and Williams 1993).
The equivalent to a withdrawal at mid-levels of the gap flow is to have the strongest
temperature differences between the upstream and downstream side and thus the
largest hydrostatically produced pressure gradient at mid-levels.
Elevated gaps, which are more commonly called “passes”, are constricted both
laterally and vertically. One extensively studied one, Brenner Pass in the Alps,
is shown in Fig. 3.29, a pseudo-3D view looking at the gap from the north.
Topography is much more complicated than in the level gaps shown so far. The
main gap, approximately 20 km wide, at 2–2.2 km MSL (“channel crest”) is cut
into the main crest line of 3 km MSL. The channel formed by the valley north of
the pass is no longer nearly straight as for the two previous strait examples but
contains tributaries and a bend and its exit drops 200 m into another valley, which
runs nearly perpendicular. Gap flows are possible in both directions, northwards and
southwards, but only northward gap flows have been studied extensively.
Three prototypical gap flows were found (Armi and Mayr 2007). The first one,
shown in Fig. 3.30, has a (nearly) mixed flowing layer on the upstream side, which
accelerates across the gap, before descending and thinning. Upstream, the flow is
isolated by a blocked layer underneath and a cap with strong stability seen in the
abrupt increase of potential temperature aloft. Downstream, the well-mixed gap

flow layer has sped up. The speed maximum lies a few hundred meters above
ground. The flow is no longer isolated underneath by a stagnant layer. Aloft,
however, a wedge of (nearly) neutral and stagnant air has formed. Its lower boundary
follows the descending gap flow whereas its upper boundary remains approximately
horizontal. The flow above the gap flow is therefore unaffected by the underlying
gap and channel topography. Since the gap flow layer is well mixed on both sides of
the pass, this case is the “single-layer prototype”.
The second prototype (second row in Fig. 3.30) differs from the first one by
having a stably stratified gap flow layer upstream, typically with a linear increase
of potential temperature with height. Upstream, the gap flow is isolated from the
surface and from the flow aloft, respectively, by neutrally stratified, (nearly) stagnant
layers.1 Downstream of the narrowest part, however, turbulent mixing takes place,
both in an internal hydraulic jump and over the rough surface, quickly turning
the stable stratification into a neutral one. The downstream part of the second
prototype is thus similar to the first one. Also, a wedge of (nearly) stagnant and
neutrally stratified air forms on top of the accelerating and descending gap flow layer
downstream of the pass. Since a thin isolating layer is already present upstream of
the pass, the thickness of the wedge is larger than for the single-layer prototype.
Another common feature of both prototypes is that the flow aloft is unaffected by
the underlying channel and pass topography and follows the horizontal isentropes.
The upstream part of the second prototype corresponds to the analytical solution
of Wood (1968), the downstream part is a single layer, making this case the
“Wood – single-layer prototype”. Wood’s solution describes a self-similar flow that
accelerates from an upstream reservoir to the control. It is the only known analytical
solution in such a setting for a continuously stratified fluid and is described and
derived in more detail in subsection 3.5.2.
The second prototype is also observed in the laboratory (Fig. 3.28c) when a sill
is added to the lateral constriction to form a pass or col. Upstream of the pass, the
stably stratified gap flow is isolated from the channel bottom and the fluid aloft by
neutrally stratified layers. The flow response downstream of the pass changes: the
gap flow layer accelerates even more, thins and rebounds in an internal hydraulic
jump at the downstream end of the sill. Further downstream, the gap flow layer is
well mixed. The depth of the upstream isolating layer on top of the gap flow layer
increases as the gap flow layer plunges downstream of the pass to form a thick
(nearly) stagnant layer. This layer both isolates the gap flow from the fluid aloft and
the fluid aloft from the underlying topography. The isopycnals remain horizontal,
like the isentropes in the atmospheric case.
The third, “hybrid” prototype (Fig. 3.30, bottom row) is so deep that the flow no
longer only crosses through the gap but also over the crest. A downslope windstorm
(or “deep foehn” as it is called in the Alps) is superimposed on the gap flow. The
1
Below the lower neutral isolating layer, the air next to the surface might be stably stratified (e.g.
nocturnally) and blocked
6 6
5 5
altitude (km MSL)
4 4
3 isolating layer 3
2 2
1 1
290 300 310 320 -10 0 10 20 30 40 290 300 310 320
6 6
5 5
altitude (km MSL)
4 4
isolating layer
3 isolating layer 3
2 2
isolating layer
1 1
290 300 310 320 -10 0 10 20 30 40 290 300 310 320
6 6
isolating layer
isolating layer
5 5
altitude (km MSL)
4 4
3 3
2 isolating layer 2
1 1
290 300 310 320 -10 0 10 20 30 40 290 300 310 320

pot. temperature (K) distance pot. temperature (K)
(km)
Fig. 3.30 Prototypes of gap flow across Brenner Pass in the central Alps. Single layer hydraulics:
shallow gap flow (first row). Shallow, stably stratified gap flow with Wood’s (1968) self similar
solution upstream turning into a single layer flow downstream (second row). Deep, stably stratified
flow through the gap and across the crest (often called “deep foehn” or “downslope windstorm”)
(third row). Soundings of potential temperature from upstream and downstream are shown to the
left and right, respectively. Arrows in the vertical cross section show their respective locations.
Topography in the sections is shown shaded, bounded by a dashed line at the valley floor.
The solid bold line above the valley floor shows the average altitude across the gap “felt” by
the flow. Isentropes are at 2 K intervals and arrows indicate along-valley speeds upstream and
downstream of the gap (Adapted from Armi and Mayr 2007. Reproduced by permission of the
Royal Meteorological Society)
flowing layer on the upstream side is stably stratified as in the “Wood – single-layer
prototype” and is isolated from the surface and from the flow aloft by a neutrally
stratified (nearly) stagnant layer. The upper isolating layer is no longer at or below
crest but clearly above it. The gap-crest flow layer plunges, accelerates and thins.
However, the internal jump occurs only within the channel so that only the lowest
portion of the deep flow can be mixed into a neutrally stratified part; the largest part
of the flowing layer remains stably stratified downstream of the pass. Again, the
upper isolating layer thickens into a deep wedge keeping the flow aloft from being
affected by the underlying topography.
Before the gap flow becomes established, all three prototypes have lower
potential temperatures in the gap flow layer on the upstream side of the pass
than in the air mass on the downstream side (Mayr and Armi 2008). Figure 3.31
shows stages in the evolution of a particular gap flow case using a sounding on
the far upstream side of the pass and at the end of the channel. Prior to the onset
of northward gap flow (i.e. towards the viewer in the accompanying topography
Fig. 3.29), potential temperatures are higher on the southern side of the Alps than on
the northern side to above the main Alpine crest as seen in the difference between the
southern and northern radiosounding (center of Fig. 3.31). No northward gap flow
occurred.2 Changes on the synoptic scale brought a warmer air mass to the north of
the Alps and a colder one to the south so that potential temperature differences
(center of Fig. 3.31) are colder south of the pass to an altitude just above the
channel crest. Higher up in the channel to about 1 km above the main Alpine
crest, however, potential temperatures are still higher on the southern side, Gap
flow had commenced a few hours prior to the soundings. With the continuation
of the synoptic development, the air mass to the south kept getting colder between
2 and 4 km MSL, i.e. channel to 1 km above the main Alpine crest, whereas the air
mass to the north remained unchanged. Consequently, the gap flow layer deepened
and extended even above the main crest, causing hybrid gap flow with superimposed
cross-crest foehn flow.
It must be emphasized that the driving mechanism behind this gap flow event was
the air mass difference on both sides of the crest. During pure gap flow at times 3 and
4, the layer on top of the descending gap flow layer moved at speeds comparable to
and even higher than the gap flow layer on the upstream side. Yet it did not descend
but remained nearly horizontal since its potential temperatures were not low enough
to descend.3
A colder air mass upstream of the gap was also noted in other observations
of gap flow around the world. On the North American Pacific coast it is mostly
cold wintertime continental air contrasting with warmer air over the Pacific Ocean
(e.g. Cameron 1931; Overland 1984; Lackmann and Overland 1989; Colman and
Dierking 1992; Bond and Macklin 1993; Finnigan et al. 1994; Jackson and Steyn
1994a; Pan and Smith 1999; Colle and Mass 2000; Collier 2002; Sharp and Mass
2
Actually a gap flow in the opposite direction – southward – happened.
3
Spatial details were observed with two aircraft flights.
4
km MSL
Alpine
3 crest
5 4 3 2 1
6 5 43 2
1 channel 5
crest 6
2 1 2 43
6
deepest
cut
1
295 300 305 310 -10 -5 0 5 10 295 300 305 310
θ Milano Δθ = θ Milano - θ Innsbruck θ Innsbruck
Fig. 3.31 Changes in the air masses upstream and downstream of the Brenner Pass and their
effects on the flow through the pass. At time 3 the potential temperature was for the first time
lower to the south (center column) and very shallow gap flow occurred. The northward gap flow
became deeper (time 4) and changed into a deep hybrid flow composed of deep foehn (“downslope
windstorm”) and gap flow at time 5. At the end (curve 6), gap flow seized as the potential
temperature difference vanished (From Mayr and Armi 2008. Reproduced by permission of the
Royal Meteorological Society)
2004; Bond et al. 2006; Neiman et al.2006). Similarly, gap flows along the Adriatic
and Black Sea coasts are driven by a colder continental air mass compared to the
warmer air mass over the Mediterranean (e.g. Bajić 1989; Grubišić 1989; Glasnovic
and Jurčec 1990; Gohm and Mayr 2005; Gohm et al. 2008; Grisogono and Belušić
2009; Alpers et al. 2009).
Sometimes cold air over the North American continent surges southwards from
where it was formed and reaches as far south as the Central American Pacific Coast,
causing gap flows in Mexico (e.g. Hurd 1929; Schultz et al. 1997; Steenburgh et al.
1998). Over Japan, where gap flows are called “dashikaze”, temperature differences
might be caused by typhoons (e.g. Arakawa 1969; Ishii et al. 2007; Mashiko 2008).
Colder air advected across the ocean can cause gap flows, e.g. on the Falkland
Islands (Mobbs et al. 2005), through the Strait of Gibraltar (Scorer 1952; Dorman
et al.1995), and through the Rhone Valley (e.g. Pettre 1982; Jiang et al. 2003).
Different mechanisms lead to colder air upstream of gap flows in the interior of
a continent (cf. Sect. 3.5.3). Examples of such observations are in Seibert (1990),
Doran and Zhong (2000), and Mayr and Armi (2008).
In many of the gap wind cases cited in the preceding paragraphs, cold continental
air of the gap flow extends over the oceans. In this case, significant air-sea
interactions can result as sensible and latent heat fluxes from the warm water surface
alter the overlying air, and in turn the cold wind affects the oceans. Examples of the
impact of strong, cold low level flow on the ocean, in this case of the bora, which
often emerges as a gap-like flow onto the Adriatic, are discussed by Orlić et al.
(1994, 2006) who describe the complex response of the sea to strong bora winds.
Enger and Grisogono (1998) use two-dimensional simulations to demonstrate that
bora flow is extended over the warm sea due to the effect of the sea surface
temperature on reducing N which allows the flow to remain supercritical for a
greater distance by postponing the hydraulic jump. Gap flow over the Gulf of
Tehuantepec has been found to induce coastal upwelling, as well as a lowering of
sea surface temperature by as much as 8ı C (e.g. Schultz et al. 1997).
3.5.2 Theory
Linearizing the governing equations about a basic state has been successfully used
to (semi)analytically describe orographic flows. However, this approach fails to
describe gap flows as shown by Pan and Smith (1999). In linear theory (e.g. Queney
1948; Klemp and Lilly 1975; Smith 1979, 1980), vertically propagating gravity
waves cause lower pressure and accelerating winds on the lee slopes of a mountain
ridge; their strength is proportional to the height of the mountain. Accordingly, the
strongest pressure anomaly occurs along the lee slopes of the mountains on either
side of the gap and not in the gap itself even when confluence at the gap entrance
is included (Pan and Smith 1999; Zängl 2002c). Linearized gap flows are always
weaker than the flow down the adjacent crests, contrary to what is often observed.
Hydraulic theory, on the other hand, can describe gap flows. Its application
to mountain flows was pioneered by Prandtl (1942) and Long (1953, 1954). It
is fully nonlinear, which is an essential characteristic of asymmetric gap flows.
Steady state is assumed in hydraulic theory. Further assumptions are made about the
vertical structure of the air/fluid either by vertically averaging layers or by requiring
self-similarity. Self-similarity means that, at any position, the density and velocity
profiles can be scaled with the height of the moving stratified layer.
Hydraulics in the atmospheric science community is often taken synonymously
with its simplest, “single layer” case of one layer of constant vertical profile of
potential temperature (or density in case of fluids). The theory of the single layer
case is described in Sect. 3.2.3.
Gap flows frequently have stably and continuously stratified rather than ho-
mogeneous, neutrally stratified air masses on the upstream side, as was shown in
Fig. 3.32 Wood’s (1968) self-similar solution for stratified flow from a reservoir through a
contraction, i.e. a level gap. Upper panel: vertical section with isopycnals and the vertical density
profiles upstream and downstream of the contraction; lower panel: plan view. Flow from left to
right
the previous section. The single layer theory must therefore be expanded. Wood
(1968) could show that an analytical hydraulic solution exists for withdrawal from
a reservoir of fluid with a known, arbitrary stratification through a level confluent-
diffluent gap as shown in Fig. 3.32. Upon withdrawal, the uppermost and lowermost
streamlines will respectively drop down and rise up as indicated in the figure. If
the volume in these wedge-shaped regions is relatively small then the fluid filling
them will originate from a relatively small range of depth in the reservoir and
be thus of approximately constant density. This implies that the uppermost and
lowermost streamlines are (approximately) also lines of constant pressure. The
analytical solution is for the general case of arbitrary density distribution. Let 0
be the density at the uppermost streamline. The density increases by a total
to the lowermost streamline over the depth of the flowing layer. At any height
z, the density is thus 0 C ı. An analytical solution exists when the density and
velocity distributions are self-similar, i.e. when they scale with the depth of the
flowing layer
at each section. Then the density profile at all sections has the form
ı
F D F .ƒ/ where F() is determined by the far-upstream density
z
distribution. Exploiting the fact of constant pressure along the streamline and using
Bernoulli’s equation for a streamline in the flow, yields the velocity, v, at an arbitrary
height
"Z Z #
.0 C ı/ v2 1 1
D2 F .ƒ/ dƒ C F .ƒ/ F .ƒ/ F .ƒ/ dƒ 2G .ƒ/
g Œzt .1/ zt 0
(3.39)
where zt is the height of the uppermost streamline. For example, with a linear
stratification, F and G become F(ƒ) D 1 ƒ and 2G ./ D 3ƒ 4ƒ2 C ƒ3 ,
respectively, and the velocity distribution is parabolic-like. The analytical solution
could be reproduced in the laboratory (Armi and Williams 1993, their Fig. 5) and
is shown here in Fig. 3.28b. Note the continuous stratification far upstream in the
reservoir. As it is withdrawn through the gap, well-mixed neutrally stratified and
stagnant layers form above and below the flowing layer, which isolate it from the
remaining fluid. Whether both isolating layers form depends on the height range
over which fluid is withdrawn downstream of the gap. When it is withdrawn at
the bottom, an isolating layer forms only at the top. The atmospheric equivalent is
shown in observations and numerical simulations of a gap flow event through Juan
de Fuca Strait in Figs. 3.27b and c. Similarly, when the fluid is withdrawn from the
top of the channel, an isolating layer forms underneath only. And when the fluid
is withdrawn from the center, both upper and lower isolating layers form as shown
in Fig. 3.28b.
For a particular flow rate, a range of densities 0 to 0 C flows from
the reservoir. As the flow rate changes, the range of densities drawn from the
reservoir changes so that the flow at the narrowest section of the gap remains critical
with respect to the longest wave mode. Note that a continuously varying density
stratification results in an infinite number of wave modes. Thus there is an infinite
number of sections which act as virtual controls in addition to the physical control of
the narrowest cross section. Benjamin (1981), in a generalization of Wood (1968),
showed that self-similar flows will tend to be realized whenever fluid is withdrawn
from a continuously stratified reservoir. With a sufficiently high extraction rate
downstream, the resulting flow is critical at the narrowest section and supercritical
downstream of it. Even the longest (and fastest) wave mode is swept downstream in
supercritical regions of the flow.
The equivalent to a reservoir of continuously stratified fluid in the atmosphere
is a different air mass on the upstream side of the mountain range. The withdrawal
of that air through a gap is started when a relatively warmer air mass is on the
downstream side. The pressure gradient force between the different air masses
substitutes for the pump, which withdraws fluid in the laboratory experiments. At
what vertical range that pressure gradient and air mass difference exist, determines
where the isolating neutral layers will form. In the case of the Alps (Wipp Valley), it
was mid-level since both upper and lower isolating layers were observed (Fig. 3.30
middle and bottom).
Upstream b0 F 22 = 2 Upstream b0 F12 = 2

2 1
layer 1
or
layer 2 or
Lower layer dominant Upper layer dominant

downstream downstream
Upstream bv Crest F 22 = 2 Upstream bv Crest F12 = 2
Lower layer dominant Upper layer dominant

downstream downstream
Fig. 3.33 Two-layer hydraulic flow across a pass (lateral and vertical contraction). Upper panel:
low to moderate flow rates with flow controlled at the narrowest part of the gap. Lower panel: high
flow rates with supercritical flow at the narrowest section. When the lower layer dominates (left
column) it remains supercritical downstream of the gap center and vice versa for an dominating
upper layer (right column). In the figures Fi is the Froude number in layer i, b0 is the location of the
primary control for the flow at the top of the hill, bv is the virtual control where the Froude number
becomes unity (Adapted from Armi and Riemenschneider 2008. Reproduced by permission of
Cambridge University Press)
Atmospheric gap flows can consist of more than one dynamically active layer.
For two layers of comparable depths and flow rates, two unique controlled asym-
metric flows are often possible (e.g. Armi and Riemenschneider 2008). They are
determined by the difference in the fluid reservoirs (or air masses in case of the
atmosphere) upstream and downstream of the gap. For relatively low flow rates,
either the lower or upper layer can be active layers. As flow rates increase, there
may no longer be the possibility of the two asymmetric solutions satisfying critical
conditions at the topographic control, the narrowest section of the (level) gap.
Instead, they merge upstream and the control moves upstream of the narrowest
section as a “virtual” control. An additional possibility of two-layer flows are
“exchange flows” with flow in opposite direction in each layer (cf. Baines 1995,
Sect. 3.11).
As was seen in Sect. 3.5.1 on observations of gap flows, the contraction in
a gap can be only lateral (“level gap”), only vertical (constant width but bottom
rises; “sill”), or a combination of lateral and vertical (“pass”). For two layer flow,
a level gap directly contacts both layers. A sill, on the other hand, only directly
contacts the bottom layer, and the response of the upper layer is indirect. Armi
and Riemenschneider (2008) found analytical solutions for two-layer flows over
a pass. Four different flow configurations are summarized in Fig. 3.33. When the
lower layer dominates (upper left), it accelerates as the interface between the two
layers first rises slightly but then descends even upstream of the crest. Compared to
a single-layer flow, the presence of the upper layer keeps the interface higher at the
center of the gap and also increases the flow rate, but both only by less than 5%. Such
a small change is currently not detectable observationally in the atmosphere. The
flow in the upper layer can even be in the opposite direction. When the upper layer
dominates (upper right part of Fig. 3.33), the vertical contraction part of the pass is
only communicated indirectly via the dynamics of the lower layer. The upper layer
depth decreases as it accelerates towards the pass and becomes supercritical. As
flow rates increase, only supercritical conditions can exist at the gap center and the
transition from the subcritical conditions in the reservoir occurs further upstream at
the virtual control (lower panel in Fig. 3.33). On its way towards the pass, the upper
layer accelerates more than the lower layer (Armi and Riemenschneider 2008).
Whether upper or lower layer dominate and remain supercritical downstream of
the gap center, depends on the matching downstream conditions.
Modeling of gap flows has spanned the whole spectrum from idealized to realistic
4D-simulations. Shallow water models are used with surprising success along with
non-hydrostatic numerical weather prediction models. Forecasting gap flows has
not attracted much research interest yet but well-established practices exist in
forecasting offices which have gap flows in their forecast region. An ingredients-
based method for forecasting gap flows is presented.
3.5.3.1 Models
The most basic models to simulate gap flows are shallow water models, which
numerically solve the single-layer hydraulic flow equations for one or two well-
mixed layers. Pan and Smith (1999) used such a model developed by Schär and
Smith (1993) to study gap flows at Unimak Island southwest of the Alaskan
Peninsula but had limited data with which to initialize and compare their model.
Jackson and Steyn (1994b) used a different hydraulic model, which also included
friction. Flow through Howe Sound could be reproduced. Horizontal pressure
gradients and turbulent friction were dominant in a force-balance analysis.
The same model as in Pan and Smith (1999) was used for more intensively
observed gap flows in the Wipp Valley with data from the Mesoscale Alpine
Programme (Gohm and Mayr 2004) and for a case study of bora flow at the Adriatic
coast (Gohm et al. 2008). Systematic parameter studies with the single-layer shallow
water model for the Brenner Pass – Wipp Valley by Gohm and Mayr (2004)
emphasized the dominant role of small-scale topographical features for details of
the gap flow. Ridges protruding into the valley downstream of the pass, a widening
or a turning of the valley can all be locations where supercritical flow returns
to a subcritical state in a dissipative, turbulent hydraulic jump. For the Brenner

Pass geometry, the vertical constriction rather than the lateral constriction was
found to be largely controlling the flow. A comparison with detailed observations
showed that the shallow water simulations included all basic features of the
gap flows.
A similarly successful replication was achieved for a bora flow at the Adriatic
coast (Gohm et al. 2008), where the gap flow jets and hydraulic jumps were
captured. At first, it is surprising that a single layer model can reproduce the essence
of gap flows in realistic settings and continuous upstream stratification – a deviation
from the neutral stratification in the numerical model. As Wood (1968) showed in
his analytical solution discussed in Sect. 3.5.2 and as Armi and Williams (1993)
showed in the laboratory, well-mixed nearly stagnant layers isolate the gap flow
from the flow aloft so that disregarding the effects of the flow aloft as is done in the
shallow water model has only small adverse affects on the simulations. Furthermore,
Durran and Klemp (1987, cf. their Fig. 8) and Farmer and Armi (1999, their Fig. 3)
showed that a continuously stratified flow over a ridge (vertical constriction), despite
its continuous stratification is qualitatively similar to a single-layer hydraulic flow,
especially when the depth of the continuously stratified system is less than one half
of its vertical wavelength (Durran and Klemp 1987). The third reason for the success
of the simple model is that turbulence near the surface, in hydraulic jump(s) and at
the upper interface of the gap flow mixes shallower gap flows from a continuous to
a neutral stratification.
Numerical simulations solving the complete set of nonlinear governing equations
have been used in varying degrees of idealization for the simulations of gap
flows. Idealization of both topography and flow/air masses will be considered first,
followed by idealization of only flow/air mass, and finally fully realistic flows.
These fully non-linear simulations with mesocale numerical models confirmed
turbulent friction identified in the hydraulic modeling of Jackson and Steyn (1994b)
as an additional key parameter for gap flows, besides the shape of the underlying
topography. Gap flow depths are of comparable magnitude to the depth of the
planetary boundary layer. Highly idealized simulations with a geostrophically
balanced frictionless background with uniform static stability leave the gap flow
direction dependent on the shape of the topography, which is not supported by
observations. For the balanced background flow parallel to an infinite ridge Sprenger
and Schär (2001) and Zängl (2002a) found flow towards lower pressure. An
isolated ridge, however, reverses the direction (Zängl 2002a). It splits the flow
on its upwind edge, leading to anticyclonic flow on the left side (looking in flow
direction) of the isolated ridge. This flow has an ageostrophic component directed
towards the mountain, thus counteracting and slightly overpowering the larger-scale
geostrophic pressure gradient. The addition of friction turns the low-level wind
counterclockwise compared to the flow aloft, i.e. towards the region of lower large-
scale pressure (Zängl 2002a; Gaberšek and Durran 2006). Gaberšek and Durran
(2006) quantified the importance of friction with volume-averaged momentum-flux
budgets.
Turbulent friction in the boundary layer produces filaments of anomalous

potential vorticity, called potential vorticity banners, at the side walls of the gap
(Ross and Vosper 2003), which are then advected further downstream. They can
be found at large distances away from the gap and can also occur in radiatively
driven downvalley flows with a lateral constriction at the terminus of the valley
(Zängl 2004).
Including the realistic topography of the Alps but keeping the geostrophically
(except in the boundary layer) balanced initial flow fields with constant static
stability throughout the troposphere, Zängl (2003) found 3D-propagation of gravity
waves excited over adjacent ridges to be crucial for the acceleration of a gap flow in
the Wipp Valley. The initial conditions, however, strongly deviate from the typical
situation of two different air masses of limited vertical extent on either side of the
gap. The limited depth leads to the self-isolation of the gap flow from the flow aloft
(cf. Sect. 3.5.2).
Including realistic initial conditions is one of the main difficulties for the
numerical modeling of realistic gap flows. Detailed observations, especially vertical
profiles in the proximity of gaps are typically sparse or none-existent. Even if
they exist, making full use of them is difficult since more sophisticated methods
than nudging for assimilating such data into mesoscale models (e.g. 4D-VAR,
ensemble Kalman filters) are just emerging. As a result, the initial state of the
mesoscale models is still dominated by the fields from global models, into which
the mesoscale models are nested. The horizontal resolution of global forecasting
models is currently (2010) at 16 km or coarser, so that model topography differs
significantly from the real topography (e.g. the wider part of the valley south
of Brenner Pass is about 0.7 km below the ECMWF T1279 topography). Initial
conditions of the models therefore lack the sharp potential temperature steps on top
of the gap flow layer, which occur often within only dozens or hundreds of meters.
The smoothing introduced by the numerics during model integration aggravates the
situation. In contrast, the sharp separation can easily be introduced to the much
simpler hydraulic models.
Gohm et al. (2004) compared model simulations of a gap flow and foehn event
in the Wipp Valley with the rich data set of MAP. They attributed considerable
discrepancies at the beginning of the event to deficiencies in the model profile
upstream of the pass, where the flowing layer (and consequently the potential
temperature step at its top) was too shallow. Given that the height of the gap flow
layer on the upstream side, the mass flux through the gap, and the “withdrawal” on
the downstream side are all related (cf. Sect. 3.5.2), a combination of deviations of
the model from the real atmosphere both upstream and downstream might have been
the cause. Later in the event, too much mass flowed across the pass and the gap flow
layer did not descend as much as observed downstream of the pass. However, most
of the prominent observable gap flow features like hydraulic jumps and cross-valley
asymmetries of the flow are so much dominated by the underlying topography, that
they were still found in the simulations. The simulation of a level gap flow event
through the Strait of Juan de Fuca (cf. Fig. 3.27) also found cross-gap asymmetries
caused by flow into the channel over one of its sides. One of the main advantages
of the fully realistic simulations is the ability to obtain a complete (time-dependent)
three-dimensional depiction of gap flows (Jackson and Steyn 1994a).
Topography must be sufficiently resolved in the model for the correct amount
of air to be withdrawn from the upstream side. Horizontal mesh widths for Wipp
Valley simulations were as fine as 0.27 km in the innermost model domain and
0.8 km in the domain that spans both sides of the gap in Gohm et al. (2004) and
Zängl et al. (2004). Reducing the 0.8 km grid spacing to 0.4 km further improved the
results (Zängl 2006). A mesh of at least approximately 1 km was also needed for gap
flow through the Columbia Gorge (Sharp and Mass 2002), with 0.44 km horizontal
spacing also improving the results.
When the flow (and the different air masses) are deep enough to “flood” the
main crest into which a gap or pass is cut, downslope windstorms (or deep foehns
as they are called in the Alps) are superimposed on the gap flows. Gap flows play
then a relatively minor role in the total mass transport but otherwise their main
characteristics remain similar (e.g. Colle and Mass 1998; Gohm et al. 2004; Zängl
et al. 2004).
Numerics and turbulence schemes are further challenges in correctly simulating
gap flows with numerical models. Reinecke and Durran (2009b) in simulating
mountain waves showed that second-order finite differencing schemes overam-
plified a standing mountain wave by about a third compared to fourth-order
schemes within the nonhydrostatic regime with typically used horizontal resolutions
of the mountain. Applied to the American Sierra Nevada, this translated to an
overestimation of downslope speeds at the surface of 20 m s1 and of vertical
velocity maxima by 10 m s1 (as seen in the study of Reinecke and Durran 2009b,
see Fig. 9.18 in Chap. 9).
Most numerical simulations of flows over mountains until recently were per-
formed using second-order or third-order schemes. The large mountain waves and
strong rebounds (often termed “wave breaking”) seen in Fig. 9.18 were common
features of these simulations and also part of a scientific dispute over flow over an
oceanic sill, where observations (Farmer and Armi 1999, 2001; Armi and Farmer
2002) could not be reconciled with numerical simulations (Afanasyev and Peltier
2001a, b). A large part of the discrepancy occurred at the top of the flowing layer,
where small scale entrainment was observed but wave breaking on a much larger
scale was modeled. Small scale entrainment was also found to be the important
mechanism at the top of a stratified flow down a slope in the laboratory (Baines
2001, 2005). Gohm et al. (2008) found a strong sensitivity to the turbulence scheme
of the top of the gap flow and the layer immediately above. An improvement of these
schemes seems to be needed. At the ground, the height of the lowest model surface
above the ground had a surprisingly large impact in sensitivity studies of gap flow
reported in Zängl et al. (2008), even more so than the type of PBL scheme.
Essential for the correct simulation of gap flows and the interactions between
the different air masses involved is that horizontal diffusion of the mass field is
truly computed horizontally, not along the terrain following coordinate surfaces
(Zängl 2002b).
3.5.4 Forecasting
Forecasting methods for gap flows have been developed in local forecasting offices
but have not yet progressed into easily accessible peer-reviewed literature. They
are typically empirical and statistically based without establishing a direct physical
connection between predictor and predictand. Some combine statistics and the
fundamental characteristics of gap flows. An example is Drechsel and Mayr (2008),
who use the frequency distribution of potential air mass difference on a model
surface and reduced pressure difference across the barrier to probabilistically
forecast the binary event of gap flow – yes or no.
Following Doswell et al. (1996) an “ingredients-based” method is proposed here,
which uses the physical understanding of gap flows. It is not regionally dependent
(gap flows occur all over the globe) but flexible enough to be easily adaptable
to local forecasting demands and to incorporate new observational or modeling
advances as they become available.
Gap flows occur due to different air masses on either side of a gap. The
resulting pressure gradient is the main driving force. The altitude range of the
air mass difference will determine the depth of the gap flow. Most forecasting
applications will be interested in gap flows reaching the surface downstream of
a gap. Unless the potential temperature of the air mass upstream of the gap is
lower or at most the same as the potential temperature of the air mass down-
stream, an asymmetric gap flow with higher wind speeds downstream will not
happen.
The ingredients to create different air masses on either side of the gap will
be examined progressing from large to small scales. On the large scale is the
quasi-horizontal differential advection of air masses. Synoptic-scale systems may
transport different air masses to opposite sides of the gap. Examples are cold fronts
and subsequent mid-level troughs typically co-located with the coldest air passing
over only one side of the obstacle (Fig. 3.34a). A complementary situation is that of
warm fronts and mid-level ridges being confined to only one side. A combination of
these two is also possible.
Another mechanism on the large scale is air mass formation under anticyclones.
Different surface characteristics on either side of the gap lead (diabatically) to
different air masses (Fig. 3.34b). An example is the wintertime formation of cold
air over the continent, which is separated by mountain ranges from relatively warm
air over the ocean. Well-known regions are the Pacific Northwest and the Adriatic
coast. Ocean-land contrasts also lead to large differences over the Falkland Islands
and create a particularly large potential temperature step on top of the flowing layer
(Mobbs et al. 2005).
Advective and diabatic processes are also at work on the mesoscale to create
sufficiently lower potential temperatures at one side of the gap to cause gap flows.
Lower pressure further downstream of the gap can lead to ageostrophic flow of
air towards the low pressure. Compensating subsidence with adiabatic temperature
increase might be sufficient to raise potential temperatures higher than the ones
a b
H
F 500
c d
q
Fig. 3.34 Forecasting gap flows: processes creating lower potential temperatures in the flowing
layer upstream than near the surface downstream. (a) Large-scale differential advection from a
cold front and upper level trough followed by a warm front. (b) Formation of air masses under
a high pressure system (cold wintertime continent, warmer ocean). (c) Ageostrophic drainage of
cold air towards low pressure caused by flow over the larger mountain range into which the gap
is embedded. (d) Differential heating caused by upstream cloud-cover and (nearly) clear skies
downstream
upstream of the gap. Causes for lower pressure further away from the gap can be an
approaching cyclone or the topographic low pressure caused by flow over the whole
mountain range, into which the gap is embedded (Fig. 3.34c).
Low and/or mid-level cloud cover on one side of the gap with nearly clear skies
on the other side will lead to much stronger diurnal warming near the surface on
the clear side of the gap, raising potential temperatures above the ones on the
cloudy side (Fig. 3.34d). Consequently, many gap flow regions have a climatological
maximum in the afternoon. At nighttime, however, the radiative cooling can lower
potential temperatures sufficiently to prevent the gap flow from reaching the ground.
Differential heating between a valley atmosphere and the adjacent (flat) forelands
will create valley wind systems (see Chap. 2). Given sufficient heating difference
and a pronounced lateral constriction, e.g. at the exit of the valley, an asymmetric
gap flow with strong acceleration downstream of the constriction can develop. It is
most pronounced for nocturnal downvalley flow and has been observed and modeled
for several Alpine valleys draining into the forelands (Müller et al.1984; Zängl 2004,
and references therein, and Chap. 2 in this volume).
Finally, at the meso-gamma scale, small-scale topographical features might
exclude gap flow from small regions. A kink in the terrain will lead to flow
separation when the potential temperature is similar to the one further underneath
(i.e. nearly neutrally stratified). Lateral flow separation will keep some regions in
the wake of the flow.
Shear-induced turbulence from gap flow shooting over a cold pool downstream
of the gap acts to increase the potential temperature step across the interface and
the thickness of the interface by mixing in air of higher potential temperature
from aloft and lower potential temperature from the cold pool, respectively. The
thickness of the induced turbulent eddies depends on the thickness of the shear
layer and the stratification of the flow, as expressed by a (gradient) Richardson
number. These eddies pair with neighboring eddies to form larger vortices (Koop
and Browand 1979), which increase the vertical extent of the mixing region at
the interface several-fold (Koop and Browand 1979, Fig. 8). When the cold pool
is shallow enough to be comparable to the size of the shear-induced eddies, these
eddies and thus the gap flow can penetrate down to the surface and cause a sudden
temperature increase there. The details are sensitive to the value of the Richardson
number. Currently no routine observational methods exist to be able to exploit this
“turbulence ingredient” for forecasting the break-in of gap flow through a cold pool.
Observations and NWP forecasts still need to be improved to make the best
use of such a conceptual, ingredients-based forecasting method. Vertical profiles
of air masses on either side of the gap are observed in very few locations of the
world only. The increase in the size of automatic weather station networks has
put more stations into exposed locations like mountain tops, which provide coarse
information about the vertical structure of the air mass. Vertical structures in NWP
analysis and forecast products are strongly smoothed versions of the actual profile.
However, sharp potential temperature steps over a few decameters only and well-
mixed layers of a few hundred meters and their exact location are crucial for gap
flow dynamics.
An upper limit to the increase of speed from a station in the bottom part of the
flowing layer upstream of the gap, vu , and a downstream station, vd can be computed
from the potential temperature difference across the top of the flowing layer, ™,
and how far the top of the layer has descended between upstream and downstream
locations (e.g. Armi and Mayr 2007), h, which yields a pressure difference p:

p D g h (3.40)
0
Alternatively the pressure difference can be taken from weather stations (as long
as they are in the flowing layers, see below) but have to be reduced to a common
altitude (cf. Mayr et al. 2002, for a discussion of reduction methods). With the
Bernoulli equation along a streamsurface:
p
v2d D v2u 2 (3.41)

This is an upper limit since it does not include the effects of surface friction, hy-
draulic jumps and the fact that the stations might not lie on the same streamsurface.
In many locations, the only diagnostic tool available is pressure observations
from valley/plain stations. Since pressure is an integral measure of the temperature
profile aloft, the pressure difference across the gap can suffice for diagnosing (and
forecasting) air mass differences and gap flows. Caution must be taken to account
for blocked layers underneath gap flows. Especially upstream of passes they can be
of substantial depth. As they do not flow across the pass, their contribution to the
pressure difference needs to be discounted. A decrease of temperature by 3 K over
a depth of 1 km increases pressure by 1 hPa.
3.6 Barrier Jets
A barrier jet is an elevated wind maxima on the windward side of a mountain

barrier, blowing parallel to the barrier. Barrier jets can occur when stably stratified
flow approaches an extra-tropical mountain barrier and is blocked by the barrier
for hours to days or longer. The blocked flow is forced to move up the barrier, and
as the low level cool air is forced against buoyancy to rise, it cools adiabatically
and creates higher pressure along the slope. The higher pressure against the barrier
acts to decelerate and block the flow. With enough time, geostrophic adjustment
can occur between the flow and the high pressure along the slope. This adjustment
causes the flow to turn left (right) in the northern (southern) hemisphere, so that the
high pressure is to the right (left) of the flow. Turbulent friction reduces the wind
near the surface, and thermal wind considerations imply the strongest jet will lie
above the largest horizontal temperature gradient. The general structure of a barrier
jet is shown in Fig. 3.35.
Fig. 3.35 Schematic diagram showing a barrier jet as a low-level flow of strong winds (under
the dashed line) parallel to the mountain barrier (flow is into the page). The flow is formed
by the upslope movement of stratified air, increasing the pressure along the slope so that the
pressure gradient force is balanced by the Coriolis force in the cross-barrier direction. The flow
is ageostrophic in the along-barrier direction, where an antitriptic balance forms with the along-
barrier pressure gradient balanced by friction and inertia. H is the mountain height, L is the
mountain half-width, and Lr is the internal Rossby radius of deformation (Lr D NH/f)
3.6.1 Observations
The across-barrier blocking, turning and acceleration of the wind has been doc-
umented in many coastal and inland mountain regions such as: north of the
Brooks range in Alaska (Schwerdtfeger 1974) and southeast Alaska (Lackmann and
Overland 1989; Overland and Bond 1993; Loescher et al. 2006; Olson et al. 2007),
southwest British Columbia (Overland and Bond 1995; Doyle and Bond 2001),
Washington State (Mass and Ferber 1990), Oregon (Braun et al. 1997), California
(Doyle 1997; Yu and Smull 2000), east of the Appalachian and Rocky Mountains
(Bell and Bosart 1988; Colle and Mass 1995), Taiwan (Li and Chen 1998; Yeh
and Chen 2003), the Colorado Front Range (Marwitz and Toth 1993), windward
of the Sierra Nevada (Parish 1982), southern Norway (Økland 1990), Antarctica
(Schwerdtfeger 1975), and as a related phenomenon, the Southerly Buster in
southern Australia (Baines 1980). A recent field experiment – the Southeast Alaska
Regional Jets (SARJET) experiment (Winstead et al. 2006) – completed over
the Gulf of Alaska in September and October of 2004 has provided new data
(aircraft and Synthetic Aperture Radar) and modeling results that have contributed
significantly to better understanding Barrier Jets.
Barrier Jets can be categorized into a few broad groups:
Classic barrier jets. In this case the onshore flow of stratified marine air onto a
western coast is blocked by coastal mountains, resulting in wind turning into an
along-barrier direction as a barrier jet near and upwind of the coastal mountains.
These conditions: stratified air impinging on a flow-perpendicular mountain barrier,
are common along mountainous coastlines on the western edges of continents –
especially of North and South America. During winter, west coasts at mid-latitudes
can be characterized by the regular passage of frontal systems across the coastal
mountains often resulting in the onshore flow of stable air ahead of a surface
warm front. The onshore (across-barrier) flow of cool marine air below the frontal
inversion can be blocked by the coastal mountains, decelerate in the across-barrier
direction and deepen, thereby creating a high pressure ridge along the mountain
barrier. In the northern hemisphere, the onshore flow turns to the left along the
barrier, both in geostrophic adjustment to this pressure ridge, and in response to
decreased along-barrier Coriolis force as the onshore wind speed drops (Smith
1979). The enhanced across-barrier pressure gradient can result in a barrier jet in
the coastal zone (e.g. Schwerdtfeger 1979; Parish 1982, 1983; Li and Chen 1998;
Overland and Bond 1993, 1995).
In a numerical simulation of a coastal jet that formed ahead of a land-falling cold
front approaching the northern California coast, Colle et al. (2002) found the block-
ing of the onshore flow resulted in coastal pressure ridging and terrain-enhanced
southerly winds that exceeded 25 m s1 (90 km h1 ). Considerable along-coast
variability in the jet was noted, with reductions in speed downwind of Cape Blanco
and Mendocino and accelerations adjacent to higher coastal topography. Numerical
model diagnostics suggest that the low level blocking of flow impinging on coastal
mountain ranges is often important for generating ageostrophic coastal jets in

California and elsewhere (e.g., Doyle 1997). Essentially, this blocking contributes
to pressure ridging upwind of the mountain barrier which then adjusts to form
an along-barrier jet. The winds in the model jet were enhanced by 45% due to
topographic blocking and peak winds of 22 m s1 were simulated (Doyle 1997).
Olson et al. (2007) discuss the structure of various barrier jets observed in southeast
Alaska during the SARJET experiment (Winstead et al. 2006), and found that
classical jets had maximum winds over 30 m s1 at the coast between 600 and
800 m above sea level, extending approximately 60 km offshore. An illustration of
a Classic Barrier Jet, over the waters of Southwest Alaska using Synthetic Aperture
Radar (SAR) is shown in Fig. 3.36a (Loescher et al. 2006). Aircraft observations
and MM5 simulations from SARJET of a classic barrier jet from Olson et al. (2007)
are shown in Fig. 3.37a and b
Hybrid barrier jets. In regions of complex terrain, such as the Pacific Northwest
region of North America, interactions between different orographically forced
mesoscale flows often occur. For example, the source of cold air along the windward
slopes of North American west coast mountains can be gap flow exiting coastal
valleys which is turned poleward by the Coriolis force as the flow undergoes
geostrophic adjustment upon exiting the gap. This is the so-called “hybrid” barrier
jet that is reported by Doyle and Bond (2001) and Loescher et al. (2006) who
discuss the similar cases of high-latitude (strong f) gap flows in British Columbia
and southeast Alaska where gap flows exiting onto a mountainous coastline develop
characteristics of terrain-parallel barrier jets. This is illustrated over southeast
Alaskan waters using SAR imagery in Fig. 3.36c (Loescher et al. 2006).
The geostrophic adjustment of a low-latitude gap flow was modeled by
Steenburgh et al. (1998). They found that the adjustment curvature was determined
by an inertial circle – in their case due to weak Coriolis forcing, the gap flow
never became a jet, however at higher latitudes since the inertial circle radius is
inversely proportional to the Coriolis parameter, the gap flow can turn rapidly
poleward, merging with the synoptic flow becoming a barrier jet. Overland and
Bond (1995) suggest that cold air supplied by gap flow from the Strait of Juan de
Fuca that separates Vancouver Island from the Olympic Peninsula on the west coast
of North America, enhanced the barrier jet observed during the Coastal Observation
and Simulation with Topography Experiment (COAST). Numerical modeling
experiments by Doyle and Bond (2001) indicate that the cold air supplied by
the gap flow was as important as the topography of Vancouver Island in enhancing
the barrier wind. Olson et al. (2007) in an analysis of barrier jets over southeast
Alaska found that hybrid jets with a maximum speed of approximately 30 m s1 at
500 m above sea level were displaced 30–40 km offshore (in contrast to classical
jets which had maximum winds over the coastline). Aircraft observations and MM5
simulations from SARJET of a hybrid barrier jet from Olson et al. (2007) are shown
in Fig. 3.37c and d. A conceptual model of the hybrid barrier jet over southeastern
Alaskan waters is shown in Fig. 3.38d (Olson et al. 2007).
Fig. 3.36 Synthetic Aperture Radar imagery of Barrier Jet surface winds over the coastal waters
of southwest Alaska (Source: Loescher et al. 2006. © American Meteorological Society. Reprinted
with permission)
Fig. 3.37 Vertical cross sections across the width of barrier jets from IOP1 – classical barrier jet
(panels a, b) and IOP7 – hybrid barrier jet (panels c, d) of the SARJET experiment in southeastern
Alaska. Panels (a) and (c) are from aircraft observations and panels (b) and (d) are from MM5
simulations (Figures from Olson et al. 2007, Figs. 6c, d and 12c, d. © American Meteorological
Society. Reprinted with permission)
Cold Air Damming. In situations where flow impinges on an eastward-facing

mountain barrier, cold air damming can occur if cold air is present to the north
(in the northern hemisphere). In this situation, the mountain may block the low-
level flow, creating higher pressure along the mountain slope and northerly flow
in which the across barrier pressure gradient is balanced by the Coriolis force and
there is an antitriptic balance between friction, inertia and the pressure gradient
in the along barrier direction so as cold air is advected to the south (northern
hemisphere). The jet is effectively trapped against the topography by the Coriolis
force and high static stability at the top of the cold air. These events have been
documented east of the Appalachian Mountains (e.g. Bell and Bosart 1988), east of
the Rocky Mountains (Colle and Mass 1995), among other locations. Marwitz and
Toth (1993) in a Cold Air Damming event along the Colorado Front Range, found
a northerly barrier jet formed by upslope flow with adiabatic cooling enhanced by
cooling from evaporation and melting of falling precipitation. The jets formed by
cold air damming on the eastern slopes of mountain barriers which advect cold air
Fig. 3.38 Conceptual model of a hybrid barrier jet over the coast of southeastern Alaska. Thick
arrows represent the gap flow, medium arrows the 1,000–1,500 m flow and thin arrows the
low-level flow. T represents temperature anomalies (Source: Olson et al. 2007. © American
southward, are somewhat analogous to the hybrid jets along the western slopes of
mountains discussed above. In the case of hybrid jets, the cold air is supplied by gap
flow of cold continental air through mainland inlets and the barrier jet advects the
cold air poleward, while in cold air damming events the cold air originates on the
barrier-jet side of the mountains and the cold air is advected equatorward.
The “Southerly Buster” of southeastern Australia is an along-barrier jet resulting
from the interaction of a midlatitude frontal system with the coastal orography. In
this case, a relatively shallow cold front approaches southeastern Australia from the
ocean and transitions into a barrier jet-type flow as the front is partially blocked
on the inland side of the mountains but accelerates along the coast of New South
Wales as the upper disturbance moves eastward across the region. A rapidly moving
along-coastal barrier jet may form, whose leading edge (the front, which resembles
a topographically trapped gravity current) is often marked by a roll cloud over the
coastal Tasman Sea (Baines 1980). Since this leading edge separates the preceding
warm offshore flow by the cool along-barrier jet, the local term “Southerly Buster”
has arisen. The situation is in some ways analogous to the Cold Air Damming
situation in North America discussed above. However there are also significant
differences since the Australian landmass ends at mid-latitudes, the cold air source
region is the cool ocean south of Australia rather than the continent, the westerly
jet in the region contributes to the Southerly Buster, and the topography of the
region is low compared to that of North America. Studies with mesoscale models
(e.g., Howells and Kuo 1988; McInnes and McBride 1993; Reid and Leslie 1999)
have attempted to isolate the various roles of season, time of day, topography,
surface friction and heating in the evolution of the Buster. The Southerly Buster is
similar to the orographic coastal jets observed and modeled over central California
by Doyle (1997).
Loescher et al. (2006) used Synthetic Aperture Radar (SAR) imagery from the
Gulf of Alaska to document the spatial and temporal variability of barrier jets
over the ocean in that region. SAR detects centimeter-scale capillary waves on
the water surface that, with some knowledge of wind direction, can be related to
wind speed. Therefore SAR imagery gives high resolution spatial images of wind
speeds over water. Figure 3.36 documents the different categories of barrier jets that
they found and suggested. Figure 3.36a shows a classic barrier jet – the flow has
an onshore component with wind speeds higher than 20 m s1 near the coastline.
Figure 3.36c depicts a hybrid barrier jet – the flow has an offshore component, and
offshore gap flow can be observed exiting the mainland fjords, which undergoes
rapid geostrophic adjustment into a coast-parallel direction upon exiting the gap.
In Fig. 3.36d, a pure gap flow case, there is no turning of the wind upon exiting
the gaps, therefore no barrier jet forms. They found that about 1/3 of all barrier
jets were observed to have a sharp transition between the barrier jet flow and the
weaker ambient flow, and called these “shock jets” – Fig. 3.36e. This form was
associated with anomalously cold, dry air over the continent (Colle et al. 2006).
Variable jets were observed 23% of the time (Fig. 3.36f) in which the along-barrier
flow was found to vary considerably along the coast – this was associated with more
convective conditions in which it was hypothesized that higher-momentum air from
aloft was transferred locally to the surface via convection. Loescher et al. (2006)
observed more barrier jets in the cool season than in the spring and summer. This
was true of all barrier jets, but especially so for hybrid jets. This is expected since
the gap flow component of hybrid barrier jets occurs mainly during the cold season
when conditions exist for a large offshore-directed low-level pressure gradient:
there is a strong contrast between the cold continental air favoring high surface
pressure and the warmer coastal air favoring lower surface pressure. They also
found a close association between percent occurrence of barrier jets and the coastal
orography, with the most frequent occurrence of barrier jets associated with the
highest topography within 100 km of the coast – this is illustrated in Fig. 3.39.
The median speed was found to be 20 m s1 for both classic and hybrid jets. The
median width, measured from the base of the terrain to the outer edge of the barrier
jet was found to be 50 km for classic jets and 60 km for hybrid jets. They found
that some hybrid jets detached from the coastline over part of their length, often by
about 10 km. Olson et al. (2007) note that barrier jets over southeast Alaska reach a
maximum strength when the onshore flow is oriented approximately 40ı from coast
parallel suggesting the influence of ambient wind direction on barrier jet intensity
and evolution.
At this point, it is useful to point out that not all jets blowing parallel to mountain
barriers are barrier jets. Coastally Trapped Wind Reversals (CTWRs) (also called
Coastally Trapped Disturbances) are a type of coastal jet that has been extensively
Fig. 3.39 Maximum terrain height found within 100 km of the Gulf of Alaska coastline compared
with the frequency of barrier jets (From Loescher et al. 2006. © American Meteorological Society.
studied along the West Coast of North America from California to British Columbia
(e.g. Reason and Steyn 1992; Bond et al. 1996; Rogers et al. 1998; Jackson et al.
1999; Nuss et al. 2000; Reason et al. 2001; and others). While there are some
apparent similarities – both phenomena result in southerly flow within a Rossby
radius of the coastal mountains – the initiation mechanism is quite different. Rather
than being forced by the onshore flow of stratified air, CTWRs are initiated by
offshore (downslope) flow at mountain crest level to the north, and the creation of
lower surface pressure offshore to the north of the disturbance, creating an along
barrier pressure gradient (Mass and Bond 1996). CTWRs also propagate from
south to north along the coast as surface ridges and are primarily a warm-season
phenomenon.
3.6.2 Theory
In its simplest form, at steady-state, a barrier jet is semi-geostrophic: it is

in geostrophic balance with the across-barrier pressure gradient, while in the
along-barrier direction, an antitriptic balance exists, in which the pressure gradient
force is balanced by inertia and friction. However this simple force balance does not
adequately address the normal situation when the atmosphere is not in steady state,
nor does it tell us much about the mechanisms or conditions under which barrier jets
form. Scale analysis of the governing equations (e.g. Smith 1979; Pierrehumbert
and Wyman 1985; Overland and Bond 1993, 1995) has characterized the general
dynamical nature of stratified flow impinging on a mountain barrier, and can help
us understand the conditions under which barrier jets form even though these
studies neglect complicating factors such as time variation, mesoscale topographic
variation, friction, and diabatic effects.
Important atmospheric properties that determine the amount of blocking and
the distance away from the barrier that blocking will occur are determined by
several dimensionless numbers: the non-dimensional mountain height (HO ), Rossby
number (Ro), Burger number (B), and the Rossby radius of deformation (Lr). The
non-dimensional mountain height, HO D NH=U , as discussed in Sect. 3.2.2, is
a measure of the importance of vertical displacement by an obstacle and tells us
whether or not the cross-barrier flow has sufficient kinetic energy to overcome the
potential energy needed to cross a mountain barrier. If HO < 1 the flow has sufficient
kinetic energy to go over the mountain, if HO > 1 the flow will tend to be blocked by
the mountain leading to upstream deceleration (Pierrehumbert and Wyman 1985).
The distance of upstream deceleration, and therefore the extent of the barrier jet, is
approximately a Rossby radius of deformation from the mountain crest (Pierrehum-
bert and Wyman 1985). The across-barrier Rossby number, Ro D U/fL is the ratio of
inertial to Coriolis forces and is a measure of the importance of the Earth’s rotation
on the flow. It is 1 for geostrophic flow, and 1 when the effect of the Earth’s
rotation is unimportant. The Burger number, B D NH/fL D RoHO , a dynamically
scaled mountain slope, defines the general hydrodynamic regime for the influence of
the mountain barrier on impinging flow. Finally, the Rossby radius of deformation,
Lr D NH/f D BL is the length scale at which rotation effects are as important as
stratification, and represents the maximum distance upstream of blocking by the
mountains. In the expressions above, U is the across-barrier (onshore) velocity up-
stream, f is the Coriolis parameter, L is the mountain half-width, H is the mountain
height, and N is the Brunt-Väisälä frequency. If the Burger number, B 1, the
mountain is effectively gently sloped and the flow is over the mountain barrier and
quasi-geostrophic. If B 1, the mountain is “hydrodynamically steep” and block-
ing is complete and does not depend on the mountain half width (Overland and Bond
1993, 1995). If 0.1 < B < 1, the flow is semigeostrophic and is modified by the slope
of the mountain, primarily over the mountain (Pierrehumbert and Wyman 1985).
For HO > 1 (i.e. when the flow is shallow compared to the mountain height),
Overland and Bond (1995) suggest that the appropriate offshore distance scale for
deceleration and the barrier jet should be Lr/Ĥ D U/f (instead of Lr), because the
vertical scale in this case is not the mountain height, H, but an inertial or gravity
height U/N found by setting HO D 1. In other words, the flow does not “feel” the
full height of the mountain when HO > 1. Overland and Bond (1995) found that
scaling arguments suggest the barrier jet strength enhancement is on the order of
the incident across-barrier speed for HO > 1 and on the order of NH for HO < 1.
The largest barrier jet response should occur when B > 1 and HO 1, since when
HO > 1 that usually means U is small, and when HO < 1, N is likely to be small
(Overland and Bond 1995). Smith (1979) discusses the importance of the width of
an elevated plateau downwind of the mountain barrier, on upstream and downstream
flow. He demonstrates that far upstream, the influence of a broad plateau is to induce
cyclonic turning of the wind due to stretching of a column of air as isentropes aloft
are displaced upwards. This causes the flow impinging on a mountain barrier to shift
to more of a barrier-parallel orientation and enhances the formation of a barrier jet
near the mountain.
The scale analysis discussed above gives general guidance on barrier jet for-
mation, but is probably not applicable to many aspects of jets created by land-
falling storms. These storms are not uniform and in steady state, are subject to
mesoscale topographic variability, diabatic and frictional effects, and hence the
scaling assumptions are violated. However idealized 2D (Braun et al. 1999) and
3D numerical simulations (Olson and Colle 2009) can provide further insight.
Idealized 2D numerical simulations conducted by Braun et al. (1999) explored the
importance of mountain half width, height, as well as the width of an elevated
plateau downstream of the mountain barrier, on barrier jet characteristics. They
found that while the upstream deceleration is determined by the mountain half width
and height, it is the downstream plateau size that determines the amount of upstream
cyclonic turning and hence the strength of the barrier jet. The barrier jet strength
also depends on the duration of onshore flow: reaching a steady value only after the
time required for the flow to cross the mountain barrier. This finding is consistent
with the theoretical work of Smith (1979) and was confirmed in the idealized 3D
simulations with the realistic terrain of the Gulf of Alaska conducted by Olson and
Colle (2009). Therefore it is expected that wide mountain barriers or mountains with
wide plateaus would have the strongest barrier jet formation.
Mesoscale numerical models have shown good skill at capturing the essential
features of barrier jets in a number of both idealized (e.g. Cui et al. 1998; Yeh and
Chen 2003; Barstad and Grønås 2005; Olson and Colle 2009) and realistic numerical
studies (e.g. Olson et al. 2007; Doyle 1997; Colle and Mass 1995). This suggests
that NWP models with sufficient spatial resolution should be able to adequately
forecast these events. With a width in the across-barrier direction of approximately
50 km, it is likely that at least 10 km horizontal resolution would be required for
accurate representation of classical barrier jets and cold air damming barrier jets.
In cases of hybrid barrier jets, the model must have adequate resolution to simulate
both the gap flow that delivers the cold air onto the coast, as well as the barrier jet
itself. Since topographic gaps in mountain barriers are typically much narrower than
the across-barrier jet scale, it seems likely that models with horizontal resolutions
of 1–2 km would be required to adequately simulate hybrid barrier jets when gap
widths are a few km. In southeast Alaska, Olson et al. (2007) successfully simulated
hybrid jets with an inner grid of 4 km horizontal resolution which was sufficient
to resolve the Cross Sound gap which is approximately 50 km wide. Routinely

operated NWP models will achieve these resolutions as computing power continues
to increase.
While NWP models will continue to improve their ability to forecast barrier
jets, much has been learned from studies that can aid in both the interpretation of
NWP, as well as the creation of rules of thumb for diagnosis and forecasting of
barrier jet events. The basic ingredients of course are that there must be a flow of
stable air toward a mountain barrier for a sufficiently long period of time before
a barrier jet can form. However this leaves several unanswered questions such as:
How high must the barrier be? Are there critical values of Ĥ or B before barrier
jets can form? What factors will determine the barrier jet width and wind speed
enhancement? How long must the across-barrier flow exist before barrier jets can
form?
A climatology of barrier jets by Loescher et al. (2006) using SAR imagery over
the Gulf of Alaska found that terrain height plays a key role in discriminating
locations where barrier jets may form. They found that barrier jets occurred most
frequently in locations where the maximum terrain heights within 100 km of the
coast exceeded 2 km (Fig. 3.38), and that the maximum topography within 100 km
of the coast was more important than the terrain height at the coast in determining
barrier jet frequency. Hybrid barrier jets in this location were associated with major
gaps in the mountain barrier as the gap flow turned to the right due to geostrophic
adjustment upon exiting the gap. The synoptic conditions conducive to barrier jet
formation in this region were investigated by Colle et al. (2006) who found that
cool season jets (without shock or variable features) were associated with a deeper
than normal upper-level trough approaching the Gulf of Alaska, and an anomalously
high ridge over the northwestern North American continent, resulting in low-level
southerlies and warm advection. They found that shock barrier jets had significant
cold anomalies at low levels over the interior, and that variable barrier jets had
weaker low-level stability that favored mixing of higher momentum air to the
surface in localized areas.
Idealized MM5 simulations using the realistic topography of southeast Alaska
with an inner grid of 6 km resolution conducted by Olson and Colle (2009) provide
insight into both the structure and synoptic characteristics useful for forecasting
both classical and hybrid barrier jets in this location and possibly elsewhere too.
They found that the broad inland plateau is essential to the strength of the barrier
jet and results in an anticyclonic pressure perturbation along the windward slopes of
mountains, especially for mountain perpendicular flow. This acts to turn the winds
into a more barrier-parallel direction as much as 500–1,000 km upstream of the
mountains. The timescale for this to develop was found to be the time required
for the flow to cross the full mountain barrier. They found that barrier jet width
tends to be narrower when the impinging flow is more perpendicular to the barrier.
Gap flow from an inland cold pool could contribute to hybrid jets when the low
level ambient flow was nearly barrier-parallel – this results in a hybrid barrier jet
of greater width, with a jet maximum shifted further offshore. When the ambient
flow was nearly barrier-perpendicular, the gap flows were reduced and only classical
barrier jets formed. The greatest wind speed enhancements (nearly two times the
component of ambient flow that is normal to the barrier) were for
• High Ĥ (2.5–3.3) consistent with other studies (Ólafsson and Bougeault 1996;
Braun et al. 1999; Petersen et al. 2003; and also with Colle and Mass 1995 for
cold air damming east of the Rocky Mountains)
• Lower wind speeds, and
• Ambient wind directions that are 30–45ı from barrier-parallel.
Olson and Colle (2009) also found that the barrier jet height was most dependent
on wind speed with strongest winds (25 m s1 ) resulting in the deepest jets (900 m).
Scaling arguments can provide guidance on the across-barrier scale, with the Rossby
radius of deformation, Lr D NH/f defining the maximum barrier jet width, and
this value being reduced by a factor of 1/HO to Lr=HO when the non-dimensional
mountain height is greater than one.
3.7 Summary
This chapter detailed flow phenomena that are dynamically forced when wind
interacts with orography. Flow approaching a mountain barrier, will pass over it,
flow through gaps or valleys that dissect it, or be blocked by the mountain and
diverted horizontally around it. As we have seen, each of these scenarios results
in different dynamically-driven wind phenomena. When air is able to pass over a
mountain barrier under specific conditions, mountain waves are generated that can
result in lee waves, rotors and downslope windstorms. Air flowing through channels
that dissect a mountain barrier can result in gap flow arising from along-channel
acceleration. If the air is blocked by a mountain barrier and diverted around it, then
a barrier jet may form. This chapter described these various phenomena.
Mountain waves are a common and important feature of atmospheric flow in
mountainous regions. Long wavelength mountain waves have a tendency to prop-
agate vertically into the middle and upper atmosphere where the waves eventually
break, causing turbulence and a drag force on the flow. Shorter wavelength waves
tend to propagate horizontally as well as vertically and wave motion appears
downwind of mountain ranges. Wave reflection, caused by height variations in
wind and stability, can enhance the downwind propagation through the formation
of trapped lee waves in some circumstances and downslope windstorms in others.
Mountain waves of sufficiently large amplitude can give rise to severe turbulence
in the troposphere. The term mountain-wave rotor is used to describe horizontally
oriented vortices which form under the wave crests. Rotors can be broadly classified
into two types. The type I rotor is probably the most common and forms beneath
the crests of trapped lee waves. Its formation is linked to boundary layer separation,
which is itself caused by the strong pressure gradient exerted on the near-surface
flow by the lee waves aloft. The rarer type II rotor resembles a hydraulic jump
and is a deep layer of severe turbulence which forms downwind of the mountain,
associated with low-level wave breaking and downslope windstorms. Its structure is
not as well defined as the type I rotor and it is less well understood. Both types of
rotor pose a significant risk to aviation safety and present a difficult challenge for
forecasting.
Downslope windstorms are a consequence of large-amplitude mountain waves
that can form downwind of a mountain barrier in which the low level winds are
accelerated on the lee slope, sometimes to two to three times the wind speed at
mountain top level. They resemble the hydraulic flow of fluid over a rock with
shallow supercritical flow corresponding to the strong downslope winds on the
lee slope which terminate in a hydraulic jump. The circumstances resulting in
this flow can arise in a few ways. If the atmosphere has an upstream layered
structure that already resembles the structure of classical hydraulic flow, then the
transition to a lower shooting flow is facilitated. If the atmosphere above the
ridge has reverse shear (cross-barrier winds decreasing with height either through
changing wind speed or direction) or decreasing stability above a lower stable layer
then this may result in a mean-state critical layer which can act to reflect upward
propagating mountain waves and decouple the lower layer, creating the conditions
for shallow supercritical flow on the lee slope. If the atmosphere has constant
stratification and the non-dimensional mountain height is sufficiently large, then
vertically propagating mountain waves may break in the upper troposphere, creating
a stagnant turbulent zone in which the cross barrier wind goes to zero resulting in a
self-induced critical layer that acts like a mean-state critical layer.
Gaps or valleys that dissect a mountain provide a means for an air mass to cross a
barrier. Potential temperatures on the upstream side have to be lower than in the air
mass downstream for asymmetric gap flows to develop where the layer, which flows
through the gap, thins and accelerates with the highest speeds occurring downstream
of the narrowest part of the gap. Gaps can be level or elevated and asymmetric
gap flows are found all over the world. Hydraulic, non-linear theory best describes
gap flows. An analytical solution exists even for continuously stratified flow, where
nearly neutrally stratified and stagnant layers form above and/or below the gap flow
layer to isolate it from the flow aloft and/or the underlying terrain (on the upstream
side). Usually it is the narrowest section of the gap, which forms a “control” for
the flow. Given a particular value of the upstream depth of the gap flow layer, only
one value of the flow rate through the control is possible when flow transits from
subcritical on the upstream side to supercritical on the downstream side, where it
will eventually adjust to the downstream air mass in a hydraulic jump. Shallow
water models are successful in numerically simulating gap flows. Full NWP models
currently still face challenges to (1) properly resolve the underlying topography, to
(2) handle turbulence near the ground and in the shear layer at the top of the gap flow
and in the hydraulic jump, and to (3) contain the details of the air mass differences
across the gap and the processes producing the hydraulic flow. An ingredients-based
forecasting method identifies processes that lead to the formation of air masses with
different characteristics on both sides of the gap.
Barrier jets are elevated wind maxima that blow parallel to a mountain barrier
on its windward side. When flow approaching a mountain barrier is blocked and
diverted around the mountain then barrier jets may occur if the flow is blocked
by the barrier for hours to days or longer. The blocked flow is forced to move up
the barrier, and cools adiabatically creating higher pressure along the slope which
acts to decelerate and block the flow. Geostrophic adjustment turns the flow left
(right) in the northern (southern) hemisphere and a low-level barrier-parallel jet is
formed between the surface and the mountain crest. Barrier jets have been observed
in many places – and can occur in the extra-tropics (i.e. where rotational effects are
important) wherever stratified flow approaches a significant mountain barrier. They
have been described as classic barrier jets along the western slopes of mountains,
as hybrids where gap flow of cold air from the east side of mountains contributes
the low-level cold air needed for stratification, as cold-air damming events along the
eastern slopes of mountains, and as resulting from the interaction of fronts as they
approach mountains.
Understanding of dynamically forced flows has progressed considerably thanks
to observational, theoretical, numerical modeling (Chaps. 9 and 10), and com-
putational advances over the past three decades. We have progressed from a
quite limited linear, two-dimensional, “dry”, constant stability/constant wind un-
derstanding of these phenomena, to the consideration of more realistic conditions.
Specifically, an understanding of the importance of effects such as non-linearity,
three-dimensionality, moisture, varying wind and stability profiles, and boundary
layer (Chaps. 2 and 5) effects, are now beginning to be more fully understood.
As knowledge has developed, the consideration of these other factors has become
increasingly important.
Acknowledgements The authors thank the volume editors (Fotini Kapodes Chow, Stephen De
Wekker and Bradley J. Snyder) for keeping us on track with their encouragement and organization,
and for all of their efforts in bringing this book to completion. We also thank the three anonymous
reviewers who provided many helpful comments that improved the chapter. We thank the authors
and organizations who allowed us to use many of the figures in this chapter. The first author would
like to acknowledge partial funding support of his research program from an NSERC Discovery
Grant.
References
Afanasyev, Y. D., and W.R. Peltier, 1998: The three-dimensionalization of stratified flow over two-
dimensional topography. J. Atmos. Sci., 55, 19–39.
Afanasyev Y.D., and W.R. Peltier, 2001a: On breaking internal waves over the sill in Knight Inlet.
Proc. R. Soc. Lond. A, 457, 2799–2825.
Afanasyev Y.D., and W.R. Peltier, 2001b: Reply to comment on the paper ‘On breaking internal
waves over the sill in Knight Inlet’. Proc. Roy. Soc. A, 457, 2831–2834.
Alpers, W., A. Ivanov, and J. Horstmann, 2009: Observations of bora events over the adriatic sea
and black sea by spaceborne synthetic aperture radar. Mon. Wea. Rev.,137, 1150–1161.
Anthes, R.A., Y.H. Kuo, D.P. Baumhefner, R.P. Errico, and T.W. Bettge, 1985: Predictability of
mesoscale atmospheric motions. Advances in Geophysics, 28B, Academic Press, 159–202.
Arakawa, S. 1968: A proposed mechanism of fall winds and dashikaze. Met. Geophys., 19,
69–99.
Arakawa, S. 1969: Climatological and dynamical studies on the local strong winds, mainly in
Hokkaidō, Japan. Geophys. Mag., 34, 359–425.
Armi, L. 1986: The hydraulics of two flowing layers with different densities. J. Fluid Mech., 163,
27–58.
Armi, L., and D. Farmer, 2002: Stratified flow over topography: bifurcation fronts and transition to
the uncontrolled state. Proc. R. Soc. London, 458, 513–538.
Armi, L., and G.J. Mayr, 2007: Stratified flow across an alpine crest with a pass: Shallow and deep
flows. Quar. J. Roy. Meteorol. Soc., 133, 459–477.
Armi, L., and U. Riemenschneider, 2008: Two-layer hydraulics for a co-located crest and narrows.
J. Fluid Mech., 615, 169–184.
Armi, L., and R. Williams, 1993: The hydraulics of a stratified fluid flowing through a contraction.
J. Fluid Mech., 251, 355–375.
Bacmeister, J.T., and R. T. Pierrehumbert, 1988: On high-drag states of nonlinear flow over an
obstacle. J. Atmos. Sci., 45, 63–80.
Baines, P.G., 1980: The dynamics of the Southerly Buster. Australian Meteorological Magazine,
28, 175–200.
Baines, P.G., 1995: Topographic effects in stratified flow. Cambridge University Press, 482 pp.
Baines, P.G., 2001: Mixing in flows down gentle slopes into stratified environments. J. Fluid Mech.,
443, 237–270.
Baines, P.G., 2005: Mixing regimes for the flow of dense fluid down slopes into stratified
environments. J. Fluid Mech., 538, 247–267.
Bajić A., 1989: Severe bora on the northern Adriatic Part 1: Statistical analysis. RASPRAVE,
24, 1–9.
Banta, R.M., L.D. Olivier, and J.M. Intrieri, 1990: Doppler lidar observations of the 9 January
1989 severe downslope windstorm in Boulder, Colorado. Preprints, Fifth Conf. on Mountain
Meteorology, Boulder, CO, Amer. Meteor. Soc., 68–89.
Barstad, I., and S. Grønås, 2005: Southwesterly flows over southern Norway – Mesoscale
sensitivity to large-scale wind direction and speed. Tellus, 57A, 136–152.
Bell, G.D., and L.F. Bosart, 1988: Appalachian cold-air damming. Mon. Wea. Rev., 116, 137–161.
Belušić, D., M. Žagar, and B. Grisogono, 2007: Numerical simulation of pulsations in the bora
wind. Quar. J. Roy. Meteorol. Soc., 133, 1371–1388.
Benjamin TB. 1981. Steady flows drawn from a stably stratified reservoir. J. Fluid Mech.,106,
245–260.
Bond, N.A., C.F. Dierking, and J.D. Doyle, 2006: Research aircraft and wind profiler observations
in Gastineau Channel during a Taku wind event. Wea. Forecasting, 21, 489–501.
Bond, N.A. and S.A. Macklin, 1993: Aircraft observations of offshore-directed flow near wide bay,
Alaska. Mon. Wea. Rev.,121, 150–161.
Bond, N.A., C. F. Mass, and J.E. Overland, 1996: Coastally trapped wind reversals along the United
States west coast during the warm season, Part I: Climatology and temporal evolution. Mon.
Wea. Rev.,124, 430–445.
Bougeault, P., P. Binder, A. Buzzi, R. Dirks, R. Houze, J.P. Kuettner, R.B. Smith, R. Steinacker, and
H. Volkert, 2001: The MAP Special Observing Period. Bull. Amer. Meteor. Soc., 82, 433–462.
Bougeault, P., and coauthors, 1993: The atmospheric momentum budget over a major mountain
range: First results of the PYREX field program. Ann. Geophys., 11, 395–418.
Braun, S.A., R.A. Houze Jr., B.F. Smull, 1997: Airborne dual-Doppler observations of an intense
frontal system approaching the Pacific Northwest coast. Mon. Wea. Rev.,125, 3131–3156.
Braun, S.A., R. Rotunno, J.B. Klemp, 1999: Effects of coastal orography on landfalling cold fronts.
Part I: Dry, inviscid dynamics. J. Atmos. Sci., 56, 517–533.
Brinkmann, W.A.R., 1974: Strong downslope winds at Boulder, Colorado. Mon. Wea. Rev., 102,
592–602.
Brown, J.M., 1986: A decision tree for forecasting downslope windstorms in Colorado. Preprints,
11th Conf. on Weather Forecasting and Analysis, Kansas City, MO, Amer. Meteor. Soc., 83–88.
Cameron, D.C., 1931: Easterly gales in the Columbia River Gorge during the winter of 1930–
1931 – Some of their causes and effects. Mon. Wea. Rev., 59, 411–413.
Casswell, S.A., 1966: A simplified calculation of maximum vertical velocities in mountain lee
waves. Meteorol. Mag, 95, 68–80.
Clark. T.L., and R.D. Farley, 1984: Severe downslope windstorm calculations in two and three
spatial dimensions using anelastic interactive grid nesting: A possible mechanism for gustiness.
J. Atmos. Sci., 41, 329–350.
Clark, T.L., and W.R. Peltier, 1977: On the evolution and stability of finite-amplitude mountain
waves. J. Atmos. Sci., 34, 1715–1730.
Colle, B.A., and C.F. Mass, 1995: The structure and evolution of cold surges east of the Rocky
Mountains. Mon. Wea. Rev., 123, 2577–2610.
Colle, B.A., and C.F. Mass, 1998: Windstorms along the Western Side of the Washington Cascade
Mountains. Part I: A High-Resolution Observational and Modeling Study of the 12 February
1995 Event. Mon. Wea. Rev., 126, 28–52.
Colle, B.A., and C.F. Mass, 2000: High-resolution observations and numerical simulations of
easterly gap flow through the Strait of Juan de Fuca on 9–10 December 1995. Mon. Wea.
Rev., 128, 2398–2422.
Colle, B.A., B.F. Smull, and M.-J. Yang, 2002: Numerical simulations of a landfalling cold front
observed during COAST: Rapid evolution and responsible mechanisms. Mon. Wea. Rev., 130,
1945–1966.
Colle, B.A., K.A. Loescher, G.S. Young, and N.S. Winstead, 2006: Climatology of barrier jets
along the Alaskan coast. Part II: Large-scale and sounding composites. Mon. Wea. Rev., 134,
454–477.
Collier, C., 2002: Developments in radar and remote-sensing methods for measuring and forecast-
ing rainfall. Phil. Trans. Roy. Soc., series A, 360, 1345–1361
Colman, B.R., and C. F. Dierking, 1992: The Taku Wind of Southeast Alaska: Its Identification
and Prediction. Wea. Forecasting, 7, 49–64.
Colson, D., 1954: Meteorological problems in forecasting mountain waves. Bull. Amer. Meteor.
Soc., 35, 363–371.
COMET, 2008: MetEd: Mountain Meteorology training module. Available online at: http://www.
meted.ucar.edu/mesoprim/mtnwave/, University Corporation for Atmospheric Research.
Cui, Z., M. Tjernström, and B. Grisogono, 1998: Idealized simulations of atmospheric coastal flow
along the central coast of California, J. Appl. Meteor., 37, 1332–1363.
Darby, L.S. and G.S. Poulos, 2006: The evolution of lee-wave rotor activity in the lee of Pike’s
Peak under the influence of a cold frontal passage: Implications for aircraft safety. Mon. Wea.
Rev., 134, 2857–2876.
Dierking, C.F., 1998: Effects of a mountain wave windstorm at the surface. Wea. Forecasting, 13,
606–613.
Doran, J.C., and S. Zhong, 2000: Thermally driven gap winds into the Mexico City basin. J. Appl.
Meteorol., 39, 1330–1340.
Dorman, C.E., R.C. Beardsley, and R. Limeburner, 1995: Winds in the Strait of Gibraltar. Quart.
J. Roy. Meteorol. Soc.,121, 1903–1921.
Doswell, C.A., H.E. Brooks, and R.A. Maddox, 1996: Flash flood forecasting: An ingredients-
based methodology. Wea. Forecasting,11, 560–581.
Doyle, J.D., 1997: The influence of mesoscale orography on a coastal jet and rainband. Mon. Wea.
Rev., 125, 1465–1488.
Doyle, J.D., D.R. Durran, C. Chen, B.A. Colle, M. Georgelin, V. Grubišić, W.R. Hsu, C.Y.
Huang, D. Landau, Y.L. Lin, G.S. Poulos, W.Y. Sun, D.B. Weber, M.G. Wurtele, and M. Xue,
2000: An intercomparison of model-predicted wave breaking for the 11 January 1972 Boulder
windstorm. Mon. Wea. Rev., 128, 901–914.
Doyle, J.D., and M.A. Shapiro, 2000: A multi-scale simulation of an extreme downslope
windstorm over complex topography. Meteor. Atmos. Phys., 74, 83–101.
Doyle, J.D., and N.A. Bond, 2001: Research aircraft observations and numerical simulations of a
warm front approaching Vancouver Island, Mon. Wea. Rev., 129, 978–998.
Doyle, J.D., and Durran, D.R. 2002. The dynamics of mountain-wave induced rotors. J. Atmos.
Sci., 59, 186–201.
Doyle, J.D., and R.B. Smith, 2003: Mountain waves over the Hohe Tauern. Quart. J. Roy. Meteor.
Soc.,129, 799–823.
Doyle, J.D., and Durran, D.R. 2007. Rotor and sub-rotor dynamics in the lee of three-dimensional
terrain. J. Atmos. Sci., 64, 4202–4221.
Doyle, J.D., and Q. Jiang, 2006: Observations and numerical simulations of mountain waves in the
presence of directional wind shear. Q. J. R. Meteorol. Soc., 132, 1877–1905.
Doyle, J.D., and C.A. Reynolds, 2008: Implications of regime transitions for mountain-wave-
breaking predictability. Mon. Wea. Rev., 136, 5211–5223.
Drechsel, S., and G.J. Mayr, 2008: Objective forecasting of foehn winds for a subgrid-scale alpine
valley. Wea. Forecasting, 23, 205–218.
Durran, D.R., and Klemp, J.B. 1982. The effects of moisture on trapped lee waves. J. Atmos. Sci.,
39 2490–2506.
Durran, D.R., 1986: Another look at downslope windstorms. Part I: The development of analogs
to supercritical flow in an infinitely deep, continuously stratified fluid. J. Atmos. Sci., 43,
2527–2543.
Durran, D.R., and J.B. Klemp, 1987: Another look at downslope winds. Part II: Nonlinear
amplification beneath wave-overturning layers. J. Atmos. Sci., 44, 3402–3412.
Durran, D.R., 1990: Mountain waves and downslope winds. Atmospheric Processes over Complex
Terrain, Meteor. Monogr., No. 45, Amer. Meteor. Soc., 59–81.
Durran, D.R., 1992: Two-layer solutions to Long’s equation for vertically propagating mountain
waves: How good is linear theory? Quart. J. Roy. Meteor. Soc., 118, 415–433.
Durran, D.R., 1999: Numerical methods for wave equations in geophysical fluid dynamics.
Springer, 465 pp.
Enger, L. and B. Grisogono, 1998: The response of bora-type flow to sea surface temperature.
Quart. J. Roy. Meteorol. Soc., 124, 1227–1244.
Epifanio, C.C., and T. Qian, 2008: Wave–turbulence interactions in a breaking mountain wave.
J. Atmos. Sci., 65, 3139–3158.
Etling, D. 1989: On atmospheric vortex sheets in the wake of large islands. Meteor. Atmos. Phys.,
41, 157–164.
Farmer, D.M., and L. Armi, 1999: Stratified flow over topography: the role of small-scale
entrainment and mixing in flow establishment. Proc. Roy. Soc. London A, 455, 3221–3258.
Farmer, D.M., and L. Armi, 2001: Stratified flow over topography: models versus observations.
Proc. Roy. Soc. A, 457, 2827–2830.
Finnigan, T.D., J.A.Vine, P.L. Jackson, S.E. Allen, G.A. Lawrence, and D.G. Steyn, 1994:
Hydraulic physical modeling and observations of a severe gap wind. Mon. Wea. Rev., 122,
2677–2687.
Gaberšek, S., and D.R. Durran, 2006: Gap Flows through Idealized Topography. Part II: Effects of
rotation and surface friction. J. Atmos. Sci., 63, 2720–2739.
Georgelin, M., E. Richard, M. Petitdidier, A. Druilhet, 1994: Impact of subgrid-scale orography
parameterization on the simulation of orographic flows. Mon. Wea. Rev.,122, 1509–1522.
Glasnovic, D., and V. Jurčec, 1990: Determination of upstream bora layer depth. Meteor. Atmos.
Phys., 43, 137–144.
Gohm, A., and G.J. Mayr, 2004: Hydraulic aspects of föhn winds in an Alpine valley. Quart. J. Roy.
Meteorol. Soc., 130, 449–480.
Gohm, A., and G.J. Mayr, 2005: Numerical and observational case study of a deep Adriatic bora.
Quart. J. Roy. Meteorol. Soc., 131, 1363–1392.
Gohm, A., G.J. Mayr, A. Fix, and A. Giez, 2008: On the onset of bora and the formation of rotors
and jumps near a mountain gap. Quart. J. Roy. Meteorol. Soc., 134, 21–46.
Gohm, A., G. Zängl, and G.J. Mayr, 2004: South foehn in the Wipp Valley on 24 October 1999
(MAP IOP 10): Verification of high-resolution numerical simulations with observations. Mon.
Weather Rev., 132, 78–102.
Grisogono, B., and D. Belušić, 2009: A review of recent advances in understanding the meso- and
microscale properties of the severe bora wind. Tellus A, 61, 1–16.
Grisogono, B. and L. Enger, 2004: Boundary-layer variations due to orographic-wave breaking in

the presence of rotation. Quart. J. Roy. Meteorol. Soc., 130, 2991–3014.
Grubišić, V., 1989: Application of the hydraulic theory in cases of bora with strong upstream flow.
RASPRAVE, 24, 21–27.
Grubišić, V. and B.J. Billings, 2007: The intense lee-wave rotor event of Sierra Rotors IOP8.
J. Atmos. Sci., 64, 4178–4201.
Grubišić, V., J.D. Doyle, J. Kuettner, S. Mobbs, R.B. Smith, C.D. Whiteman, R. Dirks, S. Czyzyk,
S.A. Chon, S. Vosper, M. Weissmann, S. Haimov, S.F.J. De Wekker, L.L. Pan, and F.K.
Chow, 2008: The Terrain-induced Rotor Experiment. A field campaign overview including
observational highlights. Bull. Amer. Meteor. Soc., 89, 1513–1533.
Grubišić, V. and Lewis, J.M. 2004. Sierra Wave Project revisited: 50 years later. Bull. Amer. Meteor.
Soc., 85, 1127–1142.
Grubišić, V. and M. Orlic 2007: Early observations of rotor clouds by Andrija Mohorovičić. Bull.
Grubišić, V., R.B. Smith, and C. Schär, 1995: The effect of bottom friction on shallow-water flow
past an isolated obstacle. J. Atmos. Sci.,52, 1985–2005.
Grubišić, V. and I. Stiperski, 2009: Lee-wave resonances over double bell-shaped obstacles. J.
Atmos. Sci., 66, 1205-1228.
Hertenstein, R.F. and J.P. Kuettner, 2005. Rotor types associated with steep lee topography:
influence of the wind profile. Tellus A, 57, 117–135.
Hoinka, K.P., 1985: A comparison of numerical simulations of hydrostatic flow over mountains
and observations. Mon. Wea. Rev., 113, 719–735.
Howells, P.A.C, and Y.H. Kuo, 1988: A numerical study of the mesoscale environment of a
southerly buster event. Mon. Wea. Rev., 116, 1771–1788.
Holmboe, J. and H. Klieforth, 1957: Investigation of mountain lee waves and the air flow over
the Sierra Nevada. Final Report. Contract No. AF19(604)-728, University of California ADNo.
133606, Dept. of Meteorology, University of California, Los Angeles, 290 pp.
Houghton, D.C., and E. Isaacson, 1968: Mountain winds. Stud. Numer. Anal., 2, 21–52.
Houghton, D.D., and A. Kasahara, 1968: Nonlinear shallow fluid flow over an isolated ridge.
Commun. Pure and Appl. Math., 21, 1–23.
Huppert, H.E., and J.W. Miles, 1969: Lee waves in a stratified flow Part 3: Semi-elliptical obstacles.
J. Fluid Mech., 35, 481–496.
Hunt, J.C.R., H. Olafsson and P. Bougeault, 2001: Coriolis effects on orographic and mesoscale
flows. Quart. J. Roy. Meteorol. Soc.,127, 601–633.
Hurd, W.E., 1929: Northers of the Gulf of Tehuantepec. Mon. Wea. Rev., 57, 192–194.
Ishii, S., K. Sasaki, K. Mizutani, T. Aoki, T. Itabe, H. Kanno, D. Matsushima, W. Sha, A. Noda,
M. Sawada, M. Ujiie, Y. Matsuura, and T. Iwasaki, 2007: Temporal evolution and spatial
structure of the local easterly wind “kiyokawa-dashi” in Japan part I: Coherent Doppler lidar
observations. J. Meteorol. Soc. Japan, 85, 797–813.
Ivančan-Picek,B. and V.Tutiš, 1995: Mesoscale bora flow and mountain pressure drag. Meteorol.Z.,
4, 119–128
Ivančan-Picek,B. and V.Tutiš, 1996: A case study of a severe Adriatic bora on 28 December 1992.
Tellus A, 48, 357–367
Jackson, P.L., and D.G. Steyn, 1994a: Gap winds in a fjord. Part I: Observations and numerical
simulation. Mon. Wea. Rev., 122, 2645–2665.
Jackson, P.L., and D.G. Steyn, 1994b: Gap winds in a fjord. Part II: Hydraulic analog. Mon. Wea.
Rev., 122, 2666–2676.
Jackson, P.L., C.J.C. Reason, S. Guan, 1999: Numerical modeling of a coastal trapped disturbance.
Part II: Structure and dynamics, Mon.Wea. Rev., 127, 535–550.
Jurčec, V., 1981: On mesoscale characteristics of bora conditions in Yugoslavia, Pure Appl.
Geophys.,119, 640–657.
Jiang, J.H., D.L. Wu, and S.D. Eckermann, 2002: Upper Atmosphere Research Satellite
(UARS) MLS observation of mountain waves over the Andes. J. Geophys. Res., 107:
10.1029/2002JD002091.
Jiang, Q., R.B. Smith, and J.D. Doyle, 2003: The nature of the mistral: Observations and modelling
of two MAP events. Q. J. R. Meteorol. Soc. 129, 857–875.
Jiang, Q., and J.D. Doyle, 2004: Gravity wave breaking over the central Alps: Role of complex
terrain, J. Atmos. Sci., 61, 2249–2266.
Jiang, Q., J.D. Doyle, and R.B. Smith, 2005: Blocking, descent and gravity waves: Observations
and modeling of a MAP northerly föhn event. Quart. J. Roy. Meteorol. Soc., 131, 675–701.
Jiang, Q., J.D. Doyle, and R.B. Smith, 2006: Interactions between lee waves and boundary layers.
J. Atmos. Sci., 63, 617–633.
Jiang, Q., J.D. Doyle, S. Wang, and R.B. Smith, 2007: On boundary layer separation in the lee of
mesoscale topography. J. Atmos. Sci., 64, 401–420.
Jiang, Q., and J.D. Doyle, 2008: Diurnal variation of downslope winds in Owens Valley during the
Sierra Rotor Experiment. Mon. Wea. Rev., 136, 3760–3780.
King, J.C., P.S. Anderson, D.G. Vaughan, G.W. Mann, S.D. Mobbs, and S.B. Vosper, 2004: Wind-
borne redistribution of snow across an Antarctic ice rise. J. Geoph. Res., 109, No. D11, D11104.
Klemp J.B., and D.K. Lilly, 1975. The dynamics of wave-induced downslope winds. J. Atmos. Sci.,
32, 320–339.
Klemp J.B., and D.K. Lilly, 1978. Numerical simulation of hydrostatic mountain waves. J. Atmos.
Sci., 35, 78–107.
Klemp, J.B., and D.R. Durran, 1987: Numerical modeling of bora winds. Meteor. Atmos. Phys.,
36, 215–227.
Klemp, J.B., W.C., Skamarock, and O. Fuhrer, 2003: Numerical consistency of metric terms in
terrain-following coordinates. Mon. Wea. Rev., 131, 1229–1239.
Koop, C.G., and F.K. Browand, 1979: Instability and turbulence in a stratified fluid with shear.
J. Fluid Mech., 93, 135–159.
Lackmann, G.M., and J. Overland, 1989: Atmospheric structure and momentum balance during a
gap-wind event in Shelikof Strait, Alaska. Mon. Wea. Rev., 117, 1817–1833.
Lane, T.P., M.J. Reeder, B.R. Morton, and T.L. Clark, 2000: Observations and numerical modelling
of mountain waves over the Southern Alps of New Zealand. Quart. J. Roy. Meteorol. Soc., 126,
2765–2788.
Lester, P.F., and W.A. Fingerhut, 1974: Lower turbulent zones associated with mountain lee waves.
J. Appl. Meteor., 13, 54–61.
Li, J., and Y.-L. Chen, 1998: Barrier jets during TAMEX. Mon. Wea. Rev., 126, 959–971.
Lilly, D.K., 1978: A severe downslope windstorm and aircraft turbulence event induced by a
mountain wave. J. Atmos. Sci., 35, 59–77.
Lilly, D.K., and J.B. Klemp, 1979: Effects of terrain shape on non-linear hydrostatic mountain
waves. J. Fluid Mech., 95, 241–261.
Lilly, D.K., and E.J. Zipser, 1972: The Front Range windstorm of January 11 1972. Weatherwise,
25, 56–63.
Liu, H., P. Olsson, and K. Volz, 2008: SAR observation and modeling of gap winds in the Prince
William Sound of Alaska. Sensors, 8, 4894–4914.
Loescher, K.A., G.S. Young, B.A. Colle, and N.S. Winstead, 2006: Climatology of barrier jets
along the Alaskan Coast. Part I: Spatial and temporal distributions. Mon. Wea. Rev., 134,
437–453.
Long, R.R., 1953: Some aspects of the flow of stratified fluids. I. A theoretical investigation. Tellus,
5, 42–57.
Long, R.R., 1954: Some aspects of the flow of stratified fluids. II. Experiments with a two-fluid
system. Tellus, 6, 97–115.
Lorenz, E.M., 1969: The predictability of a flow which possesses many scales of motion. Tellus,
21, 289–307.
Lyra, G. 1943. Theorie der stationären Leewellenstromung in freier Atmosphäre. Z. Angew. Math.
Mech., 23, 1–128.
Marwitz, J., and J. Toth, 1993: The Front range blizzard of 1990. Part I: Synoptic and mesoscale
structure. Mon. Wea. Rev., 121, 402–415.
Mashiko, W., 2008: Formation mechanism of a low-level jet during the passage of typhoon Ma-on
(2004) over the Southern Kanto District. J. Meteorol. Soc. Japan. Ser. II, 86, 183–202, URL
http://ci.nii.ac.jp/naid/110006633437/en/.
Mass, C.F., and M.D. Albright, 1985: A severe windstorm in the lee of the Cascade mountains of
Washington State, Mon. Wea. Rev., 113, 1261–1281.
Mass, C.F., and G.K. Ferber, 1990: Surface pressure perturbations produced by an isolated
mesoscale topographic barrier. Part I: General characteristics and dynamics. Mon. Wea. Rev.,
118, 2579–2596.
Mass, C.F., and N.A. Bond, 1996: Coastally trapped wind reversals along the United States west
coast during the warm season, Part II: Synoptic evolution. Mon. Wea. Rev., 124, 446–461.
Mayr, G.J., and L. Armi, 2008: Föhn as a response to changing upstream and downstream air
masses. Quart. J. Roy. Meteorol. Soc., 134, 1357–1369.
Mayr, G.J., J. Vergeiner, and A. Gohm, 2002: An Automobile Platform for the Measurement of
Foehn and Gap Flows. J. Atmos. Oceanic Tech., 19, 1545–1556.
McInnes, K.L., and J.L. McBride, 1993: Australian southerly busters. Part I: Analysis of a
numerically simulated case-study. Mon. Wea. Rev., 121, 1904–1920.
Mercer, A.E., M.B. Richman, and H.B. Bluestein, 2008: Statistical modeling of downslope
windstorms in Boulder, Colorado. Wea. Forecasting, 23, 1176–1194.
Miranda, P.M.A. and I.N. James, 1992: Nonlinear three-dimensional effects on gravity wave drag:
Splitting flow and breaking waves. Quart. J. Roy. Meteor. Soc., 118, 1057–1081.
Mobbs, S., S. Vosper, P. Sheridan, R. Cardoso, R. Burton, S. Arnold, M. Hill, V. Horlacher, and A.
Gadian, 2005: Observations of downslope winds and rotors in the Falkland Islands. Quart. J.
Roy. Meteorol. Soc., 131, 329–351.
Mohorovičić, A. 1889: Interessante Wolkenbildung über der Bucht von Buccari (with a comment
from the editor J. Hann). Meteor. Z., 24, 56–58.
Müller, H., R. Reiter, and R. Sládkovič, 1984: Die vertikale windstruktur beim merkur-
schwerpunkt ’tagesperiodische windsysteme’ aufgrund von aerologischen messungen im inntal
und im rosenheimer becken (engl.: Vertical structure of the diurnal wind system within the inn
valley and the adjacent plain: results from aerological soundings during the merkur campaign).
Arch. Met. Geophys. Biokl. B, 33, 359–372.
Nance, L. B., and B. R. Colman, 2000: Evaluating the use of a nonlinear two-dimensional model
in downslope windstorm forecasts. Wea. Forecasting, 15, 715–729.
Nappo, C. J. 2002: An introduction to atmospheric gravity waves. Academic Press, 279 pp.
Neiman, P.J., R.M. Hardesty, M.A. Shapiro and R.E. Cupp, 1988: Doppler lidar observations of a
downslope windstorm. Mon. Wea. Rev., 116, 2265–2275.
Neiman, P.J., F.M. Ralph, A.B. White, D.D. Parrish, J.S. Holloway, and D.L. Bartels, 2006: A
multiwinter analysis of channeled flow through a prominent gap along the northern California
coast during CALJET and PACJET. Mon. Wea. Rev.,134, 1815–1841.
Nuss, W.A., J. Bane, W. Thompson, C. Dorman, M. Ralph, R. Rotunno, J. Klemp, W. Skamarock,
R. Samelson, A.Rogerson, C.J.C. Reason, P.L. Jackson, 2000: Coastally trapped wind reversals:
A new level of understanding from the experiment on coastally trapped disturbances, Bull.
Ogura, Y., and N. Phillips, 1962: Scale analysis for deep and shallow convection in the atmosphere.
J. Atmos. Sci., 19, 173–179.
Økland, H., 1990: The dynamics of coastal troughs and coastal fronts. Tellus, 42A, 444–462.
Ólafsson, H., and P. Bougeault, 1996: Nonlinear flow past an elliptic mountain ridge. J. Atmos.
Sci., 53, 2465–2489.
Ólafsson, H., and P. Bougeault, 1997: The effect of rotation and surface friction on orographic
drag. J. Atmos. Sci., 54, 193–210.
Olson, J.B., B. Colle, N. Bond, and N. Winstead, 2007: A comparison of two coastal battier jet
events along the southeast Alaskan Coast during the SARJET field experiment. Mon. Wea.
Rev., 135, 3642–3663.
Olson, J.B., and B. A. Colle, 2009: Three-dimensional idealized simulations of barrier jets along
the southeast coast of Alaska, Mon. Wea. Rev., 137, 391–413.
Orlić, M., M. Kuzmić, and Z. Pasarić, 1994: Response of the Adriatic Sea to the bora and sirocco
forcing. Continent. Shelf Res., 14, 91–116.
Orlić, M., V. Dadić, B. Grbec, N. Leder, A. Marki, F. Matić, H. Mihanović, G. Beg Paklar, M.
Pasarić, Z. Pasarić, and I. Vilibić, 2006: Wintertime buoyancy forcing, circulation systems
produced in the Adriatic. J. Geophys. Res. Oceans, 111, C03S07.
Overland, J.E., 1984: Scale analysis of marine winds in straits and along mountainous coasts. Mon.
Wea. Rev., 112, 2530–2534.
Overland, J.E., and N.A. Bond, 1993: The influence of coastal orography: The Yakutat storm. Mon.
Wea. Rev., 121, 1388–1397.
Overland, J.E., N.A. Bond, 1995: Observations and scale analysis of coastal wind jets. Mon. Wea.
Rev., 123, 2934–2941.
Pan, F., and R.B. Smith, 1999: Gap winds and wakes: SAR Observations and numerical
simulations. J. Atmos. Sci., 56, 905–923.
Parish, T.R., 1982: Barrier winds along the Sierra Nevada Mountains. J. Appl. Meteor., 21,
925–930.
Parish, T.R., 1983: The influence of the Antarctic peninsula on the wind-field over the western
Weddell Sea. J. Geophys. Res., 88, 2684–2692.
Peltier, W.R., and T.L. Clark, 1979: the evolution and stability of finite-amplitude mountain waves.
Part II: Surface wave drag and severe downslope windstorms. J. Atmos. Sci., 36, 1498–1529.
Peltier, W.R., and T.L. Clark, 1983: Nonlinear mountain waves in two and three spatial dimensions.
Quart. J. Roy. Meteor. Soc.,109, 527–548.
Peltier, W.R., and J.F. Scinocca, 1990: The origin of severe downslope windstorm pulsations.
J. Atmos. Sci., 47, 2853–2870.
Peng, M.S., and W.T. Thompson, 2003: Some aspects of the effect of surface friction on flows over
mountains. Quart. J. Roy. Meteorol. Soc., 129, 2527–2557.
Petersen, G.N., J.E. Kristjánsson, and H. Ólafsson, 2003: The effect of upstream wind direction
on atmospheric flow in the vicinity of a large mountain. Quart. J. Roy. Meteorol. Soc., 131,
1113–1128.
Pettre, P., 1982: On the problem of violent valley winds. J. Atmos. Sci., 39: 542–554.
Pierrehumbert, R.T., and B. Wyman, 1985: Upstream effects of mesoscale mountains. J. Atmos.
Sci., 42, 977–1003.
Prandtl, L. 1942: Führer durch die Strömungslehre. Friedr. Vieweg & Sohn, Braunschweig, 3rd
edn.
Queney, P. 1948: The problem of airflow over mountains: A summary of theoretical studies. Bull.
Raphael, M.N., 2003: The Santa Ana winds of California. Earth Interactions, 7(8), 13 pp.
Reason, C.J.C., D.G. Steyn, 1992: The dynamics of coastally trapped mesoscale ridges in the lower
atmosphere, J. Atmos. Sci., 49, 1677–1692.
Reason, C.J.C., K. Tory, P.L. Jackson, 2001: A model investigation of the dynamics of a Coastally
Trapped Disturbance. J. Atmos. Sci., 58(14), 1892–1906.
Reed, R.J., 1981: A case of a bora-like windstorm in western Washington. Mon. Wea. Rev., 109,
2383–2393.
Reinecke, P. A., and D. R. Durran, 2009a: Initial-condition sensitivities and the predictability of
Reinecke, P. A., Durran, D., 2009b: The Overamplification of Gravity Waves in Numerical
Solutions to Flow over Topography. Mon. Wea. Rev., 137, 1533–1549.
Reid, J.J., L.M. Leslie, 1999: Modeling coastally trapped wind surges over southeastern Australia.
Part I: Timing and speed of propagation. Wea. Forecasting, 14, 53–66.
Richard, E., P. Mascart, and E.C. Nickerson, 1989: On the role of surface friction in downslope
windstorms. J Appl. Meteor., 28, 241–251.
Rogers, D.P., C.E. Dorman, K. A. Edwards, I.M. Brooks, W.K Melville, S.D. Burk, W.T.
Thompson, T. Holt, L. M Ström, M. Tjernström, B. Grisogono, J.M. Bane, W.A. Nuss, B.
M. Morley and A.J. Schanot, 1998: Highlights of Coastal Waves, 1996, Bull. Amer. Meteor.
Soc., 79, 1307–1326.
Ross, A.N., and S.B. Vosper, 2003: Numerical simulations of stably stratified flow through a
mountain pass. Quart. J. Roy. Meteor. Soc., 129, 97–115.
Sawyer, J.S., 1960: Numerical calculation of the displacements of a stratified airstream crossing a
ridge of small height. Quart. J. Roy. Meteor. Soc., 86, 326–345.
Sawyer, J.S., 1962: Gravity waves in the atmosphere as a three-dimensional problem. Quart. J. Roy.
Meteor. Soc., 88, 412–425.
Schär, C., and R.B. Smith, 1993: Shallow-water flow past isolated topography. Part I: Vorticity
production and wake formation. J. Atmos. Sci., 50, 1373–1400.
Schär, C., and D.R. Durran, 1997: Vortex formation and vortex shedding in continuously stratified
flows past isolated topography. J. Atmos. Sci., 54, 534–554.
Schär, C., D. Leuenberger, O. Fuhrer, D. Lüthi, and C. Girard, 2002: A new terrain-following vical
coordinate formulation for atmospheric prediction models. Mon. Wea. Rev., 130, 2459–2480.
Schultz, D.M., W.E. Bracken, L.F. Bosart, G.J. Hakim, M.A. Bedrick, M.J. Dickinson, K.R. Tyle,
1997: The 1993 superstorm cold surge: Frontal structure, gap flow, and tropical impact. Mon.
Wea. Rev.,125, 5–39.
Schwerdtfeger, W., 1974: Mountain barrier effect on the flow of stable air north of the Brooks
Range. Proc. 24th Alaskan Science Conf., Fairbanks, AL, Geophysical Institute, University of
Alaska Fairbanks, 204–208.
Schwerdtfeger, W., 1975: The effect of the Antarctic Peninsula on the temperature regime of the
Weddell Sea. Mon. Wea. Rev., 103, 45–51.
Schwerdtfeger, W., 1979: Meteorological aspects of the drift of ice from the Weddell Sea towards
the mid-latitude westerlies. J. Geophys. Res., 84, 6321–6327.
Scinocca, J.F., and W.R. Peltier, 1989: Pulsating downslope windstorms. J. Atmos. Sci., 46,
2885–2914.
Scorer, R., 1952: Mountain-gap winds: A study of surface wind at Gibraltar. Quart. J. Roy. Meteor.
Soc., 78, 53–61.
Seibert, P., 1990: South foehn studies since the ALPEX experiment. Meteor. Atmos. Phys.,43,
91–103.
Sharp, J., C.F. Mass, 2002: Columbia Gorge gap flow. Bull. Amer. Meteor. Soc., 83, 1757–1762.
Sharp, J., C.F. Mass, 2004: Columbia Gorge Gap Winds: Their Climatological Influence and
Synoptic Evolution. Wea. Forecasting, 19, 970–992.
Sheridan, P.F., V. Horlacher, G.G. Rooney, P. Hignett, S.D. Mobbs, and S.B. Vosper, 2007:
Lee waves and flow separation downwind of the Pennines. Quart. J. Roy. Meteor. Soc.,
133,1353–1369.
Shutts, G.J., 1997: Operational lee wave forecasting. Meteorol. Appl. 4, 23–35.
Smith, C.M. and E.D. Skyllingstad 2009: Investigation of upstream boundary layer influence on
mountain wave breaking and lee wave rotors using a large-eddy simulation. J. Atmos. Sci., 66,
3147–3164.
Smith, R.B., 1976: The generation of lee waves by the Blue Ridge. J. Atmos. Sci., 33, 507–519.
Smith, R.B., 1977: The steepening of hydrostatic mountain waves. J. Atmos. Sci., 34, 1634–1654.
Smith, R.B., 1979: The influence of mountains on the atmosphere. Adv. Geophys., Vol. 21,
Academic Press, 87–230.
Smith, R.B., 1980: Linear theory of stratified hydrostatic flow past an isolated mountain.
Tellus,32A, 348–364.
Smith, R.B., 1985: On severe downslope winds. J. Atmos. Sci., 42, 2597–2603.
Smith, R.B., 1987: Aerial observations of the Yugoslavian bora. J. Atmos. Sci., 44, 269–297.
Smith, R.B., 1989: Hydrostatic airflow over mountains. Adv. Geophys., 31, 1–41.
Smith, S.A., 2004: Observations and simulations of the 8 November 1999 MAP mountain wave
case. Quart. J. Roy. Meteor. Soc., 130, 1305–1326.
Smith, S.A., and A.S. Broad, 2003: Horizontal and temporal variability of mountain waves over
Mont Blanc. Quart. J. Roy. Meteor. Soc., 129, 2195–2216.
Smith, R.B., A.C. Gleason, and P.A. Gluhosky, 1997: The wake of St Vincent. J. Atmos. Sci., 54,
606–623.
Smith, R.B., and S. Grønås, 1993: Stagnation points and bifurcation in 3-D mountain air-flow.
Tellus, 45A, 28–43.
Smith, R.B., S. Skubis, J.D. Doyle, A.S. Broad, C. Kiemle, and H. Volkert, 2002: Mountain waves
over Mont Blanc: Influence of a stagnant boundary layer, J. Atmos. Sci.,59, 2073–2092.
Smith, R.B., and J. Sun, 1987: Generalized hydraulic solutions pertaining to severe downslope
winds. J. Atmos. Sci., 44, 2934–2939.
Smith, R.B., Jiang, Q. and Doyle, J.D. 2006. A theory of gravity wave absorption by boundary
layers, J. Atmos. Sci., 63, 774–781.
Smith, R.B., 2007: Interacting mountain waves and boundary layers. J. Atmos. Sci., 64, 594–607.
Smith, R.B., J.D. Doyle, Q. Jiang, and S.A. Smith, 2007: Alpine gravity waves: Lessons from MAP
regarding mountain wave generation and breaking. Quart. J. Roy. Meteor. Soc., 133, 917–936.
Smolarkiewicz, P.K. and R. Rotunno, 1989: Low Froude number flow past three-dimensional
obstacles. Part I: Baroclinically generated lee vortices. J. Atmos. Sci., 46, 1154–1164.
Sprenger, M., and C. Schär, 2001: Rotational aspects of stratified gap flows and shallow foehn.
Q. J. R. Meteorol. Soc., 127, 161–187.
Steenburgh, W.J., D.M. Schultz, and B.A. Colle, 1998: The structure and evolution of gap outflow
over the gulf of Tehuantepec, Mexico. Mon. Wea. Rev., 126, 2673–2691.
Tutiš,V. and B. Ivančan-Picek, 1991: Pressure drag on the Dinaric Alps during the ALPEX SOP,
Meteor. Atmos. Phys, 47, 73–81.
Vergeiner, I., 1971: An operational linear lee wave model for arbitrary basic flow and two-
dimensional topography. Quart. J. Roy. Meteor. Soc., 97, 30–60.
Volkert, H., C. Keil, C. Kiemle, G. Poberaj, J-P. Chaboureau, and E. Richard, 2003: Gravity waves
over the eastern Alps: A synopsis of the 25 October 1999 event (IOP10) combining in situ and
remote-sensing measurements with a high-resolution simulation. Quart. J. Roy. Meteor. Soc.,
129, 777–797.
Vosper, S.B., 2003: Development and testing of a high resolution mountain-wave forecasting
system. Meteor. Appl.,10, 75–86.
Vosper, S.B., 2004: Inversion effects on mountain lee waves. Quart. J. Roy. Meteor. Soc., 130,
1723–1748.
Vosper, S.B., and A.R. Brown, 2007: The effect of small-scale hills on orographic drag. Quart.
J. Roy. Meteor. Soc., 133, 1345–1352.
Vosper, S.B. and S.D. Mobbs, 1996: Lee waves over the English Lake District, Quart. J. Roy.
Meteor. Soc., 122, 1283–1305.
Vosper, S.B., P.F. Sheridan, and A.R. Brown, 2006: Flow separation and rotor formation beneath
two-dimensional trapped lee waves. Quart. J. Roy. Meteor. Soc., 132, 2415–2438.
Winstead, N.S., B. Colle, N. Bond, G. Young, J. Olson, K. Loescher, F. Monaldo, D. Thompson,
and W. Pichel, 2006: Using SAR remote sensing, field observations, and models to better
understand coastal flows in the Gulf of Alaska, Bull. Amer. Meteor. Soc., 87, 787–800.
Wood I.R., 1968: Selective withdrawal from a stably stratified fluid. J. Fluid Mech.32: 209–223.
Yeh, H.-C., and Y.-L. Chen, 2003: Numerical simulations of the barrier jet over northwestern
Taiwan during the mei-yu season. Mon. Wea. Rev., 131, 1396–1407.
Yoshino, M.M., 1972: An annotated bibliography on bora. Climatol. Notes, 10, 1–22.
Yoshino, M.M., 1976: Bora studies: A historical review and problems today. In Local Wind Bora,
M.M. Yoshino Ed., 3–18.
Yu, C.-K., and B.F. Smull, 2000: Airborne Doppler observations of a landfalling cold front
upstream of a steep coastal orography. Mon. Wea. Rev., 128, 1577–1603.
Žagar, M. and J. Rakovec, 1999: Small scale surface wind prediction using dynamical adaptation.
Tellus, 51A, 489–504.
Zängl, G., 2002a: Idealized numerical simulations of shallow föhn. Quart. J. Roy. Meteor.
Soc., 128, 431–450.
Zängl, G., 2002b: An Improved Method for Computing Horizontal Diffusion in a Sigma-
Coordinate Model and Its Application to Simulations over Mountainous Topography. Mon.
Wea. Rev.,130, 1423–1432.
Zängl, G., 2002c: Stratified flow over a mountain with a gap: Linear theory and numerical
simulations. Q. J. R. Meteor. Soc.,128, 927–949.
Zängl, G., 2003: Deep and shallow south foehn in the region of Innsbruck: Typical features and
semi-idealized numerical simulations. Meteor. Atmos. Phys., 83, 237–262.
Zängl, G., 2004: A reexamination of the valley wind system in the Alpine Inn Valley with numerical
Zängl, G., 2006: North foehn in the Austrian Inn valley: A case study and idealized numerical
Zängl, G., A. Gohm, and G. Geier, 2004: South foehn in the Wipp Valley – Innsbruck region:
Numerical simulations of the 24 October 1999 case (MAP-IOP 10). Meteor. Atmos. Phys., 86,
213–243.
Zängl, G., A. Gohm, and F. Obleitner, 2008: The impact of the PBL scheme and the vertical
distribution of model layers on simulations of Alpine foehn. Meteor. Atmos. Phys., 99, 105–128.
Zecchetto, S., and C. Cappa, 2001: The spatial structure of the Mediterranean Sea winds as revealed
by the ERS-1 scatterometer. Int. J. Remote Sensing, 22, 45–70.
Chapter 4
Understanding and Forecasting Alpine Foehn
Hans Richner and Patrick Hächler
Abstract This chapter focuses on the history, physics, climatology, forecasting

and the broad effects of Alpine foehn on human populations. In the European
Alps, foehn winds have been studied since the mid-1800s. The main focus of
the investigations was the question of why foehn winds are so warm. While it
soon became clear that adiabatic processes provide an explanation, the role of wet
adiabatic rising on the upwind side of the Alps continued to be strongly debated.
The so-called textbook theory for foehn – heat gain by wet adiabatic, forced lifting
on the upwind side followed by dry adiabatic descent in the lee – represents only
an extreme situation. Foehn occurs also with partial or complete blocking of the
upwind air mass, i.e., with limited or no heat gain by wet adiabatic expansion. The
second focus is on processes which lead to descending air masses after passing
the mountain ridge. A discussion of the most important processes shows that there
seems to be no theory which is applicable in all situations. Forecasting foehn is still
a challenge to meteorologists. While the general foehn situation can be predicted
reliably, today’s numerical models still often poorly simulate the sudden break in
of potentially devastating foehn air in the lee. Efforts to improve this must continue
because foehn storms have a significant societal impact (threat to transportation
systems and massive increase of fire danger) as several recent incidents show.
H. Richner ()
Institute for Atmospheric and Climate Science (IACETH), ETH Zurich, Universitätsstrasse 16,
Zurich CH-8092, Switzerland
e-mail: hans.richner@ethz.ch
P. Hächler
Federal Office of Meteorology and Climatology MeteoSwiss, Krähbühlstrasse 58, Zurich
CH-8044, Switzerland
e-mail: patrick.haechler@meteoswiss.ch

220 H. Richner and P. Hächler
4.1 Introduction
Foehn is a prominent meteorological phenomenon occurring in practically all

extended mountain ranges. “Foehn” is a generic term for a downslope wind that
is strong, warm, and dry; WMO (1992) defines foehn wind as a “wind [which is]
warmed and dried by descent, in general on the lee side of a mountain.” Its societal
impact is considerable, be it in a favorable manner with regard to climatology (mild
climate) or in a negative way (windstorms) with respect to aviation and traffic.
Outside the Alps, foehn storms are often called downslope windstorms.
While the term “foehn” originated in the European Alpine area, foehn winds
occur all over the world where there are extended mountain ranges. Foehn winds are
even observed in places where the mountain ridges are not that high such as in the
United Kingdom. Depending on the place, foehn winds might have a different name
such as, e.g., Chinook (North America) or Helm wind (UK). The rapid temperature
rise and the dry air of foehn type winds have also led to numerous local, descriptive
names such as “snow eater,” “grape cooker,” etc. Examples of some of these flows
are given in Chap. 3, Sect. 3.4.
This chapter continues the discussion of dynamically-driven winds from Chap. 3,
but specifically focusing on the combined history, physics, climatology, forecasting
and the broad effects of Alpine foehn on human populations. Alpine foehn presents
a significant forecasting and modeling challenge, and has been the topic of study
for close to two centuries. This chapter highlights this particular flow phenomenon
because of the breadth of its impact and the peculiar challenges associated with
forecasting. The chapter centers on recent results of the Mesoscale Alpine Program
(MAP) and on ongoing research activities mainly in Switzerland, but touches also
on activities in Austria and France. For research on foehn-type winds in other parts
of the world, see e.g. Brinkmann (1974) (Rocky Mountains), Raphael (2003) (Santa
Ana Winds), or McGowan et al. (2002) (New Zealand Alps).
The overview encompasses statistical analyses of foehn occurrence in different
Alpine regions, on the interaction of foehn flow and the cold pool, and on
current techniques for forecasting foehn-related windstorms. Figure 4.1 shows a
topographical map of the Alpine massif. A detailed list of terms and definitions is
provided in the appendices to help clarify the varying terminology used to describe
foehn events.
4.2 History of Foehn Research
For many decades, foehn was the outstanding example to explain thermodynamic
processes and the role of latent heat in the atmosphere. Hence, nearly every textbook
contains a graph similar to Fig. 4.2. Driven by the synoptic scale pressure field,
humid air is forced towards a mountain range. As it ascends by forced convection,
it cools dry-adiabatically until reaching saturation. From then on, the rise is
4 Understanding and Forecasting Alpine Foehn 221
Fig. 4.1 Topographical map of the Alpine massif. The two dashed lines show the Gotthard cross
section (western) and the Brenner cross section (eastern). Only those Alpine regions are labelled
which are referred to in the text
Fig. 4.2 Schematic description of the “textbook theory” of foehn. Diagrams similar to this one are
found in many textbooks on meteorology
wet-adiabatically until the air reaches the crest of the mountain range; as a
consequence, clouds are formed and precipitation occurs. As the air descends in the
lee of the mountain, it is heated dry-adiabatically; as consequence, the air becomes
dry and reaches temperatures that – at equal elevation – are higher than the original
Fig. 4.3 The “proof” that Hann was aware of the two types of foehn: Original illustration depicting
Hann’s (1866) foehn theory in Ficker (1920). “Nordseite”: north side; “Südseite”: south side;
“Föhn, Erwärmung”: foehn, warming; “Luft in Ruhe”: air at rest. Note that north–south is reversed
with respect to Fig. 4.2
temperature on the windward side. Because of the very clean and dry air on the lee
side, visibility is outstanding (D>foehn window1 ), and over the crest, the piled-up
clouds can be seen as the D> foehn wall (Fig. 4.2).
As beautiful as this example is, in reality foehn winds often do not follow
this classical textbook theory that is attributed to Hann (nineteenth century). In
his “Lehrbuch der Meteorologie”, Hann (1901) describes both types of foehn (see
Figs. 4.2 and 4.3) and mentions that many forms in between the two have been
observed. Seibert (2005) concludes that Hann’s explanation was seriously distorted
in the first half of the twentieth century. Particularly Austrian researchers repeatedly
questioned the validity of the textbook approach. In his book “Environmental
Aerodynamics”, Scorer (1978) wrote a chapter entitled “Foehn Fallacy” which gives
at least two explanations for the warming of air masses which do not require the heat
of condensation on the windward slope of the mountain ridge: mechanical stirring
of a stably stratified air mass, and – more important – blocking. In the first case, the
lower part becomes warmer and the upper part cools because the mixing produces a
constant potential temperature, in the second case, potentially warmer air subsides
in the lee.
Analyses of data collected during and after the Alpine Experiment ALPEX in
1982 clearly confirmed that, at least in the Brenner cross-section, there was blocked
air during most foehn cases. This is particularly true for the “century foehn” of
November 8, 1982. In the Ticino region, i.e., south of the Gotthard, however, there
was some light precipitation towards the end of this extreme foehn event (hourly
mean winds up to 35 m s1 and gusts over 50 m s1 at the station Gütsch, see
Fig. 4.7).
It is a little disturbing that not only popular publications but also modern
textbooks often present only the theory depicted in Fig. 4.2 without discussing
alternative foehn schemes.
Austria and Switzerland were clearly the hotspots for foehn research. Innsbruck
(Austria) is the place where scientists like Hann, Ficker, Hoinkes, Kuhn, Dreiseitl,
1
Terms preceded by “D>” are further explained in Appendix A.
Steinacker, Seibert, Hoinka, Mayr, and Gohm have carried out their research; in the
Swiss Alps it is mainly Billwiller, Wild, Streiff-Becker, Schüepp, Frey, Gutermann
and others who have made observations and developed their theories since the
nineteenth century. In Switzerland, research was not quite as active as in Austria.
However, in Switzerland, monks and pastors meticulously observed and recorded
foehn events, this being particularly true for the Reuss Valley. Thanks to them, a
quite homogenous time series for foehn events in Altdorf is available since 1864.
4.3 Characteristics of Foehn
4.3.1 Textbook Theory and Real Dynamics
South foehn occurs in the Alpine region when a synoptic pressure field according
to Fig. 4.4 exists. (For simplicity the following discussions is restricted to south
foehn, however, some remarks related to north foehn and west foehn are given
in Sect. 4.3.4.) Nowadays it is widely accepted that foehn winds can and do
occur without precipitation. The lower the crest height, the more likely it is that
advected air simply crosses the mountain ridge and, subsequently, descends, as
already debated by Hann, Ficker and others and as depicted in Fig. 4.3. Trajectory
analyses as well as tracer studies with ozone confirm this mechanism (Baumann
et al. 2001).
Fig. 4.4 Synoptic pressure field producing foehn in the Alps

Fig. 4.5 Suggested parameters for characterizing foehn types. A: depth of blocked air, C: crest
height, : difference in potential temperature
The fact that much foehn research was and still is being done along the rather
flat Brenner cross-section, explains why in Austria the textbook theory was more
fiercely queried than, e.g., in Switzerland. As a matter of fact, Hann (1866) already
distinguished between foehn type I or “Swiss foehn type” (for what is here called
the textbook theory) and foehn type II or “Austrian foehn type” (for the airflow
without gain of latent heat and with reduced or no precipitation on the windward
side). These two foehn types must be regarded as extremes, as anything in between
can be observed.
It must be stressed that even heavy precipitation on the upwind side is no proof
for gain of latent heat; precipitation can stem from the lower air mass that is
not part of the foehn flow. For describing a foehn case, it would be desirable to
use as additional characterizing parameters (a) the height of the internal boundary
between the lower, blocked air and the pressure driven foehn flow aloft, and (b) the
difference ™ between the potential temperature of the two air masses (see Fig. 4.5).
If sufficient information on the vertical structure is available, this parameter can be
determined by trajectory calculations. In a textbook foehn case (i.e., type I), the
value for A would be zero; in a foehn situation with total blocking on the upwind
side (i.e., type II), the value for A would correspond to the ridge height C. The
value for A could be expressed either in absolute units or in percent of the ridge
height.
There are at least two reported cases when foehn winds were observed simulta-
neously on both sides of the same mountain range (Frey 1986). This can only be
explained by assuming subsidence of significant air masses over the ridge. In this
1986 event, south winds were in excess of 40 m s1 north of the Alps!
On the other hand, it would be wrong to dismiss textbook theory completely.
Observations clearly show that heavy precipitation can and does occur on the
windward side of high mountain complexes during foehn. There are indications
that this is more often the case south of the (higher) Gotthard than south of the
(lower) Brenner massif. Although this suspicion has been brought up repeatedly
after ALPEX, there are no statistical analyses that would prove this, and no firm
statement can be made.
4.3.2 The Effect of Topography
Foehn areas are primarily defined by topography. Basically, airflow penetrates a

given valley more easily the better the valley axis is aligned with the flow, and hence
the fewer ridges further downwind hinder the flow. This situation can be described
by saying that the flow is diffluent, the degree of the diffluence being controlled
by topography. Based on these principles, the topography defines the foehn areas.
In general, major valleys perpendicular to the ridge represent the areas with the
highest foehn frequency (see Sect. 4.5).
Valleys have a strong channeling effect on the flow. Because the cross section
decreases as the flow further penetrates into the valley, the wind speeds can be higher
near the ground than they are above. As observations show, the channeling effect (in
the sense that the valley axis determines the direction of the flow) can reach up to
heights well above the crest of the mountains bordering the valley in question.
A forecaster’s rule says that foehn winds rarely descend more that 2,000 m
behind the crest over which they streamed in. For this reason, foehn is seldom
experienced in the region of Lake Thun, a tourist area just north of the Jungfrau
region in Switzerland. A sound explanation in terms of dynamics for this rule is not
available.
When talking about foehn frequency at a given location, it is important to exactly
define what is meant by saying, “There is foehn”. For some people it is sufficient
to have a foehn airflow above the location which does not reach the ground, for
others, the winds must actually touch down. During the winter period September
to March, foehn reaches the downtown area of Zurich, Switzerland, on average
only on 2–3 days, however, on 50–70 days, there is a foehn situation where the
downtown remains in the calm cold pool while a possibly strong, certainly warmer
foehn airflow exists a few hundred meters above.
In some cases it is very clear whether foehn is present or not (see next section).
However, for a foehn climatology, the definition of foehn can become crucial
and problematic. Barry (2008, p. 173) discusses this issue and gives examples
for definitions used in different foehn climatologies. The “Alpine Research Group
Foehn Rhine Valley/Lake Constance” (AGF) has adopted objective definitions for
both manual (AGF 2007) and automatic (see Sect. 4.6.1) foehn identification.
4.3.3 A Real Example of Foehn Characteristics
Figure 4.6 shows how foehn manifests itself in one of Switzerland’s important
foehn valleys, the Reuss Valley which is part of the so-called Gotthard cross-
section (Fig. 4.7). The rise of temperature, the drop in relative humidity, the onset of
high winds, and the constant wind direction occur simultaneously within minutes.
Consequently, the onset and the breakdown of foehn can easily be determined. This
is true for almost any foehn station that is located in the center of a not too broad
foehn valley.
Fig. 4.6 Example of foehn in Altdorf on January 2–7, 2008 as observed by the surface station.
The horizontal axes are time axes, date and time is indicated below the first frame. TT temperature,
UU humidity, ff wind speed, dd wind direction. The two red vertical lines mark the simultaneous
abrupt changes in the variables at the onset of the main and of a secondary short foehn phase. Note
how clearly foehn can be identified (Data kindly provided by MeteoSwiss)
Figure 4.6 also demonstrates that the breakdown of a foehn occurs usually in
several phases. Or, as one of our colleagues once put it, foehn breakdown is a
gigantic battle between two air masses, namely the foehn and the approaching cold
front. As the foehn weakens, the cold air moves in, only to be pushed back for a
limited period, a back-and-forth process which is repeated several times before the
foehn flow breaks down for good.
Fig. 4.7 The so-called Gotthard cross-section, the only south–north profile with only one ridge,
and valleys practically perpendicular to it. In the south is the Leventina, in the north the Reuss
Valley. The distance between Lugano and Altdorf is almost exactly 100 km
Table 4.1 Characteristic parameters for stations

on the upstream side, the ridge, and the lee side
for March 10, 2008, 1200 UTC (Meteorological
data kindly provided by MeteoSwiss)
Param Lugano Piotta Gütsch Altdorf
z (m) 273 1,007 2,287 449
p (hPa) 970.0 887.1 752.8 943.1
T (ı C) 6.5 2.3 4.5 13.1
™ (ı C) 8.9 11.9 18.1 17.9
Figure 4.8 shows a foehn occurrence, which proves that there are cases which
indeed follow the textbook theory. There is significant precipitation and south wind
on the upwind side. Table 4.1 shows that on the top of the ridge the potential
temperature is markedly higher than on the upwind side (for the location of the
stations see Fig. 4.7). Precipitation on the windward slope alone is not sufficient
proof for the textbook theory, it merely indicates blocking. An increase of the
potential temperature along the upwind slope, however, implies forced advection
and, consequently, a type I foehn.
For the Mesoscale Alpine Program, MAP, special efforts were made to densely
instrument an area for foehn studies in the Rhine Valley (Richner et al. 2005). For
some foehn cases, detailed investigations of the three-dimensional structure of foehn
were made (Drobinski et al. 2003). A general summary of the main results related to
airflow over and in mountains can be found in Drobinski et al. (2007). In these, the
distinction between D> “shallow foehn” and D> “deep foehn” that had evolved
during the last decades was clearly found in the life cycle of the foehn episode.
As expected, compared to simple mountain cross-sections (like the aforementioned
Gotthard cross-section), complex topographical features cause significantly higher
complexity of foehn flow; a glance at the topography shown in Fig. 4.9 makes this
statement understandable.
Fig. 4.8 A foehn case on March 10/11, 2008 with precipitation on the upwind side. The horizontal
axes are time axes, date and time is indicated below the first frame. Top frame: precipitation (RR) in
Lugano (blue) and Magadino (red); next frames show temperature (TT), humidity (UU) and wind
(ff ) for Altdorf (red), Lugano (blue), and Gütsch (green); all data are surface data. The station
Magadino lies 18 km north of Lugano, it is included to give a more representative measure of the
precipitation intensity (Data kindly provided by MeteoSwiss)
Model runs for a MAP foehn case concluded that the simulation outputs
essentially confirm the observed foehn dynamics (Jaubert and Stein 2003). However,
it is also realized that the role of passes and valleys is not fully understood yet.
Fig. 4.9 Topography around the Rhine Valley where foehn studies during MAP were conducted
(Drobinski et al. 2003, reproduced by permission of Wiley and Sons)
Fig. 4.10 TKE values in the

lee of the Alps during foehn.
Yellow: values measured by
aircraft, blue: values
computed by the French
“meso-NH” model (Lothon
2002, reproduced by
permission of Marie Lothon)
At any rate, for a reliable foehn forecast, today’s mesoscale models are not yet
sufficiently accurate, this despite the fact that they do produce realistic flow and
temperature fields. Turbulence, on the other hand, is very poorly reproduced in
model runs. Computed turbulent kinetic energy (TKE) values were orders of
magnitudes lower than those observed by aircraft (Fig. 4.10). A detailed study of
the observed turbulent properties of Alpine foehn events is found in Lothon et al.
(2003).
4.3.4 Foehn from Directions Other than South
As mentioned, Alpine foehn also occurs in the reverse direction, i.e., with a north-
to-south airflow. This north foehn is almost always connected with precipitation
in the north, i.e., north foehn does often follow the textbook theory. In addition
to the difference of the dominating mechanisms, there are additional differences
between south and north foehn. North foehn blows in the direction of the positive
north–south (i.e., meridional) temperature gradient, hence, the warming effect of the
foehn is less pronounced because the ambient temperature in the lee is also higher.
Nevertheless, because of its low humidity, north foehn regularly dries the extended
chestnut forests in southern Switzerland, and forest fires occur recurrently during
north foehn season. Strong north foehn occurs also regularly in the Trentino, in
South Tyrol, and in the Lombardy region of Italy; particularly strong events such as
the one on January 27/28, 2008, are called “foehn bollente” (bollente [Italian] D hot,
boiling).
In the Piedmont region (Northern Italy, west of Milan), foehn occurs with
westerly flow because the Alps bend here towards the south. West winds coming
from France over the Ligurian Alps and the Maritime Alps (i.e., the part of the Alps
between France and Italy, see Fig. 4.1) generate foehn with all its attributes in the
region of Turin (Musso and Cassardo 2004; Di Napoli and Mercalli 2008).
In France, very little research on foehn has been published. Buchot (1977)
compared the foehn winds blowing from northeast to west in the Tarantaise
(basically the region of the Isère Valley in Savoy) with the south foehn in Austria
and Switzerland. From a statistical analysis covering 15 years, he concluded that
there are no significant differences in the characteristics.
4.4 Foehn Dynamics
4.4.1 Leeside Motion
Probably the least understood mechanism in the flow dynamics is the behavior of
the air masses after passing the mountain ridge. Why does the air descend and not
simply continue at the same height level? This question is even more justified by
the fact that its potential temperature is in most cases higher, i.e., that a stable
stratification is present! Steinacker (2006) has compiled the different theories that
have emerged over the last century; the six theories summarized here follow largely
his work; Figs. 4.11–4.16 are slight modifications of those found in Steinacker
(2006).
Probably the first person to write about the problem of the descending air was
Wild (1901), a Swiss physicist who became interested in meteorological problems
because he realized that weather influenced his astronomical observations. Quite
some time later, Streiff-Becker (1933) came up with the so-called vertical aspiration
Fig. 4.11 Schematic representation of the vertical aspiration theory after Streiff-Becker (1933)
Fig. 4.12 Schematic representation of horizontal aspiration theory after Ficker (1931). Solid lines
represent isentropes before, dashed lines isentropes after the aspiration by the low
Fig. 4.13 Schematic representation of lee waves. Solid lines represent isentropes
Fig. 4.14 Solenoid theory according to Frey (1944): effect of the rotational term due to the non-
coincidence of temperature and pressure surfaces
Fig. 4.15 Schematic representation of the waterfall theory according to Rossmann (1950)
Fig. 4.16 Schematic depiction of the hydraulic jump theory according to Schweitzer (1953)
theory (Fig. 4.11), an explanation which builds on the ideas of Wild, that potentially
warmer air moves over colder air. After passing the ridge, it successively removes
the stagnant, potentially colder air by turbulent erosion. By this process, the
potentially warmer air from above sinks gradually into the valleys, replacing the
colder air masses.
In 1912, Ficker published the so-called horizontal aspiration theory or passive
replacement flow theory (Fig. 4.12), a concept that he slightly modified again in
1931 and 1943. It was based on an earlier hypothesis by Billwiller (1899): An
approaching low over the Atlantic sucks the near-ground air masses away, causing
the potentially warmer air above to descend. The topography prevents the advection
of air from the sides. Ficker assumed that air would flow out of the valleys just
like water would flow out of a basin. A south–north oriented pressure gradient
would cause an (ageostrophic!) near-surface flow. Hence, foehn is nothing but
a replacement flow for the air lost in the valleys due to the depression. As a
consequence, all places at a given elevation would simultaneously be subjected to
the adiabatically warmed foehn air that represents this southerly replacement flow.
Observations showed that flow over mountains caused the formation of waves
that extend well beyond the obstacle (see Sect. 3.2.2 by Jackson et al.). Lyra (1940)
and Queney (1948) were the first to present a theoretical analysis of waves in the lee
of a mountain massif. Given the streamlines that emerged from their computations,
the sinking of air masses in the lee could be regarded as a forced downslope motion
(Fig. 4.13). In their often-referenced paper, Klemp and Lilly (1975) provide a
detailed analysis of a strong downslope wind induced by lee waves such as can
occur with foehn. They state that the observed amplification can occur only “if the
upstream wind and stability profiles lie within sharply limited but plausible ranges.”
Hence, depending on temperature, humidity, and wind profile, such a mechanism
may be responsible for the downslope flow in certain cases, but will not in
general.
Based on observations, Frey (1944) developed the so-called solenoid theory
(Fig. 4.14). Measurements showed that temperature in the lee of the Alps increases
towards the north while, at the same time, the pressure falls. Hence, isothermal and
isobaric surfaces are not at the same angle, resulting in a rotational acceleration
(or solenoid). According to Frey’s theory, the kinetic energy of the moving foehn
is not only determined by the pressure gradient, but also to a substantial degree to
the strength of the rotational term; the kinetic energy is sufficient to remove the cold
pool by a plain mechanically forced replacement.
Rossmann (1950) and Schüepp (1952) saw the cause for the descent of the flow
in the D> foehn wall, i.e., in the clouds formed above the crest of the mountain
ridge (Fig. 4.15). They assumed that the air in the cloud is colder and, therefore,
denser than the air outside the cloud. This would result in a downward acceleration,
the kinetic energy to be sufficient to penetrate the lee side valleys. This explanation
became known under the descriptive name waterfall theory.
Using fluid dynamics theory, Schweitzer (1953) proposed the phenomenon of the
“stationary hydraulic jump” as an explanation for the descent of the foehn air after
passing the ridge (Fig. 4.16; see also Sect. 3.2.3 by Jackson et al.). This phenomenon
is observed (and theoretically well understood) on rivers and dams when the water
is discharged at high speed into a region of significantly lower flow speed. The result
of this transition is an abrupt rise of the slower moving fluid.
Obviously, there is no generally applicable theory. Meteorologists have to live
with the fact that various physical mechanisms are needed to explain the same
phenomenon, but occurring under different conditions. It seems that the lee wave
theory is rarely a reasonable explanation, while that hydraulic jump theory is often
a quite plausible description of the observations. Brinkmann (1971) and Hoinka
(1985a) discuss in detail how large amplitude lee waves and/or hydraulic jumps
accelerate foehn flow in valleys perpendicular to the ridge. The solenoid theory
explains quite well thermally driven wind fields such as mountain-valley wind
systems. Whether it is an explanation in the dynamically driven foehn flow is not
clear, here the rotational term could just as well be the result rather than the cause.
The waterfall theory might be useful to explain local effects, however, it can hardly
be regarded as general mesoscale phenomenon controlling the lee side flow. (For
additional theoretical discussions related to these issues, see Chap. 3 Sect. 3.2 in
this book.)
4.4.2 Dimmerfoehn
Under certain conditions, precipitating clouds may be drawn over the ridge, far
into the lee-side valleys. Nevertheless, foehn is present here. As a consequence
of the significant cooling due to evaporation, the dewpoint depression is rather
small, typically 1–7 K instead of the 10–20 K observed normally. An extreme case
happened on November 16, 2002, when there was foehn in Altdorf with 95 mm of
precipitation.
This type of foehn is called D> dimmerfoehn (“dimmerig” or “dimmrig” [Swiss
German] meaning dim, obscure). The airstream aloft shows typically cyclonic shear,
which transports strong condensation from S to N. It was first described in the first
half of last century by Swiss researchers and prompted a fierce and not always
very scientific dispute between Austrian and Swiss meteorologists. The reservation
Austrians had against this foehn type may be explained by the fact that dimmerfoehn
is very rare; there are years without any such case, and it seems that this foehn type
occurs even more rarely in Austria.
Sometimes, the expression “dimmerfoehn” is also used for situations with
southerly flow and strong haze, often caused by advected Saharan dust. Such a case
occurred on April 27, 1993; it is well documented in Burri et al. (1999). (It must be
noted that the definitions found for dimmerfoehn in different publications are not
entirely consistent! The explanation given above is based on the original description
by the Swiss meteorologist Streiff-Becker (1947), it was somewhat modified by
subsequent, more accurate observations.)
4.4.3 Interaction with Cold Pool
When foehn winds descend in the lee of a mountain ridge, they clash with much
cooler, stagnant air. If topography allows, this cold pool is usually flushed away,
typically within hours. However, there are many cases where topography prevents
the outflow of the cold pool; the Wipp and Inn Valley in Austria (with the Nordkette
as downstream obstacle), and the lower Reuss Valley in Switzerland (with the Jura
Mountains as obstacles) are prominent examples for this constellation. In addition,
mere bends in the main valley axis can prevent the cold pool from being flushed out
rapidly.
On the Swiss Plateau, i.e., the area between the Alps in the south and the Jura
Mountains in the north, often a cold pool of a few hundred meters depth remains
while foehn flows over it. In winter, such a situation can persist for up to several
days. The cold pool has a wedge-like form, the angle is typically 2ı . Of course,
air in the cold pool remains calm, while the line where the foehn flow leaves the
ground to flow over the cold air is very gusty. Over water, this border is easily seen
by a spray line and by the haze usually present in the cold air.
As the strength of foehn flow increases, it works its way downstream, gradually
forcing the cold pool back. This can occur by three possible mechanisms or any
combination thereof: (a) heating by convection, (b) erosion of the top by mixing
and entrainment, and (c) static and dynamic displacement.
During MAP, an attempt was made to directly measure the heat flux near the
internal boundary between foehn and cold air by a small aircraft. The daily mean
found was about 15 W m2 , almost exactly the same value that was measured at the
ground. Hence, heating by convection and mixing at the top of the cold pool seem
to be of similar importance for cold pool removal.
4.4.4 Waves
Figure 4.17 depicts the two types of waves which commonly occur with foehn,
namely lee waves at about crest height and above, and shear-induced gravity waves
on top of the cold pool.
Lee waves or simply mountain waves are stationary, hence they belong to the
group of standing waves. Their wavelength is – depending on the width and shape of
the mountain massif – anywhere between 5 and 10 km. Also the amplitudes depend
on the geometry; they are typically of the order of 100–200 m. The amplitudes
are largest at some intermediate level just above ridge height. (See also Chap. 3,
Sect. 3.3.)
In Alpine D> south foehn, the air has passed over the Mediterranean Sea and
picked up substantial moisture. Consequently, when the air reaches saturation by
adiabatic expansion at the wave crests, lens-shaped clouds (cumulus lenticularis or
altocumulus lenticularis, see Fig. 4.18, D> foehn cloud) may be formed. Naturally,
Fig. 4.17 The two wave types occurring with foehn: (a) above crest height the mountain lee waves
(sometimes also called D> foehn waves) and (b) on top of the cold pool shear-induced gravity
waves
Fig. 4.18 Lenticularis clouds over the Monte Rosa Massif near Zermatt in the Swiss Alps
(Reproduced by permission of Andreas Fuchs)
these clouds are also quasi-stationary. Glider pilots were the first to describe lee
waves; this phenomenon enables them to reach very high altitudes by cleverly
utilizing the strong updrafts. However, lee waves can also cause severe turbulence,
particularly when they break or when rotors are formed. Numerous accidents with
small aircraft, commercial aircraft, balloons, paragliders, etc. were caused by lee
waves or by turbulence associated with them.
For a more thorough discussion of lee waves and rotors, see the reports of a recent
study in the Sierra Nevada (Grubišić et al. 2008; Grubišić and Billings 2008), or the
Pennines (Sheridan et al. 2007); the interaction between foehn-type wind and large
amplitude lee waves is treated in detail by Beran (1967). Finally, a broad discussion
is found in Chap. 3, Sects. 3.3 and 3.4.
Fig. 4.19 Waves on cold pool in the Rhine Valley, looking SE. Foehn is from right to left, notice
the weak counter flow in the cold air mass (Reproduced by permission of Andreas Walker)
Fig. 4.20 Waves on the cold pool seen by a sodar. In this time-height diagram, the height covered
is 900 m, the total recording time about 2 h. Observations were made on October 1, 1976,
about 45 km north of Altdorf in the Reuss Valley (at the north end of Fig. 4.7 in the cold pool)
(Reproduced by permission of Werner Nater)
At the internal boundary between foehn flow and cold pool, there is a strong
shear as well as a large temperature gradient. Wind speed in the foehn air might
be typically 20 m s1 while it is virtually zero in the cold pool, the temperature
differences between the two air masses is typically 5–15 K. Shear and different
densities (due to temperature difference) cause gravity waves on top of the cold
pool by the same mechanism by which waves appear on a lake’s surface when the
wind blows over it. Waves on the cold pool can be seen when strong haze or fog is
found in the cold air mass (Fig. 4.19), however, they can be made visible indirectly,
e.g., with sodar, RASS or lidar (Fig. 4.20). These waves produce small fluctuations
in surface pressure (0.1–1 hPa). Depending on the vertical profile of temperature
and wind, their period is somewhat longer than the Brunt-Vaisala period (of the
order of 10–20 min). Unlike lee waves, gravity waves propagate with a phase speed
of roughly half the shear. Obviously, period (measured at a fixed location), phase
velocity and wavelength are interdependent as described by the wave equation.
4.4.5 The Three-Dimensionality of Foehn
While it is the aim of any science to find common mechanisms, to formulate general
laws, describe systematic behavior etc., it must not be forgotten that in reality no two
foehn cases are exactly alike. The case studies presented here were chosen more
or less arbitrarily as “typical” and recent examples. Many disputes happened only
because scientist X adjusted the theory to a specific foehn case while scientist Y
did the same, but (a) in another topography und (b) for another foehn case. Hence,
generalization must always be made with great caution. Insight into processes can
be gained both by statistical analyses of many cases and by detailed case studies.
While systematic statistical analyses of a sufficiently large number of foehn cases is
lacking, there are numerous case studies, many of them very detailed.
In the above discussions, foehn mechanisms are treated in two dimensions only.
For a more precise analysis, the third dimension must, of course, be taken into
consideration, too. However, if this is done, generally valid statements can only be
made for the particular topography and the particular meteorological situation. It is
obvious that tributary valleys do play an important role, and it is to be expected that a
kink in the valley’s axis changes the flow pattern. But each synoptic weather pattern,
and each vertical profile has its distinct effect on the dynamics and thermodynamics
on the air masses involved.
Because foehn represents an airflow more or less perpendicular to a mountain
ridge, and because there are often gaps or passes in the ridges, there is a close
connection between foehn and D> gap flow. The dynamics of foehn-related gap
flow manifests itself in a particularly clear manner along the Brenner cross section,
this because (a) the topography in the ridge area is well suited and (b) because foehn
occurs here sometimes without clouds over the ridge and rarely with heavy upwind
precipitation. Further details on D> gap flow and foehn can be found in Gohm and
Mayr (2004), in Gohm et al. (2004), and in Chap. 3, Sect. 3.5 in this book.
There are several detailed case studies on the three-dimensionality of foehn,
each giving insight into characteristic properties of the case in question. Seibert
(1990) collected south foehn studies that were made since the Alpine Experiment
ALPEX in 1982 in the eastern part of the Alpine massif. Based on these studies
she concluded that the textbook theory is not correct. As discussed above, such a
statement goes too far.
Sprenger and Schär (2001) show how – under D> shallow foehn conditions –
the complex topography of the Alpine ridge can cause flow splitting. Similar studies
were carried out by Zängl (2002). Using observations made during the Mesoscale
Alpine Program MAP, Jaubert and Stein (2003) analyzed and modeled a foehn case
that showed wave breaking and hydraulic jumps. Drobinski et al. (2003) looked at
the scale interactions by which smaller scale topography influences the development
and the characteristics of foehn. Across mountain passes, so-called gap winds occur
in foehn situations. Examples of observations and a discussion of their theory are
given in Chap. 3, Sect. 3.5.
4.5 Foehn Climatology
In Austria, a map showing foehn regions was produced. The intention was to advise
people suffering from discomfort supposedly caused by foehn (see Sect. 4.7.4).
This map (Fig. 4.21) is primarily based on the experience of forecasters and only
minimally on quantitative observations. As mentioned in Sect. 4.3.2, universal
criteria for foehn in a given area are almost impossible to define. For a case with
strong foehn, Gutermann (private communication) has produced a foehn map based
on surface temperature (Fig. 4.22). While this map has the limitation of being a case
study, it does show clearly where the foehn regions are.
Statistical analyses of data from dedicated foehn stations show that foehn
occurrence is highly variable, and that no trend is discernible in the last 140 years
(Fig. 4.23; Richner and Gutermann 2007). In addition, the regional and seasonal
variability is considerable. Figures 4.24 and 4.25 give some ideas about variability.
Observations are made three times a day, morning, noon, and evening; the numbers
refer to the number of such observations.
Fig. 4.21 Foehn regions in Austria and Bavaria. The different shadings represent the frequency
of foehn (arbitrary scale) (Reproduced by permission of the Central Institute for Meteorology and
Geodynamics [ZAMG], Vienna, Austria)
Fig. 4.22 Case study showing the main areas in Switzerland where foehn reaches the surface
(Reproduced by permission of Thomas Gutermann, data kindly provided by MeteoSwiss)
Fig. 4.23 Year-to-year variation of foehn occurrence at the station Altdorf. The vertical axis
represents the sum of daily observations in the morning, at noon, and in the evening, the heavy line
represents the 20-year running mean (Richner and Gutermann 2007, basic data kindly provided by
MeteoSwiss)
Seasonal variations in foehn frequency (upper frame of Fig. 4.25) are caused by
the changing general circulation pattern and, subsequently, by the resulting synoptic
situations. The changes in the diurnal distribution of foehn activity (lower frame of
Fig. 4.25) are the result of interactions of foehn flow with thermally driven local
wind systems, i.e., mountain-valley winds. Note that the relative frequency of foehn
at noon remains the same throughout the year.
Fig. 4.24 Comparison of the yearly foehn occurrence at different Swiss stations. Values represent
the mean yearly sum for the period 1973–1982. The value for Altdorf (ALT) for this period is 62.2
(Richner and Gutermann 2007, basic data kindly provided by MeteoSwiss)
4.6 Forecasting Problems
Foehn forecasting rests on three pillars: (a) probabilistic methods based on a

few observed or forecasted parameters, (b) on model output, and (c) (still very
important!) on the skill of experienced forecasters.
4.6.1 Probabilistic Methods
Forecasting foehn means predicting a mesoscale phenomenon based on a synoptic

situation. Such forecasts are primarily based on the pressure field at different
altitudes. Depending on the orientation of the isobars, foehn might develop in
one valley and not in another, hence, different locations would require specific
forecasting procedures. In the 1960s, Widmer developed for the foehn station
Altdorf, Switzerland, a “foehn test” that was refined by Courvoisier and Gutermann
(1971); it remains the operational tool until today. Two pressure gradients across
the Alps plus the trend of one of these are used to compute an index. If the index
is below a certain, season-dependent threshold value, the probability of foehn at
Fig. 4.25 Seasonal variation of total foehn occurrences per month (upper frame) and relative
frequencies of occurrence at the three observing times (lower frame) for the foehn station Altdorf.
Yellow: morning, red: noon, blue: evening observations (Richner and Gutermann 2007, basic data
kindly provided by MeteoSwiss)
Altdorf within the next 36 h is over 70%. The index allows also predicting the
breakdown of foehn with even a slightly higher success rate. Figures 4.26 and 4.27
give an example of the Widmer Index and its potential for forecasting a foehn case
in January 2008. Note that foehn breakdown is predicted when the Widmer index
has passed though its maximum and starts to decrease again, hence, the prediction
of the breakdown was not very good in this example.
Dürr (2008) developed an automated method for identifying, i.e., nowcasting
foehn. His procedure is based on 10-min real-time data from the automated Swiss
surface network. The most important predictors are the differences in potential
temperature between the reference station Gütsch (2,282 mASL, close to the Alpine
ridge) and the locations for which foehn should be nowcasted. For the time being,
Fig. 4.26 Example of Widmer Index for January 2008. The indices are computed from pressure
differences across the Alps as forecasted for defined gridpoints. Different lines relate to different
forecasts and pressure levels, or combinations thereof. The horizontal line represents the winter
threshold; a value above this indicates foehn at Altdorf (note inverted vertical scale!). Data used
here are the ECMWF and COSMO-7 forecasts of January 1 and 2; ECMWF forecasts for 10
days (dashed and dotted lines), COSMO-7 for 3 days (black and orange lines). COSMO-7 is
the MeteoSwiss mesoscale forecasting model for the Alpine region with 6.6 km resolution (Data
kindly provided by MeteoSwiss)
the application of this technique is restricted to locations on valley floors or near

valley exits; since July 2008, the procedure is implemented as an automated routine
diagnostic tool at MeteoSwiss.
Quite recently, Drechsel and Mayr (2008) developed an objective, probabilistic
forecasting method for foehn in the Wipp Valley (Innsbruck) based on ECMWF
model output. As predictors, they use cross-barrier pressure differences and,
additionally, the isentropic descent. A test over 3 years proved that – based on the
two variables – reliable forecasts of up to 3 days for foehn and the associated gust
winds can be made.
4.6.2 Model Forecasts
As the resolution of models is improved, the topography of the complex Alpine ter-
rain (and with this the foehn valleys) is more accurately represented. Consequently,
Fig. 4.27 Comparison of observed (via humidity) and forecast (using the Widmer index) onset of
foehn; zooming in on Fig. 4.26; January 2–5, 2008 (Data kindly provided by MeteoSwiss)
Fig. 4.28 Wind speed and temperature for the foehn stations Altenrhein (ARH, red) and Vaduz
(VAD, blue). The time series with 10-min resolution are observations (fine lines), the series with
1-h resolution (heavier lines) are model data (Hächler et al. 2011, reproduced by permission of
Klaus Burri)
there are well-founded hopes that models will provide sufficiently accurate foehn
predictions.
At MeteoSwiss, COSMO-2, a 2.2 km grid size model, has been run operationally
since February 2008. Figure 4.28 assesses its capability to analyze a foehn case
(a test case from the pre-operational phase, Burri et al. 2007). For both stations,
the analysis represents the onset of foehn too early. The increase in wind speed
is significantly below the observations, the same is even more pronounced for
temperature.
In summary it can be said that COSMO-2 forecast fields and analysis fields
overestimate the spatial foehn extension and mostly underestimate temperature and
wind speed not only at the two stations investigated here, but at most locations.
The modeled temperature gradient between valley stations and the Alpine ridge
site Gütsch (not shown here) never reached dry adiabatic conditions, this in
contradiction to the observations. In an attempt to improve the situation, not only
deterministic, but also statistical methods will be used the future. In particular,
ongoing work shows that model output statistics (MOS) are a promising tool for
improving foehn forecasting.
4.6.3 Open Problems
In practice, the skills of experienced forecasters who are familiar with the local
situations are still an indispensable prerequisite for a successful forecast. They
know from experience how a somewhat different wind direction might influence
the onset or breakdown of foehn in a given valley, how observed wind data must
be interpreted to arrive at a correct prediction. On the other hand, any tool, be it
based on probability or on model output, is a welcome and appreciated support
giving a first approximation which is subsequently modulated with the forecaster’s
experience and skill.
The synoptic chart and particularly the surface pressure pattern is one of the
most useful tools for predicting the onset of foehn. Here, the forecaster will focus
primarily on the pressure pattern and the formation of the so-called foehn knee
or foehn nose, the characteristic deformation of the isobars on the upwind side
of the Alps (Fig. 4.4). It must be stressed, however, that the formation of the
foehn knee alone is an insufficient indicator for foehn since this is purely a surface
characteristic and does not say anything about the situation aloft. In this respect, the
pressure gradient at ridge height (e.g. from 850 hPa maps) is a much more useful
parameter.
The breaking down of foehn is in most cases associated with an incoming
coldfront from the west. In these cases, forecasting the breakdown can be achieved
quite reliably by observing the progressive increase in surface pressure. If, however,
a high moves in from the east, predicting the breakdown is rather difficult, there
are no clear rules for this case. Nevertheless, it seems that further improvements in
models – both in resolution and parameterization – will improve future forecasting
of foehn.
After the Mesoscale Alpine Program MAP, a few open problems became
obvious. Among these is, as indicated, the role of the tributary valleys which is
still not well understood. This is partly due to a lack of sufficient high-density
observations, but also to a poor understanding of the interaction of air masses
coming from different valleys. There are many locations where two or even more
foehn valleys merge. Observations indicate that the flow does not necessarily merge,
but that one foehn “stream” might cross over the other. Which flow will go over the
other depends on the synoptic fields and the orography. The higher the elevation
at which the winds cross the Alps, the higher are the resulting temperatures in the
valleys due to dry stable structure of the atmosphere in the south. Thus, the warmer
foehn streak (which crossed the ridge at greater height) flows over the colder one.
This effect can be observed, e.g., in the area of the lake Walensee where airflow
down the Rhine Valley (directly towards the lake of Zurich) interacts with airflow
from the Linth Valley in the Glarus Alps. Also here, refined and higher resolving
models might soon provide better solutions.
4.7 Societal Impact of Foehn
4.7.1 Climate Impact
Foehn has serious impacts on the local climate that, in turn, influence agricultural
possibilities in foehn valleys. Thanks to foehn winds, wine can be grown in areas
where it otherwise would be impossible. Foehn storms (also called downslope
windstorms) have also a major effect on the snowmelt, an effect, which is not
particularly liked in skiing resorts.
4.7.2 Air Quality
Foehn situations provide the most spectacular views of the mountains. As a

consequence of the precipitation on the upwind side, the foehn air is usually very
clean, there is no haze, and distant objects seem to be much closer (see Burri et
al. 1999). Likewise, the air mass originating at high levels in foehn type II has
much lower aerosol concentrations than the air it replaces in the valleys. The entire
fantastic Alpine panorama can be seen from places where one normally does not
see any mountains at all. On the negative side of these stunning postcard-views,
however, are the increases in ozone concentration. Although the values seldom
reach alarm levels, foehn air easily triples existing ozone concentrations (Baumann
et al. 2001). Trajectory calculations and aircraft measurements prove that the higher
concentrations are simply due to the descending of ozone-richer air from about
4,000 m over the Mediterranean region. Figure 4.29 depicts the situation for Altdorf
for the foehn case discussed above.
Fig. 4.29 Ozone during foehn at Altdorf in January 2008 (All data taken from an air quality station
that is not collocated with the foehn station. Hence, temperature data is not exactly the same as in
Fig. 4.5.) (Data kindly provided by “inLuft”, Central Switzerland [www.in-luft.ch])
4.7.3 Fires and Traffic Accidents
The most striking danger from foehn events, however, was and still is the spreading
of fires. The warm and very dry air combined with high wind speed supports and
proliferates fires very efficiently. In the course of time, numerous towns have burned
down completely. In 1861, 600 houses of the canton capital Glarus, Switzerland,
were completely devastated during a foehn storm, and only recently, in 2001, a fire
maintained by foehn winds in excess of 15 m s1 destroyed 15 houses in Balzers,
Principality of Liechtenstein. During foehn situations, a few towns still activate a
fire watch during nights, and smoking outside houses is strictly forbidden. In some
mountain regions, it is – as a matter of principle – illegal to start fires outside
specially designated fire areas.
Foehn winds can be dangerous to flying. Professional pilots and local airports
are well aware of the problems and issue the necessary warnings. At Innsbruck
airport, e.g., special foehn procedures ensure that the most turbulence prone parts
of the valley are avoided during climb out and approach in order to enhance safety
and passenger comfort. However, when hot-air ballooning, paragliding, and similar
sports became popular during the last third of last century, the number of severe
accidents due to high winds and large shears increased significantly. Improved
training, special safety courses, and specific information and forecasts have reduced,
but not eliminated, this problem.
Cable car accidents and even train accidents can be caused by strong and gusty
foehn winds. Although all cable transportation systems are required to monitor wind
speed and to have an alarm system, gusts and shears that slip between the different
anemometers can surprise operators. Sadly, there was such an accident in January
2008 in the Jungfrau region. High winds during a foehn storm “derailed” the cable
of a double chair cable lift. First, the lift stopped, and the cable was caught in the
cable catcher, but a successive gust threw the cable out of the catcher, and the chairs
dropped to the ground. One person died and several were severely injured; mean
wind speed was about 25 m s1 which is below the alarm level of 28 m s1 . (An
almost identical incident happened 2003 in Wangs-Pizol, Switzerland. That time the
directly affected gondolas were empty and nobody was injured. The cause was again
gust winds, this time associated with a thunderstorm front.)
On February 15, 1925, a train was thrown from its track in a foehn storm
in Strobl near Salzburg, Austria. Another spectacular accident happened during a
severe foehn storm in the Jungfrau region on November 11, 1996: a double motor
coach of a narrow gauge railway was blown from its track. Its mass of 43 t could
not withstand a foehn gust of 52 m s1 ; four persons were injured, fortunately none
seriously.
On lakes, foehn storms hamper scheduled boat traffic; in extreme cases, opera-
tions have to be suspended. The Swiss town of Brunnen (which is directly in the
main axis of the Reuss Valley) has built a special “foehn harbor.” It serves two
purposes: (a) it is used by boats in storm situations as shelter, and (b) if wind and
waves still allow the scheduled ships to navigate, they can dock in this harbor, which
is better protected against the waves and gusts than the wharf in the center of the
town.
The most significant recurring damage caused by foehn winds is most probably
that to boats, piers, and shores. After severe foehn storms, pictures of loose-torn
boats lying on shores or damaged piers appear regularly in the news. The previously
mentioned “century foehn” on November 8, 1982, destroyed not only numerous
boats and yachts, but also the newly built pier of the town of Sisikon (Lake Uri).
4.7.4 Biometeorological Effects
Still much debated are biometeorological effects of foehn winds. Interestingly, it is

primarily in the Alpine area that people blame foehn winds for almost any ailment,
accident, crime, and in particular for headaches. Sferics (electromagnetic radiation
originating in the atmosphere), ion concentrations, and pressure fluctuations were
considered as possible causes of foehn-related ailments. While recent measurements
prove that neither sferics nor ion concentrations are correlated with foehn events,
pressure fluctuations induced by gravity waves on the cold pool remain a possible
link between foehn and man.
Statistical analyses of pressure fluctuations and subjective well-being show that

there is indeed a statistically significant correlation between the two. However, there
is no proof of any cause-and-effect mechanism. The study of short-term pressure
fluctuations did not take into account the actual weather situation. Since fronts also
cause high amplitude pressure fluctuations, the positive correlation might simply
reflect that people feel subjectively better when the weather is good. A direct
analysis of the frequency of headaches and the occurrence of foehn (as defined in
the Alpine weather statistics) did not produce any result.
Interestingly, on the American continent the interest in biometeorological re-
search seems to be picking up after a long phase of disinterest. There have
been several research projects dealing with Chinook winds and headache, strokes,
etc. However, so far no significant progress was made in relating ailments to
foehn-type winds or weather in general (see, e.g., Cooke et al. 2000; Field and
Hill 2002).
Acknowldgments We thank our colleagues in the “Alpine Research Group Foehn Rhine Val-
ley/Lake Constance” for their contributions and support. In addition, we acknowledge valuable
suggestions from three anonymous reviewers for improvements of our text.
Appendix A: Explanation and Definition of Foehn-Related

Terms
Alpine Foothills Foehn

A warm but humid wind on the Alpine foothills, in the lee the air is mostly calm.
This is not a foehn in the meteorological sense.
Anticyclonic Foehn
Foehn in an anticyclonic situation with a high potential height of the 500 hPa
level. Only partly cloudy in the south, and usually dry. Temperature gain in
the foehn valleys due to adiabatic sinking, flow forced by southerly pressure
gradient.
Cyclonic Foehn
Foehn in connection with a cyclonic regime, i.e. low potential height of the
500 hPa level. Causes cloudiness even in the classical foehn areas, but criteria for
foehn are well pronounced. Develops in extreme cases to dimmerfoehn.
Deep Foehn, High-reaching Foehn
The classical foehn situation with a high-reaching SW flow driven by a high over
northern Italy and a low over southern Germany; isobars show the typical foehn
knee; lee waves, lenticularis clouds, foehn wall, often also lee side rotors are present;
warm, dry winds with high speed reach the valley floors (Hoinka 1990).
Dimmerfoehn
A south foehn which does not immediately follow the topography in the lee, but
touches the surface further downwind. The mountain ridge is in clouds that extend
downwind. The comparatively calm area right downwind of the ridge is dark due to
heavy clouds, hence the name (“dimmerig” or “dimmrig” [Swiss German] means
dim, obscure). In rare cases, there is no precipitation but it is very hazy due to
Saharan dust.
Double Foehn
The simultaneous existence of north and south foehn. Double foehn is a short-
lived, very rare phenomenon which occurs when a cold high-pressure system in the
lower troposphere moves quickly over the Alpine ridge while a stormy westerly flow
is present a greater height (Frey 1986).
Flat Foehn D> Shallow Foehn
Foehn Air
The warm and dry air produced by foehn.
Foehn Bank D> Foehn Wall
Foehn Break D> Foehn Window
Foehn Clearance
At distances larger than about 50 km downwind from the Alpine ridge, foehn
winds become light, while they remain warm and dry. This leads to a significant
reduction of cloudiness, i.e., to a clearing (Hoinka 1985b).
Foehn Cloud
Lenticularis clouds in the lee of the mountain barrier. They are caused by the
lee waves associated with the foehn flow. Their orientation is parallel to the ridge,
and because the lee waves are stationary, also the foehn clouds are quasi-stationary.
Note: Lenticularis clouds are neither a prerequisite for, nor proof of foehn, they can
occur with any airflow over a barrier. The term “foehn cloud” is not be confused
with the D> foehn wall.
Foehn Cyclone
A cyclone in the lee of a mountain massif which is formed or enhanced by a
foehn process.
Foehn Gap D> Foehn Window
Foehn Island
The surface area where foehn touched down while the surrounding area is still
covered with cold air.
Foehn Knee, Foehn Nose, Foehn Wedge
The typical deformation of the isobars (or isohypses) in a foehn situation. A high-
pressure wedge lies on the windward side of the mountain range, while in the lee
a trough is formed. As a consequence, the isobars (or isohypses) bulge and take a
nose-like or knee-like form. In the German literature, “foehn knee” (“Föhnknie”) is
more commonly used, in the English literature the term “foehn nose” (Föhnnase) is
normally found.
Foehn Nose D> Foehn Knee

Foehn Pause
Foehn winds do not blow constantly, but there is varying intensity and sometimes
cessations. Such an interruption is called foehn pause (the German “Pause” means
break, pause, intermission). Foehn pauses occur predominantly during the early
morning hours before sunrise when cold air masses intrude.
Rarely, the term foehn pause is used for the region delimiting the foehn air from
the ambient air, this analogous to “tropopause”, “stratopause”, etc.
Foehn Period
This term deals with the duration or simply with the occurrence of foehn at one
or several stations. The use of the term by different research groups is very diverse,
hence, there is no generally accepted definition.
Foehn Phase
A rather general term used by Alpine foehn researchers for classifying the
different development stages of a foehn situation; the definition of the phases might
differ among different authors. The systems used by Ficker and by Streiff-Becker
(1933) differentiated three phase, the recent system introduced by Seibert (1990)
has four:
Phase I: cold air masses on both sides of the Alps
Phase II: warm advection from W or SW, development of temperature gradient and,
as consequence, pressure gradient across Alpine ridge, D> shallow foehn develops
Phase III: approaching trough changes flow to SW or S, speed and humidity in-
crease, pressure gradient reaches maximum, D> deep foehn develops, precipitation
in the upwind region
Phase IV: passing of cold front, breakdown of foehn, often formation of a cyclone
in the N.
Foehn Storm
A potentially destructive storm as result of a foehn situation. A foehn storm
is characterized by its sudden occurrence in a practically calm situation and its
intermittent nature during its onset. After the initial phase, wind speed remains high
but steady. Additionally, the fast moving air is, as a result of the foehn mechanism,
significantly warmer than the displaced air. Foehn storms belong into the category
of downslope windstorms where they are often referred to as warm downslope
windstorms.
Foehn Tongue
The area covered by foehn air flowing out of foehn valleys.
Foehn Trough
The trough in the lee of the mountain range formed dynamically by the flow
across the barrier.
Foehn Wall, foehn bank
In those situations where foehn occurs with precipitation on the windward side
of the mountain range, the clouds can be seen from the lee side as a “wall” topping
the mountain ridge. Depending on the characteristics of the foehn flow, the clouds
can be dragged over the ridge where they are dissolved while sinking. Under these
conditions, the foehn wall has the appearance of a cap.
Foehn Waves
The lee waves or mountain waves associated with a foehn situation. Sometimes
foehn waves become visible because they produce “foehn clouds”, i.e., cumulus
lenticularis. Foehn waves must not be confused with the gravity waves occurring on
the top of the cold pool.
Foehn Wedge D> Foehn Knee
Foehn Window, Foehn Gap, Foehn Break
A cloud-free gap or clearance in the lee of the mountain range over which a foehn
flow occurs. This gap is caused by the descending foehn air that is dry-adiabatically
heated and, consequently, becomes dry.
Foehnic Situation, Foehn-Like Situation
Southerly flow which does not penetrate into the valleys. Foehn wall, foehn
window, and lenticularis clouds may be present, visibility is very high.
Free Foehn
The sinking of air masses in the free atmosphere on a synoptic scale, particularly
in high-pressure areas. The vertical motion is often brought to a standstill by an
inversion that sometimes tops a fog layer. The increase in temperature and decrease
in humidity caused by the dry-adiabatic sinking leads to dissipation of clouds. This
process has nothing to do with mountain winds, and the use of the term “foehn” in
this context is strongly discouraged.
Gap Foehn (Gap Flow) D> Pass Foehn
High-Reaching Foehn D> Deep Foehn
Light Foehn
Corresponds to phase II of the Innsbruck foehn strength scale (D> foehn phase).
Pass Foehn, Gap Foehn
A pass foehn exists when the foehn criteria are met at least at one mountain or
pass station, but not at a valley station. The distinction between pass foehn and D>
valley foehn was introduced to characterize also foehn situations in which the flow
does not penetrate into the valleys. Another name for a D> foehnic situation.
Sandwich Foehn
A complex and short-lived three-layer flow situation along the Brenner cross
section, first described during the Mesoscale Alpine Program MAP (Vergeiner and
Mayr 2000): a shallow pressure-driven south foehn event below a decoupled flow
from W or even NW, below the foehn flow a very stable cold pool which cannot be
removed by the southerly flow. (In other areas, e.g., over the Swiss Plateau, such a
situation can persist for up to several days; here the term is not used.)
Shallow Foehn, Flat Foehn

A foehn with all its characteristics in the downwind side valley but without the
dominant southerly flow at high altitude, here flow can be from W or even NW.
Shallow foehn can be regarded as compensation flow between different air masses
on both sides of the mountain massif; it is often found in the lee of passes which
form gaps in the mountain barrier (Seibert 1990).
Strong Foehn
Corresponds to phase III of the Innsbruck foehn strength scale (D> foehn phase).
Talfoehn D> Valley Foehn
Valley Foehn, Talfoehn
A valley foehn is present when the criteria for foehn are fulfilled at least for one
mountain station and at least for one valley station. Used in this sense, it is not a
foehn type but a foehn phase. See also pass foehn.
Also, the term “valley foehn” is a somewhat obsolete name for deep foehn used
by Streiff-Becker. He used it to distinguish between deep foehn and Alpine foothills
foehn (the latter not being a foehn in a meteorological sense).
Appendix B: Names of Foehn-Type Winds in Different Regions
The German term Föhn originated most probably during Roman times. When the
Romans conquered the territories north of the Alps in the first century BC, they
found this stormy, warm, and dry mountain wind. From home, they knew a similar
wind, the “favonius” (meaning “the favorable”), a warm and dry wind originating in
northern Africa. Albeit the two winds are meteorologically very different, they do
have similar characteristics; hence, it was logical that the Romans gave this name
also to the warm wind they found north of the Alps. In the Raeto-Roman language
(still spoken today as Rumansch by a minority in the Swiss Alps) “favonius” became
“favuogn”, in its dialect “fuogn”. In the Old High German language “fuogn” became
“phōnno” which over time gradually developed to the German name “Föhn” (the öh
spoken as in b-i-rd or l-ea-rn).
As mentioned in Sect. 4.1, foehn winds in the meteorological sense occur
wherever large mountain massifs exist. In meteorology, the general term “foehn”
for the warm dry wind was chosen, because research started in the Alps, hence, the
local name became also the scientific label. The following list is based on Schamp
(1964), on information collected through personal contacts by Gubser (2006), and
on internet searches.
Aperwind (Alps)
Foehn in Spring which melts the snow (apertus [lat.]: open; apern [ger.]: melting
of snow).
Aspre (Western Slope of Central Massif, France)

Dry, warm easterly winds blowing from the Central Massif over the Garonne
plain, also called “Lou Cantalié”.
Appenzeller Föhn (Eastern Switzerland)
Regional foehn in the foothills of the Alps around Appenzell, St. Gallen, and
Lake Constance. No southerly wind component at the Alpine ridge, but locally
favorable foehn conditions. Occurs usually with westerly winds and falling pressure
north of the Alps.
Austru (Walachia, Romania)
Foehn downwind of the Balkan Massif and the Transylvanian Alps from W
to SW; occurs predominantly in winter, the associated clearing leading to severe
radiation frost.
Autan (Southern France)
The Foehn in the lee of Cevenne Mountains and the Pyrenees from SE to E.
Autan Blanc (Southern France)
Particularly strong D> Autan with cloudless sky; can reach the Atlantic coast
south of the Gironde.
Autan Noir (Southern France)
An D> Autan followed by rain steered by a low over the Bay of Biscay.
Berg Wind (South Africa)
Hot and dry foehn-like wind that blows from the highlands down to the coastal
plateau; particularly Southwest Africa.
Bohorok, Also Bokorot (Eastern Sumatra)
Dry wind blowing from the Karo Plateau down to the plains of northeast Sumatra,
a passat wind warmed by the foehn mechanism.
Boulder Windstorm (Boulder, U.S.A.)
A windstorm occurring basically with D> Chinook in Boulder, CO.
Bregenzer Fallwind (Bay of Bregenz, Austria)
An easterly to northeasterly foehn-type wind blowing form the Gebhardsberg and
Pfänder, also called “Ostföhn” (east foehn), “Falscher Föhn” (wrong foehn, because
it comes from the unusual direction), or “Pfänderwind”.
Broebroe, Brubru (Sulwesi, Indonesia)
A gusty foehn-type easterly wind, part of the east monsoon near Makassar on the
west coast of the island.
Canterbury Northwester D> Northwester
Cape Doctor (Capetown, South Africa)
Maritime air masses, heated by a foehn-type process but still cold, flowing from
False Bay over Capetown towards Table Bay.
Chanduy (Mexico)
Foehn-type wind in the area of Guayaquil, Mexico; occurs primarily during the
dry season in the afternoon.
Chinook (Rocky Mountains, U.S.A.)
Warm and dry foehn-type west wind on the eastern slope of the Rocky Moun-
tains. (In coastal regions of Oregon and Washington also a warm, humid sea breeze
from SW passing over coastal areas, has nothing to do with foehn.) The American
analogue to Alpine foehn, for details of the mechanism see Brinkmann (1973).
Chinook Arch
A striking feature of the D> Chinook, a band of stationary stratus clouds caused
by air undulations over the mountains due to orographic lifting.
Falscher Föhn D> Bregenzer Fallwind
Favogn, Favuogn (Grisons, Switzerland)
Romansh name for foehn.
Gending (Java)
Foehn winds over the northerly plains of Java during the SE monsoons.
Ghilbi (Libya)
Basically a hot desert wind often connected with sand storms. However, when
passing over coastal mountain ranges, there is an additional foehn effect.
Great Wind (Inner Asia)
Foehn-type NE wind on the SW slopes of the Alatau.
Guggifoehn (Bernese Oberland, Switzerland)
The Alpine foehn in the region of the Lauberhorn mountain.
Halny Wiatr (Poland)
A south foehn in the High Tatra.
Helm Wind (Northern England)
A strong foehn-type wind from the Cross Fell, characterized by cloud caps on
the mountain peaks, hence the name.
Himmelsbesen D> Sky Sweeper
Ibe (Western China)
Strong wind through a gap, the Dzungarian Gate (separating the depressions of
Lake Balkhash [Kazakhstan] and Lake Ala-Kul [Kyrgyzstan]). The wind is a gap
wind similar to foehn.
Jauk (Klagenfurt Basin, Austria)
South foehn over the Karawanken mountains.
Kâchan (Sri Lanka)
Foehn-type wind connected with the SW monsoon in the easterly parts of Sri
Lanka. The Tamil speaking population calls it “solaha-kâchan”, i.e., dry, burning
monsoon.
Kumbang, Koembang (Java)

SE wind near Tijriban and Tegal on the north coast, the foehn effect is due to the
Pembarisan mountains.
Lenzbote (Alpine countries)
Name of the spring foehn melting the snow (“Lenzbote” [ger.]: spring messen-
ger).
Levanto (Canary Islands)
Foehn-type E wind particularly in the Orotava valley on Teneriffa.
Livas (Greece)
Any foehn-type wind. The name “Livas” (plural: “Lives”) is derived from “Lips”,
a warm SW wind from the direction of Libya (which is a warm wind but not a foehn-
type wind).
Ljuka (Carinthia, Austria)
Local name of foehn, most likely derived from the Slovenian word “jug” meaning
south wind.
Llebetg, Llebejado (Roussillon and Eastern Pyrenees, France)
Arabic or Catalan name of a foehn-type SW wind on the northern slopes of the
eastern Pyrenees.
Lou Cantalié D> Aspre
Megas (Greece)
A foehn-type SW wind blowing from Parnassus towards Boeotia, particularly in
the plain of the dried-up Kopais lake.
Mikuni-Oroshi (Japan)
The “downwind of Mikunitoge” is a foehn-type W to NW wind in the Tone valley
near Maebashi (eastern coast).
Moazagotl (Riesengebirge, Germany)
The name of a peculiar cloud showing the lee waves. It is formed in the prefrontal
stage of a foehn cyclone.
Canterbury Northwester D> Northwester
Montana Monsoon (Montana, U.S.A.)
Popular name for the D> Chinook in the prairies of Montana.
Northwester, Canterbury Northwester (New Zeeland)
A warm and dry northwestly wind that reaches Canterbury Plain on South Island.
Norder, Norther or Nortes (Gulf of Mexico, Central America)
Actually a cold stormy wind, however, when blowing from the southeast it
passes over the Isthmus of Tehuantepec, the blocking on the upwind side causes
precipitation while in the lee, on the side of the Pacific foehnic clearing occurs.
Ostföhn D> Bregenzer Fallwind

Pacific Wind (Colorado, U.S.A.)
A name sometimes used for the Chinook.
Pfänderwind D> Bregenzer Fallwind
Puelche (Southern Chili)
Foehn wind in the Andes, in Argentina the same wind is called Zonda.
Rotenturm Wind, Talmac Wind (Transylvania, Romania)
A foehn-type south wind near Sibiu (Hermanstadt) blowing from the southern
Carpatians over the Rotenturm Pass into the Transylvanian Basin; sometimes also
called Talmac Wind after the town Nagy-Talmacs near Sibiu.
Santa Ana (California, U.S.A.)
Northeastern foehn wind in the basin of Los Angeles, named after the river and
the pass.
Sky Sweeper (Majorca, Spain)
Foehn-type NW wind over the coastal mountain ranges of Majorca, Spain.
Note that among seamen the term “Sky Sweeper” (or its German translation
“Himmelsbesen”) signifies any northeasterly wind in northerly latitudes. It causes a
very clear view after “sweeping the clouds from the sky.”
Solano (Spain)
Foehn-type wind from south to southeast. (The name is also used for easterly
winds from the Mediterranean Sea bringing monsoon-like precipitation.)
Talmac, Talmesch Wind D> Rotenturm Wind
Tenggara (Speimonde Archipel near Sulawesi, Celebes)
Föhn-type winds during easterly monsoons in the wake of the southern peninsula
of Sulawesi.
Toureillo (Arriège, Southern France)
South foehn from the Pyrenees in the Arriège Valley.
Tsinias (Aegean Sea)
Foehn wind from the southern cliffs of the islands in the Aegean Sea, name is
mainly used on the island of Tinos.
Türkenwind (Northern Tyrol, Austria; Rhine Valley, Switzerland)
Foehn accelerates the ripening of corn (and other crop). In this region, corn is
also called Türkenkorn (Turkish grain, instead of the German “Mais”, hence the
name Türkenwind [Turkish wind]).
Vent d’Espagne (Southern France)
West to southwest foehn-type wind from the Pyrenees when a depression moves
in north of them.
Wambraw, Wambru (New Guinea)

Southwesterly foehn-type wind over the Geelvink Bay in northwestern New
Guinea during summer monsoon.
Wasatch (Utah, U.S.A.)
Easterly Chinook in the valleys of the Wasatch mountains.
Zephyr (Colorado, U.S.A.)
Another name for Chinook.
Zonda (Argentina)
Foehn wind in the Andes, in Chili the same wind is called Puelche.
References
AGF, 2007: http://www.agf.ch

Barry, R.G., 2008: Mountain Weather and Climate. 3 rd edition, Cambridge University Press.
506 pp.
Baumann, K., H. Maurer, G. Rau, M. Piringer, U. Pechinger, A. Prévôt, M. Furger, B. Neininger,
and U. Pellegrini, 2001: The influence of south Foehn on the ozone distribution in the Alpine
Rhine valley – results from the MAP field phase. Atmos. Environ., 35, 6379–6390.
Beran, D.W., 1967: Large amplitude lee waves and chinook winds. J. Appl. Meteorol., 6,
865–877.
Billwiller, R., 1899: Über verschiedene Entstehungsarten und Erscheinungsformen des Föhns.
Meteorol. Z., 16, 204–215.
Brinkmann W.A.R., 1971: What is Foehn? Weather, 26, 230–239.
Brinkmann, W.A.R., 1973: A climatological study of strong downslope winds in the Boulder area.
NCAR, Coop. Thes. 27, Univ. of Colorado, INSTAAR Occasional Pap. No. 6, 229 pp.
Brinkmann, W.A.R., 1974: Strong downslope winds at Boulder. Mon. Weather Rev., 102, 592–602.
Buchot, C., 1977: Le fœhn en Haute-Tarentaise. Rev. géographie alpine, 65, 257-276 l.
Burri K., P. Hächler, M. Schüepp, and R. Werner, 1999: Der Föhnfall vom April 1993. Arbeits-
bericht 196, MeteoSchweiz, 193 pp.
Burri, K., B. Dürr, Th. Gutermann, A. Neururer, R. Werner, and E. Zala, 2007: Foehn Verification
with the COSMO Model. Poster, Int. Conf. Alpine Meteorol., June 4 to 8, 2007, Chambéry,
France.
Cooke, L.J., M.S. Rose, and W.J. Becker, 2000: Chinook winds and migraine headache. Neurology,
54, 302.
Courvoisier, H.W. und Th. Gutermann, 1971: Zur praktischen Anwendung des Föhntests von
Widmer. Arbeitsbericht MeteoSchweiz, 21. (in German, out of print).
Di Napoli, G. and L. Mercalli, 2008: Il vento (capitolo 27). In: Il Clima di Torino. Società
Meteorologica Subalpina, Bussoleno, 629–660. ISBN 978-88-903023-4-3.
Drechsel S. and G. Mayr, 2008: Objective Forecasting of Foehn Winds for a Subgrid-Scale Alpine
Valley. Wea. Forecasting, 23, 205–218.
Drobinski, P., C. Haeberli, E. Richard, M. Lothon, A.M. Dabas, P.H. Flamant, M. Furger, and R.
Steinacker, 2003: Scale interaction processes during the MAOP IOP 12 south foehn event in
the Rhine Valley. Q.J.R. Meteorol. Soc., 129, 729–753.
Drobinski P, R. Steinacker, H. Richner, K. Baumann-Stanzer, G. Beffrey, B. Benech, H. Berger,
B. Chimani, A. Dabas, M. Dorninger, B. Dürr, C. Flamant, M. Frioud, M. Furger, I. Gröhn,
S. Gubser, Th. Gutermann, C. Häberli, E. Häller-Scharnhost, G. Jaubert, M. Lothon, V. Mitev,
U. Pechinger, M. Piringer, M. Ratheiser, D. Ruffieux, G. Seiz, M. Spatzierer, S. Tschannett, S.

Vogt, R. Werner, and G. Zängl, 2007: Foehn in the Rhine Valley during MAP: A review of its
multiscale dynamics in complex valley geometry. Q.J.R. Meteorol. Soc., 133, 897–916.
Dürr, B., 2008: Automatisiertes Verfahren zur Bestimmung von Föhn in Alpentälern. Arbeitsbericht
223, MeteoSchweiz, 22 pp.
Ficker, H., 1920: Der Einfluss der Alpen auf die Fallgebiete des Luftdruckes und die Entstehung
von Depressionen über dem Mittelmeer. Meterol. Z., 55, 350–363.
Ficker, H., 1931: Warum steigt der Föhn in die Täler herab? Meteorol. Z., 48, 227–229.
Field, T.S. and M.D. Hill, 2002: Weather, Chinook, and Stroke Occurrence. Stroke, 33, 1751–1758.
Frey, K., 1944: Zur Entwicklung des Föhns. Verh. Schweiz. Naturforsch. Ges., 124. Jahresvers.
Sils. 90–93.
Frey, K., 1986: The simultaneous occurrence of North and South Foehn during a westerly flow
over Central Europe. Meteor. Atmos. Phys., 34, 349–366.
Gohm, A. and G. Mayr, 2004: Hydraulic Aspects of foehn winds in an Alpine Valley, Quart. J.R.
Meteorol. Soc., 130, 449–480.
Gohm, A., G. Zängl, and G.J. Mayr, 2004: South Foehn in the Wipp Valley on 24 October 1999
(MAP IOP 10): Verification of High-Resolution Numerical Simulations with Observations.
Mon. Wea. Rev., 132, 78–102.
S.A. Cohn, S. Vosper, M. Weissmann, S. Haimov, S.F.J. De Wekker, L.L. Pan, and F.K. Chow,
2008: The Terrain-Induced Rotor Experiment. Bull. Amer. Meteor. Soc., 89, 1513–1533.
Grubišić, V. and B.J. Billings, 2008: Climatology of the Sierra Nevada Mountain-Wave Events.
Mon. Wea. Rev., 136, 757–768.
Gubser, S., 2006: Wechselwirkung zwischen Föhn und planetarer Grenzschicht. Dissertation Nr.
16286, ETH Zürich, 121 pp.
Hächler, P., K. Burri, B. Dürr, T. Gutermann, A. Neururer, H. Richner und R. Werner, 2011: Der
Föhnfall vom 8. Dezember 2006 – Eine Fallstudie. Arbeitsbericht 234, MeteoSchweiz, 52 pp.
Hann, J., 1866: Zur Frage über den Ursprung des Föhns. Z. Österr. Ges. Meteorol., 1, 257–263.
Hann, J., 1901: Lehrbuch der Meteorologie. Tauchnitz, Leipzig, 805 pp.
Hoinka, K.P., 1985a: Observations of the airflow over the Alps during foehn. Quart. J.R. Meteorol.
Soc.,111, 199–224.
Hoinka, K.P., 1985b: What is a Foehn Clearance. Bull. Am. Meteor. Soc., 66, 1123–1132.
Hoinka, K.P., 1990: Untersuchungen der alpinen Gebirgsüberströmung bei Südföhn. Forschungs-
bericht der DLR, FB 90–30, 186 pp.
Jaubert, G. and J. Stein, 2003: Multiscale and unsteady aspects of a deep foehn event during MAP.
Q.J.R. Meteorol. Soc.,129, 755–776.
Klemp, J.B. and D.R. Lilly, 1975: The Dynamics of Wave-Induced Downslope Winds. J. Atmos.
Sci., 32, 320–339.
Lothon, M., 2002: Etude phénoménologique du foehn dans la vallée du Rhin dans le cadre de
l’expérience MAP (Mesoscale Alpine Programme). Thèse, Université Paul Sabatier, Toulouse,
260 pp.
Lothon, M., A. Druilhet, B. Bénech, B. Campistron, S. Bernard, F. Saı̈d, 2003: Experimental
study of five foehn events during the Mesoscale Alpine Programme: from synoptic scale to
turbulence. Q.J.R. Meteorol. Soc.,129, 2171–2193.
Lyra, G., 1940: Über den Einfluss von Bodenerhebungen auf die Strömung einer stabil
geschichteten Atmospäre. Beitr. Phys. frei. Atmos., 26, 197–206.
McGowan H.A., A.P. Sturman, M. Kossmann, and P. Zawar-Reza, 2002, Observations of foehn
onset in Southern Alps, New Zealand, Meteorol. Atmos. Phys., 79, 215–230.
Musso A. and C. Cassardo, 2004: Climatologia del foehn in Piemonte. Nimbus, 31–32, 40–45.
Queney, M. P., 1948: The problem of airflow over mountains: A summary of theoretical studies.
Raphael, M.N., 2003: The Santa Ana Winds of California. Earth Interact., 7, 1–13.
Richner, H. and Th. Gutermann, 2007: Statistical analysis of foehn in Altdorf, Switzerland.
Extended Abstracts Volume 2, Int. Conf. Alpine Meteorol., June 4 to 8, 2007, Chambéry, France,
457–460.
Richner, H., K. Baumann-Stanzer, B. Benech, H. Berger, B. Chimani, M. Dorninger, P. Drobinski,
M. Furger, S. Gubser, Th. Gutermann, C. Häberli, E. Häller, M. Lothon, V. Mitev, D. Ruffieux,
G. Seiz, R. Steinacker, S. Tschannett, S. Vogt, and R. Werner, 2005: Unstationary aspects of
foehn in a large valley, part I: operational setup, scientific objectives and analysis of the cases
during the special observing period of the MAP subprogramme FORM., Meteorol. Atmos.
Phys., 92, 255–284.
Rossmann, F., 1950: Über das Absteigen des Föhns in die Täler. Ber. deutsch. Wetterd. US-Zone,
12, 94–98.
Schamp, H., 1964: Die Winde der Erde und ihre Namen. Franz Steiner Verlag, Wiesbaden, 94 pp.
Schüepp, W., 1952: Die qualitative und quantitative Bedeutung der Föhnmauer. Meteorol. Rdsch.,
5, 136–138.
Schweitzer, H., 1953: Versuch einer Erklärung des Föhns als Luftströmung mit überkritischer
Geschwindigkeit. Arch. Meteor. Geophys. Bioklimatol., A5, 350–371.
Scorer, R.S., 1978: Environmental Aerodynamics. Wiley & Sons, 488 pp.
Seibert, P., 1990: South foehn studies since the ALPEX experiment. Meteorol. Atmos. Phys., 43,
91–103.
Seibert, P., 2005: Hann’s Thermodynamic Foehn Theory and its Presentation in Meteorological
Textbooks in the Course of Time. From Beaufort to Bjerknes and Beyond, Algorismus, Issue
52, 169–180; ISBN 978-3936905-13-7.
Sheridan, P.F., V. Horlacher, G.G. Rooney, P. Hignett, S.D. Mobbs, and S.B. Vosper, 2007:
Influence of lee waves on the near-surface flow downwind of the Pennines. Q.J.R. Meteorol.
Soc., 133, 1353–1369.
Sprenger, M., and C. Schär, 2001: Rotational aspects of stratified gap flows and shallow foehn.
Quart. J.R. Meteorol. Soc., 127, 161–187.
Steinacker, R., 2006: Alpine foehn – a new verse to an old song. Promet, 32, 3–10 (in German).
Streiff-Becker, R., 1933: Die Föhnwinde. Viertelj.schr. Naturforsch. Ges. Zürich, 78, 66.
Streiff-Becker, R., 1947: Der Dimmerföhn. Viertelj.schr. Naturforsch. Ges. Zürich, 92, 195–198.
Vergeiner J. and G. Mayr, 2000: Case study of the MAP-IOP “Sandwich” foehn on 18th October
1999. MAP Newsletter, 13, 36–37.
Wild, H., 1901: Über den Föhn und Vorschlag zur Beschränkung seines Begriffs. Denkschr.
Schweiz. Naturf. Ges., 38, 99 p. plus appendix.
WMO (ed.), 1992: International Meteorological Vocabulary. WMO No. 182, World Meteorologi-
cal Organization, Geneva, Switzerland, 784 pp.
Zängl, G., 2002: Idealized numerical simulations of shallow föhn. Quart. J.R. Meteorol. Soc., 128,
431–450.
Chapter 5
Boundary Layers and Air Quality
in Mountainous Terrain
Douw G. Steyn, Stephan F.J. De Wekker, Meinolf Kossmann,

and Alberto Martilli
Abstract In this chapter, the general problem of air pollution in mountainous

terrain is discussed. Terrain effects on the spatial distribution and temporal vari-
ability of air pollutants are described including specific effects encountered in
stable- and convective boundary layers and under dynamically driven conditions.
The uses of numerical models, scale models, and observational studies are described
as tools for understanding air pollution in mountainous terrain. Specific integrated
regional studies of air pollution conducted in the Lower Fraser Valley, Canada and
in the Upper Rhine Valley, Germany are used to illustrate the terrain effects. The
chapter concludes with a discussion of approaches to forecasting air pollution in
mountainous terrain.
5.1 Introduction
In very simple terms, once emitted, the concentration of air pollutants in the
atmospheric boundary layer is determined by a combination of chemical (production
and destruction) and dispersion processes (the combined effects of dilution by mean
D.G. Steyn ()

Department of Earth and Ocean Sciences, The University of British Columbia,
Vancouver, BC, Canada
e-mail: dsteyn@eos.ubc.ca
S.F.J. De Wekker
Department of Environmental Sciences, University of Virginia, Charlottesville, VA, USA
e-mail: dewekker@virginia.edu
M. Kossmann
Climate and Environment Consultancy, Deutscher Wetterdienst, Offenbach am Main, Germany
e-mail: Meinolf.Kossmann@dwd.de
A. Martilli
CIEMAT Unidad de Contaminacion Atmosferice, Madrid, Spain
e-mail: alberto.martilli@ciemat.es

262 D.G. Steyn et al.
Fig. 5.1 Schematic of air pollution in a valley, illustrating elevated temperature inversion, multiple
pollution sources, pollution plumes and cities with industrial, commercial and residential buildings.
The country breeze is a thermally driven wind resulting from the horizontal temperature gradient
between an urban area and the surrounding country (Modified from Elsom 1987)
wind advection and cross-wind diffusion). In flat terrain, or over the oceans or lakes,
the dispersion processes are limited in the vertical by the depth of the atmospheric
boundary layer, and in the horizontal only by wind speed, wind trajectory sinuosity,
and the intensity of crosswind turbulence. In the near field (generally within 10 km
from a pollution source), diffusion effects dominate while the pollution is still in
a defined plume, while further downwind, boundary layer depth and mean wind
speed become dominant. In mountainous terrain the matter becomes much more
complicated. The atmospheric boundary layer responds to topographic forcing, and
most of these meteorological variables are strongly altered, generally in ways which
result in more severe pollution than would be expected from the same emissions
in flat locations. These conditions are particularly severe in valley bottoms, along
mountain slopes and in mountain basins, as indicated in Fig. 5.1. Many of the
mechanisms which affect air quality in mountainous terrain have been described
in detail in Chaps. 2 and 3, so references are provided here as appropriate.
Humans settle in mountains for economic, strategic, historical or political
reasons. In the year 2000, 720 million people (12% of world population) lived in
mountainous areas, and 663 million of these people lived in developing or transition
countries (Huddlestone et al. 2003), where restrictions on emissions of pollutants
are generally less stringent than elsewhere. It is thus important to understand
the fate and distribution of pollutants in mountainous environments. The history
of pollution episodes shows that many notably polluted locations are in valleys
or mountain basins, or topographically constrained coastal plains. These include
examples like the Plateau of Switzerland, the Po Plain in Italy, The Los Angeles
Basin in USA, Mexico City (also see Chap. 2), the Lower Fraser Valley in Canada
(also see Chap. 8), and the Rhine Valley in Germany. The latter two examples will be
examined as interesting case studies in this chapter, which will touch on particular
atmospheric meteorological phenomena that are important in limiting atmospheric
dispersion in mountainous regions.
5 Boundary Layers and Air Quality in Mountainous Terrain 263
5.2 Boundary Layer Processes Relevant for Air Quality

in Mountainous Terrain
In locations of flat, homogeneous terrain, the boundary layer depth is constant in

space, and varies temporally due to mechanically or thermally induced mixing or
due to stabilization by radiative loss from the surface. In convective conditions,
the boundary layer depth undergoes strong diurnal variation, as shown in Fig. 5.2.
The overall result is that the boundary layer depth is identical to the mixed layer
depth – the depth to which surface emitted substances (including air pollutants) will
be mixed. In regions of complex, mountainous terrain, boundary layer depth varies
strongly in both time and space. In general, boundary layer height will mirror terrain,
but with suppressed amplitude. Local, topographic scale, up- and down-slope flows
(as discussed in Chap. 2, Sect. 2.3) will enhance turbulent transport by adding small-
scale advection to thermally driven convective motions. Most notably, thermally
driven upslope flows will add substantially to the vertical transport of pollutants.
In many cases the net effect will be a transport of pollutants well above the locally-
determined convective boundary layer (CBL) depth to form a regional pollution
layer, often visible as a haze layer blanketing the mountains. This will contrast
strongly with times under which strong surface cooling results in a surface based
temperature inversion and very little turbulent mixing, and with katabatic winds
carrying pollutants or clean air down into valleys and basins. Clearly, different large-
scale weather conditions produce strongly varying pollution potential everywhere,
but particularly in the mountains. This section will elaborate on pollution in stable,
convective and dynamically driven boundary layers in mountainous terrain.
Fig. 5.2 Conceptual model of the diurnal evolution of the convective boundary diurnal in flat
terrain, showing various sub-layers. NBL is nocturnal boundary layer
Fig. 5.3 Schematic representation of processes affecting the dispersion of air pollutants in a stably
stratified boundary layer over complex terrain. (1) drainage flow, (2) slope flow detrainment,
(3) urban heat island circulation, (4) terrain induced flow splitting and convergence, (5) mesoscale
subsidence, (6) regional wind systems, (7) gravity waves, (8) intermittent turbulence (Jerome Fast,
Pacific Northwest National Laboratory, operated by Battelle for the U.S. Department of Energy.
5.2.1 Processes in Stable Boundary Layers
Episodes of severe pollution are frequently associated with poor dispersion

conditions occurring in stable boundary layers. While this association is found
everywhere, mountain induced processes can lead to particularly poor air quality.
A schematic illustration of nighttime air pollution dispersion processes over
mountainous terrain is presented in Fig. 5.3. Most prominent is the negative effect
of cold air pools forming in basin and many valleys during nighttime, as described
in detail in Chap. 2, Sect. 2.4. Such cold air pools are characterized by very stable
stratification, very low wind speeds, and weak or intermittent turbulent mixing.
With human settlements and hence anthropogenic emission sources being mostly
located at low terrain elevations, e.g. at the bottom of basins or along valley floors,
pollutants emitted into cold air pools tend to accumulate to high concentrations.
Particularly hazardous air quality conditions are likely to occur in persistent (multi-
day) cold air pools (Whiteman et al. 2001).
Under synoptically quiescent conditions, thermally driven slope- and valley
winds (Chap. 2) are very important as they maintain some degree of ventilation
in valleys and basins. However, flow convergence associated with cold air draining
from surrounding terrain elevations towards cold pools may favor the occurrence
of stagnant conditions with weak winds (Kossmann and Sturman 2004; Corsmeier
et al. 2006).
Down-slope and down-valley wind systems typically advect relatively unpolluted
air from rural surroundings at higher terrain elevations towards the densely settled
and therefore more polluted valley and basin sites (Banta et al. 1997; Raga et al.
1999; Baumbach and Vogt 1999; Poulos and Zhong 2008). This phenomenon
often results in the increase of concentrations of secondary pollutants such as
ozone because of an influx of precursor substances stored in the residual layer
(Löffler-Mang et al. 1997; Raga et al. 1999; Salmond and McKendry 2002).
Furthermore, down-slope or down-valley winds might re-circulate pollutants that
have been carried away from the emission sources by daytime up-slope or up-valley
winds. An example of mountain and coastal effects on pollutant re-circulation is
described in Sect. 5.4.1 for the Lower Fraser Valley, BC (also see Chap. 8).
In many cases the application of the topographical amplification factor concept
(see Chap. 2, Sect. 2.4) is useful in distinguishing between pooling and draining
valleys, and for a better understanding of the occurrence of valley/basin inversion
break-up (McKee and O’Neal 1989; Sakiyama 1990; Whiteman et al. 1996).
However, forecasting stable boundary-layer processes for air quality applications
still represents a major challenge. Reasons for this include the high spatial model
resolution required to resolve the most important processes and the difficulties with
turbulence parameterization under very stable conditions. Stable boundary-layer
processes in mountainous terrain have also been recognized as important in pro-
viding an understanding of carbon dioxide budgets, which are essential for climate
change monitoring and projections (Sun et al. 1998; Eugster and Siegrist 2000).
5.2.2 Processes in Convective Boundary Layers
Figure 5.4 presents a conceptual model of the diurnal development of the CBL over
mountainous terrain (Fiedler et al. 1987; Whiteman et al. 2000). During the night
and in the early morning hours, cold air accumulates in the valleys, resulting in
greater atmospheric stability in valleys than over the mountain ridges (Fig. 5.4a).
After sunrise, this leads to a slower development of the CBL in the valleys than
over the mountain ridges (Fig. 5.4b). Development of the CBL in valleys is further
suppressed by compensatory sinking motions as a result of mass removal by upslope
flows (e.g. Whiteman 1982; De Wekker 2008; see also Chap. 2). After the inversion
in the valleys has broken up the CBL approximately follows the terrain (Fig. 5.4c).
Further development of the CBL depends on factors such as the total sensible heat
input, the ambient stability and the strength of the synoptic scale sinking motions.
This often leads to parallel CBL growth in the valley and surrounding mountain,
and maintains the terrain-following CBL structure up to early afternoon. By late
afternoon, due to faster CBL development over the valley, the CBL top may become
a roughly horizontal surface (Fig. 5.4d).
This general conceptual model has important implications for air pollution in
mountain regions. Air pollutants that have accumulated in the valleys at night dis-
perse upward during the morning hours. In deep valleys, breakup of the temperature
inversion can take many hours and on some days does not occur, especially in
the winter when incoming radiation is small. Once the surface-based temperature
inversion breaks up, dispersion can occur in several ways, transport by upslope flows
and convective mixing in the growing CBL being the most important of these. Due
to these transport and mixing processes, mountain top locations often experience
increased levels of air pollution after the break-up of surface based temperature
inversions in adjacent valleys (e.g. Seibert et al. 1998).
Fig. 5.4 Conceptual model of the daytime boundary layer development over mountainous terrain.
(a) before sunrise, (b) early morning, (c) at noon, (d) late afternoon. Light grey shading represents
topography. Dark grey shading illustrates strongly stable stratification in nocturnal cold air pools.
Arrows denote horizontal and vertical flows. The wavy lines depict the upper limit of the surface
generated convection. Vertical profiles of potential temperature are shown at several locations along
the cross section (After Fiedler et al. 1987. Reprinted with permission)
The presence of mesoscale circulations and their interactions with the ambient
flow and with boundary layer convection complicates boundary layer structure in
mountainous terrain relative to that found over flat terrain. Very often, multiple
elevated temperature inversions are found in the lower atmosphere over mountains
leading to an ambiguous mixing height. In addition, aerosol layers are often found
above these temperature inversions, making the common definition of mixing
heights in flat terrain inoperable in complex terrain (De Wekker et al. 2004).
Mechanisms that can explain the presence of aerosol (and other pollutants) above
the CBL top are summarized in Fig. 5.5 (De Wekker et al. 2004). In this figure,
aerosol layer (AL) heights are relatively uniform and reach a maximum altitude
of about 4 km in the afternoon. The CBL heights are considerably lower than AL
heights and show more spatial variability. CBL heights tend to follow the terrain
more closely than AL heights, although the extent to which the CBL height follows
5
ha
4
height (km asl)
zi 3 4
1 + 2
3
1
2 5
2
SOUTH NORTH
0
0 10 20 30 40 50
horizontal distance (km)
Fig. 5.5 Conceptual diagram of mid-afternoon CBL in mountainous terrain (indicated by the
shaded grey area). The curve labelled zi is the CBL height and the curve labelled ha is the
AL height. The depicted mechanisms are (1) mountain venting, (2) cloud venting, (3) advective
venting, (4) advection of aerosols from airmasses elsewhere and (5) plume convection that does
not overshoot the CBL top (After De Wekker 2002. Reprinted with permission)
the terrain decreases during the day. The mechanisms contributing to deeper AL
than CBL heights and indicated in Fig. 5.5 are the following: (1) mountain venting,
(2) cloud venting, (3) advective venting, and (4) horizontal advection of aerosols
from airmasses upwind of the area of interest. Mountain venting and cloud venting
can occur simultaneously when an orographically induced thermal reaches lifting
condensation level. All of these mechanisms have been observed to occur in a
variety of field studies. For an overview, see Kossmann et al. (1999).
As noted, observations and model results show that the extent to which the CBL
height follows the terrain typically decreases during the day (Lenschow et al. 1979;
Fiedler 1983; De Wekker et al. 2004). Stull (1998) derived an equation (applicable
under free convection conditions) for the tendency of the CBL top to become more
horizontal in the course of the day by applying the mass conservation equation for a
CBL over hilly terrain. The results showed that advection, entrainment and friction
contribute to making the CBL follow the terrain, while gravitational forces tend
to level the top of the CBL. If there is more than just a light synoptic wind, the
advection term was found to be dominant in making the top of the CBL follow the
terrain. A major shortcoming of Stull’s theory is that it does not take into account
the effect of thermally-driven flows.
5.2.3 Dynamical Processes
The dynamic modification of a larger scale airflow passing over mountainous

terrain plays an important role in the transport, accumulation, and diffusion of air
pollutants. In addition to mean wind speed and direction, turbulence intensities
and structure are also subject to dynamical modification by the terrain. Dynamical
dispersion processes in mountainous terrain include flow over and around hills
and mountains, airflow channeling in valleys, barrier winds, gravity waves, flow
separation, and the formation of traveling or standing eddies with horizontal or
vertical axes (see Chap. 3). Turbulence intensities are often enhanced by flow
separation because of mechanically generated turbulence in wind shear at separation
zones, while highest wind speeds generally occur over exposed convex terrain
formations such as mountain ridges or due to airflow funneling through terrain
constrictions. Lowest mean wind speeds are usually observed at sheltered valley
or leeside locations.
Chapter 3 provides a detailed description of dynamical airflow processes over
mountainous terrain. Furthermore, the current understanding of the well-known
foehn winds (a strong downslope wind) is described in detail in Chap. 4. Foehn
winds are important for the turbulent erosion of cold air pools in valleys and basins
(Petkovsek 1992; Zhong et al. 2003). Hence, the onset of foehn winds often ends
periods of poor air quality. A useful tool to determine whether air pollutants emitted
at a certain height on the windward side of a mountain are likely to impinge on
the mountain slope or pass over a mountain is the dividing streamline concept
(see Chap. 3, Sect. 3.4.2). This concept is based on considerations of the kinetic
and potential energy of the approaching flow via the Froude number (Chap. 3,
Sect. 3.2.3). The Froude number can also be used to give a good indication of
the characteristics of gravity waves possibly developing on the leeward side of a
mountain.
The wind climate in mountain valleys is often dominated by airflow channeling
along the valley axis (Kalthoff and Vogel 1992; Smedman et al. 1996; Eckman
1998; Cogliati and Mazzeo 2006). A detailed discussion of mechanisms generating
airflow channeling is given in Chaps. 2 and 3. The consequences of such channeling
on plume dispersion are demonstrated in Fig. 5.6, which illustrates the dispersion
of plumes in a valley and over the adjacent plain. Over the plain, plumes from
neighboring sources only have intersecting centerlines if the regional wind blows
parallel to the connecting line between the sources. Plumes from sources located
along a valley, however, can be superimposed onto each other for almost all regional
wind directions, because airflow channeling leads to a bimodal wind direction
climatology in valleys. A consequence is that poor air quality situations resulting
from plume interference are far more likely to occur in valleys than over a plain.
Even more complex dispersion conditions occur in curved valleys, or at valley
bifurcations, where airflow channeling is likely to generate zones of converging or
diverging along-valley winds (Enger et al. 1993; Kossmann et al. 2002a; Kossmann
and Sturman 2003; Dobrinski et al. 2006).
Interaction of channeled along-valley winds and airflow across a mountain
barrier potentially causes strong wind shear and often results in eddy formation.
An example of such a dynamically-induced eddy with a horizontal axis is the
rotor which is associated with the existence of lee waves. A favorable location for
dynamical rotor formation, airflow channeling, and dust storm development is the
Owens Valley in California (Grubišić et al. 2008; Zhong et al. 2008). A lidar scan of
Fig. 5.6 Idealised plumes

from two sources within and
two sources outside an
north–south oriented valley.
Circles in red show the
location of sources. Dashed
lines indicate the valley
sidewalls. VG is the
geostrophic wind vector and
V0 is the channeled surface
wind vector in the valley. The
top panel shows terrain
elevations along a vertical
cross section, ‚(z) is the
potential temperature profile
aerosol backscatter intensity across the Owens Valley during strong westerly flows
across the Sierra Nevada Mountains is depicted in Fig. 5.7. The convergence of the
aerosol-free westerly surface winds and of channeled aerosol-filled southerly winds
along the Owens Valley leads to a lifting of aerosols up to 3 km above valley floor
in this situation.
The formation of eddies with a vertical axis and re-circulation of air pollutants
within these eddies is quite common over mountainous terrain. For example, the
Fresno Eddy shown schematically in Fig. 5.8 is generated by the interaction of local
mountain winds in the San Joaquin Valley and a regional low-level jet, which is
triggered by marine air intrusions in the San Francisco Bay area. Re-circulation
over strong emission sources and hence accumulation of air pollutants within the
Fresno Eddy is a very important mechanism for the build-up of high photochemical
pollution levels in the area and in adjacent National Parks (Lin and Jao 1995; Bao
et al. 2008). Such topographically induced eddies with vertical axes can occur under
a wide range of atmospheric stratifications. Observations and modeling studies
Fig. 5.7 Range Height Indicator (RHI) (vertical) lidar scan of aerosol backscatter intensity
captured during the T-REX campaign in the Owens Valley around 0015 UTC 3 March 2006
(After De Wekker and Mayor 2009). The azimuth angle of the scan is 282ı . The white arrow
directed toward the right indicates westerly flow down the east side of the Sierra Nevada slopes.
This relatively clean air undercuts the aerosol-laden southerly flow in the Owens Valley, which has
an easterly component, indicated by the white arrow directed toward the left. The lifting generated
by the near surface flow convergence and the resulting recirculation is illustrated by the black
arrows. The recirculation of the aerosol-laden air lasted for several minutes, as inferred from the
animation of RHI scans. Range rings are drawn at every 2 km distance from the lidar (© American
Fig. 5.8 Formation of eddies with vertical axes in the Central Valley of California (Bao et al.
2008). Recirculation and trapping of polluted air causes poor air quality in the area and often in
surrounding National Parks (© American Meteorological Society. Reprinted with permission)
indicate that conditions with a stably stratified atmosphere and moderate ambient
winds are particularly favorable for the formation of these eddies, which have been
discovered at many different locations around the world. Examples include the
Melbourne Eddy or Spillane Eddy (Spillane 1978; McGregor and Kimura 1989),
the Kanto Plain Eddy (Harada 1981; Kimura 1986), the Graz City Eddy (Oettl
et al. 2001), the Schultz Eddy (Zaremba and Carroll 1999; Bao et al. 2008), and
the Denver Cyclone (Wilczak and Christian 1990; Levinson and Banta 1995).
5.2.4 Interactions and Other Processes
Under most synoptic conditions air pollution dispersion over mountainous terrain
is complex because multiple thermally and dynamically generated processes occur
simultaneously. For example, nighttime down-slope winds may interact with gravity
waves on the lee side of mountains (Poulos et al. 2000), or airflow along valleys
may be characterized by layers of opposing wind directions with either dynamically
channeled winds or thermally driven mountain winds dominating in each layer
(Schmidli et al. 2009). During daytime, atmospheric stability at valley and basin
sites can be strongly modified by the advection of mixed layers formed over
neighboring mountain ranges (Arritt et al. 1992; Stensrud 1993). The advection of
elevated mixed layers and anabatic winds favor the formation of elevated pollution
layers (Kossmann et al. 1999; Frioud et al. 2003; Henne et al. 2004) that may
become subject to long-range transport. Down-mixing of these pollution layers due
to growing CBLs can significantly enhance surface concentrations of air pollutants
(Neu et al. 1994; McKendry et al. 1997).
In most mountain areas surface cover is not uniform. Water surfaces and urban
areas are often present in valley bottoms, for example. These surface heterogeneities
generate thermally driven wind systems and roughness effects which are superim-
posed on mountain induced processes, often resulting in enhanced dispersion (Lu
and Turco 1994; McGowan and Sturman 1996; Bischoff-Gauß et al. 1998; Millan
et al. 1997; McKendry and Lundgren 2000; Ohashi and Kida 2002; Troude et al.
2002; Kalthoff et al. 2004; Bastin et al. 2004; de Foy et al. 2005).
Figure 5.9 shows a situation from central Japan where sea/land breezes and multi-
scale orographically induced circulation systems cause a complex dispersion pattern
of air pollutants emitted at the coast in Tokyo and in neighboring cities on the
Kanto Plain (Kurita et al. 1990). The pollutants are carried inland during daytime
by the sea breeze and anabatic valley and slope winds. Particularly strong heating
in the central mountain basins causes the formation of a thermal low (Kuwagata
and Sumioka 1991) and triggers the onset of plain-to-basin winds (Kimura and
Kuwagata 1993). By the evening, the air pollutants emitted at the coast arrive in
the central mountain basins of Japan where local emissions are low. Air quality
in these remote regions therefore degrades significantly in the evening. During the
night some of the polluted air is transported back towards the coast and possibly
offshore by katabatic winds and land breezes.
Fig. 5.9 Vertical cross section of pollutant transport processes across Central Japan, from the
coastal regions to the mountainous inland region. Broken line denotes the average altitude of the
mountainous central region. Vertical hatching a depression in sea level pressure, PA polluted air,
TL thermal low, CL convergence line, UW upper level wind (only shown for 0600 Japan Standard
Time (JST)), LB land breeze, MW mountain wind, VW valley wind, ESB extended sea breeze, LSW
large scale wind toward the thermal low (Kurita et al. 1990) (© American Meteorological Society.
In densely populated mountain basins with higher emissions than in the sur-
rounding region, plain-to-basin winds are important because they advect relatively
unpolluted air into the basin. However, the advection of cold air into the basin
associated with the plain-to-basin wind might also cause strong reductions in
mixing depth inside basins (Regmi et al. 2003). These effects of plain-to-basin
winds on atmospheric stability and on the transport of clean or polluted air across
mountain ranges or mountain passes have made these winds the subject of several
investigations (De Wekker et al. 1998; Fast and Zhong 1998; Whiteman et al. 2000;
Kossmann et al. 2002b; Zawar-Reza and Sturman 2006).
5.3 Research Tools
Research tools typically employed in the study of air pollution in complex,

mountainous terrain include observations, numerical models and physical (scale)
models. While these tools are distinct, they are generally used in concert to provide
maximum information from a combination of tools and approaches.
5.3.1 Observations
Many technical and logistical challenges exist when attempting to make meteo-
rological measurements in mountainous terrain. Sites are often difficult to access
and/or electrical power is not available. All those challenges recur when adding
measurements of pollutants to meteorological measurements. In addition, the
representativeness of single measurements in mountainous terrain is often for much
smaller areas than corresponding flat terrain measurements because of the large
spatial variability in pollution fields. Mountain top measurements are an exception,
as under certain conditions they are representative of a regional or even larger scale.
The solutions to the limited representativeness of air pollution measurements in
mountainous terrain are exactly those for meteorological measurements – remote
sensing, very dense networks and aircraft borne sensors. Aircraft measurements in
mountainous terrain are made extraordinarily difficult by restricted and crowded
airspace in mountains. Ultimately, the solutions to these difficulties involve a mix
of fixed sensors monitoring the temporal changes (at many locations to also sample
the spatial variability), with mobile platforms to fill in the spatial gaps, and remote
sensing systems to provide broad coverage and data in inaccessible places. Chapter 8
provides a thorough discussion of these observational approaches. Examples of
these solutions are provided throughout this chapter in descriptions of various field
studies.
5.3.2 Numerical Modeling
Difficulties involved in making atmospheric measurements in complex terrain

would imply that much reliance would be placed on modeling as an adjunct to
measurement. In many ways this has happened, but experience has shown that mod-
eling in extremely steep, mountainous terrain has its own set of difficulties. Many of
these difficulties are outlined in Chaps. 9, 10, and 11. Undoubtedly, the overriding
advantage of numerical modeling of air pollution in complex terrain is that model
output is available at locations where measurement would be impractical or even
impossible. This allows investigation of the spatial structure of pollution fields,
rather than the apparent structure, represented by the distribution of monitoring
sites, which are only in locations that allow instrument operation. In many cases,
this is driven by the availability of electrical power to drive the instruments. The
major weaknesses of numerical models of air quality in complex terrain are similar
to those of meteorological models in complex terrain that are used to drive them
(see Chap. 10). Of paramount concern must be a consideration of mesoscale model
grid resolution in relation to the scale of terrain-induced motions. In an ideal world,
the model grid must be designed so that resolved scales of motion are smaller than
dominant topographically induced scales of pollution advection and dispersion. In
far too many cases, this ideal has to be compromised because of the computational
cost of fine resolution grids, or because of the scales of motion being smaller than
the finest resolvable scale of the mesoscale model being used. The technique of
large-eddy simulation has been developed in recent years to allow application to real
complex topography. However, successful application of this technique in weather
and air quality forecasting is far from being realized. Furthermore, an improvement
in spatial resolution in numerical models does not always imply improvements in
simulations of atmospheric processes (see Chap. 10, Sect. 10.2.1). More research is
certainly needed in this area.
5.3.3 Physical Modeling
Physical scale models (either wind tunnels or water tanks) have been used to
study a wide variety of atmospheric phenomena, including many in the field of
air pollution meteorology. Most commonly, wind tunnel studies focus on plume
dispersion, often in complex terrain, or around buildings (Plate 1982). Water tank
studies are primarily used to study density driven phenomena (primarily anabatic or
katabatic winds) in complex terrain (Chen et al. 1996). In many cases these studies
are motivated by an interest in the topographic-scale (therefore smaller mesoscale)
advection of pollutants in regions of complex terrain. Common target phenomena
are the venting of pollutants out of valleys in anabatic flows, or conversely, the
pooling of pollutants in low-lying areas because of katabatic flows.
In common with all physical scale modeling of flow phenomena, the major
challenge is ensuring that dimensionless governing parameters in the scale model
match those in the full-scale world. This generally means that it is not possible to re-
alistically model both turbulent diffusion and mesoscale advection simultaneously.
These difficulties are compensated for by the fact that tank experiments are highly
repeatable and permit measurements that are impossible in the field, and allow easy
control and manipulation of external governing variables.
An example of a tank study of an air pollution phenomenon in complex terrain
is provided by Reuten et al. (2007) who examine pollutant trapping in a growing
morning boundary layer along a steep sided valley wall. The technique has large
untapped potential for studies that are locally specific, as well as generic. Some
more examples of tank studies are provided in Chap. 2, Sect. 2.1.4.
5.4 Examples from Integrated Studies of Air Pollution

in Complex Terrain
The range and complexity of processes discussed so far indicate that local and
regional air quality in mountainous areas could be severely compromised because
of boundary layer meteorological processes found there. The need for scientifically
supported air quality management strategies, and the development of air quality
forecast tools has resulted in integrated, comprehensive air quality studies in many
mountainous areas. These studies all capture, to varying degrees the processes
discussed in isolation in this chapter.
Understanding the effects of these processes on air quality is an important part
of the air quality forecasts that have become increasingly popular and available
in the last few years (Eder et al. 2010). Even though the level of sophistication
of air quality models has increased, their performance in mountainous regions
such as the Appalachian Mountains is poor (Eder et al. 2006). The incorporation
of sophisticated air quality models in daily air quality forecasting follows the
same process as the incorporation of numerical weather prediction models in daily
weather forecasting. This process includes detailed evaluations of the performance
of these forecast models. Model evaluations in mountainous terrain are particularly
challenging for reasons mentioned earlier but they are extremely important to
provide guidance to air quality forecasters. There are different ways in which
such guidance for forecasters could be envisioned. One example would be to
quantitatively assess the effects of terrain on atmospheric processes in case studies
where these effects are shown to contribute to poor forecasts. These effects,
such as modified boundary layer heights, wind channeling, thermally-driven flows,
stagnation, etc., could then be incorporated into the daily forecasts. The detailed
model evaluation that is at the core of improved forecast guidance requires: (1)
significant interactions between operational forecasters and researchers, and (2)
integrated, comprehensive air quality studies in mountainous areas. Such studies
have been performed in the past, primarily driven by the need for scientifically
supported air quality management strategies.
In order to provide a glimpse into the fascinating complexity that is encountered
when some of these processes occur simultaneously, or in sequence, we provide
brief summaries of two examples of mountainous regions whose air quality has
been studied in some detail. In undertaking this exercise, it must be recognized that
human activities of notable size (and therefore resulting in substantial emissions),
are seldom found on mountains, but rather in valleys, basins or plateaus surrounded
by mountains. Our two examples will typify this.
5.4.1 Lower Fraser Valley, BC
The Lower Fraser Valley (LFV) in British Columbia, Canada is home to the city of
Vancouver and its surrounding municipalities, with a total population of roughly
2.2 million people. The valley is roughly triangular, with all three sides being
approximately 100 km in length. The western margin of the valley is the inland
waterway of the Strait of Georgia (Fig. 5.10). The LFV is flat bottomed, bordered to
the north by the Coast Mountains, which rise steeply to an elevation of 2,200 m ASL,
and to the south, by the slightly lower Cascade Mountains of Washington State,
USA. The valley is subject to mild, wet winters with long periods under the influence
Fig. 5.10 Map of the Lower Fraser Valley showing typical ozone plumes. (a) 1988 (pre controls),
(b) 2003 (post controls)
of weak, occluded frontal systems, interspersed by often deep, cold, dry anticyclonic
conditions (Oke and Hay 1994). In summertime, the LFV comes under the influence
of an eastward extension of the Pacific High, resulting in warm, dry conditions,
and very light synoptic winds. Emissions sources in the LFV are primarily due
to the transportation sector, concentrated in the northwest portion of the valley,
there being very little primary industry in the region. Wintertime air pollution is
largely oxides of nitrogen (NOx) and particulate matter (PM), and is closely linked
to transportation corridors. The most important summertime air pollution concern in
the LFV is ozone and secondary PM (Steyn et al. 1997). The occurrence and spatio-
temporal variability of summertime ozone and PM episodes is strongly influenced
by the presence of interacting land/sea breeze circulations, and valley- and slope-
flow systems. The strong recirculation of pollutants in these topographically driven
flows and the presence of very shallow convective boundary layers (Steyn and Oke
1982; Hägeli et al. 2000) lead to far more severe pollution episodes than might
be expected, given the relatively small population and absence of heavy industry.
Contributing to the severity of the episodes is vertical, horizontal and day-to-day
Fig. 5.10 (continued)
recirculation of pollutants as indicated in Fig. 5.11. These three recirculation modes

are driven by a complex interplay of slope flows, land/sea breezes and boundary
layer mixing. These quasi-closed recirculation pathways are characteristic of the
LFV summertime ozone episodes which occur under very weak synoptic forcing,
and contribute to day-to-day buildup of pollutants within an episode. Evidence for
the formation of the elevated polluted layer above the CBL by topographically
induced advection of polluted air is given in Fig. 5.12. Chapter 8, Sect. 8.2 shows
some of the meteorological, and therefore pollutant distribution complexities that
arise from exchange of air between side valleys and the major LFV. Undoubtedly,
pollutants stored in these side valleys by day are recirculated by down-valley flows
at night.
Figure 5.10 indicates typical ozone plumes before and after a program of NOx
and VOC emission controls. The overall shape of both plumes is strongly affected
by topographic barriers, and sea breeze channeling. The eastward (downwind)
displacement of the more recent plume is hypothesized to be a result of changing
NOx /VOC ratios resulting in a less reactive chemical mixture under essentially
unchanged wind speeds (Ainslie and Steyn 2007).
Fig. 5.11 Three modes of pollutant recirculation in the Lower Fraser Valley. (a) Pollutants emitted
into the morning sea breeze are advected inland during the day and return by the landward arm of
the land breeze at night. (b) Pollutants emitted into the evening land breeze area advected seaward
during the night and return by the seaward arm of the sea breeze by day. (c) Pollutants emitted into
the sea breeze by day are advected landward (solid heavy arrow) and carried aloft by upslope flows
into the return flow, and are subsequently carried downward by subsidence (open arrow), entrained
into the convective boundary layer and mixed downwards (small curved arrows)
5.4.2 Upper Rhine Valley
The Upper Rhine Valley located in south western Germany, eastern France, and
northern Switzerland is a densely populated and industrialised area where complex
terrain meteorology and air quality issues have been of great interest for many years
(e.g. Fiedler and Prenosil 1980; Fiedler et al. 1987; Fiedler 1992a, b; Fiedler and
Zimmermann 1993; Fiedler and Borell 2000).
In September 1992, the TRACT (Transport of Air Pollutants over Complex
Terrain) field campaign was conducted in the Upper Rhine Valley providing a
dataset with a high temporal and spatial resolution of the CBL over mountainous
terrain (Fiedler 1992a; Fiedler and Borell 2000). Measurements from rawinsonde,
tethersonde, research aircraft and sodar (see Fig. 5.13) on a cross section through the
Black Forest Mountains were used to determine the spatial and temporal behavior of
the CBL (De Wekker 1995; Kossmann et al. 1998; Kalthoff et al. 1998). The terrain
is characterized by a sharp transition from the Rhine Valley to the Black Forest
region, with a height difference of roughly 1,000 m over a horizontal distance of
Fig. 5.12 Polluted boundary layer in LFV indicated by lidar backscatter superposed on profiles
of potential temperature (white solid line) and specific humidity (white dashed line) and inferred
AL height (dark line) measured on July 26, 2001 from 1314 to 1355 Pacific Standard Time. The
mountain slope can just be seen rising to the right (north) of the figure. Sodar data indicate winds
with a slight northerly component above the CBL top (Reuten et al. 2005) (© Springer. Reprinted
with permission)
about 10 km. The analyses showed that during the morning hours, the mixed layer
top followed the underlying terrain but generally tended to become more level,
i.e., less terrain-following, over the course of the day. In summary, three types of
afternoon CBL behavior were identified (Fig. 5.14). In the first type (typified by
panel 1), the mixed layer over the Rhine Valley reached heights of about 2,000 m
and the top of the mixed layer was relatively level over the investigation area. In
the second type (typified by panel 3), boundary layer development in the Rhine
Valley was delayed by the presence of fog, and the afternoon mixed layer stayed
well below the mountain height. Consequently, the top of the mixed layer more or
less followed the steep western slope of the Black Forest region in the afternoon.
In the third type (typified by panel 2), the behavior was intermediate between the
other types, with mixed layer heights over the Black Forest region a few hundred
meters higher than over the Rhine Valley. A common feature of all three types
of behavior is the presence of a relatively horizontal capping inversion over the
Black Forest region. The mixed layer top obviously does not follow the individual
small-scale ridges and valleys in the afternoon. Overall, smaller afternoon CBLs
were observed over the mountain top site Hornisgrinde (the highest elevation in
the investigation area) than in the Rhine Valley and the Black Forest region east of
the Hornisgrinde. Applying one-dimensional mixed layer growth rate models for a
valley and the mountain top site showed that for the valley site, relatively good cor-
respondence was found with observations but that the mixed layer growth rates for
Fig. 5.13 Vertical cross sections of potential temperature, specific humidity and horizontal wind
vectors from the foothills of the Vosges Mountains (left) through the Upper Rhine Valley (center)
to the Black Forest Mountains (right) at around 1300 Central European Summer Time (CEST)
of 11 September 1992 (De Wekker 1995). White lines in the upper two panels represent balloon
ascent paths or aircraft flight tracks
Fig. 5.14 Schematic

representation of three
different types of afternoon
mixed layer height (zi )
behaviour along a
cross-section from the Upper
Rhine Valley to the northern
Black Forest derived from
observations of the TRACT
field campaign (De Wekker
et al. 1997)
the mountain top site were clearly overestimated by these models (De Wekker 1995;
Kossmann et al. 1998). De Wekker (1995) and Binder (1997) proposed the use of an
effective sensible heat flux for use in growth rate models that implicitly takes into
account the effects of advection, orography shape and other effects that influence
CBL growth over mountain ridges.
The large density of surface-based meteorological stations during TRACT
allowed the identification of major wind systems in the Upper Rhine Valley and
in some of the adjacent side-valleys (Fig. 5.15). The wind systems in the side
valleys show a pronounced diurnal cycle with upvalley (downvalley) and upslope
(downslope) winds during the day (night). The influence of these flows on the ozone
concentrations during TRACT was studied by Löffler-Mang et al. (1997). Stations
in side-valleys of the upper Rhine Valley often showed less nocturnal ozone removal
(compared to stations on flat terrain) or even a clear secondary ozone maximum in
the night due to down-valley advection of ozone-rich and weakly polluted air from
the mountains.
60
50
40
y in km
3020
10
0
0 10 20 30 40 50 60 70
x in km
Daytime wind direction 10%

frequency distribution Height in m MSL
Fig. 5.15 Frequency distribution of daytime wind direction in the period of the TRACT field
campaign (7–21 September 1992), showing channeling and slope flows in the Upper Rhine Valley
(Kossmann 1998)
The behavior of ozone and it precursors was also investigated in the Schauinsland
Ozone Precursor Experiment (SLOPE) in 1996 in the southern part of the Upper
Rhine Valley near Freiburg (Kalthoff et al. 2000; Fiedler et al. 2000). Meteorological
and chemical measurements were carried out at several stations along the expected
transport path between Freiburg in the Rhine Valley and Schauinsland, a mountain
top location in the Black Forest. Results showed that not only mixing layer growth,
but also flow splitting and mountain venting processes, induced on the slopes and
in the side valleys play an important role in explaining the dispersion of pollutants
up the Black Forest Mountains. Kossmann et al. (1999) showed similar results for
the northern part of the upper Rhine Valley. Another important result from SLOPE
is the strengthening and lowering of an elevated inversion caused by the advection
associated with a thermally driven upvalley wind system. This process confined the
air pollutants to a shallow layer near the surface and degraded the air quality. Similar
interactions between thermally driven wind systems and CBL depth have been noted
by Kossmann et al. (1998) and De Wekker (2008).
The along-valley flow in the Upper Rhine Valley, which is evident in Fig. 5.15,
is mainly due to pressure-driven channeling (Wippermann 1984; Gross and
Wippermann 1987; Kalthoff and Vogel 1992). This type of channeling has been
given much attention in the case of the upper Rhine Valley, and has important
consequences for the dispersion of air pollutants (see Sect. 5.2.3).
5.5 Conclusion
Throughout this chapter we have stressed that air quality in regions of complex,
mountainous topography is frequently (and quite often severely) degraded because
of meteorological processes and phenomena related to the topography. While we
have chosen only two regions as examples to illustrate this, there exist many other
locations that experience degraded air quality because of topographic effects. The
difficulties of both atmospheric measurement and modeling in complex terrain
mean that there will always be substantial uncertainty over air pollution problems
in complex terrain. This uncertainty translates into difficulties in both forecasting
pollution, and providing scientific guidance to policy makers.
The current practice for routine air quality forecasts is to employ numerical
models operating over regional and continental scales, as exemplified by Kang et al.
(2009). The horizontal resolution of such forecasts is generally 10–20 km, a scale at
which most of the phenomena discussed in this chapter are simply not resolved. This
means that anyone wanting to forecast air quality in a mountainous region will have
to take the forecast provided by model output, and interpret or modify it taking into
account the phenomena and effects covered in this chapter. As these models develop
and increase their resolution, and their ability to explicitly capture the phenomena
discussed in this chapter, such forecaster intervention will become less important. It
is unlikely however that it will ever become unnecessary.
It must be emphasized that the quality of an air pollution forecast based on
numerical model output will depend on a number of factors. The first among these
will be the requirement of accurate model initialization and boundary conditions
for both meteorology and atmospheric chemistry. These may be provided through
data assimilation schemes, but chemical data assimilation requires measurements of
many chemical species, which generally are not measured with sufficient detail. The
second requirement is for accurate spatially, temporally and chemically resolved
emissions data. Such data are generally not available in sufficient resolution. This
difficulty is in most cases the major factor limiting the accuracy of air quality
forecasts. The third requirement is for detailed and chemically realistic models of
aerosol chemistry and physics. This matter is a topic for much current research,
and therefore a source of much uncertainty in the context of operational air quality
forecasts. These three requirements are seldom uniformly met, and so efforts to
determine the source of deficiencies in air quality forecasts are made especially
difficult because of interactions between multiple sources of model weaknesses. It is
often unclear whether inadequacies in air quality models are due to meteorological,
chemical or emissions (or all three) weaknesses.
An additional difficulty arises because numerical models needed for air quality
forecasts must perform both meteorological and chemical transformation calcula-
tions, resulting in a great increase in computational demand over pure meteorolog-
ical modeling. The result often is that the spatial resolution is sacrificed in order to
allocate computational resources to the chemical modeling. The resulting reduction
in spatial resolution can result in mountain induced meteorological processes being
inadequately resolved.
Another consequence of the uncertainty surrounding air quality in mountainous
terrain is that most studies of air pollution in such terrain combine both observations
and modeling as approaches to building understanding. Both approaches will always
be needed, with observations to tie model output to measured reality, and models
to fill in the unmonitored spaces. Many phenomena become evident only through
observation, and then subsequently, are better understood through the use of models,
which then become the most important tools for making predictions, and exploring
scenarios. Models and observations are tied in an upward spiraling process of
development, and nowhere is this as evident as in studies of air pollution in
complex terrain. In this way, the increasing demands of air quality forecasting will
inspire further development of new parameterizations of meteorological processes,
which will in turn make possible more accurate air quality forecasts in complex,
mountainous terrain
Acknowledgements The authors would like to thank the three anonymous reviewers and the book
editors for their helpful comments that improved this chapter. We also would like to thank Prof.
Franz Fiedler from the ‘Institute for Meteorology and Climate Research’ of the ‘Karlsruhe Institute
of Technology’ (Germany) for his permission to use Fig. 5.4, Dr. Jerome Fast for his permission to
use Fig. 5.3, and Eric Leinberger who drafted Figs. 5.1 and 5.11.
References
Ainslie, B and D.G. Steyn, 2007: Spatiotemporal Trends in Episodic Ozone Pollution in the Lower
Fraser Valley, B.C. in Relation to Meso-scale Atmospheric Circulation Patterns and Emissions.
J. Appl. Meteorol. 46 (10), 1631–1644.
Arritt, R.W., J.M. Wilczak, and G.S. Young, 1992: Observations And Numerical Modeling Of An
Elevated Mixed Layer. Mon. Wea. Rev.,120, 2869–2880.
Bastin, S., P. Drobinski, A. Dabas, P. Delville, O. Reitebuch, and C. Werner, 2004: Impact of
the Rhône and Durance valleys on sea-breeze circulation in the Marseille area, Atmospheric
Research 74, 303–328.
Banta, R.M., P.B. Shepson, J.W. Bottenheim, K.G. Anlauf, H.A. Wiebe, A. Gallant, T. Biesenthal,
L.D. Olivier, C-J. Zhu, I.G. McKendry, and D.G. Steyn, 1997: Nocturnal Cleansing Flow In
Tributary Valley. Atmos. Env. 31, 2147–2162.
Bao, J.-W., S.A. Michelson, P.O.G. Persson, I.V. Djalalova, and J.M. Wilczak, 2008: Observed And
WRF-Simulated Low-Level Winds In A High Ozone Episode During The Central California
Ozone Study. J. Appl. Meteorol. Climatol., 47, 2372–2394.
Baumbach, G. and U. Vogt, 1999: Experimental Determination Of The Effect Of Mountain-Valley
Breeze Circulation On Air Pollution In The Vicinity Of Freiburg. Atmos. Env. 33, 4019–4027.
Binder, H.-J., 1997: Tageszeitlich Und Räumliche Entwicklung Der Konvektiven Grenszicht Über
Stark Gegleidertem Gelände. Dissertation, Institut für Meteorologie und Klimaforschung,
Karlsruhe, 310p.
Bischoff-Gauß, I., N. Kalthoff, and F. Fiedler, 1998: The Impact Of Secondary Flow Systems On
Air Pollution In The Area Of Sao Paulo. J. Appl. Meteorol., 37, 269–287.
Chen, R.-R., N. S. Berman, D. L. Boyer, and H. J. S. Fernando, 1996: Physical model of diurnal
heating in the vicinity of a two-dimensional ridge, J. Atmos. Sci., 53(1), 62–85.
Cogliati, M.G. and N.A. Mazzeo, 2006: Air Flow Analysis In The Upper Rio Negro Valley
(Argentina). Atmos. Res., 80, 263–279.
Corsmeier, U., M. Kossmann, A.P. Sturman, and N. Kalthoff, 2006: Temporal Evolution Of Winter
Smog Within A Nocturnal Boundary Layer At Christchurch, New Zealand. Meteorol. Atmos.
Phys., 91, 129–148.
de Foy, B., E. Caetano, V. Magana, A. Zitacuaro, B. Cardenas, A. Ratema, R. Ramos, L.T. Molina,
and M.J. Molina, 2005: Mexico City Basin Wind Circulation During The MCMA-2003 Field
Campaign. Atmos. Chem. Phys., 5, 2267–2288.
De Wekker, S.F.J., 1995: The Behaviour Of The Convective Boundary Layer Height Over
Orographically Complex Terrain. MS thesis University of Karlsruhe, Germany/Wageningen
Agricultural University, The Netherlands, 74 pp.
De Wekker, S.F.J., Kossmann, M., Fiedler, F., 1997: Observations Of Daytime Mixed Layer Heights
Over Mountainous Terrain During The TRACT Field Campaign. 12th AMS Symposium
on Boundary Layers and Turbulence, 28 July – 1 August 1997, Vancouver, B.C., Canada,
498–499.
De Wekker, S.F.J., S. Zhong, J.D. Fast, and C.D. Whiteman, 1998: A Numerical Study Of
The Thermally Driven Plain-To-Basin Wind Over Idealized Basin Topographies. J. Appl.
Meteorol., 37, 606–622.
De Wekker, S.F.J., 2002: Structure and Morphology The Convective Boundary Layer in Mountain-
ous Terrain. PhD thesis The University of British Columbia, Canada, 191 pp.
De Wekker, S.F.J., D.G. Steyn, and S. Nyeki, 2004: A Comparison Of Aerosol Layer- And
Convective Boundary Layer Structure Over A Mountain Range During STAAARTE’97.
De Wekker, S.F.J., 2008: Observational And Numerical Evidence Of Depressed Convective
Boundary Layer Heights Near A Mountain Base. J. Appl. Meteorol. Climatol., 47, 1017–1026.
De Wekker, S.F.J. and S. D. Mayor, 2009: Observations of atmospheric structure and dynamics in
Meteorol. climatol., 48, 1483–1499.
Dobrinski, P., S. Bastin, J. Dusek, G. Zängl, and P.H. Flamant, 2006: Flow Splitting At The
Bifurcation Between Two Valleys: Idealized Simulations In Comparison With Mesoscale
Alpine Programme Observations. Meteorol. Atmos. Phys., 92, 285–306.
Eckman, R.M., 1998: Observations And Numerical Simulations Of Winds Within A Broad
Forested Valley J. Appl. Meteorol., 37, 206–219.
Eder, B., D. Kang, R. Mathur, S. Yu, and K. Schere, 2006: An operational evaluation of the Eta-
CMAQ air quality forecast model. Atmos. Env. 40, 4894–4905.
Eder, B., D. Kang, S. Trivikrama Rao, R. Mathur, S. Yu, T. Otte, K. Schere, R. Wayland, S. Jackson,
P. Davidson, J. McQueen, and G. Bridgers, 2010: Using National Air Quality Forecast
Guidance to Develop Local Air Quality Index Forecasts. Bull. Amer. Met. Soc. 2010; 91:
313–326.
Elsom, Derek, 1987: Atmospheric Pollution: Causes, Effects, and Control Policies. Blackwell,
Oxford, UK, 1987, 319 p.
Enger, L., D. Koracin, and X. Yang, 1993: A Numerical Study Of Boundary-Layer Dynamics In A
Mountain Valley. Bound.-Layer Meteor., 66, 357–394.
Eugster, W. and F. Siegrist, 2000: The Influence Of Nocturnal CO2 Advection On CO2 Flux
Measurements. Basic Appl. Ecol., 1, 177–188.
Fast, J. D. and S. Zhong, 1998: Meteorological Factors Associated With Inhomogeneous Ozone
Concentrations Within The Mexico City Basin. J. Geophys. Res., 103, 18927–18946.
Fiedler, F.,1983: Einige Charakteristika der Strömung in Oberrheingraben Wissenschafliches
Berichte des Meteorologischen Instituts der Universität Karlsruhe. Nr. 4, 1130123.
Fiedler, F., Adrian, G., Hugelmann, C.P., 1987: Beobachtete Phänomene Während des TULLA-
Experimentes [Phenomena Observed During the TULLA Experiment], 3 rd PEF Statuskollo-
quium, 10–12 March 1987 Karlsruhe, KfK-PEF Vol. 12, Issue 2, 347–365.
Fiedler, F., 1992a: TRACT Operational Plan, Institut für Meteorologie und Klimaforschung,
Karlsruhe, 66p.
Fiedler, F., 1992b: Das Regio-Klima Projekt: Wie Regeln die Natürlichen Energieumsetzungen das
Klima in Einer Region. KfK-Nachr., Jahrg. 24, 3/92, 125–131.
Fiedler, F. and Prenosil, 1980: Das Mesoklip-Experiment, mesoskaliges Klimaprogramm im Ober-
rheintal. Wissenschaftliche Berichte des Meteorologischen Instituts der Universität Karlsruhe,
Nr.1.
Fiedler, F. and Zimmermann, H., 1993: General Report of TRACT, an Overview. in: Annual report
1992 EUROTRAC subproject TRACT, International Scientific Secretariat, Fraunhofer-Institut
für Atmosphärische Umweltforschung, Garmisch-Partenkirchen, 3–34.
Fiedler, F. and P. Borell, 2000: TRACT: Transport of Air Pollutants over Complex Terrain.
In; S, Larsen et al. (Eds.), Exchange and Transport of Air Pollutants over Complex Terrain
and the Sea, Springer-Verlag, 223–268.
Fiedler, F., I. Bischoff-Gauß, N. Kalthoff, and G. Adrian (2000), Modeling Of The Transport
And Diffusion Of A Tracer In The Freiburg-Schauinsland Area, J. Geophys. Res., 105(D1),
1599–1610.
Gross, G., and F. Wippermann, 1987: Channeling and Countercurrent in the Upper Rhine Valley:
Numerical Simulations. J. Appl. Meteorol., 26, 1293–1304.
Frioud, M., V. Mitev, R. Matthey, C. Häberli, H. Richner, R. Werner and S. Vogt, 2003: Elevated
Aerosol Stratification Above The Rhine Valley Under Strong Anticyclonic Conditions, Atmos.
Env. 37, 1785–1797.
Grubišić, V., Doyle, J.D., Kuettner, J., Mobbs, S., Smith, R.B., Whiteman, C.D., Dirks, R., Czyzyk,
S., Cohn, S.A., Vosper, S., Weissmann, M., Haimov, S., De Wekker, S.F.J., Pan, L.L., and Chow,
F.K., 2008: The Terrain-Induced Rotor Experiment – A Field Campaign Overview Including
Observational Highlights. Bull. Amer. Met. Soc., 89, 1513–1533.
Hageli, P., D.G. Steyn and K. Strawbridge, 2000: Spatial and Temporal Variability of MLD and
EZT. Boundary Layer Meteorol., 97 (1), 47–71.
Harada, A., 1981: An Analysis Of The Nocturnal Cyclonic Vortex In The Planetary Boundary
Layer Of The Kanto Plains. J. Met. Soc. Japan 59, 602–610.
Henne, S., M. Furger, S. Ny eki, M. Steinbacher, B. Neininger, S.F.J. De Wekker, J. Dommen,
N. Spichtinger, A. Stohl, and A. S. H. Prévôt, 2004: Quantification Of Topographic Venting Of
Boundary Layer Air To The Free Troposphere. Atmos. Chem. Phys., 4, 497–509.
Huddlestone, B., E. Ataman, P. de Salvo, M. Zanetti, M. Bloise, J. Bel, G. Franceschini and L. Fè
d’Ostiani, 2003: Towards a GIS-Based Analysis of Mountain Environments and Populations.
FAO, UN, Rome. 26p.
Kalthoff, N. and B. Vogel, 1992: Counter-Current And Channelling Effect Under Stable Stratifica-
tion In The Area Of Karlsruhe. Theor. Appl. Climatol., 45, 113–126.
Kalthoff, N., C. Kottmeier, J. Thürauf, U. Corsmeier, F. Sad, E. Fréjafon and P.E. Perros, 2004:
Mesoscale Circulation Systems And Ozone Concentrations During ESCOMPTE: A Case Study
From IOP 2b. Atmos. Res., 74, 355–380.
Kalthoff, N., H.-J. Binder, M. Kossmann, R. Vögtlin, U. Corsmeier, F. Fiedler, and H. Schlager,
1998: Temporal Evolution And Spatial Variation Of The Boundary Layer Over Complex
Terrain. Atmos. Env., 32, 1179–1194.
Kalthoff, N., V. Horlacher, U. Corsmeier, A. Volz-Thomas, B. Kolahgar, H. Geiß, M. Möllmann-
Coers, and A. Knaps (2000), Influence Of Valley Winds On Transport And Dispersion
Of Airborne Pollutants In The Freiburg-Schauinsland Area, J. Geophys. Res., 105(D1),
1585–1597.
Kang, D., R. Mathur, and S. Trivikrama Rao, 2009: Assessment of bias-adjusted NAQFC PM2.5 air
quality forecasts over the continental United States during 2007 Geosci. Model Dev. Discuss.,
2, 1–32.
Kimura F., 1986: Formation Mechanism of the Nocturnal Mesoscale Vortex in Kanto Plain. J. Met.
Soc. Japan, 64, 857–869.
Kimura, F. and T. Kuwagata, 1993: Thermally Induced Wind Passing From Plain To Basin Over A
Mountain Range. J. Appl. Meteorol., 32, 1538–1547.
Kossmann, M. and A.P. Sturman, 2003: Pressure-Driven Channeling Effects In Bent Valleys.
J. Appl. Meteorol., 42, 151–158.
Kossmann, M., 1998: Einfluss Orographisch Induzierter Transportprozesse Auf Die Struktur Der
Atmosphärischen Grenzschicht Und Die Verteilung Von Spurengasen. Dissertation, Physics
Faculty, University of Karlsruhe, Germany, 193 pp.
Kossmann, M., R. Vögtlin, U. Corsmeier, B. Vogel, F. Fiedler, H. -J. Binder, N. Kalthoff, and
F. Beyrich, 1998: Aspects Of The Convective Boundary Layer Structure Over Complex Terrain.
Atmos. Env., 32, 1323–1348.
Kossmann, M., Whiteman C.D. and Bian X., 2002a: Dynamic Airflow Channeling Over The Snake
River Plain, Idaho. Proceedings Tenth Conference on Mountain Meteorology, 17–21 June
2002, Park City, Utah, Amer. Meteor. Soc., 360–363.
Kossmann, M., A.P. Sturman, P. Zawar-Reza, H.A. McGowan, A.J. Oliphant, I.F. Owens, and
R.A. Spronken-Smith, 2002b: Analysis of the Wind Field and Heat Budget in an Alpine Lake
Basin during Summertime Fair Weather Conditions. Meteorol. Atmos. Phys., 81, 27–52.
Kossmann, M. and A.P. Sturman, 2004: The Surface Wind Field During Winter Smog Nights In
Christchurch And Coastal Canterbury, New Zealand. Int. J. Climatol., 24, 93–108.
Kossmann, M., U. Corsmeier, S.F.J. de Wekker, F. Fiedler, R. Vögtlin, N. Kalthoff, H. Güsten, and
B. Neininger, 1999. Observations Of Handover Processes Between The Atmospheric Boundary
Layer And The Free Troposphere Over Mountainous Terrain. Contrib. Atmos. Phys. 72,
329–350.
Kurita, H., Ueda, H., and Mitsumoto, S., 1990: Combination Of Local Wind Systems Under Light
Gradient Wind Conditions And Its Contribution To Long-Range Transport Of Air Pollutants.
J. Appl. Meteorol., 29, 331–348.
Kuwagata, T. and M. Sumioka, 1991: The Daytime Heating Process Over Complex Terrain In
Central Japan Under Fair And Calm Weather Conditions. Part III: Daytime Thermal Low And
Nocturnal Thermal High. J. Met. Soc. Japan, 69, 91–103.
Lenschow, D. H., B. B. Stankov, and L. Mahrt, 1979: The Rapid Morning Boundary-Layer
Transition. J. Atmos. Sci., 36, 2106–2124.
Levinson, D.H. and R.M. Banta, 1995: Observations Of A Terrain-Forced Mesoscale Vortex And
Canyon Drainage Flows Along The Front Range of Colorado. Mon. Wea. Rev., 123, 2029–2050.
Lin, Y.-L. and I.-C. Jao, 1995: A Numerical Study Of Flow Circulations In The Central Valley Of
California And Formation Mechanisms Of The Fresno Eddy. Mon. Wea. Rev., 123, 3227–3239.
Löffler-Mang, M., M. Kossmann, R. Vögtlin, and F. Fiedler, 1997: Valley Wind Systems And Their
Influence On Nocturnal Ozone Concentrations. Contr. Atmos. Phys., 70, 1–14.
Lu, R. and R.P. Turco, 1994: Air Pollution Transport In A Coastal Environment. Part I: Two-
Dimensional Simulation Of Sea-Breeze And Mountain Effects. J. Atmos. Sci., 51, 2285–2308.
McGowan, H.A. and A.P. Sturman, 1996: Interacting Multi-Scale Wind Systems Within An Alpine
Basin, Lake Tekapo, New Zealand. Meteor. Atmos. Phys. 58, 165–177.
McGregor, J.L. and F. Kimura, 1989: Numerical Simulation Of Mesoscale Eddies Over Melbourne.
Mon. Wea. Rev., 117, 50–66.
McKee, T.B. and R.D. O’Neal, 1989: The Role Of Valley Geometry And Energy Budget In The
Formation Of Nocturnal Valley Winds. J. Appl. Meteorol., 28, 445–456.
McKendry, I.G., D. G. Steyn, J. Lundgren, R. M. Hoff, W. Strapp, K. Anlauf, F. Froude, J. B.
Martin, R.M. Banta and L.D. Olivier, 1997: Elevated Ozone Layers And Vertical Down-Mixing
Over The Lower Fraser Valley, BC. Atmos. Env., 31, 2135–2146.
McKendry, I.G. and J. Lundgren, 2000: Tropospheric Layering Of Ozone In Regions Of Urbanized
Complex And/Or Coastal Terrain: A Review. Prog. Phys. Geog. 24, 329–354.
Millan, M.M., R. Salvador, E. Mantilla, and G. Kallos, 1997: Photooxidant Dynamics In The
Mediterranean Basin In Summer – Results From European Research Project. J. Geophys.
Res., D102, 8811–8823.
Neu, U., T. Künzle, and H. Wanner, 1994: On The Relation Between Ozone Storage In The
Residual Layer And Daily Variation In Near-Surface Ozone Concentration – A Case Study.
Oettl, D., R.A. Almbauer, P.J. Sturm, M. Piringer, and K. Baumann, 2001: Analysing The
Nocturnal Wind Field In The City Of Graz. Atmos. Env., 35, 379–387.
Oke, T.R. and J.E. Hay, 1994: The Climate of Vancouver. B.C. Geographical Series, Number
50, 84p.
Ohashi, Y. and H. Kida, 2002: Effects Of Mountains And Urban Areas On Daytime Local-
Circulations In The Osaka And Kyoto Regions. J. Met. Soc. Japan, 80, 539–560.
Petkovsek, Z., 1992: Turbulent Dissipation Of Cold Air Lake In A Basin. Meteorol. Atmos.
Phys., 47, 237–245.
Plate, E., 1982: Engineering Meteorology. Elsevier, Amsterdam, 740 p.
Poulos, G.S, J.E. Bossert, T.B. McKee, and R.A. Pielke, 2000: The Interaction Of Katabatic Flow
And Mountain Waves. Part I: Observations And Idealized Simulations. J. Atmos. Sci., 57,
1919–1936.
Poulos, G. and S. Zhong, 2008: An Observational History Of Small-Scale Katabatic Winds In
Mid-Latitudes. Geog. Compass 2, 1–24.
Raga, G.B., D. Baumgardner, G. Kok, and I. Rosas, 1999: Some Aspect Of Boundary Layer
Evolution In Mexico City. Atmos. Env. 33, 5013–5021.
Regmi, R.P., T. Kitada, and G. Kurata, 2003: Numerical Simulation Of Late Wintertime Local
Flows In Kathmandu Valley, Nepal: Implication For Air Pollution Transport. J. Appl. Meteo-
rol., 42, 389–404.
Reuten, C., D.G. Steyn, K.B. Strawbridge, and P. Bovis, 2005: Observed trapping of pollutants in
upslope flow systems. Bound.-Layer Meteorol., 116 (1), 37–61.
Reuten, C., D.G. Steyn, and S. E. Allen. 2007: Water Tank Studies of Atmospheric Boundary
Layer Structure and Air-Pollution Transport in Upslope Flow Systems. J. Geophys. Res., 112,
D11114.
Sakiyama, S.K., 1990: Drainage Flow Characteristics And Inversion Breakup In Two Alberta
Mountain Valleys. J. Appl. Meteorol., 29, 1015–1030.
Salmond, J. A. and I.G. McKendry, 2002: Secondary Ozone Maxima In A Very Stable Nocturnal
Boundary Layer: Observations From The Lower Fraser Valley, BC. Atmos. Env. 36. 5771–5782.
Schmidli, J., G.S. Poulos, M.H. Daniels, and F.K. Chow, 2009: External Influences On Nocturnal
Thermally Driven Flows In A Deep Valley. J. Appl. Meteorol. Climatol., 48, 3–23.
Seibert, P., H. Kromp-Kolb, A. Kasper, M. Kalina, H. Puxbaum, D.T. Jost, M. Schwikowski, and
U. Baltensperger, 1998: Transport of polluted boundary layer air from the Po Valley to high-
alpine sites. Atmos. Env., 32, 3953–3965.
Smedman, A.S., H. Bergström, and U. Högström, 1996: Measured And Modelled Local Wind
Fields Over A Frozen Lake In A Mountainous Area. Contr. Atmos. Phys. 69, 501–516.
Spillane, K., 1978: Atmospheric Characteristics On High Oxidant Days In Melbourne. Clean Air
(Aust) 12, 50–56.
Stensrud, D.J. 1993: Elevated Residual Layers And Their Influence On Surface Boundary-Layer
Evolution. J. Atmos. Sci., 50, 2284–2293.
Steyn D. G., and T. R. Oke, 1982: The Depth Of The Daytime Mixed Layer At Two Coastal Sites:
A Model And Its Validation. Bound.-Layer Meteor., 24, 161–180.
Steyn, D.G., J.W. Bottenheim, and R.B. Thomson, 1997: Overview Of Tropospheric Ozone In The
Lower Fraser Valley, And The PACIFIC ‘93 Field Study. Atmos. Env., 31 (4), 2025–2036.
Stull, R.B., 1998: Introduction to Boundary Layer Meteorology. Kluwer, Dordrecht. 666 p.
Sun, J., R. Desjardins, L. Mahrt, and I. MacPherson, 1998: Transport of carbon dioxide, water
vapor, and ozone by turbulence and local circulations, J. Geophys. Res., 103(D20), 25,
873–25,885.
Troude, F., E. Dupont, B. Carissimo, and A. Flossmann, 2002: Relative Influence Of Urban And
Orographic Effects For Low Wind Conditions In The Paris Area. Bound.-Layer Meteor., 103,
493–505.
Whiteman, C. D., 1982: Breakup Of Temperature Inversions In Deep Mountain Valleys: Part I.
Observations. J. Appl. Meteorol., 21, 270–289.
Whiteman, C.D., T.B. McKee, and J.C. Doran, 1996: Boundary Layer Evolution Within A
Canyonland Basin. Part 1. Mass, Heat, And Moisture Budgets From Observations. J. Appl.
Meteorol., 35, 2145–2161.
Whiteman, C.D., S. Zhong, X. Bian, J.D. Fast, and J.C. Doran, 2000: Boundary Layer Evolution
And Regional-Scale Diurnal Circulations Over The Mexico Basin And Mexican Plateau.
J. Geophys. Res., 105, 10081–10102.
Whiteman, C.D., S. Zhong, W.J. Shaw, J.M. Hubbe, and X. Bian, 2001: Cold Pools In The
Columbia Basin. Weather and Forecasting 16, 432–447.
Wilczak, J.M. and T.W. Christian, 1990: Case Study Of An Orographically Induced Mesoscale
Vortex (Denver Cyclone). Mon. Wea. Rev., 118, 1082–1102.
Wippermann F., 1984: Air Flow Over And In Broad Valleys: Channelling And Counter-Current.
Contr. Atmos. Phys., 57, 92–105.
Zaremba, L.L. and J.J. Carroll, 1999: Summer Wind Flow Regimes Over The Sacramento Valley.
J. Appl. Meteorol., 38, 1463–1473.
Zawar-Reza, P. and A.P. Sturman, 2006 Two-Dimensional Numerical Analysis Of A Thermally
Generated Mesoscale Wind System Observed In The Mackenzie Basin, New Zealand.
Australian Meteorol. Mag. 55, 19–34.
Zhong, S., X. Bian, and C.D. Whiteman, 2003: Time Scale For Cold-Air Pool Breakup By
Turbulent Erosion. Meteorolog. Zeit. 12, 229–233.
Zhong, S., J. Li, C.D. Whiteman, X. Bian, and W. Yao, 2008: Climatology Of High Wind Events
In Owens Valley, California. Mon. Wea. Rev., 136, 3536–3552.
Chapter 6
Theory, Observations, and Predictions
of Orographic Precipitation
Brian A. Colle, Ronald B. Smith, and Douglas A. Wesley
Abstract There have been rapid advances in the understanding of orographic

precipitation during the past two decades given the advent of high resolution
mesoscale modeling and numerous field studies around the world. This chapter
begins with introducing the fundamental ingredients, processes, and scaling param-
eters for orographic precipitation. If the low level air is stable, terrain flow blocking
can occur, which shifts the precipitation enhancement upwind of the barrier. Low-
level diabatic cooling from evaporation and melting can enhance the near-surface
stability and flow blocking. In contrast, latent heating within the cloud reduces
the effective stability and the potential for flow blocking, which in turn results in
stronger vertical motions over the ridges and increases the riming and precipitation
fallout.
For less stable flows, there are mountain gravity waves over the mountain that
can change the depth and upstream extent of the orographic cloud, and lead to
precipitation enhancements over individual windward ridges. Barrier dimension is
also important, with wide barriers favoring precipitation enhancement upwind of
the crest and more efficient water vapor removal from the ambient flow. If the ridges
are very narrow, the mountain waves decay with height and have less impact on the
precipitation. For unstable flow, the upslope flow can trigger convective systems or
quasi-stationary banded precipitation.
The various orographic processes are illustrated using several field study and
modeling results. The prediction of orographic precipitation is discussed using a
B.A. Colle ()

School of Marine and Atmospheric Sciences, Stony Brook University/SUNY, Stony Brook,
NY 11794-5000, USA
e-mail: brian.colle@stonybrook.edu
R.B. Smith
Geology and Geophysics Department, Yale University, New Haven, CT 06520-8109, USA
D.A. Wesley
Compass Wind, 1730 Blake St., Ste. 400, Denver, CO 80202, USA

292 B.A. Colle et al.
1.33-km grid spacing simulation and sensitivity results from a linear orographic
precipitation model for a flooding event over southwest Washington in early
November 2006. The chapter concludes with a summary and areas for future work.
6.1 Introduction
Orographic precipitation can be defined as the modification of rain, snow, and other
hydrometeors resulting from the interaction of moist flow with topography. As
noted by Sawyer (1956), the three important factors for orographic precipitation
are: (1) the interaction of the ambient flow with terrain, (2) cloud microphysical
processes, and (3) larger-scale atmospheric circulation. For enhanced precipitation
growth to occur over the windward slope (Fig. 6.1), there is often a synergistic
relationship between the large-scale ascent associated with an approaching baro-
clinic or tropical cyclone, and the moist dynamical flow (upslope) response, which
is dependent on the magnitude of the upstream flow orientated perpendicular to the
mountain, moisture, and stratification (Fig. 6.1). The microphysical processes and
the time scale for available ice/water growth within the cloud can also modify the
precipitation distribution. Atmospheric models and operational forecasters need to
consider all of these factors in order to forecast precipitation accurately in areas of
steep terrain.
There has been a rapid advance in the understanding of orographic precipitation
with the application of high-resolution models and new observational field datasets
collected over various topographic barriers around the world (Fig. 6.2). During the
last few decades these field studies have quantified various orographic precipitation
processes for a broad spectrum of ambient environments and barrier dimensions
(Table 6.1). For example, the Cascade Project in the early 1970s examined oro-
graphic airflow and microphysics over the Washington Cascades using aircraft
(Hobbs et al. 1973; Hobbs 1975). During the 1980s, the Sierra Cooperative Pilot
Project (SCPP) studied the airflow and precipitation processes over the Sierra
Nevada Mountains of California using both ground-based radar and aircraft in situ
sensors (Reynolds and Dennis 1986).
The mesoscale flow and precipitation structures of storms approaching the U.S.
West coast were also studied during the Coastal Observations and Simulation with
Topography (COAST) experiments in December 1993 and 1995 (Bond et al. 1997)
and in the California Land-Falling Jets (CALJET) and Pacific Land-Falling Jets
(PACJET) experiments in 1998 and 2000–2002, respectively (Ralph et al. 1999;
Neiman et al. 2002). Using airborne dual-Doppler radar data around the Olympic
Mountains in COAST, Colle and Mass (1996) showed that three-dimensional winds
can be synthesized from airborne Doppler radar over steep terrain. This helped put
the precipitation in context with the three-dimensional flow, which motivated other
subsequent orographic field experiments to use this airborne Doppler technology.
For example, the IMPROVE II (Improvement of Microphysical PaRameterization
through Observational Verification Experiment II) field program investigated the
6 Theory, Observations, and Predictions of Orographic Precipitation 293
Fig. 6.1 Schematic diagram of the elements of orographic precipitation: upstream water vapor
flux (Fw ), large-scale ascent, windward ascent, streamlines of theta-E, snow and graupel fallout,
and windward precipitation within a distance P
Fig. 6.2 Location of orographic precipitation field projects around the world that involved radar,
aircraft, and/or isotope measurements in areas of terrain (shaded). See Table 6.1 for a detailed list
of the experiments and their focus
Table 6.1 Major orographic precipitation field studies around the world (shown on Fig. 6.2) that utilized radar, aircraft, and/or isotope measurements
294
Number Project title Location Year Objective

1 N/A South Wales 1970s Precipitation enhancement over small hills
2 Sierra – Cooperative Project Sierra Mountains late 1970s–1980s Precipitation evolution, thermodynamics, and
microphysics over a wide barrier with
landfalling storms
3 Taiwan Area Mesoscale Experiment Taiwan 1987 The effects of orography on the Mei-Yu front and
(TAMEX) on mesoscale convective systems
4 Hawaiian Rainband Experiment (HARP) Hawaii 1990 Diurnal variability of precipitation around
isolated topography
5 Coastal Observation and Simulation with Olympic Mountains 1993 Precipitation and kinematic evolution over
Topography (COAST) isolated barrier
6 Winter Icing Storms Project (WISP) Colorado Front 1994 Ice nucleation, super-cooled water, and
Range precipitation evolution along the Front Range
7 Southern Alpine Precipitation Experiment New Zealand Alps 1996 Understanding the processes through which the
(SALPEX) Southern Alps (narrow steep barrier)
influence precipitation
8 Mesoscale Alpine Programme (MAP) Alps 1999 Linkage of moist dynamics (blocked and
unblocked flow) with precipitation
distribution and processes
9 California Landfalling Jets Experiment California coastal 1998, 2000–2001 Coastal precipitation enhancement, warm rain
CALJET/Pacific Jets (PACJET) mountains processes, and atmospheric rivers
10 Improvement of Microphysical Central Oregon 2001 Precipitation processes and microphysics, and
Parameterization through Cascades the role of gravity waves on precipitation
Observational Verification Experiment
(IMPROVE-2)
11 Intermountain Precipitation Experiment Wasatch Mountains 2001 Precipitation processes, diabatic impacts, and
(IPEX) microphysics for narrow barrier
12 Southern Andes Project Southern Andes 2005 Air mass transformation and isotope analysis
13 Convective Orographic Precipitation Southwest Germany 2007 Study orographically-induced convective
Experiment (COPS) precipitation
B.A. Colle et al.
microphysics and orographic cloud structures for frontal systems crossing the
Oregon Cascades (Stoelinga et al. 2003). Baroclinic storm systems and moist
southerly flow impinging on the Alps were documented in autumn 1999 during the
Mesoscale Alpine Programme (MAP, Bougeault et al. 2001; Rotunno and Houze
2007). Meanwhile, the impact of more narrow and steep barriers on orographic
precipitation, such as the Wasatch in Utah (IPEX, Schultz et al. 2002) and southern
Andes (Smith and Evans 2007) have also been investigated. More tropical envi-
ronments associated with three-dimensional island geometries were investigated
in Taiwan during TAMEX (Kuo and Chen 1990) and Hawaiian Rainband Project
(HARP) (Chen and Nash 1994). More recently, convective initiation over terrain was
investigated over southwest Germany during the Convective and Orographically-
induced Precipitation Study (COPS) experiment (Wulfmeyer et al. 2008).
Significant loss of life and disruptions to the local economy can occur when
flash flooding, avalanches, and transportation delays occur in association with heavy
orographic precipitation as outlined in Chap. 1. There have been several flash
convective flood events in areas of steep terrain, such as the southern slopes of
the European Alps (Lin et al. 2001), southern Appalachians (Pontrelli et al. 1999),
and the Colorado Front Range (Petersen et al. 1999). Along the mountainous U.S.
West Coast, a plume of subtropical moisture associated with a landfalling baroclinic
wave can interact with the steep coastal terrain to produce flooding over Pacific
Northwest (Colle and Mass 2000; Lackmann and Gyakum 1999) and the California
coast (Ralph et al. 2003). Meyers et al. (2003) and Poulos et al. (2002) investigated
a heavy snow and blizzard event over the Front Range of the Rocky Mountains in
1997 that severely disrupted transportation during the 24-h snowfall event.
Understanding the fine-scale precipitation structures over a topographic barrier
is also important for hydrologic streamflow and river modeling. For example,
Westrick et al. (2002) illustrated that the streamflow forecasts over the Washington
Olympics and Cascades were sensitive to the observed and simulated precipitation
distributions on the ridge and valley scale. Ralph et al. (2003) also showed that
the peak river flows during CALJET were highly sensitive to the exact location of
the windward enhancements and rain shadows. Landfalling hurricanes often create
heavy rainfall and flooding in areas of terrain, such as hurricane Agnes along the
Appalachians (Carr and Bosart 1978), and hurricane Dean over the mountainous
island of Dominica (Smith et al. 2009b).
Mesoscale atmospheric models, such as the Penn State – National Centers for
Atmospheric Mesoscale Model (MM5; Grell et al. 1994) and Weather Research and
Forecasting model (WRF; Skamarock et al. 2005), have been shown to realistically
predict precipitation structures when run at <5 km grid spacing (Mass et al. 2002),
since the terrain (upslope) forcing for relatively wide barriers (Sierras, Cascades,
etc : : : ) can be satisfactorily represented at these grid spacings (Chap. 10 of this
book; Colle and Mass 2000; Colle et al. 2000). However, there are uncertainties
and deficiencies in model microphysics (Garvert et al. 2005b; Colle et al. 2005a,
b), upstream initialization uncertainties (McMurdie and Mass 2004), and intrinsic
predictability limits on the mesoscale precipitation over terrain (Hohenegger et al.
2006).
There have been a few review summaries of orographic precipitation in recent

years given the rapid progress of this field (Roe 2005; Smith 2006; Rotunno and
Houze 2007; Lin 2005). Most of these previous write-ups have focused on the
theory (Smith 2006), one particular field study (Rotunno and Houze 2007), or
have emphasized one particular aspect of orographic precipitation. In this chapter,
we synthesize results from several field studies and compare with recent theory,
and then attempt to subsequently link the results back to the forecast process.
Chapter 7 highlights many of the microphysical issues associated with orographic
precipitation. Section 6.2 will summarize some of the key processes of orographic
precipitation, which will incorporate results from several field programs and high
resolution model simulations. In Sect. 6.3, some of the latest forecast tools for
orographic precipitation will be described, ranging from nested models to linear
downscaling models. These tools will be applied for a major flooding event over the
Pacific Northwest on 5–7 November 2006, in which 25–50 cm of rain fell over the
Cascade mountains of southwest Washington, resulting in a 1 in 100-year flood in
some locations. The predictability of orographic precipitation will be discussed in
general and related to the November 2006 flooding event. Finally, some of the future
areas of research in orographic precipitation will be discussed.
6.2 Key Processes in Orographic Precipitation
6.2.1 Water Vapor Flux, Terrain-Forced Ascent,

and Vapor Condensation
Three important concepts in orographic precipitation (OP) are water vapor flux,
terrain-forced ascent, and vapor condensation. The horizontal water vapor flux (FW )
is the rate at which water vapor is carried horizontally past a point in a layer above
the earth’s surface. This horizontal vector quantity can be computed from a balloon
sounding according to
ZzTOP ZTOP
P
1
FEW D qV UE d z D qV UE dp; (6.1)
g
0 P0
with units kg m1 s 1 . In (6.1), qV is the specific humidity (kg/kg), U is the

wind speed normal to the barrier (m s1 ), ¡ is the air density (kgm3 ). In the special
case of uniform wind with height, FEW is the product of the wind speed and the
precipitable water integrated in the column, i.e. FEW D UE PWAT where PWAT is
given in units kg m2 D mm of liquid. If the air is saturated, qV D qVSAT .T; p/
in (6.1).
While it is not yet the custom to routinely compute FEW for each balloon sounding
(e.g. as is done for CAPE), this practice would call attention to the importance
of this variable for orographic precipitation, as highlighted in previous studies

of orographic precipitation (Smith et al. 2003; Lin and Colle 2009; Smith et al.
2009).Other things being equal, OP is roughly proportional to FEW . When a model
underestimates FEW , it is likely to underestimate OP as well. For example, if the
forecast cross barrier wind of 20 m s1 has an error of 5 m s1 (25%), the OP will
also be in error of 25%. This is why it is imperative for a model to accurately
simulate the moisture and cross barrier flow correctly in order to have a chance to
properly forecast the OP amounts.
For the case of a fully saturated atmosphere, the moist adiabatic lapse rate forms
a lower bound on the column integrated water and the water vapor flux. In this
case there is a unique relationship between surface temperature, surface saturation
specific humidity, and the column integrated water vapor. If the earth’s surface is at
p D 1,000 hPa, these are
qVSAT0 0:0038 exp.0:068T / (6.2a)
PWAT 9 exp.0:088T / (6.2b)
where T is the surface temperature in Celsius. It is remarkable how rapidly the

precipitable water and the water vapor flux increase with environmental tempera-
ture. With a wind speed of 10 m s1 and a temperature of 0ı C, (6.2b) gives FW D
U PWAT D 90kgm1 s 1 while a temperature of 20ı C gives FW D 523 kgm1 s 1 .
The equivalent moist layer depth (Dw ) also increases with temperature for a moist
adiabatic profile. From (6.2a, 6.2b)
PWAT
DW D 2400 exp.0:02T /; (6.2c)
qVSAT0
and from (6.1), the water vapor flux can be expressed as FEW D qVSAT0 DW UE .
The second important concept is terrain-forced ascent. When the low-level flow
is not blocked, ascent or descent along the local terrain slope can be approximated
by the product of the horizontal wind speed and the terrain slope (Fig. 6.1).
w D UE rh.x; y/ (6.3)
where h.x; y/ is the terrain description. In practice, formula (6.3) is slightly

ambiguous. First, what wind should we use for UE ; the ten meter wind, the 925 hPa
wind, or the wind at some other level? Second, what slope should we use, the local
terrain slope or some average slope over a 10 km patch? Third, we remain ignorant
of how the vertical motion will change with height. It might increase with height if
the horizontal wind speed does, or decrease with height if the wind speed decreases
aloft or if the streamlines flatten out.
Putting aside these questions for now, we can introduce the thermodynamics of
moist ascent. When air is forced upwards, it moves to lower pressure, expands, and
cools. The maximum amount of water in the vapor state qVSAT .T; p/ decreases at a
rate d qVSAT =d zjPARCEL. The column integrated condensation is then
Z1
d qVSAT
C.x; y/ D w jPARCELd z (6.4)
dz
0
If w is independent of height and the sounding is moist adiabatic, the integral

(6.4) reduces to
C.x; y/ D w qVSAT0 ; (6.5)
where qVSAT0 is the saturation specific humidity at the ground (6.2a). C(x,y) has the
same units .1 kg m2 s 1 D 1 mm s 1 D 3600 mm hr 1 / as rain rate P(x,y)
assuming all the condensed water dropped instantly to the ground.
Combining (6.3) and (6.5)
C.x; y/ D qVSAT0 UE ı rh.x; y/ (6.6)
which is the simplest possible model of orographic precipitation (Colton 1976; Rhea
1978). With a temperature of T D 0ı C , D 1 kg m3 , U D 10 m s 1 and a
slope of S D 0.05, (6.2a, 6.6) give C D 0:0019 kg m2 s 1 D 6:8 mm hr 1 :
For the northern California coastal mountains during CALJET (Table 6.1),
Neiman et al. (2002) showed that the upslope flow is highly correlated (R 0.9,
where R is the correlation coefficient) with the hourly rain rate when the low-level
flow is not blocked (Fig. 6.3a). The rain rate is correlated best with the average
wind just above the mean mountain top (1 km above MSL for the California
coastal mountains), but this correlation drops to near 0 at 3 km above MSL. Falvey
and Garreaud (2007) also found strong correlation between precipitation and the
upstream wind and moisture flux over a 10-year period in central Chile. They further
emphasized that the positive correlation between rainfall and moisture flux is largely
determined by the variability in the wind velocity.
Although the rate of condensation above sloping terrain is related to the
horizontal water vapor flux and the terrain slope, more knowledge is needed of the
vertical motion above the terrain. Also, there are other factors that determine the
percent of the condensed water that is converted to hydrometers and falls out (i.e.
precipitation efficiency). For example, during conversion and fallout, condensed
water is advected downwind. If the condensed water is carried to the lee of the
barrier, a fraction of the falling water may evaporate before reaching the ground.
One useful measure of orographic precipitation for an entire mountain range
is the total water removed from the airstream (PT ). Integrating precipitation rate
downwind
Fig. 6.3 Cross section perspective of the orographic rainfall distribution during (a) unblocked and
(b) blocked low-level flow (From Neiman et al. 2002). The representative coastal profiles of wind
velocity (full barb D 10 kts) and correlation coefficients (based on the magnitude of the upslope
flow at the coast versus the rain rate in the coastal mountains) shown on the left. The variable h in is
the scale height of the mountain barrier. The spacing between the rain streaks in (b) is proportional
to rain intensity. The symbol “˝” within the blocked flow in (b) portrays a terrain-parallel barrier
jet (© American Meteorological Society. Reprinted with permission)
Z1
PT D Pds (6.7)
1
where ds is the increment of distance parallel to the wind direction. This quantity
can be normalized by the incoming water vapor flux (FW ) to form the Drying
Ratio (DR),
PT
DR D (6.8)
FW
If, as (6.6) suggests, the precipitation is proportional to FW , the DR may be

independent of FW , and thus is a useful quantity to compare different mountain
ranges. Several methods for estimating DR from conventional data and models are
discussed by Smith et al. (2003) and Smith (2006) and involve calculating the water
budget across the mountain range using estimates of the surface precipitation and
incoming and outgoing water vapor fluxes. For example, one can calculate the DR
for the upslope region of a barrier by selecting a boxed region along the windward
slope, and then calculate the average Fw entering the box along the edges over some
time period divided by the total precipitation that reaches the surface within the
box (Lin and Colle 2009). DR can also be estimated using stable isotopes of water
obtained from streams and sapwood (Smith et al. 2005; Smith and Evans 2007). The
relative concentration of deuterium to hydrogen isotopes decreases across a barrier,
since the heavy isotopes are progressively rained out. Thus, the DR can be estimated
by a ratio of these isotopes upwind and downwind of the barrier according to:
1=.˛1/
DR D 1 Rp =Rp0 ; (6.9)
where Rp0 and Rp are the isotope ratios in precipitation upwind and downwind of the
barrier, and ’ is the fractionation factor (Smith 2006). Smith’s studies have yielded
estimates of DR of 35% for the Alps, 44% for the Oregon Cascades, and 50%
for the southern Andes.
Another way to quantify microphysical efficiency is to determine the characteris-
tic residence time for all water and ice generated aloft to fall as precipitation (Smith
et al. 2003). The residence time (RT) is defined within the box region as:
RT D Total ice and water amount aloft=Precipitation rate (6.10)
Typical RTs vary between 1,500 and 2,500 s (Smith et al. 2003; Colle 2008; Lin
and Colle 2009). Residence time typically decreases as the ice/snow becomes more
rimed, which increases the terminal fall speed of the particles. The RT has been
applied to determine how efficient bulk microphysical schemes are in removing
precipitation aloft (Lin and Colle 2009). For example, a scheme that has more
graupel than snow aloft, and thus larger fallspeeds, will have a smaller RT.
6.2.2 The Role of Stability
The amount of forced ascent over an orographic barrier is strongly dependent on the
static stability around the mountain. For nearly saturated conditions, a moist static
stability (Nm ) should be diagnosed upstream of the mountain, which is given by
(6.11) (Durran and Klemp 1982),

g dT Lq g dqw
Nm2 D C m 1C (6.11)
T dz RT 1 C qw dz
where qv and qc are the mixing ratios of water vapor and cloud water, respectively;
qw D qv C qc is the total water mixing ratio; R is the ideal gas constant for dry air;
L is the latent heat of evaporation; and m is the moist adiabatic lapse rate, since
it includes latent heat of condensation and the vertical gradient of liquid water.
A useful parameter to determine the mountain flow response is the non-dimensional
mountain height (inverse Froude number), as given by:
M D Nm hm =U (6.12)
in which hm is the height of the terrain and U is the upstream flow velocity
approaching the terrain averaged below crest level. Jiang (2003) found that the
upstream deceleration ratio (u/U), in which u is the perturbed flow, collapses onto a
single curve for a critical threshold for flow blocking (M 1.2) when M is computed
with the moist static stability frequency. This illustrates that the character of the
windward flow response can still be determined even when one includes moisture.
The parameter M in (6.12) can be useful to determine the flow response over the
barrier if Nm does not vary as the flow ascends the mountain; however, the moist
static stability (Nm ) can change as a result of differential latent heating, radiation,
moisture depletion over the windward ridges, changes in the vertical gradients
of liquid water, and melting processes. Recent idealized studies by Reeves and
Rotunno (2008) showed that a simple substitution of Nm for N in (6.12) for moist
flows does not provide reasonable predictions of the steady-state flow response for
all cases. They show that including other parameters in the non-dimensional phase
space (lifted condensation level and freezing level height) can better determine the
flow perturbations and therefore the precipitation changes over the barrier. There
are also horizontal gradients in the moisture that complicate the orographic flow
and precipitation response. For example, Rotunno and Ferretti (2001) showed that
the temporal and spatial distribution of precipitation for the 1994 Piedmont flood
was dependent on horizontal gradients in the moisture, which helped create a
convergence boundary between blocked and unblocked flow. Galewsky and Sobel
(2005) showed that localized regions of latent heating were important in reducing
the effective stability and thus enhancing the uplift over the mountain. Future
research is needed to determine the impacts of moisture gradients on orographic
precipitation, which clearly can lead to some predictability problems.
Fig. 6.4 Three-hour average cross section of (a) reflectivity (shaded in dBZ) and (b) radial
velocity (shaded in m s1 ) for a blocked flow case during MAP (IOP8 – 21 October 1999). (c)
and (d) Same as (a) and (b) except for the unblocked IOP2b event (Adapted from Medina and
Houze 2003. © John Wiley and Sons Publishing. Reprinted with permission)
6.2.2.1 Flow Blocking
Chapter 3 highlights several dynamical mountain flows, such as flow blocking and
terrain-forced ascent and decent. Several field studies have showed the importance
of the moist dynamical flow response over a barrier on the orographic precipitation
distribution, since the sloping ascent within this blocked flow can represent an
“effective” barrier that extends upwind of the windward slope. However, upstream
flow blocking also reduces the effective slope of the barrier (Marwitz 1980),
therefore reducing the intensity of the upward motion and precipitation over the
windward slope. For example, for the nearly two-dimensional Sierra Mountains,
Marwitz (1987b) and Rauber (1992) noted an upstream shift of precipitation and
high liquid water contents associated with the upstream convergence between the
ambient westerly flow and the terrain-parallel (southerly) barrier jet within the
blocked flow. For the flow blocking events, Neiman et al. (2002) showed little corre-
lation between the orographic precipitation enhancement over the California coastal
range during CALJET and the upstream flow near the surface (Fig. 6.3b). Sinclair
et al. (1997) and Wratt et al. (1996) found much less spillover of precipitation into
the lee of the New Zealand Alps during SALPEX flow blocking events.
The analysis of the IPEX and MAP field studies support the conceptual model for
precipitation enhancement versus flow blocking proposed by Neiman et al. (2002) in
Fig. 6.3. For example, Fig. 6.4 shows a 3-h vertical composite of radar reflectivities
and radial velocities across the Alps from southeast (right) to northwest (left) during
the MAP project for IOP-2b and IOP8 as highlighted in Medina and Houze (2003).
When the flow is blocked (Fig. 6.4a, b), there is gradual ascent over the windward
ridges and a vertical shear layer over the windward slope. Houze and Medina
(2005) hypothesized that the turbulence within this shear layer may enhance the
accretional and aggregational growth of precipitation over the windward slope. The
maximum precipitation is located upstream of the topography below a well-defined
bright band at 2-km ASL. In contrast, for unblocked flow (Fig. 6.4c, d), there is
much less vertical wind shear over the terrain and there is strong ascent over the
windward ridges, which favors a precipitation maximum over the first ridge along
the windward slope, with less precipitation over subsequent ridges as moisture is
depleted.
A similar flow blocking impact on orographic precipitation was observed over
the narrow Wasatch Mountains in northern Utah during IPEX IOP3 (Cox et al.
2005). For this event the orographic precipitation enhancement extended about
20–30 km upstream of the Wasatch crest (Fig. 6.5), which was the result of a low-
level convergent boundary between the nearly terrain-parallel blocked flow and the
ambient low-level southwesterlies upstream. The upstream flow ascended up and
over the blocked flow in both the observations and a model simulation of the event
at 1.33-km grid spacing in Colle et al. (2005) (Fig. 6.5b, c).
The upstream flow blocking associated with cold air damming events also favors
enhanced precipitation upstream of the mountain. For example, Wesley et al. (2010)
and Doesken (2003) highlighted a snow storm (50–100 year event) to the east of
the Colorado Front Range on 17–19 March 2003 that resulted from the interaction
of low-level easterly flow from a developing cyclone to the south of Colorado
with cold air damming along the Front Range (Fig. 6.6a). Adjacent to the terrain
there was low-level northerly flow that extended a few tens of kilometers to the
east of the mountain (Fig. 6.6b). Moist low-level easterly to northeasterly flow
ascended the low-level cold air trapped against the foothills, resulting in heavy
precipitation over and just east of the foothills. In addition, the snowfall distribution
within the relatively cold northerly flow was dependent on local, mesoscale terrain
variations oriented east–west (Fig. 6.6a), with several distinct precipitation shadows
documented for terrain features 0.2–0.3 km in height.
6.2.2.2 Unblocked Stable Flow
As the non-dimensional mountain height (M) decreases to <1 and it is stratified

(Nm > 0), there is less flow blocking and more gravity (mountain) waves over the
windward ridges and the crest, which can also impact on the precipitation distri-
bution. Early observational studies of the impact of gravity waves on precipitation
enhancement focused over the crest and immediate lee (Bruintjes et al. 1994) or on
the generation of supercooled water over windward ridges (Reinking et al. 2000).
More recently, Garvert et al. (2007) utilized an airborne Doppler radar dataset over
the Oregon Cascades during IMPROVE-2 to show how a barrier scale vertically
Fig. 6.5 (a) KMTX reflectivity (2,265 m MSL) at 1830 UTC 12 Feb 2000. (b) Barrier-normal
WP-3D tail-mounted radar reflectivity (shaded) cross section along the AB blue line in (a). (c)
Same as (a) except showing the MM5 simulated reflectivity, wind circulation in cross section AB,
and wind speed parallel to the section (solid every 2 m s1 ) (Adapted from Colle et al. 2005b.
propagating gravity wave can impact the vertical motion and enhance the snow
production at upper levels over the windward slope as a frontal system crossed over
the Cascades (Fig. 6.7a). Colle (2004) duplicated this response in two-dimensional
idealized model simulation using a relatively wide barrier (50 km half width).
Garvert et al. (2007) also showed that there were gravity waves over the narrow
windward ridges which enhanced the cloud water and riming production (Fig. 6.7b),
which likely increased precipitation fallout. Numerical experiments showed that the
gravity waves over the windward ridges increased the net precipitation over the
windward slope by 10–20% for the 13–14 December IMPROVE-2 event (Garvert
et al. 2007). For other cases in IMPROVE-2 with weaker (15 m s1 ) flow and
stratification, the gravity waves were weaker and the Cascade ridges produced a net
precipitation enhancement of <10% (Colle et al. 2008). Colle (2008) showed using
2-D idealized simulations that this average precipitation enhancement is maximized
when there is strong cross barrier flow and the ridge spacing is relatively small
(20 km), since there is less time for subsidence drying within the valleys and
the mountain waves become more evanescent, which favors a simple upward and
downward motion couplet over each ridge. In addition, small ridge spacing is shown
to have a synergistic effect on precipitation over the lower windward slope, in which
an upstream ridge helps to increase the precipitation over the adjacent downwind
ridge. Using an idealized two-ridge system of moist flow past the peaks oriented
along the prevailing wind direction, Jiang (2007) showed that the precipitation
efficiency is generally larger over the first/lower peak. On the other hand, for
relatively high terrain (hm 800 m), ice-phased hydrometers induced by the lee-
wave from the first peak could seed the cap clouds over the second/major peak and,
therefore, increase the precipitation efficiency over the major peak. Colle (2008) and
Jiang (2007) showed that there is little net precipitation enhancement by windward
ridges for large non-dimensional mountain height of M 1.2 (small moist Froude
numbers of Fr 0.8), since flow blocking limits the flow up and over each ridge.
6.2.2.3 Moist Neutral Flow
Moist neutral flow has also been noted for orographic precipitation events going
back to the work of Douglas and Glasspoole (1947) and Sawyer (1956, see his
Fig. 6.2). Ralph et al. (2005) showed that the atmospheric rivers approaching the
U.S. West Coast had near moist neutral stratification up to about 3 km ASL during
the PACJET/CALJET field projects. This profile was favored in the pre-cold frontal
portion of the storm, and the weak stratification favors rapid ascent and heavy
precipitation along the coastal range of California as the low-level jet interacts
with the barrier. Moist neutral soundings were also observed during the MAP IOPs
(Rotunno and Feerretti 2003). These field studies have renewed interest in studying
these moist neutral flows for orographic precipitation events.
Miglietta and Rotunno (2005, 2007) used idealized simulations to show if
saturation can be maintained, then small values of Nm lead to zero blocking and
long vertical wavelengths. These results are an extension of the idealized stable
Fig. 6.6 (continued)

experiments of Colle (2004), except that Miglietta and Rotunno’s initial sounding
includes the tropopause which limits the vertical wavelength. Their idealized
experiments suggest that it is difficult to maintain a state of saturation everywhere
given the orographic flow response. For example, they found a midlevel disturbance
that propagates upstream and downstream of the mountain, which originated in the
unsaturated air in the lee side of the barrier as the flow adjusted to the mountain.
As a result, the static stability Nm can switch from near zero to 0.01 s1 internally
depending on the local saturation, and this location cannot be prescribed a priori.
This switching from saturated to unsaturated, with feedback to the local stability
is a fundamental nonlinearity. For example, if the moisture is depleted too rapidly
upstream, this can lead to a stratified-flow response that is too strong in the lee side
flow (now unrealistically dry). This non-linearity has ramifications on predictability
and needs to be explored more in the future.
6.2.2.4 Orographic Convection
Orographic convection can occur as a result of diurnal heating and cooling around a
barrier as well as the release of conditional or potential instability from upslope flow
or mountain gravity waves. Most of the rainfall maxima over the continents occur
during the daytime (Kikuchi and Wang 2008), but nocturnal maxima can occur
close to or downstream major mountain ranges as a result of the local upslope and
downslope flows. For example, the organized convective cores can occur at night
along the edge of the Himalayas (Romatschke and Houze 2010), which is likely
from the convergence of the monsoonal flow at the base of the mountain range and
the nocturnal downslope flow. Gravity waves triggered by heating over a mountain
barrier can propagate away from the mountain and trigger convective clouds farther
from the mountain later in the day (Mapes et al. 2003).
Even within a broader precipitation shield over the windward slope there is
often intermittent orographic precipitation as a result of upslope enhancement and
triggering of embedded convective cells (Smith et al. 2003; Fuhrer and Schär 2005;
Colle et al. 2008). For example, Fig. 6.8 shows a Hovmoller plot (i.e., as a function
of time and distance from the radar) from upstream of the Oregon coastal range
eastward to the central Oregon Cascades from 1900 UTC 4 December to 1200
UTC 5 December 2001 during the IMPROVE-2 experiment (Colle et al. 2008).
J
Fig. 6.6 (a) Snowfall (m) over the central Rocky Mountain region from 16 to 20 March 2003.
Purple-shaded areas represent >1.5 m of accumulation. (b) Vertical cross-section, oriented
approximately E-W and spanning about 575 km (350 mi), showing ™e (K), winds (kts), and omega
(bs1 ) from a 6-h forecast from the Eta model initialized at 18 UTC 18 March. The vertical
axis represents pressure (mb). The cross section location is shown in the inset on the upper right
in (b). Red regions correspond to omega values of 10 to 15 bs1 . Area inside the orange oval
indicates the blocked region. Wind barbs (full barb D 10 kts) in light blue are plotted to depict
wind direction only on a planar surface, e.g. a flag or barb pointing straight upwards on the figure
indicates a northerly wind direction
Fig. 6.7 (a) Average east–west cross section across the Cascades showing the 1.33-km MM5 snow
(qs) mixing ratios (g m3 ) overlaid on airborne dual-Doppler derived radar reflectivity (dBZ; color
shading, key at right). The modeled CLW and qs shown were averaged in time from 2300 to 0100
UTC 13–14 Dec 2001 (forecast hours 10–13) using model output at 15-min intervals. (b) A north–
south cross sections along over the windward slope over the Cascades showing (a) airborne dual-
Doppler derived vertical velocity (solid m s1 ) and reflectivities (color shaded) (From Garvert
Fig. 6.8 Hovmöller-type plot showing the maximum SPol reflectivity (color shaded every
2.5 dBZ) above 2.0 km MSL from 1900 UTC 4 December to 1200 UTC 5 December west to east
across the Oregon coastal range (left) and Cascades (right) (From Colle et al. 2008. © American
Beginning around 1900–2100 UTC 4 December, cells developed over the coastal
range and propagated eastward over the Cascades, weakening as they passed over
the east side of the coastal range as a result of subsidence in the lee of this modest
barrier. After 0000 UTC 5 December there was a gradual increase in the amount
of precipitation spilling over the Cascade crest as the cross barrier flow increased
(not shown). With the surface trough’s approach between 0200 UTC 5 December
and 0700 UTC 5 December, the precipitation intensified over the Cascades and
became more widespread, while new clusters of cells continued to be enhanced over
the coastal range and then advanced eastward over the Cascades. The reflectivity
sequence seen during the 4–5 December case has similarities to the convective
structure evinced during MAP IOP2b (Asencio et al. 2005; Smith et al. 2003;
Georgis et al. 2003; Rotunno and Houze 2007) as southerly and statically unstable
airflow ascended the Apennines, a coastal barrier south of the Alps. In the lee of
the Apennines, subsidence weakened the cells, which then drifted northward and
became embedded in the stratiform, orographically-forced cloud over the Alps and
increased their intensity over local peaks (Asencio et al. 2005; Medina and Houze
2003; Smith et al. 2003).
In the case of conditionally unstable flows over idealized two-dimensional

mesoscale terrain, different convective regimes have been identified (Chu and Lin
2000; Chen and Lin 2005; Miglietta and Rotunno 2009). In the first regime,
convection propagates upstream together with the downdraft-generated density
current, while in the second, quasi-stationary convection over the mountain crest
or its nearby slopes is attained. Upstream propagating convective systems have
been shown to trigger heavy precipitation upstream of the mountain (Grossman
and Durran 1984; Smolarkiewicz et al. 1988). Quasi-stationary rain bands can
be triggered by relatively small mountains either from upslope flow or lee side
flow convergence (Cosma et al. 2002; Anquetin et al. 2003). Kirshbaum and
Durran (2005) showed that random noise added to idealized smooth topography
can trigger quasi-stationary precipitation bands, particularly when there is weak
convective instability and increasing low-level winds without much directional
turning. Kirshbaum et al. (2007a, b) showed that gravity waves may also be
important in triggering these convective precipitation bands. Kirshbaum and Smith
(2008) found that as convective cells developed over the windward slope as
the surface temperature is raised in their idealized simulations, there was not a
corresponding increase in the drying ratio over the barrier.
More organized and severe convection can also be modified by terrain. For
example, Letkewicz and Parker (2010) investigated the conditions necessary of
mesoscale convective systems to move over the Appalachian Mountains. They
found that for storms crossing the mountains there was more instability and less
convective inhibition in the lee. Idealized simulations by Frame and Markowski
(2006) showed that the modifications of the convective cold pool by terrain can lead
to discrete propagation of the storms across the terrain. Wasula et al. (2002) showed
how the complex terrain of the Northeast U.S. can modify the distribution of severe
convection. In particular, the channeled low-level flow within major river valleys
over the Northeast U.S. is shown to increase the low-level shear and thus may favor
tornado development (Lapenta et al. 2005; Bosart et al. 2006). For example, an
EF3 tornado in Great Barrington, Massachusetts, on 29 May 1995 was spawned
by a supercell that intensified while crossing over the Hudson River Valley (Bosart
et al. 2006).
6.2.3 Role of Barrier Dimensions and Geometry
The orographic flow response and resulting precipitation distribution is also sen-
sitive to the barrier geometry and dimensions. When the Rossby number is <1,
(Ro D U/f a, where f is the Coriolis parameter and a is the half-mountain width), the
earth’s rotation effect should be taken into account, since it can modify the flow and
the resulting precipitation distribution. This includes relatively wide barriers such as
the Alps, Sierras, and Rockies. For example, Jiang (2006) showed using idealized
simulations of a concave barrier to represent the Alps that the Coriolis force veered
the flow to the right and created a precipitation asymmetry over the barrier. When
the flow is blocked for a long barrier (along-barrier Rossby number < 1), the flow
deceleration and barrier jet will extend upstream of the barrier approximately a
Rossby radius of deformation (lr D Nm hm /f ). Reeves and Rotunno (2008) showed
that the southerly barrier jet along the Sierras can enhance the precipitation over the
ridges that protrude to the west of the crest.
For high Rossby number flows (Ro 1), the width of the terrain determines
in part whether the gravity wave over the barrier will be vertically-propagating or
evanescent. This is determined by the ratio of Nm to the intrinsic frequency of flow
passing over the ridge (U/a). When 2 U/a < Nm , the intrinsic frequency of the flow
up and over the mountain is less than the oscillation supported by the atmosphere
and the waves are vertically-propagating at a vertical wavelength given linear theory
of 2 U/Nm . Colle (2004) showed that the vertical wavelength of the mountain
wave plays an important role on the precipitation distribution across a barrier. For
a relatively wide barrier (a 50 km) and moderate stability (Nm 0.01 s1 ), as the
cross barrier flow increases from 10 to 30 m s1 , the vertical wavelength increases
from 8 to 15 km (Fig. 6.9). As a result, the upward motion over the windward slope
deepens and extends farther upwind from the slope. This in turn shifts the snow field
and resulting surface precipitation upstream of the windward slope by 25–50 km
even in the unblocked flow regime (Figs. 6.9c and 6.10a). These modeling results
suggest that a large and wide barrier may actually be able to seed itself with ice aloft
when upwind tilting mountain waves exist.
In contrast, when the moist stability is reduced to Nm 0.005 s1 for a 2,000 m
mountain, the vertical wavelength is increased, but there is a weaker mountain wave
and vertical motions, so there is little precipitation enhancement upwind of the
windward slope (Fig. 6.10b). Rather, with the weaker mountain wave subsidence
there is more precipitation spillover into the lee of the barrier. There is also more
lee side precipitation spillover for a more narrow mountain. When the barrier half
width is decreased from 50 to 25 km in the (Colle 2004) idealized experiments, the
orographic cloud extends upstream of the windward slope (Fig. 6.10c), but there is
less time for the hydrometeors to fall out over the windward slope given the narrower
barrier. Also, more precipitation spills into the lee of a narrow barrier when the cross
barrier flow is increased.
The geometry of the barrier is also important. For example, although Ralph et al.
(2005) observed nearly moist neutral conditions well offshore of the California coast
using Dropsonde data, Neiman et al. (2002, 2004) found many cases of orographic
precipitation with stable blocked flow along the California coast. Neiman et al.
(2006) showed that cooler and more stable continental air can be advected through
the gaps in the coastal range, such as the Pentaluma gap in central California, and
thus can explain some of this discrepancy between the offshore and near coast
stabilities in this region.
When the flow is blocked, an isolated three-dimensional (circular) barrier will
create an arc-shaped area of upward motion and precipitation over the lower wind-
ward slope (Jiang 2003). A lee-side convergence zone can form when the windward
flow splits around the barrier and converges in the lee. A well documented example
is the Puget Sound convergence zone in the lee of the Olympic Mountains of western
Fig. 6.9 Cross section of potential temperature (solid every 10 K), wind vectors, snow (gray
dashed), graupel (black dashed), and rain (solid) mixing ratios every 0.08 kg kg1 starting at
0.04 kg kg1 for the Nm D 0.01 s1 , a D 50 km, hm 1,500 m, and a cross barrier flow of (a)
10 m s1 , (b) 20 m s1 , and (c) 30 m s1 (From Colle 2004. © American Meteorological Society.
Fig. 6.10 Accumulated precipitation (6–12 h) versus wind speed (m s1 ) for a set of 2-D experi-
ments with a 1,500 m mountain with (a) Nm D 0.01 s1 and a 50 km, (b) Nm D 0.005 s1 and
a D 50 km, and (c) Nm D 0.01 s1 and a D 25 km (From Colle 2004. © American Meteorological
Washington State (Mass 1981), which can enhance precipitation dramatically, such
as the heavy snow event in Seattle, WA on 18–19 December 1990 (Ferber et al.
1993). Lee side precipitation enhancement has also been documented for the big
island of Hawaii (Chen and Nash 1994). For more concave-shaped barriers, such as
European Alps, the easterly blocked flow at low-levels enhances the precipitation
just upstream of the western part of this hook-shaped barrier (Rotunno and Ferretti
2003).
6.2.4 Diabatic and Cloud Microphysical Impacts
There is an important synergy between the cloud microphysical processes, moist

flow, and thermodynamics that can impact the orographic precipitation distribution
(Fig. 6.1). Numerous studies have shown that diabatic effects can modulate the
Fig. 6.11 Doppler-derived down-valley airflow (arrows in m s1 ) resulting from diabatic cooling
processes along the southern side of the Alps (From Steiner et al. 2003. © John Wiley and Sons
Publishing. Reprinted with permission)
strength of the terrain flow blocking. As mentioned above, latent heating within
the cloud reduces the effective stability and non-dimensional mountain height (M),
which can reduce the potential for flow blocking. This in turn results in stronger
vertical motions over the ridges and increases the riming and precipitation fallout
(Medina and Houze 2003; Galewsky and Sobel 2005). This in turn increases
(decreases) the drying ratio (residence time) over the barrier as compared to a barrier
with a few windward ridges or more flow blocking (Colle 2008). For example,
during the partially-blocked 13–14 December 2001 IMPROVE-2 case, riming was
observed over the windward slopes of the Cascades (Garvert et al. 2007), which
was important in the observed and simulated microphysical budgets of precipitation
growth during this event (Woods et al. 2005; Colle et al. 2005a).
Diabatic cooling at low-levels also impacts orographic precipitation. Marwitz
(1987a) showed that melting effects over the windward Sierra Nevada increased
the low-level stratification, which favored increased flow blocking and an increased
barrier jet. Colle et al. (2002) also showed that the low-level flow blocking upstream
of the California coastal range was enhanced by low-level evaporative cooling by
precipitation. During IPEX IOP3 on 12 February 2000 (Colle et al. 2005b), the low-
level diabatic cooling from evaporation and melting increased and the near-surface
stability, which led to enhanced low-level flow blocking and upstream extent of the
enhanced precipitation to the west of the Wasatch Mountains. Steiner et al. (2003)
and Bousquet and Smull (2003) showed that diabatic cooling from evaporation and
melting over the Alps during MAP created a northerly down-valley flow of 5–10 m
s1 within the Po River Basin (Fig. 6.11). The ambient southerly flow approaching
the stable air exiting the Po Valley was lifted over this northerly outflow to create
a broad area of precipitation enhancement upstream of the barrier. For the 17–19
March 2003 Colorado Front Range case (Wesley et al. 2010), large amounts of
melting hydrometeors in moderate to heavy precipitation, with a distinct melting or
bright band near the surface, contributed strongly to the blocked flow structure and
upstream or eastward enhancement of precipitation (not shown).
During the Cascade Project, Hobbs et al. (1973) showed that the amount of
riming was important in determining the fallout of the precipitation over a barrier,
with less rimed particles advecting more over the crest to the lee of the Cascades
(Fig. 6.1). The time scale for microphysical growth and fallout can be compared with
the advective time scale for flow to cross a barrier (a/U) in order to determine the
precipitation distribution and efficiency. If precipitation fallout is rapid with riming
(2–4 m s1 ), then the microphysical time scale is relatively short and even relatively
narrow barriers can have large precipitation rates and drying ratios. On the other
hand, snow falling at 1 m s1 from 2 km above the mountain slope will drift
downwind 20 km, and thus will quickly fall into the lee of a 10-km wide barrier.
A fraction of this lee precipitation will evaporate or sublimate, thus decreasing the
drying ratio, since the surface precipitation decreases while the incoming horizontal
moisture flux does not change.
The time scale for conversion from cloud water to rain in pure water clouds
is 15–30 min (Houze 1993), which would imply that stable flow over a narrow
mountain (a < 10–20 km) or a small hill does not have enough time to produce
much precipitation enhancement. However, in the presence of a widespread frontal
cloud, precipitation can become enhanced via a “seeder-feeder” mechanism over
relatively small (50–200 m) hills (Bergeron 1949, 1960; Storebo 1976). During this
process, which is highlighted in more detail in Chap. 7, precipitation particles falling
from aloft (seeder clouds) grow via accretional growth as they collect small cloud
droplets within the low-level (feeder) cap cloud over a hill (Houze 1993; Bergeron
1965). The “classic” seeder-feeder mechanism described by Browning et al. (1975)
is one in which the low-level feeder cloud has sufficient moisture and vertical
motion to form cloud-sized particles but not precipitation-sized particles. Hence,
precipitation would not occur without the presence of ice particles falling from the
seeder cloud into the feeder cloud and growing by accretion of cloud particles within
the feeder cloud.
Precipitation enhancement over relatively small hills has been observed over
south Wales (Browning 1980; Hill et al. 1981), central Pennsylvania (Barros and
Kuligowski 1998), and southern New England (Passarelli and Boehme 1983; Colle
and Yuter 2007). During these events there is typically strong and moist cross-
barrier flow, which creates a liquid or cloud ice layer just over the hill, while an
approaching frontal system generates the snow aloft. For example, Colle and Yuter
(2007) presented a case in which the precipitation was enhanced by a factor of two
over the 25–75 m hills of Long Island on 1 December 2006. By averaging the Upton,
NY WSR-88D radar for the 1300–1500 UTC 1 December period (Fig. 6.12a), it
was seen that there was precipitation enhancement (>30 dBZ) elongated west–
east across central Long Island (LI) and coastal Connecticut (CT). The maximum
Fig. 6.12 Two-hour reflectivity average (color shaded every 1.5 dBZ) at 0.5 km MSL from (left)
the Upton, NY, WSR-88D radar (white circle) from 1300 to 1500 UTC 1 Dec 2004. (b) Cross
section AB showing the 2-h average of the reflectivities (color shaded in dBZ) (left) observed
from the Upton, NY, WSR-88D radar at 1300–1500 UTC 1 Dec 2004 (From Colle and Yuter 2007.
reflectivity over southeast CT was generally where the coastal terrain was the
steepest. In contrast, there was a localized reflectivity minimum (<27 dBZ) over
southern LI Sound at this level. For cross section A–B during this 1300–1500 UTC
period (Fig. 6.12b), there was a precipitation maximum below 2.0 km MSL over
both LI and coastal CT. Colle and Yuter (2007) used mesoscale model simulations at
1.33-km grid spacing to show that a vertical gravity wave circulation over LI helped
to enhance the precipitation, with the hills and differential frictional at the coast
having a similar contribution to the total precipitation enhancement and the gravity
wave circulations below 3 km. A sensitivity simulation without an ice/snow cloud
above 3 km MSL in the model revealed that the seeder-feeder process enhanced
surface precipitation by about a factor of 4 (not shown).
6.2.5 Large Scale Influences
As a baroclinic wave crosses over a topographic barrier, the character of the

orographic precipitation changes as the ambient large-scale flow, stability, and mois-
ture evolve. For example, using vertically-pointing radar data during IMPROVE-2,
Medina et al. (2007) developed a conceptual model of the commonly observed
progression of precipitation features as extratropical cyclones traverse mountainous
terrain over the Pacific Northwest (Fig. 6.13). As a cyclone approaches, the
precipitation echo first appears aloft and descends towards the surface (Leading
Edge Echo or LEE stage). During the time of maximum large-scale ascent, the
precipitation consists of a vertically-continuous layer extending from the windward
slope up to 5–6 km MSL that persists for several hours. During this period,
there is often a Double Maximum Echo (DME stage) over the windward slopes.
One maximum is associated with the melting-induced bright band. A secondary
maximum (1.0–2.5 km above the bright band) results from or is enhanced by
interaction of the baroclinic system with the terrain. Finally, the Shallow Convection
Echo (SCE) stage is associated with low echo tops in the more unstably stratified
later portion of the storm, which is typically post-frontal.
Figure 6.14 shows a conceptual model of orographic precipitation for a variety
of different barrier types and large scale flows. For a north–south elongated barrier,
such as along the U.S. West Coast (Fig. 6.14a), heavy orographic precipitation
occurs when moist onshore flow develops ahead of an approaching baroclinic wave.
During these events there is sometimes a plume of water vapor extending from the
subtropics to mid-latitudes, sometimes referred to as an “atmospheric river” (White
et al. 2003; Ralph et al. 2004, 2005; Neiman et al. 2008; Smith et al. 2009a). Over
the Pacific Northwest, atmospheric rivers are often referred to as the “Pineapple
Express” (Lackmann and Gyakum 1999; Colle and Mass 2000) because the moist
flow typically originates near Hawaii. Extreme flooding events can occur when the
low-level jet required by hydrostatic and semigeostrophic balance ahead of a front
transports moisture towards the mountains (Fig. 6.14a) (Lin et al. 2001; Rotunno
and Ferretti 2001; White et al. 2003; Ralph et al. 2004, 2005). For example, along
the West Coast, the barrier jet just upstream of the Sierras has been shown to be
important in transporting low-level moisture northward (Galewsky and Sobel 2005).
The orographic precipitation response to a baroclinic wave is different over
the interior western U.S. or lee of a large mountain range. For example, over the
Intermountain western U.S., Steenburgh (2003) showed that prefrontal precipitation
over the Wasatch Mountains becomes increasingly convective as low theta-e
air aloft moves over northern Utah (not shown), while cold frontal passage is
accompanied by a convective line and a stratiform precipitation region. In the
lee of the Rockies (Fig. 6.14b), heavy upslope precipitation typically occurs as a
developing surface low to the south transports moist low-level flow from the Gulf
over an area of blocked low-level flow against the Front Range. A similar situation
can occur along the eastern side of the Appalachian Mountains when there is an
approaching extratropical or tropical storm from the south or western Atlantic.
Fig. 6.13 Schematic illustration from Medina et al. (2007) showing the typical reflectivity
structures observed in the (a) LEE, (b) DME, and (c) SCE periods of midlatitude Pacific cyclones
as they progress toward the terrain of the Oregon Cascade Range. The solid contours enclose
areas of moderate reflectivity, while the shading indicates areas of increased reflectivity. The stars
indicate snow and the ellipses indicate rain. The speckled area shows the orography. The arrows
represent updrafts (© American Meteorological Society. Reprinted with permission)
Fig. 6.14 Conceptual model for orographic precipitation for the north–south elongated barriers
of the (a) U.S. West Coast and (b) lee of the Rockies and the Appalachians, and the east–west
orientated barriers such as (c) the Alps. The low-level winds are highlighted with the arrows and
the terrain helps determine the enhanced upslope precipitation regions
During this situation, there is moist upslope easterly flow over windward (eastern)
Appalachians, which can create additional precipitation and flooding as the cyclone
moves northward along the coast, such as during hurricane Agnes in 1972 (Carr and
Bosart 1978). Mountainous tropical islands are also vulnerable to flooding when
there is a hurricane or typhoon that passes just south or north of the island, such as
what occurred when hurricane Dean passed just south of Dominica and produced
strong easterly upslope flow (Smith et al. 2009b) as well as for several typhoon
events near Taiwan (Yu and Cheng 2008). Finally, an east–west orientated barrier,
such as the Alps, will produce heavy orographic precipitation when an approaching
baroclinic wave from the west results in moist southerly near-neutral flow towards
the barrier (Rotunno and Ferretti 2001; Lin et al. 2001). The area upwind (south) of
the Alps also has a concave shape, which has been shown to enhance rain within the
concavity area (Schneidereit and Schär 2000; Jiang 2006).
6.3 Models and Predictability
There has been a rapid advance in the understanding of orographic precipitation

during the past several years with the application of various models to simulate
precipitation down to 1-km horizontal grid spacing. One of the main goals of
running models at more cloud resolving scales (<5-km grid spacing) has been to
improve quantitative precipitation forecasts (QPFs). However, accurate QPF in areas
of steep terrain is still a challenge given the complex dynamical, thermodynamical,
and cloud microphysical processes in these regions. This section will highlight
some of the models that can be used better understand the theory of orographic
precipitation as well as forecast these events.
6.3.1 Linear Model for Orographic Precipitation
Section 6.2 above highlighted that the primary ingredients for orographic pre-
cipitation include: (1) ascent over the barrier (e.g., upslope, gravity waves, and
large-scale), (2) moisture, and (3) microphysical growth and conversion. One way
to link this theory with operations is to combine these ingredients into relatively
simple yet realistic models. Smith and Barstad (2004) developed a linear steady-
state theory for orographic precipitation. The reader is referred to Smith et al.
(2005) and Smith (2006) for a more detailed description of their linear model for
orographic precipitation, as only a relatively brief description is given here. The
model can be solved analytically, yet it still considers realistic terrain and physical
properties that can be easily understood. The linear model results will be compared
with a more complex mesoscale model for an orographic flooding event in the next
section.
The linear model considers many of the important ingredients for orographic
precipitation, namely the airflow dynamics and cloud physics and how these
processes change as the barrier dimension is varied. The model assumes linear
flow dynamics as well as horizontally-uniform background flow, properties, and
background precipitation. Thus, the model cannot include nonlinear effects on
precipitation, such as flow blocking (Fr < 0.8), unstable conditions, or complex
variations in microphysics aloft. Linear mountain wave theory is used to solve
analytically the three-dimensional airflow over complex terrain, while a linear cloud
physics representation that includes cloud time scales for hydrometeor formation
and fallout is used to obtain the precipitation field. The conversion to a falling
hydrometeor occurs using a time scale (£c ), and they fall to the ground given a time
scale of £f . These time scales can vary from 200 to 2,000 s, which needs to be put in
context with the advective time scale that it takes a parcel to travel from the foothills
to the crestline aloft (£ D a/U). For example, for a barrier with a half width of 10 km
and a wind (U) of 10 m s1 , the £ D 1,000 s. Thus, if £f D 2,000 s, then there is
not enough time for all the condensed water to precipitate before reaching the lee,
while a £f D 200 s favors all the conversion and fallout to occur before reaching the
crest. Thus, a narrow ridge, or a large £f will advect generated cloud water from
the windward side to the lee, where it may evaporate before it is converted into
hydrometeors.
The model is derived using two steady-state advection equations for atmospheric
water:
qN c .x; y/
U rqc D S.x; y/ ; (6.13)
c
qNc .x; y/ qNr .x; y/
U rqr D ; (6.14)
c f
where qc (x,y) and qr (x,y) are the vertically-integrated cloud water and hydrometeor
densities (kgm2); U is the horizontal wind speed averaged over a region of interest,
which can be separated into the eastward (U) and northward (V) components; and
x, y are the Cartesian coordinates. The term S in (6.13) is the source term, defined as
the vertically-integrated condensation rate generated by the supersaturation of water
vapor by ascent:
Z 1
Cw
S.x; y/ D w.x; y; z/e z=Hw dz: (6.15)
Hw 0
In (6.15), Cw is the uplift sensitivity factor that depends on the surface humidity
and lapse rate, Hw is the ambient moist layer thickness, and w(x,y,z) is the terrain-
forced vertical air velocity:
Cw D 0 =; (6.16)
R T 2
Hw D ; (6.17)
L
where ” is the environmental lapse rate, ¡v0 is the surface water vapor density, is
the average moist adiabatic lapse rate, T is the surface air temperature in Kelvin, L
is the latent heat of vaporization, and Rv is the gas constant for vapor.
Linear mountain wave theory is used to obtain the vertical variations in vertical
velocity with altitude using a kinematic lower boundary condition determined from
the horizontal wind and the local terrain gradient:
w.x; y; 0/ D U:rh.x; y/: (6.18)
Unlike the slab model discussed in Sect. 6.2.1 above, this linear model allows
for the decay in vertical velocity with height caused by vertical-tilting mountain
waves and lateral flow around three dimensional obstacles, which in turn can lead to
a decrease in condensation rates aloft. The ascent makes S positive, resulting in an
increase in cloud water downwind. The conversion term in (6.13) acts to decrease
cloud water and convert to rain water at a rate of qc /£c , and the rain fall out in (6.14)
at a rate of qf /£c .
As shown in Smith and Barstad (2004), Equations 6.13–6.15 can be transformed

into Fourier space and after some algebraic manipulation, one can obtain an
expression for the Fourier transform of the distribution of precipitation rate, P(k,l):
SO .k; l/
PO .k; l/ D ; (6.19)
.1 C i ¢c /.1 C i ¢f /
Where S(k,l) is the Fourier transform of the source term in (6.15); k and l are
the horizontal components of the wavenumber; and ¢ D Uk C Vl is the intrinsic
frequency. The transformed source term is given as:
O l/
Cw i ¢ h.k;
SO .k; l/ D ; (6.20)
.1 imH w /
where h(k,l) is the Fourier transform of the terrain and m is the vertical wave number
controlling the depth and tilt of the forced air ascent:
1=2
Nm2 ¢ 2
mD .k C l /
2 2
(6.21)
¢2
By combining (6.19) and (6.20), one can obtain a single equation for the linear
precipitation model:
O l/
Cw i h.k;
PO .k; l/ D : (6.22)
.1 imH w /.1 C i c /.1 C i f /
Terrain is the only gridded variable needed to solve (6.22), and precipitation
is dependent on many physical properties, such as the surface temperature, wind,
stability, moist layer thickness, and cloud delay times (i.e., £c). The first term in the
denominator in (6.22) describes how the airflow dynamics modifies the source term,
while the second and third terms describe the advection of condensed water during
conversion and fallout, respectively. If £f and £c D 0 and the uplift sensitivity factor
Cw D ¡qv , then the solution reduces to the raw upslope model in (6.6) of Sect. 6.2.1
above.
To compute the precipitation field (P) in x, y space, an inverse Fourier transform
can be applied:
Z Z
P .x; y/ D max O
P .k; l/e i.kxCly/
dk dl C P1 ; 0 : (6.23)
The additional precipitation from the background synoptic-scale lift is repre-

sented by adding a P1 term to (6.23). Any negative precipitation values in (6.23)
by the model in decent regions are truncated and represent leeside evaporation.
An advantage of the linear model is that it can represent many of the processes
in orographic precipitation using a relatively simple formulation. There are also a
limited number of input parameters necessary to run, such as: P1 , U, V, T, Hm ,
£c , £f , which makes it easy to implement and run high resolution simulations with
any desktop computer. The disadvantages are that the linear model cannot include
non-linear processes such as flow blocking and gravity wave breaking, and it applies
rather simple microphysical time scales. The linear precipitation model has been
applied to the Oregon coastal terrain and Cascades (Smith et al. 2005), southern
Andes (Smith and Evans 2007), and Iceland (Crochet et al. 2007). The linear model
will also be applied to a case study in Sect. 6.3.3 to explore some of these issues.
6.3.2 Mesoscale Model Capabilities
With additional computational capabilities in recent years, mesoscale models with

horizontal grid spacing of <10-km are now used to predict orographic precipitation
(Benoit et al. 2002; Mass et al. 2002; Richard et al. 2007; among others). It has
been shown that mesoscale models at less than 5-km grid spacing are capable of
generating realistic orographic flows and precipitation structures (see Chap. 10;
Bruintjes et al. 1994; Colle and Mass 1996, 2000; Gaudet and Cotton 1998; among
others). The horizontal resolution needed to accurately simulate the precipitation
depends on the scale of the orographic barrier and ambient conditions. For example,
a relatively wide barrier, such as the California Sierras (L 100 km), requires
horizontal grid spacings of 4 km to realistically predict the precipitation amplitude
over the windward slope and crest (Grubisić et al. 2005). However, for a relatively
narrow barrier, such as the Wasatch Mountains of Utah (L 10 km), at least 1-km
grid spacing is needed (Colle et al. 2005b). For intermediate barriers, such as
the Oregon and Washington Cascades (L 50 km), 4-km grid spacing can still
produce many realistic structures (Colle et al. 2000); however, to properly simulate
the precipitation over the 10–20 km wide narrow ridges and valleys, about 1-km grid
spacing is still needed (Garvert et al. 2007; Colle et al. 2008). All of these previous
studies looked at mainly stratiform precipitation enhancements over terrain. If the
precipitation is more convective and cellular in structure, then less than 4-km grid
spacing may be needed even for a relatively wide barrier (Colle and Zeng 2004).
Since these mesoscale models must be run at <5 km grid spacing in order
to realistically simulate many terrain-forced circulations, the physical processes
are partially resolved at the grid scale. Therefore, the parameterization of clouds
and precipitation depend on the bulk microphysical parameterization (BMPs) in
the models (e.g. Lin et al. 1983; Rutledge and Hobbs 1983, 1984; Reisner et al.
1998; and Thompson et al. 2004, 2008). These schemes use empirically and
theoretically derived sources, sinks, and exchange terms in order to explicitly predict
the hydrometeor type, mixing ratios, and number concentrations for various ice and
water species. Previous studies have shown that orographic precipitation is sensitive
to parameters with the BMPs, such as the snow fallspeeds (Colle and Mass 2000;
Colle et al. 2005a), degree of riming (Lin and Colle 2009), and slope intercept for
snow (Colle and Zeng 2004; Colle et al. 2005a). Chapter 10 highlights the setup
and initialization of these models, while Chaps. 7 and 11 discuss some of the model
microphysical and predictability issues, respectively.
Forecast realism in a mesoscale model with higher resolution does not necessar-
ily translate to improved accuracy in quantitative precipitation forecasts QPFs. For
many studies (Colle et al. 2000; Colle and Mass 2000; Grubisić et al. 2005), there
was little (<10%) increase in model accuracy when reducing the horizontal grid
spacing below 10 km over the Washington Cascades and the Sierras. Some of this
may be attributed to using a relatively sparse precipitation gauge network to validate
the model, which cannot verify the precipitation variability in the higher resolution
domains; however, there are other predictability problems as well, since model
QPF depends on the interplay between many different components of a numerical
weather prediction model (Richard et al. 2007). Model orographic precipitation
errors can originate from uncertainties in the model initialization, especially low-
level moisture over upstream data sparse regions, error growth of the evolving
synoptic-scale forcing, growing temperature, winds, and moisture errors toward
the barrier, model uncertainties in the physical parameterizations, especially the
boundary layer and bulk microphysical parameterizations, and the lateral boundary
conditions in limited area models.
Many studies have highlighted moisture and model deficiencies in simulating
orographic precipitation. Some global positioning system (GPS) precipitable water
vapor (PWV) data assimilation studies highlighted the large sensitivity of QPF
to the model water vapor fields (Falvey and Beavan 2002; Marcus et al. 2007).
Colle et al. (1999) separated the precipitation errors on wind directions and found
the MM5 tended to overforecast precipitation shadowing in the lee of major
orographic barriers when there was significant cross-barrier flow. More recently, two
IMPROVE-2 case studies (Garvert et al. 2005a, b; Lin and Colle 2009) highlighted
snow overprediction aloft and surface precipitation overprediction in the immediate
lee of the Cascades in several of the BMPs in mesoscale models. In contrast, Colle
et al. (2005) highlighted snow underprediction aloft over the Wasatch Mountains of
Utah. Evaluating four cool seasons (2004–2007) of high-resolution precipitation
observations and the Penn-State – National Centers for Atmospheric Research
Mesoscale model (MM5; Grell et al. 1994) output at 4-km grid spacing over the
western Olympic Mountains, Minder et al. (2008) found that the MM5 performed
well in simulating the small-scale precipitation pattern over the ridges and valleys
when averaged over a season, but there were major errors for individual storms.
They suggested that some of these errors may not be related to deficiencies in
model resolution or microphysics alone, since errors from synoptic scale, mesoscale
flow and stability, boundary layer processes, land surface processes, and cloud
microphysics may all contribute to the QPF errors.
The precipitation error growth for a cloud resolving model over terrain can be an
order of magnitude larger than synoptic-scale errors (Hohenegger and Schär 2007).
Relatively small initial condition and boundary errors can propagate through the
domain due to both sound and gravity waves and then can amplify non-linearly
through the triggering and growth of perturbations in areas of convective instability.

Thus, this makes the model forecast much less useful 6–12 h after initialization
(Hohenegger et al. 2006). Thus, this argues for the use of model ensembles for
orographic precipitation. However, there has been limited high-resolution opera-
tional ensemble forecasts of terrain forced flows and precipitation given the large
computational expense necessary to run a limited-area ensemble at <5-km grid
spacing. There are also issues with what perturbations to use in complex terrain and
how to assimilate observations when there is a mismatch between the model and
actual terrain height. Recently, Hohenegger et al. (2008) applied a high-resolution
(2.2 km grid spacing) numerical weather prediction model in an ensemble mode (ten
members) for a 2005 flooding event over the European Alps. The ensemble included
perturbed initial and lateral boundary conditions, and neglected the model physical
uncertainties. The high resolution ensemble was found to be comparable and
sometimes even better than a corresponding coarser resolution (10-km) ensemble
for much of the flooding event. During the first 12-h of the event, the spread in the
precipitation forecasts was dominated by the growing initial condition errors, while
afterwards errors from the coarser resolution domain influenced the forecast through
the lateral boundaries.
6.3.3 The 5–7 November 2006 Flooding Event of the Pacific

Northwest
During an intense atmospheric river (“Pinnapple express”) event on 5–7 November

2006, 15–20 in. (38–51 cm) of precipitation fell over the mountainous terrain of
the Pacific Northwest. Since the freezing levels were relatively high (3–4 km ASL),
most of the precipitation fell as rain, and there was no snow pack to absorb any of
the precipitation. As a result, there was major flooding and debris flows at many high
terrain areas. For example, Mount Rainier Park in southwest Washington State (MR
on Fig. 6.15a) experienced the worst flooding in over 100 years, which closed the
park for several days, and made several of the roads impassable for several months
(Fig. 6.15b). This extreme event represents a nice example of “bridging the gap”
between research and operations by looking at some of the forecast ingredients that
favored heavy rainfall, while exploring some of the predictability issues. Some of
the models discussed above were integrated for this event for this book.
The MM5 (Version 3.7) was used with 12, 4, and 1.33 km domains nested within
a 36-km domain covering a wide region of the eastern Pacific and Pacific Northwest
(not shown). The model was run twice daily during the 1200 UTC 1 November 2006
to 1200 UTC 8 November 2006 period starting at 0000 and 1200 UTC, with initial
and lateral boundary conditions derived from the 6-hourly NCEP Global forecast
system (GFS) model analyses. Four dimensional data assimilation (FDDA, analysis
nudging; Stauffer and Seaman 1990) was used during the first 12 h of each run
in the 36 and 12-km domains. The MM5 was one-way nested down to the 4 and
Fig. 6.15 (a) Terrain over Washington and Oregon with rain gauge sites (X) and other observa-
tional facilities (sounding at SLE and UIL, radar at Portland [RTX]). Circle represents the 150 km
radar range and the rectangle represents a portion of the 1.33 km model domain shown in Fig. 6.16.
The dashed line is the cross section used for the radar Hovmoller plot in Fig. 6.17. (b) Photo of the
road washout in Mount Rainier National Park as a result of the 2–7 November 2006 flooding event
(Photo in (b) is reprinted with permission courtesy of Mike Gauthier, National Park Service)
1.33-km, with these two domains starting at 0600 and 1800 UTC (hour 6 of the
12-km domain). By nudging the outer domains the strategy was to reduce the
large-scale errors as much as possible in order to investigate the performance of
the mesoscale forecasts within the inner nests. The last 12 h of the simulation
(6–18 h for the 4 and 1.33-km domains) were used for the model precipitation
totals for simulations between 1200 UTC 1 November 2006 and 1200 UTC 8
November 2006. The observed precipitation evolution was obtained using numerous
precipitation gauges over the Pacific Northwest as well as the WSR-88D radar at
Portland, OR (RTX) (Fig. 6.15a).
The model physics included the ETA PBL (Janjić 1994), Thompson micro-
physics (Thompson et al. 2008). The modified Kain Fritsch cumulus scheme (Kain
and Fritsch 1993; Kain 2004) was used in the 36 and 12-km domains. Other physics
included the Dudhia (1989) short and long wave radiative transfer, the MM5 simple
slab land surface model. Initial conditions for snow cover were provided by the
Rapid Update Cycle (RUC) analysis (with 20 km horizontal spacing), while SST
and soil temperature and moisture were from GFS analyses.
A well-defined atmospheric river was present during the November 2006 flood-
ing event (Neiman et al. 2008), as shown by the precipitable water vapor derived
from SSMI satellite at 1800 UTC 6 November 2006 (Fig. 6.16a). The interaction of
moist southwesterly flow with the West Coast topography fits within the conceptual
model presented in Fig. 6.14a. This sub-tropical moisture was transported northward
from near Hawaii to the Pacific Northwest as a result of a large-scale 500-mb trough
over the central Pacific (not shown). As compared to the SSMI, the 36-km MM5
realistically simulated the plume of water vapor at 1800 UTC 6 November 2006
(Fig. 6.16b), with only a slight 2–3 mm overprediction of integrated water vapor.
As noted above, it is important that the model accurately simulate the upstream
moisture, flow, and stability for these heavy orographic precipitation events. Fig-
ure 6.17 shows the observed and simulated time series of the winds in the vertical
and water vapor at Salem, OR (SLE on Fig. 6.12a) from 0000 UTC 2 November to
0000 UTC 9 November. From 3 to 7 November, there was generally southerly flow
near the surface veering to southwesterly flow above 1-km ASL. The southwesterly
flow was over 25 m s1 near crest level (1–2 km ASL) on 6 November 2006 and
the moisture deepened around this time with the arrival of the atmospheric river
(Fig. 6.16). The model winds and moisture were close to the observed, with a slight
wind overprediction (2–5 m s1 ) in the lowest kilometer. The freezing level was
between 3 and 4 km for much of this several day period (not shown), which favored
more warm rain processes at low-levels. The model also accurately predicted the
moist stability (not shown), which was slightly more stable than moist neutral
(Nm D 0.0025 s1 ) during much of the event. The non-dimensional mountain height
(M) was 0.28 (where U 18 m s1 , Nm 0.0025 s1 , and hm 2,000 m), so
little flow blocking is anticipated for this case. However, there was still low-level
flow channeling in the Willamette Valley between the coastal range and Cascades
as shown by the southerly winds at SLE in the lowest 1-km ASL (Fig. 6.17).
The storm total precipitation from a portion of the 1.33-km MM5 domain from
2 to 8 November 2006 is shown in Fig. 6.18 as well as the fraction of observed
Fig. 6.16 Precipitable water vapor (PWV, shaded in mm) from (a) SSM/I and (b) 36-km MM5 at
0000 UTC 7 November 2006
Fig. 6.17 Time height plot showing the winds (m s1 , one long bar represents 5 m s1 ) and water
vapor (color contoured in gm3 ) at SLE from (a) observed, and (b) 4-km MM5 from 0000 UTC
02 Nov to 0000 UTC 09 Nov 2006
Fig. 6.18 Mean daily precipitation (color shaded in cm) during the 02–08 Nov 2006 case for
a portion of the 1.33-km domain. Numbers overlaid is the model precipitation divided by the
observed (red >1.5, black <0.7, gray in between). Terrain is contoured every 0.6 km in black.
The green asterisk is the location of June Lake for the precipitation time series in Fig. 6.20
at the SNOTEL and Cooperative Observer gauge sites (forecast amount/observed

amount). The precipitation increased by a factor of 10 from the upstream valley
to the Cascades and the model was within 20% of the observed at many sites.
There was a sharp gradient in precipitation near the Cascade crest and across deep
valleys, since the warm rain processes favor more rapid precipitation fallout and less
spillover into the lee of the ridges.
Figure 6.19 shows a Hovmoller plot of the maximum observed reflectivity versus
time above 2-km ASL on 7 November. It illustrates the non-steady aspect of the
precipitation as noted in other heavy precipitation studies (Smith et al. 2003; Colle
et al. 2008). There were cellular precipitation areas that propagated eastward from
the Pacific Ocean, which became enhanced slightly over the coastal range and
windward Cascades. The amount of enhancement is less than what one might expect
given the precipitation differences between the valleys and Cascades. This is likely
the result of most of the precipitation growth occurring beneath the radar beam
in the lowest 1–2 km. Meanwhile, there was a sharp precipitation gradient in the
immediate lee of the Cascade crest.
Fig. 6.19 Hovmoller plot of observed reflectivity maximum above 2-km ASL (from RTX on
Fig. 6.12a) versus hours on 7 November 2006 for the southwest to northeast section across the
coastal range and Cascades (see dashed line on Fig. 6.15a)
One of the peak precipitation locations for the flooding event was June Lake,
WA over the southwest Washington Cascades (* on Fig. 6.18), where 900 mm of
liquid precipitation fell between 2 November and 8 November 2006. The 1.33-km
MM5 realistically simulated the evolution of the precipitation amounts from 1 to 7
November (Fig. 6.20), with the exception of underpredicting the precipitation early
on 7 November, since the main precipitation band slid southward too fast in the
model. The 4-km MM5 precipitation was close to the 1.33-km, suggesting a point
of diminishing returns from 4- to 1.33-km grid spacing in this area of the Cascades.
Unlike previous studies in this region (Colle et al. 1999, 2005; Garvert et al. 2005a),
there was little model overprediction, which may be the result of the dominance of
warm rain processes over the Cascades. These previous studies have suggested that
the overprediction aloft and at the surface may be related to deficiencies in the ice
microphysical processes.
This case study suggests that mesoscale models can help forecasters anticipate
flooding events when the upstream winds, moisture, and stability are well predicted.
Thus, a forecaster has to gauge the uncertainty in these parameters before using
the mesoscale model guidance. There are also simple (linear precipitation) models
Fig. 6.20 Six-hour precipitation (in mm) from June Lake, WA observations as well as for the
4-km and 1.33-km MM5 from 1 to 7 November 2007. The location of June Lake is shown by the
green asterisk in Fig. 6.18
described above, so how well can they do in predicting the precipitation? Figure 6.21
shows a comparison between the 1.33-km MM5 and the linear model described
above (Smith and Barstad 2004) for a 6-h period during the peak precipitation
period of the storm between 0000 and 0600 UTC 7 November 2006. The linear
model was run using a low-level stability (Nm 0.0025 s1 ), surface moisture
(qv 0.0112 kg kg1 ), wind profiles (248ı at 26 m s1 ), a vertical moisture scale
height of 5,300 m, and a background (large-scale) precipitation rate of 5 mm h1
from the MM5 at 0000 UTC 7 November for a point just upstream of the Oregon
coast (X on Fig. 6.21a). A microphysical cloud and fall speed time scale (£c and £f )
of 1,000 s was used for the linear model, which was the recommended value for the
Oregon region in Barstad and Smith (2005).
The 1.33-km MM5 produced areas of heavy precipitation (9–12 cm) over many
of the south facing slopes of the Cascades, while many of the adjacent valleys had
less than 5 cm. The mean flow upstream of the coastal range was southwesterly, but
channeling of the flow from south to north within the Willamette Valley near the
surface (c.f. Fig. 6.17) favored more upslope flow along the south-facing slopes.
The heaviest precipitation fell over the lower windward slopes of the Cascades
rather than the crest given the relatively high freezing levels (3 km), which favored
more warm rain processes and more rapid fallout over the first peak of rapid ascent.
The sharp gradient in precipitation over the terrain is also the result of warm rain
processes. The linear model produced a similar magnitude over the Cascades, but
the spatial coverage of orographic precipitation enhancement is much broader than
the MM5. The linear model precipitation is shifted more to the upper windward
slope, which suggests that the microphysical time scale (£c 1,000 s) is too long
for this relatively high freezing level case with rapid warm rain processes. There is
Fig. 6.21 Six-hour precipitation (shaded in cm) between 0000 UTC and 0600 UTC 7 November
2006 for the (a) 1.33-km MM5 and (b) linear precipitation model. The available observed values
are given by the black numbers (in cm). The X in (a) marks the location of the model sounding
used to initialize the linear model
also 30% more precipitation in the linear model over the Cascades, which is
consistent with a larger drying ratio in the linear model (35% versus 30% in MM5).
This reduction in the MM5 was likely the result of more evaporation of cloud
water and rain in the lee of the small scale ridges. The streaks in the simulated
MM5 precipitation upstream of the coast also suggest that the precipitation was less
steady in the MM5 (and observations in Fig. 6.19) than the prescribed background
precipitation rate in the linear model. Overall, the linear model reproduced some of
the general patterns of the precipitation enhancement; however, veering in the low-
level winds, non-steady aspects of the precipitation, and more rapid fallout made the
predictions less accurate quantitatively.
An outstanding question is what are the most sensitive atmospheric parameters
in forecasting this heavy precipitation event? In other words, how do relatively
small changes in wind, stability, and microphysical time scales impact the model
precipitation forecast? This was tested by changing these parameters in the linear
model for the 0000–0600 UTC 7 November 2006 period. When the microphysical
time scale (£c and £f ) was decreased from 1,000 to 500 s (Fig. 6.20a), this
dramatically increased the precipitation over the lower slope over the Cascades
by 4–6 cm, which is more similar in location with the MM5. However, the linear
model precipitation is now overpredicted by 50% in many areas over the windward
slope (not shown). A small change in the wind direction from 248ı in the control
to 228ı also created some fairly large changes in precipitation (3–5 cm) over the
windward slope areas (Fig. 6.20b). This is the result of downstream changes in the
precipitation shadowing and hydrometeor advection as the low-level flow is rotated.
The linear model precipitation is less sensitive to 10% reduction in both the wind
speed and moisture. Precipitation is reduced by 1–2 cm (or 10–15%) over some of
the windward slope areas. Increasing the stability by a factor of 2 (Nm 0.005 s1 )
also decreases precipitation over the upper windward slope and crest by 2–5 cm
and increases the precipitation upstream of the Cascades by 1–2 cm. A larger
Nm creates a shorter vertical wavelength (2 U/Nm ) and larger upstream buoyancy
response, thus there is more subsidence over the crest and the precipitation is shifted
more upstream of the barrier. Overall, the largest sensitivity is associated with the
microphysical time scales, and there are also relatively large sensitivities to the wind
direction and ambient stability (Fig. 6.22).
6.4 Summary and Future Directions
There has been substantial progress in our understanding of orographic precipitation

during the last few decades. This progress has been accelerated through the use of
field study datasets, numerical models, and the development of new linear theory
for orographic precipitation. This chapter has highlighted some of these advances
by synthesizing some of these major findings. Past studies have illustrated the
complexity of orographic precipitation on all scales, from the importance of long-
range moisture transport (e.g. atmospheric rivers) to small-scale gravity waves over
Fig. 6.22 (a) Precipitation difference (shaded in cm) between a linear model run using a £ D 500 s
(TAU500) and the control run (£ D 1,000 s) (TAU500 – CTL) for the 0000–0600 UTC 7 November
2006 period. (b) Same as (a) except for a linear model using a wind direction of 228ı rather than
248ı in the CTL. (c) Same as (a) except for a linear run with the wind speed and moisture decreased
by 10% relative to the CTL. (d) Same as (a) except the moist stability increased from 0.0025 s1
in the CTL to 0.005 s1
the windward ridges. Thus, it will continue to be important to link the synoptic
scale evolution with the moist dynamics (gravity waves, flow blocking, etc.) and the
precipitation processes (riming, seeder-feeder, etc.).
The bridge from research to operations has been constructed through the develop-
ment of conceptual models, simplified (upslope) models to understand precipitation
generation, and improved numerical models run at higher resolution. This chapter
presented conceptual models for large-scale flow patterns favoring heavy orographic
precipitation, such as a landfalling baroclinic wave interacting with an atmospheric
river (e.g., U.S. West Coast), cyclogenesis to the east of a north–south barrier (e.g.
Rockies) and to the southwest of a west–east orientated barrier (e.g. Alps).
Mesoscale models have been shown to realistically reproduce many of the
orographic precipitation structures when run down to less than 4-km grid spacing,
with more realism for the ridge-valley scale using grid spacings of 1-km grid
or less. This was highlighted in this chapter by looking at the 2–7 November
2006 flooding event over the Pacific Northwest at 1.33-km grid spacing. A high-
resolution model run, together with a conceptual model for heavy precipitation, can
increase forecaster confidence of a pending major weather event. However, these
same models have precipitation biases over terrain, which are likely related in part
to deficiencies and uncertainties in their bulk microphysical parameterizations and
boundary layer schemes. Some of these bulk microphysical parameterization issues
are highlighted in Chaps. 7 and 10 in this book.
A model that can realistically reproduce orographic precipitation may not
necessarily be useful operationally if its timing, placement, and magnitude of the
precipitation are often poorly forecast. The predictability of orographic precipitation
needs more fundamental research, namely how do small changes in flow, stability,
moisture, and synoptic forcing change the precipitation forecasts over terrain? This
chapter illustrated some of these issues by re-running the linear model for oro-
graphic precipitation with slightly different wind, moisture, and stability soundings
for the November 2006 flooding event. For complex terrain environments, such as
the Oregon Cascades, orographic precipitation appears to be more sensitive to small
errors in wind direction than the stability and moisture.
Given the orographic precipitation uncertainties due to relatively small changes
in ambient conditions, this argues for an ensemble modeling approach in areas
of high terrain. The challenge has been the lack of computational power to run
a large ensemble at grid spacings of <4-km in areas of terrain. Some research
case studies have begun exploring some of these sensitivities to initial conditions
and physics in areas of complex terrain; however, more work is needed. It is
not clear what initial condition perturbation strategies are best for short range
(0–24 h) ensembles over terrain given the complex (unbalanced) flows in the
regions. The background physics and observational errors may be larger than
those ensembles currently running over less complex terrain. More work is also
needed to incorporate additional radar and surface observations into the ensemble
data assimilation approaches. Finally, verification of the ensemble output needs to
go beyond simple point verification for orographic precipitation, but rather more
analysis on the ensemble ability to reliably reproduce the precipitation structures
over terrain and their magnitudes. The ensemble effort should be linked with other
societal forecast issues and applications over complex terrain, such as ensemble
hydrological forecasting.
There are many other scientific questions associated with orographic precipi-
tation that need to be addressed. It has been hypothesized that turbulence within
the planetary boundary layer over the windward slope can enhance aggregational
and accretional growth (Houze and Medina 2005). This hypothesis needs to be
tested with large-eddy simulation (LES) and additional field data. These same LES
models should also be used to investigate the horizontal and vertical mixing over
terrain. Much attention has been given to flow blocking and gravity waves in these
models over terrain, but more work is needed in the verification and improvement
of operational model PBL physics. It is still possible that some of the model
precipitation errors may also be attributed to deficiencies in PBL processes (e.g.,

sub-grid scale frictional drag) over the windward slope (Garvert et al. 2005b).
These PBL issues can lead to errors in gravity wave structures and ascent and flow
over the terrain (e.g., too much lee side spillover). Analysis of the COPS (2007)
experiment will help improve our understanding of the interaction of the boundary
layer and convective initiation in areas of terrain. More research is also needed on
orographic severe convective storms. For example, one area of emerging research is
the interaction of convective systems with steep terrain and ridge-valley structures.
Much of our theoretical understanding on orographic precipitation is for 2-D
and simple 3-D barrier geometries. Additional comprehensive idealized simulations
are needed, looking at the impact of small-scale ridges, terrain bends, upstream
terrain, moisture gradients, and microphysical impacts in three dimensions. The
precipitation model based on linear theory should be enhanced to include more
sophisticated microphysics, thermodynamic profiles, and perhaps some non-linear
processes (e.g. flow blocking). More observations and numerical studies are needed
to better understand the transport and changes in water vapor over a large barrier
with a series of complex ridges and valleys.
Additional field studies are needed to improve the models for orographic precipi-
tation given their microphysical and PBL deficiencies. To improve the microphysics,
it will require the next generation observational sensors that can measure ice mass
directly (rather than derived from the obtained size distributions as in Garvert et al.
2005b), as well as some measure of the degree of riming. It is also important to
supplement these field studies with longer-term in situ ground-based observations
from vertically-pointing radars, particle imagers (for size distributions), and density
and habit measurements of snow.
It is clear that the scientific field of orographic precipitation has matured over the
last few decades, and that there are many exciting and new problems to tackle given
the advancing technology. Future progress will continue to require a collaborative
effort between those using models and observations as scientific tools and those
interested in using the data for forecasting and various hydrologic applications.
Acknowledgements The authors would like to thank the Editors (Fotini Katopodes Chow,
Stephan F.J. De Wekker, and Bradley J. Snyder) for their invitation to write this chapter as well as
their help in organizing this whole effort. We would also like to thank two anonymous reviewers for
their many helpful comments. Contributing author Colle acknowledges the support of the National
Science Foundation under grants ATM-0094524, ATM-0450444, and ATM-0908288.
References
Anquetin, S., F. Minsicloux, J.-D. Creutin, and S. Cosma, 2003: Numerical simulation of
orographic rainbands, J. Geophys. Res., 108(D8), 8386.
Asencio, N., J. Stein, M. Chong, and F. Gheusi, 2005: Analysis and simulation of local and regional
conditions for the rainfall over the Lago Maggiore Target Area during MAP IOP 2b. Quart. J.
Roy. Meteor. Soc., 129, 565–586.
Barros, A. P., and R. J. Kuligowski, 1998: Orographic effects during a severe wintertime rainstorm
in the Appalachian Mountains. Mon. Wea. Rev., 126, 2648–2672.
Barstad, I., and R. B., Smith, 2005: Evaluation of an orographic precipitation model. J. Hydrome-
teor., 6, 85–99.
Benoit, R., C. Schär, P. Binder, S. Chamberland, H.C. Davies, M. Desgagné, C. Girard, C. Keil,
N. Kouwen, D. Lüthi, D. Maric, E. Müller, P. Pellerin, J. Schmidli, F. Schubiger, C. Schwierz,
M. Sprenger, A. Walser, S. Willemse, W. Yu, and E. Zala, 2002: The real-time ultrafinescale
forecast support during the Special Observing Period of the MAP. Bull. Amer. Meteor. Soc., 83,
85–109.
Bergeron, T., 1949: The problem of artificial control of rainfall on the globe. Part II, The coastal
orographic maxima of precipitation in autumn and winter. Tellus, 1, 15–32.
Bergeron, T., 1960: Operation and results of “Project Pluvius.” Physics of Precipitation, Geophys.
Monogr., No. 5, Amer. Geophys. Union, 152–157.
Bergeron, T., 1965: On the low-level redistribution of atmospheric water caused by orography.
Suppl. Proc. Int. Conf. Cloud Phys. Tokyo, pp. 96–100.
Bond. N. A., and Coauthors, 1997: The Coastal Observation and Simulation with Topography
(COAST) experiment. Bull. Amer. Meteor. Soc., 78, 1941–1955.
Bosart L. F. et al., 2006: Supercell tornadogenesis over complex terrain: The great Barrington,
Massachusetts, tornado on 29 May 1995. Wea. Forecasting. 21, 897–922.
Bougeault P., Coauthors, 2001: The MAP Special Observing Period. Bull. Amer. Meteor. Soc., 82,
433–462.
Bousquet O., and B. F. Smull, 2003: Airflow and precipitation fields within deep Alpine valleys
observed by airborne Doppler radar. J. Appl. Meteor., 42, 1497–1513.
Browning, K. A., 1980: Structure, mechanism and prediction of orographically enhanced rain
in Britain. Orographic Effects in Planetary Flows, R. Hide and P. W. West, Eds., GARP
Publication Series, Vol. 23, 85–114.
Browning, K. A., C. W. Pardoe, and F. F. Hill, 1975: The nature of orographic rain at wintertime
cold fronts. Quart. J. Roy. Meteor. Soc., 101, 333–352.
Bruintjes, R. T., T. L. Clark, and W. D. Hall, 1994: Interactions between topographic airflow and
cloud/precipitation development during the passage of a winter storm in Arizona. J. Atmos.
Sci.,51, 48–67.
Carr, F. H., and L. F. Bosart, 1978: A diagnostic evaluation of rainfall predictability for Tropical
Storm Agnes, June 1972. Mon. Wea. Rev., 106, 363–374.
Chen, S.H., and Y.L. Lin, 2005: Effects of moist Froude number and CAPE on a conditionally
unstable flow over a mesoscale mountain ridge. J. Atmos. Sci., 62, 331–350.
Chen, Y.-L., and A. J. Nash, 1994: Diurnal variations of surface airflow and rainfall frequencies on
the island of Hawaii. Mon. Wea. Rev., 122, 34–56.
Chu, C.M., and Y.L. Lin, 2000: Effects of orography on the generation and propagation of
mesoscale convective systems in a two-dimensional conditionally unstable flow. J. Atmos. Sci.,
57, 3817–3837.
Colle, B. A., 2004: Sensitivity of orographic precipitation to changing ambient conditions and
terrain geometries: An idealized modeling perspective. J. Atmos. Sci., 61, 588–606.
Colle, B. A., 2008: Two-dimensional idealized simulations of the impact of multiple windward
ridges on orographic precipitation. J. Atmos. Sci., 65, 509–523.
Colle, B. A., and C. F. Mass, 1996: An observational and modeling study of the interaction of
low-level southwesterly flow with the Olympic Mountains during COAST IOP 4. Mon. Wea.
Rev.,124, 2152–2175.
Colle B. A., and C. F. Mass, 2000: The 5–9 February 1996 flooding event over the Pacific
Northwest: Sensitivity studies and evaluation of the MM5 precipitation forecasts. Mon. Wea.
Rev., 128, 593–617.
Colle, B.A., B.F. Smull, and M.-J. Yang, 2002: Numerical simulations of a landfalling cold front
observed during COAST: Rapid evolution and responsible mechanisms. Mon. Wea. Rev., 130,
1945–1966.
Colle, B. A., and Y. Zeng, 2004: Bulk microphysical sensitivities and pathways within the MM5 for
orographic precipitation. Part II: Impact of different bulk schemes, barrier width, and freezing
level. Mon. Wea. Rev.,132, 2802–2815.
Colle B. A., and S. E., Yuter, 2007: The impact of coastal boundaries and small hills on the
precipitation distribution across southern Connecticut and Long Island, New York. Mon. Wea.
Rev., 135, 933–954.
Colle, B. A., Y. Lin, S. Medina, and B. Smull, 2008: Orographic modification of convection and
flow kinematics by the Oregon coastal range and Cascades during IMPROVE-2. Mon. Wea.
Rev., 136, 3894–3916.
Colle, B. A., K. J. Westrick, and C. F. Mass, 1999: Evaluation of the MM5 and Eta-10 precipitation
forecasts over the Pacific Northwest during the cool season. Wea. Forecasting., 14, 137–154.
Colle, B. A., C. F. Mass, and K. J. Westrick, 2000: MM5 precipitation verification over the Pacific
Northwest during the 1997–1999 cool seasons. Wea. Forecasting., 15, 730–744.
Colle B. A., M. Garvert, J. B. Wolfe, C. F. Mass, and C. P. Woods, 2005a: The 13–14 December
2001 IMPROVE-2 event. Part III: Simulated microphysical budgets and sensitivity studies.
J. Atmos. Sci., 62, 3535–3558.
Colle B. A., J. B. Wolfe, W. J. Steenburgh, D. E. Kingsmill, J. A. W. Cox, and J. C. Shafer, 2005b:
High-resolution simulations and microphysical validation of an orographic precipitation event
over the Wasatch Mountains during IPEX IOP3. Mon. Wea. Rev., 133, 2947–2971.
Colton, D. E., 1976: Numerical simulation of the orographically induced precipitation distribution
for use in hydrologic analysis. J. Appl. Meteor., 15, 1241–1251.
Cosma, S., E. Richard, and F. Miniscloux, 2002: The role of small-scale orographic features in the
spatial distribution of precipitation. Quart. J. Roy. Meteor. Soc. 128, 75–92.
Cox, J.A.W., W.J. Steenburgh, D.E. Kingsmill, J.C. Shafer, B.A. Colle, O. Bousquet, B.F. Smull,
and H. Cai, 2005: The Kinematic Structure of a Wasatch Mountain Winter Storm during IPEX
IOP3. Mon. Wea. Rev., 133, 521–542.
Crochet, P., T. Johannesson, T. Jónsson, O. Sigurdsson, H. Björnsson, F. Palsson, I. Barstad, 2007:
Estimating the spatial distribution of precipitation in Iceland using a linear model of orographic
precipitation. J. of Hydrometeor., 8, 1285–1306.
Doesken, N., 2003: Just what the drought doctor ordered: a summary and observations of the
March 17–20, 2003 snowstorm, a climatologist’s view. Colorado Climate: Water Year 2003–4,
No. 1–4, 13–15.
Douglas, C.K.M., Glasspoole, J., 1947: Meteorological conditions in heavy orographic rainfall in
the British Isles. Quart. J. Roy. Meteor. Soc., 73, 11–38.
Dudhia, J., 1989: Numerical study of convection observed during winter monsoon experiment
using a mesoscale two-dimensional model. J. Atmos. Sci., 46, 3077–3101.
Durran, D. and J. Klemp, 1982: On the effects of moisture on the Brunt-Väisälä frequency.
J. Atmos. Sci., 39, 2152–2158.
Falvey M., and J. Beavan, 2002: The impact of GPS precipitable water assimilation on mesoscale
model retrievals of orographic rainfall during SALPEX’96. Mon. Wea. Rev., 130, 2874–2888.
Falvey, M., and R. Garreaud, 2007: Wintertime Precipitation Episodes in Central Chile: Associated
Meteorological Conditions and Orographic Influences. J. Hydrometeorol., 8, 171–193.
Ferber, G.K., C.F. Mass, G.M. Lackmann, and M.W. Patnoe (1993): Snowstorms over the Puget
Sound Lowlands. Wea. Forecasting., 8, 481–504.
Frame, J. and P. Markowski, 2006: The interaction of simulated squall lines with idealized
mountain ridges. Mon. Wea. Rev., 134:1919–1941.
Fuhrer, O., and C. Schär, 2005: Embedded cellular convection in moist flow past topography.
J. Atmos. Sci., 62, 2810–2828.
Galewsky, J. and A. Sobel, 2005: Moist Dynamics and Orographic Precipitation in North-
ern and Central California during the New Year’s Flood of 1997. Mon. Wea. Rev., 133,
1594–1612.
Garvert M. F., B. A. Colle, and C. F. Mass, 2005a: The 13–14 December 2001 IMPROVE-2 event.
Part I: Synoptic and mesoscale evolution and comparison with a mesoscale model simulation.
J. Atmos. Sci., 62, 3474–3492.
Garvert M. F., C. P. Woods, B. A. Colle, C. F. Mass, P. V. Hobbs, M. P. Stoelinga, and J. B. Wolfe,

2005b: The 13–14 December 2001 IMPROVE-2 event. Part II: Comparison of MM5 model
simulations of clouds and precipitation with observations. J. Atmos. Sci., 62, 3520–3534.
Garvert, M. F., B. F. Smull, and C. F. Mass, 2007: Multiscale mountain waves influencing a major
orographic precipitation event. J. Atmos. Sci., 64, 711–737.
Gaudet, B., and W.R. Cotton, 1998: Statistical characteristics of a real-time precipitation forecast-
ing model. Wea. Forecasting, 13, 966–982.
Georgis, J-F, F. Roux, M. Chong, and S. Pradier, 2003: Triple-Doppler radar analysis of the heavy
rain event observed in the Lago Maggiore region during MAP IOP2b. Quart. Roy. Meteor. Soc.,
129, 495–522.
Grell, G. A., J. Dudhia, and D. R. Stauffer, 1994: A description of the fifth-generation Penn
State/NCAR Mesoscale Model (MM5). NCAR Tech. Note NCAR/TN-398 C STR, 138 pp.
[Available from National Center for Atmospheric Research, PO Box 3000, Boulder, CO
80307.].
Grossman, P. A., and D. R. Durran, 1984: Interaction of low-level flow with the Western Ghat
Mountains and offshore convection in the summer monsoon. Mon. Wea. Rev., 112, 652–672.
Grubisić V., R. K. Vellore, and A. W. Huggins, 2005: Quantitative precipitation forecasting of
wintertime storms in the Sierra Nevada: Sensitivity to the microphysical parameterization and
horizontal resolution. Mon. Wea. Rev.,133, 2834–2859.
Hill, F. F., K. A. Browning, and M. J. Bader, 1981: Radar and raingauge observations of orographic
rain over South Wales. Quart. J. Roy. Meteor. Soc., 107, 643–670.
Hobbs, P. V., 1975: The nature of w artificial inter clouds and precipitation in the Cascade
Mountains and their modification by seeding. Part I: Natural conditions. J. Appl. Meteor., 14,
783–804.
Hobbs, P.V., R. C. Easter, and A. B. Fraser, 1973: A theoretical study of the flow of air and fallout of
solid precipitation over mountainous terrain: Part II: Microphysics. J. Atmos. Sci., 30, 813–823.
Hohenegger, C., and C. Schär, 2007: Predictability and error growth dynamics in cloud-resolving
models. J. Atmos. Sci., 64, 4467–4478.
Hohenegger, C., D. Lüthi, and C. Schär, 2006: Predictability mysteries in cloud-resolving models,
Mon. Wea. Rev., 134, 2095–2107.
Hohenegger, C., A. Walser, W. Langhans, and C. Schär, 2008: Cloud-resolving ensemble simula-
tions of the August 2005 Alpine flood. Q. J. R. Meteor. Soc., 134, 889–904.
Houze, R. A. Jr, 1993: Cloud Dynamics. Academic Press, 573 pp.
Houze, R. and S. Medina, 2005: Turbulence as a mechanism for orographic precipitation
enhancement. J. Atmos. Sci., 62, 3599–3623.
Janjić, Z. I., 1994: The step-mountain eta coordinate model: Further developments of the
convection, viscous sublayer, and turbulence closure schemes. Mon. Wea. Rev., 122, 927–945.
Jiang, Q., 2003: Moist dynamics and orographic precipitation. Tellus, 55A, 301–316.
Jiang, Q,, 2006: Precipitation over concave terrain. J. Atmos. Sci., 63, 2269–2288.
Jiang, Q., 2007: Precipitation over multiscale terrain, Tellus A, 59, 321–335.
Kain J., and M. Fritsch, 1993: Convective parameterization for mesoscale models: The Kain–
Fritsch scheme. The representation of cumulus convection in numerical models, Meteor.
Monogr., No. 24, Amer. Meteor. Soc., 165–170.
Kain, J. S., 2004: The Kain–Fritsch convective parameterization: An update. J. Appl. Meteor., 43,
170–181.
Kikuchi, K., and B. Wang, 2008:, Diurnal precipitation regimes in the global tropics. J. Climate, 21,
2680–2696.
Kirshbaum, D.J., and D.R. Durran, 2005: Atmospheric Factors Governing Banded Orographic
Convection. J. Atmos. Sci., 62, 3758–3774.
Kirshbaum, D.J., G.H. Bryan, R. Rotunno, and D.R. Durran, 2007a: The triggering of orographic
rainbands by small-scale topography. J. Atmos. Sci., 64, 1530–1549.
Kirshbaum, D.J., R. Rotunno, and G.H. Bryan, 2007b: The spacing of orographic rainbands
triggered by small-scale topography. J. Atmos. Sci., 64, 4222–4245.
Kirshbaum, D. J. and R. B. Smith, 2008: Temperature and moist-stability effects on midlatitude

orographic precipitation. Q. J. R. Meteorol. Soc., 134, 1183–1199.
Kuo, Y.H., and G.T.J. Chen, 1990: The Taiwan Area Mesoscale Experiment (TAMEX): An
Overview. Bull. Amer. Meteor. Soc., 71, 488–503.
Lackmann, G. M. and J. R. Gyakum, 1999: Heavy cold-season precipitation in the Northwestern
United States: Synoptic climatology and an analysis of the flood of 17–18 January 1986. Wea.
LaPenta, K.D. et al., 2005: A multiscale examination of the 31 May 1998 Mechanicville,
New York, F3 tornado: Wea. Forecasting, 20, 494–516.
Letkewicz, C. E., and M. D. Parker, 2010: Forecasting the maintenance of mesoscale convective
systems crossing the Appalachian Mountains. Wea. Forecasting, 25, 1179–1195.
Lin, Y.-L., 2005: Dynamics of orographic precipitation. Yearbook of Science & Technology,
McGraw Hill Companies, 248–250.
Lin, Y., and B. A. Colle, 2009: The 4–5 December 2001 IMPROVE-2 event: Observed micro-
physics and comparison with the WRF model. Mon. Wea. Rev., 137, 1372–1392.
Lin Y. L., R. Farley, and H. D. Orville, 1983: Bulk parameterization of the snow field in a cloud
model. J. Climate Appl. Meteor., 22, 1065–1092.
Lin, Y.L., S. Chiao, T.A. Wang, M.L. Kaplan, and R.P. Weglarz, 2001: Some Common Ingredients
for Heavy Orographic Rainfall. Wea. Forecasting, 16, 633–660.
Mapes, B. E., T. T. Warner, and M. Xu, 2003: Diurnal patterns of rainfall in northwestern South
America, Part III: Diurnal gravity waves and nocturnal convection offshore, Mon. Weather
Rev., 131, 830–844.
Marcus, S., J. Kim, T. Chin, D. Danielson, and J. Laber, 2007: Influence of GPS Precipitable Water
Vapor Retrievals on Quantitative Precipitation Forecasting in Southern California. J. Appl.
Meteor., 46, 1828–1839.
Marwitz, J., 1980: Winter storms over the San Juan Mountains. Part I: Dynamical processes.
J. Appl. Meteor., 19, 913–926.
Marwitz, J. D., 1987a: Deep orographic storms over the Sierra Nevada. Part I: Thermodynamic
and kinematic structure. J. Atmos. Sci., 44, 159–173.
Marwitz, J. D., 1987b: Deep orographic storms over the Sierra Nevada. Part II: The precipitation
processes. J. Atmos. Sci., 44, 174–185
Mass, C.F., 1981: Topographically forced convergence in western Washington State. Mon. Wea.
Rev., 109, 1335–1347.
Mass, C.F., D. Ovens, K. Westrick, and B. A. Colle, 2002: Does increasing horizontal resolution
produce better forecasts? Bull. Amer. Meteor. Soc., 83, 407–430.
McMurdie, L., and C. Mass, 2004: Major numerical forecast failures over the Northeast Pacific.
Medina, S. and R. Houze, 2003: Air motions and precipitation growth in alpine storms. Quart.
J. Roy. Meteor. Soc., 129, 345–371.
Medina, S., E. Sukovich, and R. A. Houze, Jr., 2007: Vertical structures of precipitation in cyclones
crossing the Oregon Cascades. Mon. Wea. Rev., 135, 3565–3586.
Meyers, M.P., J.S. Snook, D.A. Wesley, and G.S. Poulos, 2003: A Rocky Mountain storm. Part II:
The forest blowdown over the west slope of the Northern Colorado Mountains – observations,
analysis, and modeling. Wea. Forecasting, 18, 662–674.
Miglietta, M.M., and R. Rotunno, 2005: Simulations of moist nearly neutral flow over a ridge.
J. Atmos. Sci., 62, 1410–1427.
Miglietta, M.M., and R. Rotunno, 2007: Further results on moist nearly neutral flow over a ridge.
J. Atmos. Sci., 63, 2881–2897.
Miglietta, M. M., and R. Rotunno, 2009: Numerical simulations of conditionally unstable flows
over a mountain ridge. J. Atmos. Sci., 66, 1865–1885.
Minder, J.R., D.R. Durran, G.H. Roe, and A.M. Anders, 2008: The climatology of small-scale
orographic precipitation over the Olympic Mountains: Patterns and processes. Quart. J. Roy.
Meteor. Soc., 134, 817–839.
Neiman P. J., F. M. Ralph, A. B. White, D. E. Kingsmill, and P. O. G. Persson, 2002: The statistical
relationship between upslope flow and rainfall in California’s coastal mountains: Observations
during CALJET. Mon. Wea. Rev., 130, 1468–1492.
Neiman P. J., P. O. G. Persson, F. M. Ralph, D. P. Jorgensen, A. B. White, and D. E. Kingsmill,
2004: Modification of fronts and precipitation by coastal blocking during an intense landfalling
winter storm in southern California: Observations during CALJET. Mon. Wea. Rev., 132,
242–273.
Neiman, P.J., F.M. Ralph, A.B. White, D.D. Parrish, J.S. Holloway, and D.L. Bartels, 2006:
A multiwinter analysis of channeled flow through a prominent gap along the northern California
coast during CALJET and PACJET. Mon. Wea. Rev., 134, 1815–1841.
Neiman, P. J., F. M. Ralph, G. A. Wick, J. D. Lundquist, M. D. Dettinger, 2008: Meteorological
characteristics and overland precipitation impacts of atmospheric rivers affecting the West
Coast of North America based on eight years of SSM/I satellite observations. J. Hydrome-
teor., 9, 22–47.
Passarelli, R. E., Jr., and H. Boehme, 1983: The orographic modulation of pre-warm front
precipitation in southern New England. Mon. Wea. Rev., 111, 1062–1070.
Petersen W. A., and Coauthors, 1999: Mesoscale and radar observations of the Fort Collins flash
flood of 28 July 1997. Bull. Amer. Meteor. Soc, 80, 191–216.
Pontrelli M. D., G. Bryan, and J. M. Fritsch, 1999: The Madison County, Virginia, flash flood of
27 June 1995. Wea. Forecasting, 14, 384–404.
Poulos, G. S., D. A. Wesley, J. S. Snook and M. P. Meyers, 2002: A Rocky Mountain storm – Part
I: The blizzard – observations, dynamics and modeling. Wea. And Forecasting, 17, 955–970.
Ralph, F. M., O. Persson, D. Reynolds, W. Nuss, D. Miller, J. Schmidt, D. Jorgensen, J. Wilczak,
P. Neiman, J.-W. Bao, D. Kingsmill, Z. Toth, C. Velden, A. White, C. King, and J. Wurman,
1999: The California Land falling Jets experiment (CALJET): Objectives and design of a
coastal atmosphere-ocean observing system deployed during a strong El Niño. 3rd Symp.
Integrated Observing Systems, 10–15 Jan. 1999, Dallas, TX, AMS 78–81.
Ralph, F.M., P.J. Neiman, and R. Rotunno, 2005: Dropsonde observations in low-level jets over the
northeastern Pacific Ocean from CALJET-1998 and PACJET-2001: Mean Vertical-Profile and
Atmospheric-River Characteristics. Mon. Wea. Rev., 133, 889–910.
Ralph F. M., P. J. Neiman, G. A. Wick, and C. S. Velden, 2004: Satellite and CALJET aircraft
observations of atmospheric rivers over the eastern North Pacific Ocean during the winter of
1997/98. Mon. Wea. Rev., 132, 1721–1745.
Ralph, F.M., P.J. Neiman, D.E. Kingsmill, P.O.G. Persson, A.B. White, E.T. Strem, E.D. Andrews,
and R.C. Antweiler, 2003: The impact of a prominent rain shadow on flooding in California’s
Santa Cruz Mountains: A CALJET case study and sensitivity to the ENSO cycle. J. Hydrome-
teor., 4, 1243–1264.
Rauber, R., 1992: Microphysical structure evolution of a central Sierra Nevada orographic cloud
system. J. Appl. Meteor., 31, 3–24.
Reeves, H. D., and Rotunno, R., 2008: Orographic flow response to variations in upstream
humidity. J. Atmos. Sci., 65, 3557–3570.
Reinking, R. F., J. B. Snider, and J. L. Coen, 2000: Influence of storm embedded orographic gravity
waves on cloud liquid water and precipitation. J. Appl. Meteor., 39, 733–759.
Reisner J., R. M. Rasmussen, and R. T. Bruintjes, 1998: Explicit forecasting of supercooled liquid
water in winter storms using the MM5 mesoscale model. Quart. J. Roy. Meteor. Soc., 124,
1071–1107.
Reynolds, D.W., and A.S. Dennis, 1986: A Review of the Sierra Cooperative Pilot Project. Bull.
Rhea, J. O., 1978: Orographic precipitation model for hydrometeorological use. Ph.D. dissertation,
Colorado State University, Atmospheric Science Paper 287, 198 pp.
Richard E., A. Buzzi, and G. Zängl, 2007: Quantitative precipitation forecasting in the Alps: The
advances achieved by the Mesoscale Alpine Programme. Quart. J. Roy. Meteor. Soc., 133,
831–846.
Roe G. H., 2005: Orographic precipitation. Annu. Rev. Earth Planet. Sci., 33, 645–671.
Romatschke, U., and R. A. Houze, Jr., 2010: Extreme summer convection in South America.
J. Climate, 23, 3761–3791.
Rotunno, R., R. Ferretti, 2001: Mechanisms of intense alpine rainfall. J. Atmos. Sci., 58,
1732–1749.
Rotunno, R., R. Ferretti, 2003: Orographic effects on rainfall in MAP cases IOP2b and IOP8.
Quart. J. Roy. Meteor. Soc., 129, 373–390.
Rotunno, R., and R. A. Houze, Jr., 2007: Lessons on orographic precipitation from the Mesoscale
Alpine Programme. Quart. J. Roy. Meteor. Soc., 133, 811–830.
Rutledge S. A., and P. V. Hobbs, 1983: The mesoscale and microscale structure and organization of
clouds and precipitation in midlatitude cyclones. VIII: A model for the “seeder-feeder” process
in warm-frontal rainbands. J. Atmos. Sci., 40, 1185–1206.
Rutledge, S. A., and P. V. Hobbs, 1984: The mesoscale and microscale structure and organization
of clouds and precipitation in midlatitude cyclones. XII: A diagnostic modeling study of
precipitation development in narrow cold-frontal rainbands. J. Atmos. Sci., 41, 2949–2972.
Sawyer J. S., 1956: The physical and dynamical problems of orographic rain. Weather, 11,
375–381.
Schneidereit, M. and C. Schär, 2000: Idealized numerical experiments of Alpine flow regimes and
southside precipitation events. Meteorol. Atmos. Phys., 72, 233–250.
Schultz, D.M., and Coauthors, 2002: Understanding Utah Winter Storms: The Intermountain
Precipitation Experiment. Bull. Amer. Meteor. Soc., 83, 189–210.
Sinclair M. R., D. S. Wratt, R. D. Henderson, and W. R. Gray, 1997: Factors affecting the
distribution and spillover of precipitation in the Southern Alps of New Zealand – A case study.
J. Appl. Meteor., 36, 428–442.
Skamarock W.C., J.B. Klemp, J. Dudhia, D.O. Gill, D.M. Barker, W. Wang, and J.G. Powers, 2005:
A description of the Advanced Research WRF, Version 2. NCAR Tech. Note., NCAR/TN-
468 C STR, 88 pp. [Available from UCAR Communications, P.O. Box 3000, Boulder, CO
80307].
Smith, B. L., S. E. Yuter, P. J. Neiman and D. E. Kingsmill, 2009a: Water vapor fluxes and
orographic precipitation over Northern California associated with a land-falling atmospheric
river. Mon. Wea. Rev., accepted.
Smith, R. B., Q. Jiang, M. Fearon, P. Tabary, M. Dorninger, and J. Doyle, 2003: Orographic
precipitation and air mass transformation: An alpine example. Quart. J. Roy. Meteor. Soc., 129,
433–454.
Smith, R.B., 2006: Progress on the theory of orographic precipitation. Chapter 1 in Special Paper
398: Tectonics, Climate, And Landscape Evolution, Geological Society of America, Boulder,
Colorado.
Smith R. B., I. Barstad 2004: A Linear Theory of Orographic Precipitation. J. Atmos. Sci., 61,
1377–1391.
Smith, R.B., and J.P. Evans, 2007: Orographic Precipitation and Water Vapor Fractionation over
the Southern Andes. J. Hydrometeor., 8, 3–19.
Smith R. B., I. Barstad and L. Bonneau, 2005: Orographic precipitation and Oregon’s climate
transition. J. Atmos. Sci., 62, 177–191.
Smith R, Schafer P., Kirshbaum D., Regina E., 2009b: Orographic enhancement of precipitation
inside Hurricane Dean. J. of Hydrometeor. 10, 820–831.
Smolarkiewicz, P.K., R.M. Rasmussen, and T.L. Clark, 1988: On the dynamics of Hawaiian cloud
bands: Island forcing. J. Atmos. Sci., 45, 1872–1905.
Stauffer, D. R., and N. L. Seaman, 1990: Use of four-dimensional data assimilation in a limited
area mesoscale model. Part I: Experiments with synoptic-scale data. Mon. Wea. Rev., 118,
1250–1277.
mountain winter storm cycle. Wea. Forecasting, 18, 1018–1036.
Steiner, M. O. Bousquet, R.A. Houze Jr., B.F. Smull, M., 2003: Air flow within major Alpine river
valleys under heavy rainfall. Q. J. R. Meteorol. Soc., 129, 411–431.
Storebo, P. B., 1976: Small scale topographic influences on precipitation. Tellus, 28, 45–59.
Stoelinga, M., and Coauthors, 2003: Improvement of microphysical parameterization through

observational verification experiment. Bull. Amer. Meteor. Soc., 84, 1807–1826.
Thompson, G., R. M. Rasmussen, and K. Manning, 2004: Explicit forecasts of winter precipitation
using an improved bulk microphysics scheme. Part I: Description and sensitivity analysis. Mon.
Wea. Rev., 132, 519–542.
Thompson, G., P.R. Field, R. M. Rasmussen, and W.D. Hall, 2008: Explicit forecasts of winter
precipitation using an improved bulk microphysics scheme. Part II: Implementation of a new
snow parameterization. Mon. Wea. Rev., 136, 5095–5115.
Wasula, A. C. et al., 2002: The influence of terrain on the severe weather distribution across interior
eastern New York and western New England. Wea. Forecasting, 17, 1277–1289.
Wesley, Douglas, M. Meyers, and G. Poulos, 2010: The structure and evolution of the Front Range
blizzard of 17–29 March 2003. Submitted to Nat. Wea. Digest.
Westrick, K.J., P. Storck, and C.F. Mass, 2002: Description and evaluation of a hydrometeorologi-
cal forecast system for mountainous watersheds. Wea. Forecasting, 17, 250–262.
White, A. B., P. J. Neiman, F. M. Ralph, D. E. Kingsmill, and P. O. G. Persson, 2003:
Coastal orographic rainfall processes observed by radar during the California land-falling jets
experiment. J. Hydrometeor., 4, 264–282.
Wratt, D.S., Ridley, R.N., Sinclair, M.R., Larsen, H.R., Thompson, S.M., Henderson, R., Austin,
G.L., Bradley, S.G., Auer, A., Sturman, A.P., Owens, I.F., Fitzharris, B.B., Ryan, B.F., and
Gayet, J-F. (1996). The New Zealand Southern Alps Experiment. Bull. Amer. Met. Soc. 77,
683–692.
Woods, C.P., M.T. Stoelinga, J.D. Locatelli, and P.V. Hobbs, 2005: Microphysical Processes and
Synergistic Interaction between Frontal and Orographic Forcing of Precipitation during the 13
December 2001 IMPROVE-2 Event over the Oregon Cascades. J. Atmos. Sci., 62, 3493–3519.
Wulfmeyer, V., A. Behrendt, H.-S. Bauer, C. Kottmeier, U. Corsmeier, A. Blyth, G. Craig, U.
Schumann, M. Hagen, S. Crewell, P. Di Girolamo, C. Flamant, M. Miller, A. Montani, S.
Mobbs, E. Richard, M.W. Rotach, M. Arpagaus, H. Russchenberg, P. SchlÃssel, M. KÃnig,,
V. Gartner, R. Steinacker, M. Dorninger, D.D. Turner, T. Weckwerth, A. Hense, and C.
Simmer, 2008: The Convective and Orographically-induced Precipitation Study: A Research
and Development Project of the World Weather Research Program for improving quantitative
precipitation forecasting in low-mountain regions. Bull. Amer. Meteor. Soc., 89, 1477–1486
Yu, C.K., and L.W. Cheng, 2008: Radar observations of intense orographic precipitation associated
with typhoon Xangsane (2000). Mon. Wea. Rev., 136, 497–521.
Chapter 7
Microphysical Processes Within Winter
Orographic Cloud and Precipitation Systems
Mark T. Stoelinga, Ronald E. Stewart, Gregory Thompson,

and Julie M. Thériault
Abstract One of the most challenging aspects of understanding and forecasting

weather in the mountainous environment is the phenomenon of orographic pre-
cipitation. While it is the dynamics of orographic flow that yield upward motion
and condensation, it is the field of cloud microphysics that addresses the question
“How does the liquid or solid water mass thus formed produce particles that fall to
the surface?” This chapter addresses this question by reviewing general principles
of cloud and precipitation microphysics, as well as aspects of this discipline that
are specific to winter-time midlatitude orographic storms. It is intended to serve
as an overview of the subject, to provide a sound scientific understanding to the
forecaster who faces the challenge of predicting weather in mountainous locations.
Topics covered include precipitation particle growth processes, interactions between
orographic flows and cloud physics, the “seeder-feeder” mechanism, orographic
storm life cycles, physical processes within the solid–liquid transition region, depth
and density of accumulating snow, and the representation of cloud physics in
numerical models used to forecast winter orographic storms.
M.T. Stoelinga ()

3TIER, Inc., 2001 6th Avenue, Suite 2100, Seattle, WA 98121, USA
e-mail: mstoelinga@3tier.com
R.E. Stewart
Department of Environment and Geography, University of Manitoba, 220 Sinnott Building,
70A Dysart Road, Winnipeg, MB R3T 2N2, Canada
e-mail: ronald stewart@umanitoba.ca
G. Thompson
National Center for Atmospheric Research, Boulder, CO 80307, USA
e-mail: gthompsn@ucar.edu
J.M. Thériault
Département des sciences de la Terre et de l’atmosphère, Université du Québec à Montréal,
Case postale 8888, Succursale Centre-ville, Montréal, Québec, H3C 3P8 Canada
e-mail: theriault.julie@uqam.ca

346 M.T. Stoelinga et al.
7.1 Introduction
Orographic precipitation can be defined as any precipitation that is either modified

or entirely produced by the interaction of atmospheric processes with hilly or
mountainous terrain. The acute need to understand and accurately predict oro-
graphic precipitation stems from a variety of societal impacts, including flood
warning and impact mitigation, highway safety, avalanche forecasting, recreational
or competitive winter sports in the mountain environment, and the effects of climate
variability on mountain-derived water resources.
Orographic precipitation is usually understood to refer to enhancement of pre-
cipitation, although orographic modification of precipitation often involves regions
of both enhancement and depletion relative to what would occur in the absence of
orography. The phenomenon of orographic precipitation is also usually understood
to result from vertical motions forced by ambient airflow interacting with orography,
with perhaps the notable exception being afternoon convection initiated by heating
of mountain peaks. As noted by Sawyer (1956) and in Chap. 6 of this book,
three important factors for orographic precipitation are: (1) large-scale ambient
atmospheric circulation, (2) the interaction of the ambient flow with terrain, and
(3) cloud microphysical processes. It is difficult to cleanly separate these three
factors, because they can operate synergistically. Nevertheless, it is somewhat nat-
ural to focus separately on the dynamics of airflow over orography, which attempts
to answer the question “Why is upward motion (and hence, saturation) occurring
in particular locations?”; and cloud microphysics, which attempts to answer the
question “How does the liquid or solid water mass thus formed produce particles
that fall to the surface?”. The former question is addressed by other chapters in
this book, most notably Chap. 6. The latter question is the subject of the present
chapter. It is intended to serve as an overview of cloud microphysical processes
within the orographic environment, to provide a sound scientific understanding to
the forecaster who faces the challenge of short- to medium-range prediction of
weather in mountainous locations.
The topic of this chapter could be defined broadly enough to cover everything
from winter mountain snow storms, to summer convection over high terrain, to
the enhancement of precipitation arising from interaction of land-falling hurricanes
with coastal terrain. Since each of these could easily yield a chapter (or more) of
material by themselves, the scope of the present chapter must necessarily be limited.
Considering that this book arose in part as an outgrowth of preparations for weather
forecasting in support of the 2010 Olympic Winter Games in Vancouver, Canada,
it seems natural to limit the scope of the present chapter primarily to cold-season
precipitation in midlatitude mountain ranges. As such, the subject of deep unstable
warm-season convection within orographic environments is largely excluded. This
exclusion is intentional, since that topic encompasses a broad array of phenomena,
yet from a microphysical perspective, there is probably little difference between
deep convection over orography and away from it.
7 Microphysical Processes Within Winter Orographic Cloud and Precipitation. . . 347
Cloud microphysics encompasses a wide array of processes that describe the

growth, interaction, and fallout of all hydrometeor types within clouds. Several
classic textbooks provide excellent resources for this sub-discipline of atmospheric
science, ranging from a single chapter in Wallace and Hobbs’ (2006) popular survey
text, to full graduate-level texts entirely devoted to the topic (e.g., Rogers and Yau
1989; Young 1993; Pruppacher and Klett 1997). The reader is assumed to have had
at least limited exposure to cloud microphysical principles at a level commensurate
with, for example, Chap. 6 of Wallace and Hobbs (2006).
Within the orographic environment, there is nothing entirely unique about micro-
physical growth processes or hydrometeor types compared to other environments.
A falling snow particle does not behave fundamentally differently simply because
it grew over complex topography. However, there are certain characteristics of the
orographic environment that lead to distinct, repeatable microphysical phenomena
that a forecaster of orographic weather should become familiar with. Five important
characteristics of the cold-season midlatitude orographic environment are:
1. Vertical velocities of typically around 1 m s1 . These values lie in an intermediate
range between those found in large-scale baroclinically generated stratiform
precipitation systems (a few 10s of cm s1 ) and mesoscale deep convective
clouds (>5 m s1 ).
2. A limited time scale for precipitation to develop that is determined by the
duration of Lagrangian ascent, roughly W/U, where W is the mountain half width
and U is the cross-barrier flow. For half widths ranging from 20 to 60 km, and
cross-barrier flows ranging from 10 to 30 m s1 , this time limit ranges from 10
to 100 min.
3. Moist stability typically ranging from moderately moist stable to weakly moist
unstable. Upward motion is primarily forced, but shallow embedded convective
elements can be important.
4. Low melting levels, ranging from 1 km above crest level to well below crest
level. As a result, cold, warm, and mixed-phase microphysics all potentially play
important roles for precipitation growth.
5. Separate but interacting contributions from orographically forced upward motion
and baroclinically forced upward motion, leading to the so-called “seeder-feeder”
phenomenon.
Each of these characteristics has important implications for the microphysical
behavior of cold-season midlatitude orographic weather systems, which in turn
directly affects the weather experienced at the ground. These characteristics and
associated microphysical implications will be elaborated upon throughout the
chapter.
Most of what we know about the microphysical behavior of orographic weather
systems derives from two avenues of research: observational field studies and
numerical modeling. Topics discussed in this chapter rely heavily upon both
sources. Much insight has been gained in the past four decades from field studies
such as the Cascade Project (Hobbs 1975), the Sierra Cooperative Pilot Project
(SCPP, Reynolds and Dennis 1986), the Winter Icing and Storms Project (WISP,
Rasmussen and Coauthors 1992), the Mesoscale Alpine Program (MAP, Bougeault
and Coauthors 2001)1, the California and Pacific Land-falling Jets Experiments
(CALJET, Ralph and Coauthors 1999; and PACJET, Ralph et al. 2005a), the
Intermountain Precipitation Experiment (IPEX, Schultz and Coauthors 2002), the
Improvement of Microphysical Parameterization through Observational Verification
Experiment (IMPROVE, Stoelinga and Coauthors 2003), and various weather
modification-focused studies in the intermountain western United States (Cooper
and Saunders 1980; Rauber et al. 1986; Sassen et al. 1990; Bruintjes et al. 1994).
Concurrently, modeling studies have abounded, employing idealized configurations
in two and three dimensions, as well as real-weather case studies using advanced
numerical weather prediction (NWP) models and either bulk or explicit microphys-
ical schemes, to explore microphysical processes and sensitivities in orographic
precipitation. The modeling studies often use input conditions that were measured
in the above-listed field studies.
It should be pointed out that most of the above studies occurred in the western
half of the United States, which naturally leads to the question: Are the ideas
reviewed in this chapter generally applicable to less studied ranges elsewhere
in the world, such as the Andes or Ural Mountains? We proceed with cautious
optimism that the answer is “yes”, based on the replicability of many of the
results discussed herein within different U.S. orographic environments, and await
confirmatory studies from the understudied orographic regions of the world. Finally,
we should alert the reader of an intentional bias in this review toward more recent
literature (from the past 15–20 years), since this book is in some ways an update
of the 1990 book entitled “Atmospheric Processes Over Complex Terrain”, which
provides a very good review of the state of knowledge at that time.
The remainder of the chapter is organized as follows. Section 7.2 reviews
microphysical processes and phenomena that have been observed in orographic
environments. Section 7.3 expands the view to examine interactions between cloud
microphysics and the kinematic and dynamical features related to airflow over
topography. Section 7.4 reviews the theory, observations, and modeling of the
“seeder-feeder” process. In Sect. 7.5, the microphysical evolution of orographic
precipitation within the context of the life cycle of passing baroclinic systems
is discussed. Section 7.6 describes the transition zone between liquid and solid
precipitation that is a ubiquitous forecasting problem in many orographic envi-
ronments, and reviews the important physical processes that affect the location
and characteristics of this important zone. Section 7.7 discusses the properties of
accumulating snow at the ground and the factors that determine snow depth and
density. Section 7.8 focuses on the representation of cloud microphysics in NWP
models with specific attention to orographic precipitation. Finally, a summary and
discussion of future directions of research and forecasting are presented in Sect. 7.9.
1
The MAP project was carried out during Autumn 1999 and included cases in which deep
convection was a significant contributor to precipitation. However, many of the principles learned
from MAP carry over to the cold season and are replicated by other strictly cold-season orographic
field studies that generally exclude deep convection.
7.2 Microphysical Processes Within Winter Orographic

Precipitation Systems
7.2.1 Precipitation Growth Processes
An important consequence of the “intermediate” range of vertical velocities

achieved in orographic environments (typically on the order of 1 m s1 ) is that
all known processes of precipitation growth are active, and potentially important.
These processes are reviewed in this section, and their relative importance within
midlatitude orographic winter storms is also discussed.
7.2.1.1 Growth of Rain by Collision and Coalescence
Also known as “warm rain microphysics”, this process involves exclusively liquid
droplets that grow initially by condensation after the cloud has achieved water
saturation. Importantly, it can occur at temperatures above or below freezing,
with the latter situation involving supercooled liquid water. The liquid droplets
subsequently collide, coalesce, and form larger droplets that increase their fall speed
relative to the original cloud droplet population, leading to a positive feedback
whereby larger drops collect more and more cloud droplets. The key to the collision-
coalescence (CC) process is getting it started. The process generally requires at least
some droplets to grow by condensation to a size of 40 m diameter before the onset
of the CC process is seen (A. Rangno, personal communication). Three primary
factors influence the rapidity of the onset of CC:
1. Rate of adiabatic liquid water production: This is essentially equal to the rate of
decrease of an ascending air parcel’s saturation mixing ratio. This rate is larger
for warmer temperatures and for larger ascent rate.
2. Cloud condensation nuclei (CCN) concentration: Relatively “clean” air masses,
with CCN concentrations of order 10s cm3 (typically unpolluted air masses of
a maritime source region) distribute the available condensed liquid water onto
fewer droplets, causing them to grow relatively rapidly by condensation, and
to achieve the necessary size threshold for the onset of CC earlier. In contrast,
“dirty” air masses with CCN concentrations of order 100s cm3 (typically air
masses of a continental source region with varying degrees of pollution or natural
dust) distribute the available liquid among a large number of droplets, slowing
their condensational growth and delaying the onset of CC.
3. Spectral broadening. Various processes cause the initially very narrow size distri-
bution of cloud droplets to broaden so that the large-size “tail” of the distribution
more rapidly reaches the threshold size for CC onset. These processes include
turbulence (leading to increased collisions or differential supersaturation), dif-
ferential evaporation due to cloud edge mixing of dry air, enhancement of
supersaturation and growth of some drops near cloud top due to radiational
cooling, and ultra-giant aerosols giving a growth “head start” to some droplets.
See Rasmussen et al. (2002) for a discussion of the radiative hypothesis, and
Beard and Ochs (1993) for a discussion of the other mechanisms.
An extreme case in which the first two factors are maximized is the flow
of moist tropical air against mountainous island barriers such as the Hawaiian
Islands. The steep terrain produces a large vertical ascent rate, warm moist air
results in large adiabatic liquid water production, and the maritime air mass with
low CCN concentrations results in rapid growth of a small number of droplets.
Precipitation production in this environment has been estimated to occur on a time
scale of 10 min (Woodcock 1975). In winter midlatitude orographic environments,
adiabatic liquid water production is less and CCN concentrations higher than in the
Hawaiian example, but pure CC-generated precipitation still occurs. A particular
phenomenon found to occur relatively frequently on the west coast of the United
States is “non-brightband (NBB) rainfall” (White et al. 2003; Kingsmill et al.
2006), where the name derives from the signature in vertically pointing radar scans
(Fig. 7.1). The lack of a melting-induced brightband2 (e.g., in left 1/3 of the time
series, compared to the obvious brightband and particle acceleration resulting from
melting between 1600 and 1730 UTC) indicates that precipitation does not result
from melting snow particles, but rather from exclusively CC-grown liquid drops
within strongly uplifted moist flow over the coastal topography. A characteristic
of NBB rainfall is a size distribution with fewer large drops compared to BB
rainfall, yielding larger rain rates for the same radar reflectivity (and consequently
an important issue for radar-based quantitative precipitation estimation [QPE]).
Another important phenomenon involving CC-grown precipitation over mountains
is the development of supercooled drizzle, which can lead to dangerous aircraft
icing conditions, as well as ice accumulation on structures on the ground, such
as ski lifts and trams. Ikeda et al. (2007) examined cases observed over the
Oregon Cascades, and found that dangerous levels of supercooled drizzle can
occur in environments of low CCN concentrations (20–30 cm3 ). The high vertical
velocities generated by the orographic lift are able to produce enough liquid water to
sustain CC growth, even in the presence of moderate ice particle concentrations of
1–2 L1 and cold temperatures (T < 15ı C), in contrast to weaker ascent situations
where the presence of any ice causes depositional growth to dominate cloud water
production and CC growth (Rasmussen et al. 2002). Maritime orographic envi-
ronments are favored for supercooled drizzle due to the low CCN concentrations,
although the phenomenon has also been observed at continental locations such as
Colorado.
2
A “brightband” is an approximately horizontal band of enhanced radar reflectivity in a vertical
cross section through a region of stratiform precipitation, caused by the melting of falling snow
or graupel particles below the 0ı C level. The melting particles tend to aggregate more readily into
larger particles, and also develop a liquid “skin”. Both of these effects result in increased reflection
of radar energy and the development of the brightband (Fabry and Zawadzki 1995).
Fig. 7.1 Time–height sections of (a) Doppler vertical velocity (m s1 ) and (b) equivalent radar
reflectivity factor (dBZe), recorded by an S-band profiler within the coastal mountains of California
during the indicated time period. The hourly rainfall accumulations (mm) are listed in (a) (From
White et al. 2003. © American Meteorological Society. Reprinted with permission)
Recent observational studies by Rosenfeld and Givati (2006) and Jirak and
Cotton (2006) have indicated that the increase in pollution emissions upwind
of western United States mountain ranges have resulted in reduced orographic
precipitation during the last century. The results support the hypothesis that higher
CCN concentrations increase droplet concentrations, decrease their mean size, and
suppress the CC process – the “second aerosol indirect effect”. Idealized modeling
experiments by Lynn et al. (2007) and Muhlbauer and Lohmann (2008) with aerosol
effects included support the hypothesized mechanism.
Finally, it should be noted that collectional growth of cloud water by raindrops is
frequently an important process even when the rain was not initiated by the CC
process. Rain formed from the melting of snow or graupel that passes through
the 0ı C level can grow significantly by additional collection of cloud water before
reaching the surface (Yuter and Houze 2003; Colle et al. 2005a), especially in cases
with a high melting level.
7.2.1.2 Growth of Ice Particles by Deposition
Depositional growth of ice particles involves transfer of water mass directly from
the vapor to the ice phase. The onset of ice in clouds (“glaciation”) accelerates
precipitation growth, for several reasons. Since the equilibrium vapor pressure with
respect to ice is lower than that with respect to water, ice particles grow from the
vapor phase at a lower humidity than water droplets. Additionally, ice particles
are far less numerous than cloud droplets [due to ice nuclei (IN) concentrations
generally being several orders of magnitude lower than CCN concentrations]. The
combination of these two properties results in ice particles being able to grow
rapidly to precipitable size strictly from the vapor phase, which cloud droplets
cannot do. Within certain ranges of upward motion and particle concentration (see
Korolev 2007), the Wegener-Bergeron-Findeisen process can occur, in which the
humidity within a liquid and ice cloud drops below water saturation but remains
above ice saturation, allowing ice particles to grow at the expense of nearby
water droplets. Additionally, in mixed-phase clouds, ice particles can increase
their mass by collecting cloud droplets, even more efficiently than do raindrops
(Zängl 2008; Kirshbaum and Smith 2008). (This process is described in more detail
in Sect. 7.2.1.4.) All of these effects combine so that glaciated clouds generally
precipitate much more readily than entirely liquid clouds.
However, as with the growth of precipitation by CC, the bottleneck in the growth
of precipitating ice particles is initiation. Creation of the first ice particles in a cloud
requires either homogeneous freezing of droplets at extremely cold temperatures
(T < 40ı C), or heterogeneous freezing on IN. In a liter of air, typically 100,000
aerosols might act as CCN, whereas only 1 might act as an IN. The ability of
an aerosol particle to act as an IN is dependent on the chemical composition of
the aerosol and its ability to mimic the crystal lattice structure of water. At warm
temperatures (10 to 0ı C), only a very limited number of effective ice nuclei
are available, but this number increases with decreasing temperature and with
increasing humidity. Consequently, clouds with tops warmer than 5ı C usually
do not produce ice particles.3 Purely orographic clouds in the midlatitude winter
environment commonly exhibit cloud top temperatures in the range of 25 to 5ı C,
which is usually cold enough to produce ice, but occasionally not (hence the wide in-
terest in the middle to late twentieth century for the potential to enhance orographic
precipitation through artificial cloud seeding). Whether or not ice is produced in
orographic clouds with warm cloud-top temperatures (>10ıC) is strongly depen-
dent on droplet concentrations, with continental clouds (higher droplet concentra-
tions) being more resistant to ice initiation (A. Rangno, personal communication).
3
See, for example, Meyers et al. (1992) for a discussion of ice nucleation mechanisms and
dependence on temperature, and relative humidity. Also, Rangno and Hobbs (2005) show a useful
graph (their Fig. 14) depicting the likelihood of ice developing in clouds as a function of cloud-
top temperature. The curves shown are not specifically for orographic clouds, but are probably
generally applicable.
Fig. 7.2 Images of snow particles of different habit types: (a) radiating assemblage of plates,
(b) broad-branch, (c) bullet rosette, (d) dendrite, (e) radiating assemblage of dendrites (note
blurry branches protruding out of page), (f) light-to-moderately rimed plate, (g) capped column,
(h) aggregate of needles, (i) column. Sizes of images were not available. Sizes probably range
from 0.5 to 3 mm (From http://www.snowcrystals.com. © Kenneth G. Libbrecht. Reprinted with
permission)
On the other hand, orographic clouds that occur in conjunction with deep baroclinic
weather systems with very cold cloud-top temperatures (<20ı C) typically produce
ample ice throughout their depth.
Once ice is initiated, depositional growth of particles proceeds in the form of
different shapes or “habits”. Basic habits include needles, columns, plates, dendrites,
and bullets (see Fig. 7.2), although many variations on these basic types are
observed. Magono and Lee’s (1966) well-known classification scheme identifies
over 80 different particle types, and subsequent studies such as Bailey and Hallett
(2004, 2009) have added more. Each particle type grows under specific conditions
of temperature and humidity, and has unique characteristics of particle mass and
fall speed (as a function of size), radiative properties, and other attributes that affect
particle behavior in important ways. Figure 7.3a shows the wide variety of terminal
fall speeds of snow particles of different types, taken from measurements in the
Cascade Mountains (McCormick 2009). For a particle that is 4 mm in diameter, the
speeds for different particle types range from 0.7 to 3.0 m s1 .
Of particular interest are dendrites (Fig. 7.2d and e), which are known to grow
at water saturation within the approximate temperature range of 13ı C to 17ı C.
McCormick (2009) used ground-based observations of snow particle types in the
Cascade Mountains to ascertain the mass fractions of habit categories that fell at the
observation location over the course of two winter seasons. He found that dendritic
types comprised 38% of the total mass of depositionally grown snow that reached
the surface. The only other study from which a similar quantity can be extracted,
that of Power et al. (1964), also found a large climatological mass fraction of
dendritic types (52%) over the course of two winter seasons,4 although this was
at a flat-terrain location (Montreal, Canada). Auer and White (1982) found that a
disproportionately large fraction of heavy snow storms in Wyoming were comprised
largely of dendritic crystals, and that in these storms, the level of maximum ascent
corresponded with the dendritic growth temperature range.
Considering that dendrites grow only at water saturation, and in a fairly narrow
temperature range, it may seem unlikely that they would comprise such a large
fraction of the observed mass of snow at the ground. However, McCormick listed
several factors that might contribute to the high fraction. First, the temperature range
at which dendrites grow corresponds to the temperature range of highest absolute
excess of water vapor above ice saturation at water-saturated conditions. This means
that for a given ascent rate, this is the temperature at which the maximum forcing
for depositional growth is possible before the achievement of water saturation
limits further supersaturation with respect to ice. Second, the typical height of
the 15ı C level (the midpoint of the dendritic range) is at 3–4 km altitude in
Cascade Mountain winter storms, which often corresponds to the level of maximum
ascent in such storms5 . Therefore, the dendritic growth zone often coincides with
updrafts that are strong enough to raise the humidity to water saturation. Third,
moisture supply generally maximizes in the lowest atmospheric levels, whereas
active IN concentrations are most plentiful at the coldest temperatures, so the
dendritic growth zone may occupy an optimal overlap of both sufficient moisture
supply and sufficient ice particle concentrations to take up the large moisture supply.
Finally, dendritic crystal structure is well suited to maximize depositional growth.
4
Power et al.’s Table 1 was used to arrive at this number. Only dominant crystal types were
reported, so considerable uncertainty should be attached to this number.
5
This estimate is based on experience examining mesoscale model simulations or forecasts of
orographic storm systems in the Pacific Northwest. In a forecast environment, vertical cross
sections of vertical air velocity in mesoscale forecast model output could be used to predict the
level of maximum ascent in a particular storm, and the temperature at that level, to help determine
the likelihood of dendritic growth.
Fig. 7.3 Best-fit power-law curves of terminal fall speed versus particle diameter, obtained from
video disdrometer and microscope measurements taken near Snoqualmie Pass (winter 2006–2007)
and Stevens Pass (winter 2007–2008) in the Cascade Mountains. Each curve is derived from 15 min
of observations when observed particle types were consistently of one habit type and degree
of riming (see legend in each plot): (a) various particle habit types; (b) radiating assemblages
of dendrites of increasing degrees of riming (From McCormick 2009. © Hafen McCormick.
Fig. 7.4 Visibility versus snowfall measured in Marshall, Colorado on (a) 29 Mar 1995 and (b) 10
April 1995. The lines on the log-log plots correspond to the theoretical calculations for the crystal
types observed during each of the two events, including dry snow, dendrites (P1e), elementary
needles (N1a), heavily rimed dendrites (R3b), rimed plates (R1c), and plates with sector-like
extensions (P1b) (From Rasmussen et al. 1999. © American Meteorological Society. Reprinted
with permission)
The lengthy branches extend rapidly away from the center of the growing crystal
and into the environment where humidity is higher, and the large surface area of
the branched structure allows the crystals to quickly dissipate latent heat produced
through deposition, thereby maintaining large supersaturation with respect to ice.
Particle habit types have many important practical forecasting implications.
Rasmussen et al. (1999) showed that visibility can vary significantly depending only
on the type of particles within the falling snow. When snowfall occurs consistently
as one type, such as dendrites (Fig. 7.4a), visibility exhibits a relatively clean linear
relationship with liquid equivalent precipitation rate, and agrees with that predicted
by theory for dendrites. However, when the snowfall is a mixture of many particle
types, visibility (both observed and theoretically predicted) varies widely for a
given precipitation rate (Fig. 7.4b). Particle habit type also affects properties of new
snow accumulation on the ground, such as density and shear strength. Density is
an important factor in predicting new snow depth (discussed further in Sect. 7.7);
and density and shear strength of new snow are both important factors that affect
the stability of recently accumulated snowpack and direct-type avalanche hazard
(Casson 2009).
7.2.1.3 Growth of Ice Particles by Aggregation
As with liquid droplets, collisions of ice particles can cause them to collect into
larger particles or “aggregates”, such as the needle aggregate illustrated in Fig. 7.2h.
One or more of three hypothesized mechanisms are generally thought to explain
why colliding ice crystals adhere to each other: mechanical interlocking, in which
the irregularities in the crystals become entangled, particularly with dendritic types
(e.g., Rauber 1987); sintering, in which water vapor migrates toward crevices
formed by contact points, due to curvature effects on saturation vapor pressure; and
pressure melting, in which the pressure produced by the collision briefly melts the
contact point, adhering the crystals upon refreezing. “Sticking efficiency” is thought
to increase with temperature, maximizing near 0ı C. See Hobbs (1965) or Field et al.
(2006) for reviews of the mechanisms of ice particle aggregation.
Aggregation can lead to the growth of particles up to several cm in diameter
under ideal conditions, such as just above and within the melting layer, contributing
to the familiar radar “brightband” at that location. The propensity of dendrites
to aggregate with their intricate interlocking arms is also the likely explanation
for the “secondary reflectivity maximum” that has been observed 2.5 km above
the melting-layer brightband in radar observations of orographic precipitation
(Houze and Medina 2005; Kingsmill et al. 2006; Medina et al. 2007). This height
corresponds approximately to the dendritic growth region, assuming a typical lapse
rate of 6ı C km1 .
Within dense populations of very small ice crystals (<300 m, Stickel 1982),
aggregation can help accelerate the growth of particles to precipitable size. However,
other studies indicate that the aggregation process is not of primary importance for
precipitation enhancement within storms. Cotton et al. (1986) performed numerical
experiments testing model sensitivity to rather drastic changes in aggregation
collection efficiency, and found modest (˙15%) changes in precipitation rate at
various locations within a region of complex terrain. The reasons for this lack
of sensitivity are two-fold. First, because they are far less numerous than cloud
droplets (by several orders of magnitude), ice particles typically grow to precipitable
size relatively rapidly via deposition without requiring the assistance of collisions.
Once a precipitating population of ice particles develops, aggregation rearranges
the mass among particles that were already precipitating, but does not convert non-
precipitating water substance into precipitating water substance, as occurs when
nonprecipitating cloud droplets are collected by rain, snow, or graupel. Second,
aggregation does not substantially change important particle properties. Fall speeds
of snow particles, whether single crystals or aggregates, generally increase with size
only up to 1 mm, beyond which their speed vs. size curve rapidly flattens out at
values ranging from 0.8 to 1.5 m s1 (e.g., Locatelli and Hobbs 1974; Mitchell
1996). Bulk densities of snow particles are also rather insensitive to aggregation
(Mitchell et al. 1990).
7.2.1.4 Growth of Ice Particles by Riming
Updrafts driven by orographic ascent are often capable of producing substantial

liquid water, even in the presence of ice particles. When ice particles fall through
supercooled cloud droplets, many of the droplets collide with the particle and freeze
onto it, resulting in growth of the particle by riming. An example of a rimed plate-
like ice crystal can be seen in Fig. 7.2f. The ability of any crystal to collide with
and capture droplets is a function not only of its shape, but also of its fall speed, and
the size of the collected droplets. Collection efficiency (defined as the fraction of
swept-out droplets that adhere to the collector particle) increases from near zero to
near one as the diameter of the collected droplets increases from 10 to 30 m (Wang
and Ji 2000; Lowenthal et al. 2011). There is also a minimum size for the collector
particle before efficient riming can begin, ranging from 100 to 300 m in maximum
dimension, depending on particle habit type (Reinking 1979). As more and more
droplets are captured by an ice particle, the particle density increases as droplets
fill in the open spaces within the crystal structure. This metamorphosis produces
first lightly, then moderately, and then densely rimed snow particles, and finally
graupel-like snow (a particle that still has the approximate shape of a snow crystal
or aggregate, but structural details are completely obscured). Graupel, a particle
that is mostly rime ice, but with interstitial air pockets, does not typically form
as the “next step” for a particle that has grown to substantial size via deposition
and riming as described in the sequence above, but rather, forms from a very
small ice kernel that grows rapidly and exclusively by riming (Reinking 1975).
Vigorous deep convective updrafts eventually produce hailstones (large, nearly solid
ice hydrometeors), but these are uncommon in winter orographic storms. The fall
speed of rimed particles is larger than that of unrimed (Zikmunda and Vali 1972;
Locatelli and Hobbs 1974), and has been shown to increase monotonically with
degree of riming (Hanesch 1999; Barthazy and Schefold 2006; McCormick 2009),
as can be seen in Fig. 7.3b for aggregates of radiating assemblages of dendrites.
An important consequence for orographic storms is the reduced downwind distance
that more heavily rimed particles travel before reaching the ground, which effects
the cross-barrier distribution of precipitation. This was illustrated by Hobbs et al.
(1973), who calculated trajectories of particles falling over the Cascade Mountains
under different assumptions of particle growth (Fig. 7.5). Particles grown primarily
by deposition fall slower and are carried into the lee, whereas particles starting at the
same location but grown primarily by riming fall faster and reach the ground on the
windward slope. As with the warm-cloud CC process, riming is sensitive to aerosol
concentrations (natural and anthropogenic), because higher aerosol concentrations
yield higher cloud droplet concentrations, smaller mean drop sizes, reduced riming
efficiency, and less precipitation growth due to riming. Borys (2000) and Borys et al.
(2003) demonstrated with measurements over Northern Colorado, and Saleeby et al.
(2009, 2011) demonstrated with model simulations in the same area, that the aerosol
effect results in enhanced spill-over of frozen precipitation into the lee of ridges,
reducing the windward precipitation and increasing the leeward.
7.2.1.5 Secondary Ice Production
Secondary ice production refers to the production of ice particles through processes
other than initiation by IN. Hobbs and Rangno (1985, 1990) and Rangno (2008)
provide a detailed review of the hypothesized mechanisms for secondary ice
Fig. 7.5 Calculated hydrometeor trajectories within simulated airflow across an idealized 2D
mountain range, originating at points A and B and growing by deposition and riming. Trajectories
are based on hydrometeors grown from ice particle concentrations of 1 (solid line), 25 (dashed
line), and 100 (dotted line) particles per liter. Fewer ice particles (solid) leads to formation of more
cloud water, dominance of riming growth, and faster fall out; whereas more ice particles (dotted)
leads to less cloud water, dominance of depositional growth, and slower fall out (From Hobbs et al.
1973. © American Meteorological Society. Reprinted with permission)
production. These include the rime-splintering process (Hallett and Mossop 1974);
droplet shattering (Hobbs and Alkezweeny 1968); and fragmentation of existing ice
particles (Brewer and Palmer 1949). Rime-splintering and droplet shattering both
arise when liquid drops freeze from the outside in, causing a build-up of pressure
in the liquid core of the freezing drop, which ultimately explodes and produces a
large number of new small ice particles. In rime-splintering, cloud droplets undergo
the outside-in freezing process upon collection by frozen precipitation. In droplet
shattering, liquid precipitation drops freeze after ice nucleation or contact with
small ice crystals. Fragmentation involves the break-up of delicate ice particles
due to collisions with other ice particles. Hobbs and Rangno (1990) questioned the
evidence that any of these mechanisms can explain the degree of ice enhancement
observed in many clouds, and leave open the possibility that an as yet unidentified
process may be contributing.
Secondary ice production is important not for the additional mass of ice
produced, but rather, for the multiplication of the number of ice particles present,
which can open the “bottleneck” for growth of ice by deposition or riming in ice-
supersaturated conditions. The potential for secondary ice production is particularly
important for purely orographic clouds, which often contain supercooled liquid
water, but whose cloud top temperatures are not quite cold enough to possess
sufficient numbers of IN for significant primary ice production. Several studies have
found evidence of secondary ice production in midlatitude winter orographic clouds.
Both rime-splintering and crystal fragmentation were suggested to be important
mechanisms for ice enhancement observed by Hobbs and Atkinson (1976) and
Woods et al. (2008a) in the Cascades; and by Marwitz (1987b), Rauber (1987),
and Demoz et al. (1993) in the Sierras.
7.2.1.6 Relative Importance of Different Growth Mechanisms
A review of the literature on the prevalence of the growth processes described above
within midlatitude winter orographic storms indicates that all growth processes are
important within this environment. The key characteristic of these storms, which
enables the inclusion of all growth processes, is the vertical velocity. By contrast,
non-orographic synoptic-scale weather systems are usually deep enough to produce
plentiful ice, but often not vigorous enough to produce significant liquid water
in the presence of ice, leading to a dominance of depositional growth above the
melting level, with perhaps modest additional CC growth below it. At the other end
of the spectrum, vigorous deep convection produces so much liquid water even in
the presence of ice, that collectional growth processes (riming and CC) dominate.
In orographic systems, upward motion is in an intermediate range where vapor
loss above the melting level is often roughly equally partitioned between direct
deposition onto snow particles, and condensation onto cloud droplets which can
subsequently be collected by the falling snow (or by rain below the melting level,
the contribution of which increases with increasing melting level height).
This view is largely confirmed by studies that have specifically looked at the
relative importance of different precipitation growth processes in orographic storms,
both from observational and modeling perspectives. Vertically pointing radar has
been a valuable tool for deducing microphysical processes and particle types within
orographic precipitation systems. Yuter and Houze (2003) examined the radar
characteristics of orographic precipitation within the southern Alps, and found
collection of cloud water by ice particles above the melting level and by raindrops
below the melting level to be approximately of equal importance to precipitation
growth. White et al. (2003) analyzed 3 months of radar data gathered during heavy
orographic precipitation events in northern California during an abnormally wet
winter, and found that approximately 28% of the precipitation occurred during “non-
brightband” periods, in which growth is almost entirely through the CC process,
with ice playing no role. The remainder of the time, precipitation resulted from snow
particles growing by deposition and riming above the melting level, and additional
CC growth of raindrops beneath it, although relative amounts were not quantified.
Both of these studies were in relatively warmer climates with melting levels
between 2 and 3 km above sea level, leaving considerable distance between the
melting level and the upslope terrain for hydrometeors to grow further by collection
after having melted. In colder climates (or at higher elevations), much of the
precipitation reaches the surface as snow, leaving two primary growth processes:
deposition and riming. The relative contributions of these two processes are difficult
to extract from observations, but a handful of studies have attempted to do this based
on formvar imaging, stereo-microscopic observations and chemical analysis of snow
particles reaching the surface, and/or precipitation and snow density measurements
(Harimaya and Sato 1989; Mitchell et al. 1990; Dixon et al. 1994; Kalina and
Puxbaum 1994; Borys et al. 2003; McCormick 2009; and Lowenthal et al. 2011).
These studies ranged in length from 3 h during a single storm (Harimaya and Sato
1989), to nearly continuous 15-minute observations during the majority of two full
winter seasons (McCormick 2009). The percentage of total snow mass produced
by riming (as opposed to vapor deposition), referred to as the “rime mass fraction”
(RMF), ranged from 17%–60% in individual storms within these studies, but the
average RMF among the three longest lasting of the studies (Mitchell et al. 1990,
McCormick 2009, and Lowenthal et al. 2011) was 38%.
A series of modeling studies by Colle and collaborators has provided valuable
information on the relative importance of growth processes within simulated
orographic storm systems (Colle and Zeng 2004a, b; Colle et al. 2005a, b). In
these studies, Colle et al. ran mesoscale model simulations of orographic precip-
itation over various midlatitude barriers and under various winter environmental
conditions, and tracked the transfer of mass by all processes represented in the
microphysical scheme used. If all processes that lead to growth of hydrometeors
are classified into either CC, depositional, or riming growth, and averaged over the
nine “pathway diagrams” shown in those four papers, the percentage contributions
to precipitation growth are 36% for deposition, 33% for riming, and 31% for CC,
which is uncannily close to a 1:1:1 ratio. Of the three, the most variable is CC,
which is related to the fact that it occurs mostly (though not entirely) below the
melting level, and therefore is quite sensitive to the depth of the T > 0ı C layer. In
high melting level scenarios, such as in the southern Alps, CC can be dominant,
whereas in ranges where there often is no melting level, such as the Wasatch, CC
is often nonexistent. On the other hand, the growth of frozen precipitation seems to
consistently divide roughly evenly between deposition and riming in the Colle et al.
simulations. Variability does occur, related to microphysical sensitivities to barrier
and upstream flow characteristics (discussed further in the next section). The mean
1:1 ratio of deposition to riming growth in the Colle et al. simulations is somewhat
lower than the average observed value of 1.6 :1 (RMF D 38%, as described above).
This difference may indicate fundamental flaws in the representation of depositional
and/or riming growth of snow in models. However, the disagreement could also
arise from differences in observed and modeled initial states, errors in observational
derivation methods, or lack of observational representativeness in space or time.
In spite of this disagreement, the picture that emerges from both observations and
modeling results is that the winter orographic environment is a microphysically rich
environment in which all three major precipitation growth processes (deposition,
riming, and CC) are important.
7.2.2 Precipitation Loss Processes
The two mechanisms by which precipitation is lost to the atmosphere before

reaching the ground are sublimation and evaporation. In the orographic environment
Fig. 7.6 East-to-west cross section, across the Cascade Mountains in Oregon, of mesoscale
model-simulated snow mixing ratio (heavy contours, g kg1 ) overlaid on airborne dual-Doppler
derived radar reflectivity (dBZ; color shading, key at right) during the 13–14 December 2001 storm
studied during the IMPROVE-2 field project. The simulated mixing ratio shown was averaged in
time over the period 2300 to 0100 UTC 13–14 December 2001 (model forecast hours 11–13).
Heavy arrow shows direction of airflow (From Garvert et al. 2007. © American Meteorological
under conditions of cross-barrier flow, these typically occur in dramatic fashion

leeward of the crest, as the vertical air velocity quickly switches from ascending
to descending. A vivid depiction of the depletion of precipitation falling on the lee
side of the Oregon Cascades can be seen in Fig. 7.6 (from Garvert et al. 2007).
There is a rapid transition from large precipitation mixing ratios above the crest to
near zero values in the immediate lee. This rapid transition is often visible from a
vantage point on the lee side of a mountain range looking toward the crest, in the
form of a foehn-wall cloud, appearing as a cloud boundary which does not advance.
Because snow usually occupies a deeper and higher layer of the atmosphere
than rain, and also falls more slowly, it is more susceptible to being transported
over the crest and subjected to lee-side subsidence drying than is rain, reducing
precipitation efficiency6 when cross-barrier flow is strong, particularly with a low

melting level (Colle 2004). The depletion of precipitation in the lee does not occur
instantaneously, as some time is required to sublimate or evaporate the cloud and
precipitation hydrometeors remaining in the flow over the crest. The amount that
does survive in the lee and reaches the ground, referred to in the literature as
spill-over precipitation (Myers 1962), is dependent on several characteristics of the
barrier and atmospheric conditions (to be discussed in the next section).
Another scenario in which sublimational or evaporational loss of precipitation
plays a role is when precipitation that is generated aloft (either through orographic
or synoptic-scale processes) falls through a dry air mass in the lower troposphere
(Steenburgh 2003). This scenario is often important in continental mountain
environments such as the Intermountain West region of the United States. Less
precipitation reaches the ground over the lower terrain, because the particles fall
through a deeper dry layer than those falling over the high terrain. The effect can
be exacerbated if the dry layer experiences orographic lift (J. Steenburgh, personal
communication). Even if the lift is not strong enough to produce cloud, it will raise
the relative humidity over the crest, and further reduce the rate of evaporative or
sublimational depletion over the higher terrain compared to the valleys.
7.3 Interaction of Microphysics with Kinematic

and Dynamic Aspects of Orographic Flows
Chapter 6 provides a detailed review of the variety of ways in which atmospheric

flow responds to orographic barriers, and how these dynamical mechanisms are
altered by characteristics of the terrain geometry and the incoming flow. Many
of these dynamical features and responses have important interactions with cloud
microphysics and the development of precipitation, and it is these interactions
(rather than the dynamical features themselves, already thoroughly reviewed in
Chap. 6) that are the subject of this section.
7.3.1 Simple Terrain-Following Flow and Fundamental

Time Scales
As a starting point, it is useful to consider orographic precipitation in a very simple

model, in which air flow is forced up and over a topographic barrier in essentially
terrain-following stream lines, as encapsulated in various early models that were
6
Precipitation efficiency is defined as the ratio of total precipitation reaching the ground within a
specified area and time period, to the total condensate produced in the volume of the atmosphere
above that area and within that time period.
Fig. 7.7 Schematic diagrams of microphysical processes within orographic precipitation. In all
panels, thin lines with arrows represent airflow; stars are unrimed snow particles. Dark gray
regions are areas of cloud liquid water; coated stars are rimed snow particles; circles are raindrops
of various sizes; the heavy dashed line is the melting level on the windward side of the barrier. (a)
Purely orographic precipitation under conditions of unblocked flow. “U” and “W” are the ambient
wind strength and mountain half width, respectively. (b) Orographic precipitation under conditions
of blocked flow. Light gray region bounded on top by the thin dashed line is the pool of blocked,
stable air, and the gray arrow represents down-valley (reverse) flow which is enhanced by melting
and evaporation of falling precipitation, and in turn enhances the pool of blocked air. (c) Mixed-
phase seeder-feeder process. Large open arrows represent large-scale ascent in middle to upper
troposphere
intended either for conceptual understanding or basic forecasting (e.g., Bader and
Roach 1977; Rhea 1978; Smith 1979). This process is depicted schematically
in Fig. 7.7a. Approximately terrain-following airflow leads to windward ascent,
producing both liquid water cloud (primarily at lower altitudes) and ice particles
(primarily at higher altitudes). The ice particles grow by deposition and fall into the
lower liquid cloud, growing either by further deposition or by riming. They reach the
ground as snow at upper elevations (depending on the height of the melting level),
and at lower elevations, they pass through the melting level and reach the ground as
rain. At the crest, remaining precipitation particles flow over the crest and are then
subject to lee subsidence and drying prior to reaching the ground.
The effectiveness of the system in actually producing precipitation is controlled
by three fundamental time scales, as pointed out by Jiang and Smith (2003) and
Smith (2003). On the one hand, the time available for precipitation to form is
limited by the advective time scale, £a , which is the length of time an air parcel
spends in the ascending air stream, or roughly W/U, where W is the mountain half
width and U is the cross-barrier flow. On the other hand, the time required for
formation of precipitation is given by the conversion time scale, £c , which is the
time required for microphysical processes to convert initial condensate into falling
particles; and the fallout time scale, £f , which is the time required for precipitable
particles, once formed, to reach the ground. If £c C £f is larger (smaller) than £a ,
precipitation efficiency will be low (high). Several observational and modeling
studies have attempted to quantify £c , but not surprisingly, have arrived at very
different values. Barstad and Smith (2005) found values ranging from 0 to 1,000 s
for orographic precipitation in three different geographic regions (the Coastal Range
of California, the Wasatch Mountains in Utah, and the Alps); Sinclair (1994)
found that 1,000 s provided the best correlation between modeled and observed
raingauge measurements in his model of orographic precipitation in New Zealand;
and Smith et al. (2005) found that 1,800–2,400 s was appropriate for the Cascade
Mountains in the state of Oregon. The wide range of conversion time scales is a
result of the different pathways for conversion described in the previous section
(e.g., condensation and CC growth of rain; purely depositional growth of snow;
depositional growth of snow augmented by cloud water production and riming),
and the fact that the dominant pathway and its intensity are nonlinearly dependent
on the size and shape of the orographic barrier and on atmospheric conditions.
Furthermore, £f is also highly variable, since it depends on the height above the
terrain at which precipitation forms, and on the terminal fall speed of the precip-
itating particles. Many studies (e.g., Nakaya 1954; Brown 1970; Zikmunda 1972;
Zikmunda and Vali 1972; Locatelli and Hobbs 1974; Kajikawa 1976; Heymsfield
and Kajikawa 1987; Kajikawa 1989) have measured the fall speeds of precipitating
particles of various types, and find that the speeds vary due to three characteristics:
size, crystal habit type, and degree of riming (with the latter two characteristics
applying only to frozen precipitation). As seen in Fig. 7.3a, all frozen particle types
yield increasing speed for increasing size, although most particles’ curves “flatten
out” at sizes larger than a few mm (except for graupel). Among habit types, dendritic
particles are the slowest falling for a given size, due to their larger aerodynamic
profile for the same mass. Rimed particles fall faster than unrimed particles of
the same type and size (Fig. 7.3b). Graupel particles, which form in environments
of rapid riming growth compared to depositional growth, fall significantly faster
than depositionally grown crystals or aggregates thereof, reducing £f , as depicted in
Fig. 7.5. Roe and Baker’s (2006) modeling study also demonstrated the importance
of accounting for the vertical distribution of precipitation formation, which adds
horizontal spread to the precipitation distribution at the ground, because particles
forming higher above the terrain fall farther downwind.
7.3.2 Effects of Mountain Geometry and Flow Characteristics

on Microphysical Processes
In spite of the complexity of microphysical growth pathways and their nonlin-

ear dependence on the physical characteristics of the mountain and impinging
atmospheric flow, several studies have provided basic insights into the relationships
between these characteristics and the growth and distribution of precipitation across
a mountain barrier.
Mountain height exerts a roughly linear effect on precipitation rate, but does not
significantly alter the distribution of precipitation or its efficiency (Colle 2004), with
the exception of the effect of mountain height on flow blocking, discussed below.
Mountain width, on the other hand, strongly influences the microphysical evolution
of orographic precipitation. A narrower barrier creates a steeper upslope zone
and more vigorous condensation (for the same mountain height and flow speed).
As such, the dominant cold-cloud microphysical pathway shifts from deposition-
dominant growth to condensation and riming-dominant growth as the barrier gets
narrower (Colle and Zeng 2004b). However, a narrow barrier also reduces the
advective time scale, £a , allowing less time for microphysical conversion and fallout,
and leading to a greater fraction of precipitation being carried into the lee (rather
than falling out on the windward slope), more loss of precipitation in the lee to
sublimation, and generally lower precipitation efficiency (Colle 2004).
The speed of the cross-barrier flow also has important implications for precip-
itation distribution and efficiency. Neiman et al. (2002) examined several coastal
California storms and found that the flow at crest height correlates strongly with
overall precipitation rate. However, stronger flow also results in a shift of the
precipitation maximum up to the crest or even into the lee (Sinclair et al. 1997; Roe
and Baker 2006). This is particularly true for slower falling frozen precipitation,
whereby stronger cross-barrier flow carries more of the snow into the lee where a
larger portion is lost to sublimation before reaching the ground, leading to lower
precipitation efficiency (Colle 2004).
The height of the melting level relative to the mountain crest affects the
precipitation distribution primarily through the relative fall speeds of rain droplets
(fast) versus graupel (slower) or snow (slowest). When the melting level is high,
precipitation growth is largely through accretion of cloud water by fast falling
raindrops, and downwind transport of precipitation is minimized, resulting in a
precipitation maximum on the windward slope or near the crest. As the melting
level descends, more of the precipitation grows as snow or graupel, and is carried
farther downwind toward the lee (Colle 2004; Colle and Zeng 2004b; Zängl 2007;
Minder et al. 2008). One modeling study even showed that with the melting level
intersecting the terrain, a double maximum can occur, with maxima on both the
windward and lee slopes (Zängl 2008). What is somewhat less clear is the relative
efficiency of warm rain versus mixed-phase growth processes. Colle (2004) found
that a higher melting level increased the precipitation efficiency due to more
prominent warm rain accretional growth; whereas Kirshbaum and Smith (2008)
found the opposite, that a higher melting level yielded less efficient precipitation
because riming growth was more efficient than warm rain accretional growth.
Yuter and Houze (2003) observed both processes to be occurring in an Alpine
orographic storm and concluded both were necessary for highly efficient orographic
precipitation.
7.3.3 Effects of Blocking on Microphysical Processes
Chapter 6 discusses at length the conditions that lead to blocked flow along the lower
windward slopes of an orographic barrier. This dynamical response has important
interactions with cloud microphysical processes that affect precipitation growth and
its distribution across the barrier. When blocked flow occurs, the pool of low-level
blocked air acts essentially as an extension of the barrier upstream to the incoming
flow (Fig. 7.7b). In a sense, the barrier is widened and given a shallower slope. The
widening of the effective barrier increases the advective time scale, £a , allowing
more time for hydrometeors to develop and fall out on the windward slope, generally
shifting the precipitation maximum down the windward slope (Peterson et al. 1991;
Doyle 1997; Sinclair et al. 1997; Houze et al. 2001; Colle 2004; Colle et al. 2005b),
and increasing the precipitation efficiency (Peterson et al. 1991). Because the ascent
up the effective slope is more gradual than it would be if it followed the actual
terrain slope, upward motion is less vigorous, and the pathway for growth of frozen
particles tends more toward deposition, as depicted by the unrimed snow particles
in Fig. 7.7b; in contrast, in unblocked flow (Fig. 7.7a), airflow is directed up the
steeper terrain slope, leading to more vigorous ascent, production of supercooled
cloud water, and more riming growth (Medina and Houze 2003; Pujol et al. 2005;
Lascaux et al. 2006; Rotunno and Houze 2007).
There is also the potential for an interaction in the other direction, whereby
microphysical processes enhance the pool of stable air along the windward slope,
thereby enhancing the blocking of the low-level cross-barrier flow (depicted by
the reverse-flow arrow near the windward terrain in Fig. 7.7b). Marwitz (1987a)
first described this process for a Sierra Nevada orographic storm. Steiner et al.
(2003) and Bousquet and Smull (2003) observed reverse flow down the valleys
on the windward side of the barrier during an Alpine orographic precipitation
event that exhibited blocked flow. Steiner also concluded that the down-valley flow
resulted from evaporative or melt cooling of falling precipitation particles. Colle
et al. (2005b) observed a similar phenomenon during a blocked flow regime in an
orographic winter storm in the Wasatch Mountains of Utah, and a model sensitivity
study demonstrated the important role that diabatic cooling played in enhancing the
blocked flow.
7.3.4 Effects of Gravity Waves on Microphysical Processes
Colle (2004) and Chap. 6 in this book provide detailed descriptions of the different
gravity wave responses that can be elicited with different mountain or flow
characteristics such as mountain width, moist stability, and vertical wind shear. The
gravity wave response to flow over terrain can interact with precipitation growth
and fallout processes to enhance or diminish precipitation. Under the simplified
assumption of terrain-following flow (or with an evanescent mountain wave), there
is a fundamental lack of correspondence between the ascent region, which extends
9.0 U30
7.0
5.0
km
3.0
1.0
100 (km) 200 300
Fig. 7.8 Vertical cross section through an idealized 2D simulation of orographic precipitation,
with 30 m s1 cross-barrier flow and a moist Brunt-Vaisala frequency of 0.01 s1 . Shown are
potential temperature (solid every 10 K), wind vectors, snow (gray-dashed), graupel (black-
dashed), and rain (solid) mixing ratios every 0.08 g kg1 starting at 0.04 g kg1 (From Colle
2004, experiment U30. © American Meteorological Society. Reprinted with permission)
vertically above the windward slope, and the trajectories of growing precipitation
particles, which tilt substantially in the windward direction. In contrast, a deep,
vertically propagating gravity wave can enhance precipitation growth in several
ways compared to a simple terrain-following flow structure. First, it tilts the ascent
region upwind, bringing it into greater overlap with the sloped trajectories of falling
precipitation particles, yielding longer particle residence times in the growth zone
(Grabowski 1989; Colle 2004). Additionally, it slows the cross-mountain horizontal
wind (Banta 1990), also leading to longer particle residence times within the ascent
zone. Finally, it extends the updraft zone deep into the troposphere (Colle 2004),
allowing for a deeper region of particle growth (Fig. 7.8). However, it should
be noted that this is for an “ideally” configured gravity wave. The gravity wave
response can also inhibit precipitation growth if the vertical wavelength is short,
resulting in a rapid vertical transition from ascent to descent above the windward
slope, and reduction in the depth of the microphysical growth zone (Colle 2004;
Kunz and Kottmeier 2006).
7.3.5 Interactions Between Small-Scale Cellularity

and Microphysics
Houze and Medina (2005) recalled Smith’s (1979) wonderment at how oro-
graphic precipitation systems are seemingly so efficient at converting condensate to
precipitation, considering the typically short time period available for microphysical
conversion and particle fallout to operate. Houze and Medina hypothesized, based
on Doppler radar and other observations of orographic storms in both the Alps and
Cascade Mountains, that microphysical conversion of condensate to precipitation-
sized particles was aided by small-scale cells of enhanced upward motion embedded
within the larger barrier-scale upslope motion. They specifically attributed the
cellularity they observed to turbulence produced within a layer of enhanced vertical
wind shear, but more generally, small-scale cellularity can result from a variety
of processes, including release of embedded convective instability (Hobbs 1975;
Fuhrer and Schär 2005; Kirshbaum and Smith 2008) and flow over small-scale
ridges and valleys (Garvert et al. 2007; Colle 2008). Figure 7.9 depicts each of
these three mechanisms through examples from the literature. The common element
in each case is a cellular ensemble of updrafts and downdrafts embedded within
the barrier-scale ascending airstream. The updrafts rapidly raise the humidity up to
water saturation, even in the presence of ice particles, and produce additional cloud
water that contributes to collectional growth of either raindrops or ice particles.
Additionally, the enhanced humidity could also increase the number of ice nuclei
available for ice initiation if ice is not already present, and the additional cloud
water could create conditions for secondary ice production through the rime-
splintering process discussed previously. There is some degree of compensatory
depletion of precipitation within the downdraft branches of the cellular circulation,
but Fuhrer and Schär (2005) showed that the net effect is positive in the case
of embedded convection, indicating a nonlinear relationship between vertical air
motion and precipitation growth/depletion. Similarly, Garvert et al. (2007) and
Colle (2008) found a net positive effect in the case of terrain-induced cellularity.
The shear-induced cellularity mechanism has not yet been tested with models, so a
quantification of the enhancement of precipitation is not known.
It should be noted that specific circumstances can reduce the precipitation
enhancement effect of cellularity. Kirshbaum and Smith (2008) found no net
enhancement of precipitation resulting from convective cellularity embedded in
orographic precipitation, primarily due to detrainment of hydrometeors into dry air
above the cloud top in their simulations. This effect would likely be reduced if the
convective cells are fully embedded within the orographic cloud. Colle et al. (2008)
found enhancement by terrain-induced cellularity in a Wasatch Mountain case, but
less so than in the analogous Cascade study (Colle 2008), due to a lower melting
level, and more stable conditions producing a weaker vertical velocity response to
the flow over the small-scale ridges.
7.4 The Seeder-Feeder Process
Thus far, this chapter has focused primarily on the direct production of condensate
and precipitation from the vertical motions induced by flow over topography.
However, in many (or even most) cases of orographic precipitation, there is a
Fig. 7.9 Examples of three hypothesized mechanisms that lead to small-scale cellularity within
orographic precipitation systems. (a) Shear-induced turbulence (From Houze and Medina 2005.
© American Meteorological Society. Reprinted with permission). Shown is a time-height plot
of Doppler vertical velocities from a vertically pointing radar during the 13–14 December 2001
storm observed during IMPROVE-2, showing cellular ascending motion above the melting level
(near 1.7 km height), with a typical time scale of 5 min. (b) Embedded convective cells
(From Kirshbaum and Smith 2008. © Royal Meteorological Society. Reprinted with permission).
Idealized simulations of moist unstable flow (left-to-right) over topography produces convective
cellularity embedded within the larger orographic cloud, as depicted by vertical “towers” of
enhanced mixing ratios of cloud water (gray shading), cloud ice (thin dotted black lines), snow
(thin dashed black lines), rain (thick solid black lines), and graupel (thick solid charcoal-grey
lines). (c) Cellularity induced by flow over small-scale terrain fluctuations (From Garvert et al.’s
2007 observations and simulations of the 13–14 December 2001 orographic storm over the Cascade
Mountains in Oregon. © American Meteorological Society. Reprinted with permission). The south-
to-north cross section is along the barrier over the windward slope, and shows the small-scale
terrain structure associated with ridges and valleys perpendicular to the main barrier. Flow is from
the southwest, so that it has a component both across the fine-scale terrain (left to right) and up
the main barrier (out of the page). Shown are airborne radar reflectivity (color fill) overlaid with
simulated vertical motion (upward solid, downward dashed contours)
contribution to the condensate production both from the orographically induced

motions, and from a larger-scale atmospheric circulation unrelated to the orography,
either in the form of a mobile midlatitude baroclinic weather system with attendant
frontal circulations, or deep convective precipitating clouds. In the case of baroclinic
weather systems, the simultaneous presence of large-scale and orographic upward
motion is usually not a coincidence, as midlatitude storms provide dynamics for
large-scale ascent, enhanced lower tropospheric horizontal flow necessary to create
orographic lift, and enhanced moisture in the lower troposphere. Of particular
interest is the case in which the barrier in question is small, with orographic
lifting confined to the lower troposphere over a small horizontal area. Under such
circumstances, the advective time scale is much shorter than the conversion or
fallout timescales and, in principle the orographic cloud is incapable of producing
precipitation by itself. However, when “seeding” precipitation particles generated
by the large-scale storm fall through the orographic “feeder” cloud, the orographic
cloud condensate is efficiently collected and removed, resulting in more precipita-
tion than the sum of what might have been produced separately by the large-scale
storm and the orographic cloud. This “seeder-feeder” (SF) scenario was postulated
by Bergeron (1965), and was first studied in detail in the region of the Welsh
Hills (500 m high and 30 km wide) during strong orographically enhanced
precipitation events (Browning et al. 1974, 1975; Hill et al. 1981). These studies
generally found that the heavy orographic rain over the Welsh Hills could be
sufficiently accounted for by Bergeron’s SF hypothesis. Companion studies using
simple SF models by Storebo (1976), Bader and Roach (1977), Gocho (1978), and
Richard et al. (1987) quantitatively confirmed the sufficiency of the SF process to
account for observed rainfall rates. A common theme in these studies was the finding
that the precipitation rate at the ground was not very sensitive to the rate of seeder
input, as long as some seeding was present. The primary control on the precipitation
rate was the strength of the low-level flow, and by inference, the rate of production
of liquid water in the low-level orographic feeder cloud.
Subsequent studies added important refinements to the warm-rain SF hypothesis.
Modeling studies by Carruthers and Choularton (1983), Richard et al. (1987),
and Robichaud and Austin (1988) noted that when wind drift of raindrops was
included in the model, SF enhancement was reduced. Other studies found that
mountain-induced gravity waves significantly altered the vertical velocity, and
hence condensate production, in both the seeder and feeder clouds (Bradley 1985;
Richard et al. 1987; Grabowski 1989). Colle and Yuter (2007) also found that when
low hills are combined with a coastal contrast in surface friction (Long Island, in
their case), the frictional convergence and ascent can be of equal importance to the
orographic lift in supplying the feeder cloud with condensate.
The SF process has also been implicated as an important factor in deeper,
mixed-phase orographic storm systems over larger mountain ranges, as depicted
schematically in Fig. 7.7c. In this case, the seeding particles take the form of
snow particles generated by large-scale upward motion in the middle to upper
troposphere, which then fall through the orographic cloud and experience enhanced
growth, leading to enhanced precipitation. The classic study by Sassen et al. (1990)
over the Tushar Mountains in Utah showed that a steadily generated barrier-level
cloud of supercooled liquid water was inefficient at producing precipitation on its
own, but was periodically depleted by passing mesoscale precipitation bands, whose
ice particles fell through the liquid cloud, growing by riming, and also producing
secondary ice through rime-splintering. A similar situation was reported over the
Cascade Mountains in Oregon by Woods et al. (2005), in which strong prefrontal
cross-barrier flow produced an orographic cloud rich in supercooled liquid water,
which was largely scavenged by falling ice particles when a precipitation band
associated with an upper-level cold front passed over the barrier. And in a unique
study of a storm over a multi-ridge barrier in Arizona, Bruintjes et al. (1994) and
Reinking et al. (2000) found that gravity wave updrafts that were generated above
and ahead of one ridge produced ice particles that enhanced the seeding of low-level
orographic liquid cloud on the windward slope of the next ridge downstream.
Unlike in warm-rain situations, mixed-phase orographic storms provide multiple
pathways for the water mass within the orographic feeder cloud to be transferred
to precipitation particles. In a modeling study of the mixed-phase SF process,
Choularton and Perry (1986) found that for long hills, the “feeder” process is
actually in the form of increased depositional growth of ice particles, rather than
through production of supercooled liquid water and riming. This is because the
orographic updraft is weak enough such that the ice particles can take up the
available excess water vapor (above ice saturation) within the upslope flow before
water saturation is ever reached. However, for shorter, steeper hills, they found
that the updraft is stronger, water saturation is achieved, cloud water forms, and
particle growth is primarily via riming (as depicted in Fig. 7.7c). It may well be
that in some cases, both processes occur simultaneously: precipitation ice particles
grow by deposition due to supersaturation with respect to ice, and by collection
of supercooled cloud water that may also be present. However, we are not aware of
observations of precipitation particles whose structure suggests simultaneous riming
and depositional growth.
There is a satisfying conceptual simplicity to the SF process, particularly in the
assumption that a clean separation exists between the large-scale, upper-level seeder
cloud that is independent of the orography, and the low-level purely orographic
liquid water cloud that is incapable of, or at least highly inefficient at, producing
precipitation on its own. This conceptual simplicity has likely contributed to the
many studies that use the SF idea to describe orographic precipitation growth they
observe. And yet, others have pointed out that the separation between the seeder
and feeder clouds is not at all clear, and is often nonexistent (Roe 2005; Minder
et al. 2008). Several studies have described cases in which the orographically
induced upward motion influences the seeding cloud (e.g., the gravity wave-
related studies mentioned above), or even results in “self-seeding”, by a number
of different mechanisms. In Colle’s (2004) 2D simulations with full dynamics and
bulk microphysics, all the various experiments produced significant precipitation, in
the absence of any externally imposed seeding cloud. In particular, simulations with
deep, upstream tilted gravity waves produced significant snow mass through a deep
layer and well upstream of the barrier (Fig. 7.8), effectively seeding the lower-level
liquid water cloud near the barrier. Self-seeding is also facilitated by a relatively cold
environment, such that even if the orographic cloud is relatively shallow, cloud tops
can be cold enough to have sufficient ice nuclei present for ice particle initiation,
such as in the steady, purely orographic storm in the Sierra Nevada described by
Rauber (1992). Additionally, the cellularity mechanisms previously discussed can
boost both primary and secondary ice production within the orographic cloud.
Browning et al. (1974) found that even over the relatively small Welsh Hills,
the presence of mid-level convective instability provided a mechanism for “self-
seeding”, whereby even small orographic ascent perturbations aloft and upwind of
the hills were sufficient to release the instability, with the resultant convective cells
acting rapidly and vigorously to seed the lower orographic cloud with raindrops.
In summary, although many cases of orographic precipitation are well described
by the seeder-feeder model, one should be careful not to force every case into this
paradigm.
7.5 The Evolution of Microphysical Processes

Within Synoptic-Orographic Storm Life Cycles
While it is instructive to consider the microphysical mechanisms of orographic

precipitation under approximately steady flow conditions, actual midlatitude winter
orographic storms evolve through a life cycle that is closely connected to the
life cycle of the synoptic scale weather disturbance in which they are embedded.
A number of observational studies have documented the microphysical evolution
associated with the passage of synoptic-scale storm systems and associated fronts
and air masses, particularly in the western United States, where several field studies
have focused on storm system life cycles over mountain ranges from the Pacific
Coast to the Continental Divide, including the Cascades (Hobbs 1975), Sierra
Nevada (Heggli et al. 1983; Heggli and Reynolds 1985; Heggli and Rauber 1988;
and Reynolds and Kuciauskas 1988), Rocky Mountains (Cooper and Saunders
1980; Rauber et al. 1986; Rauber and Grant 1986), Wasatch (Rauber and Grant
1987; Steenburgh 2003), and Tushars (Long et al. 1990; Sassen et al. 1990). A
similar set of studies has also described the synoptic and microphysical life cycle of
orographic storms in Japan (Kusunoki et al. 2004, 2005). Many of these ranges are
approximately north–south oriented orographic barriers, interacting with cyclonic
storm systems approaching from the west or southwest. Although the typical
conditions vary greatly from case to case and from range to range, a basic life cycle
emerges in terms of characteristic structural and microphysical phases associated
with the approach, passage, and aftermath of a deep synoptic-scale cold front
or occluded front. Several studies have developed different naming conventions
for these phases (Hobbs 1975; Marwitz 1980; Long et al. 1990; Kusunoki et al.
2004; Medina et al. 2007), but they are essentially variations on the same theme.
Some cases are characterized by a single major frontal passage (e.g., Hobbs 1975),
whereas in other cases, there may be a distinct upper cold frontal passage, followed
by a transition zone, and then a separate lower-tropospheric cold or occluded frontal
passage (e.g. Heggli and Reynolds 1985; Long et al. 1990; Steenburgh 2003).
The typical evolution begins in the prefrontal environment, as high clouds
associated with the approaching frontal system move into the area. Cloud bases
gradually lower as the main frontal precipitation band approaches. Initial precip-
itation is often comprised of crystals of very cold (high-altitude) origin such as
bullets and assemblages of plates, some of which may reach the ground. As the
frontal precipitation band approaches the barrier, cross-barrier winds aloft increase,
and low-level flow is often blocked due to moderately stable air. As the frontal
precipitation band crosses the barrier, ice particles are generated over a deep layer
and wide range of temperatures, with precipitation crystal types ranging from colder
types (growing at temperatures colder than 20ı C) such as assemblages of plates
and sectors, bullets, and sideplanes, to dendrites (growing at water saturation at
temperatures from 13ı C to 17ı C). Liquid water in the upper-level cloud is
typically not prevalent, and depositional growth is dominant. Low-level orographic
enhancement of precipitation depends on the strength, direction and moisture
content of low-level cross barrier flow, which can vary considerably. For example,
frontal precipitation bands moving west-to-east over the Washington Cascades
(Hobbs 1975) are often accompanied by low-level easterly (reverse) flow, resulting
in cloud water production and riming east of the crest, and drying conditions that
inhibit riming west of the crest. In other cases, flow at all levels is strongly across
the barrier (west-to-east) during frontal passage, producing a substantial feeder
cloud that contributes additional riming growth to the frontal precipitation over the
windward slope and at the crest (e.g., Woods et al. 2005).
As the front moves past the windward slope, the transition to postfrontal
conditions begins. Several important changes in the airflow and microphysics occur
over the windward slope, and these have been observed in most of the studies listed
above, with minor variations. The wind becomes cross-barrier (west-to-east) at all
levels, and is usually characterized by neutral or even slight instability, leading to
unblocked upslope flow and a robust orographic liquid water cloud. However, cloud
tops descend considerably compared to the prefrontal or frontal environment, due
to large-scale subsidence aloft behind the cold front. This renders the orographic
cloud relatively shallow, extending <2 km above crest level. Ice particles form
less readily than in the much deeper frontal cloud band due to warmer cloud
tops, but their formation may be assisted by embedded convective cells and terrain
induced cellularity as discussed previously, or by secondary ice production. Particle
types range from dendritic types, to warmer types such as columns (T > 10ı C)
and needles (T > 10ı C and water saturation). In mountain ranges closer to a
maritime environment, CCN concentrations are lower and moisture is plentiful, so
that the postfrontal orographic cloud is ideal for riming growth both in terms of the
quantity of cloud water and the larger size of the cloud droplets. In continental
ranges, droplet concentrations are higher and moisture content lower, so that
riming growth is not as effective in the postfrontal cloud, unless it is embedded
with relatively vigorous convection. In general, the postfrontal period exhibits the
greatest orographic enhancement of precipitation (where “enhancement” is defined

as the ratio of precipitation over the barrier compared to that which falls over the
flat terrain upstream).
In the case where there is an upper cold front in an occluded type of structure,
the intervening zone between the upper cold front and the surface front shares
characteristics of the prefrontal and postfrontal environments. The low-level flow
and moisture content may be similar to that in the prefrontal environment, including
the possibility for partially blocked flow, but cold air has moved in aloft, often
resulting in unstable stratification aloft and embedded convection within the upslope
cloud. This is followed eventually by surface frontal passage and conditions as
described above in the postfrontal environment.
7.6 The Solid–Liquid Transition Region
When the melting level intersects the terrain in midlatitude orographic precipitation
systems, the mountains can experience rain and snow simultaneously, with snow at
upper elevations and rain below. The location of the rain-to-snow transition region
along mountainsides, or rain-snow boundary, has long been recognized as an impor-
tant forecasting challenge (e.g., Bergeron 1949; Lumb 1961; Ralph et al. 2005b).
In this section, some of the fundamental issues related to the orographic transition
region are briefly reviewed, including the factors that determine the location of the
transition over sloped terrain, and the types of precipitation encountered there. A
special case, that of freezing rain in the presence of multiple 0ı C levels (e.g., Bell
and Bosart 1988) is a complex phenomenon that is uncommon and also not unique
to orographic environments, and so will not be addressed here.
7.6.1 Types of Precipitation
Strictly speaking, the transition region is the area bounded by all rain on one side
and all snow on the other. This means that the region itself is composed of semi-
melted particles that may be intermixed with all liquid particles. In particular, the
types of precipitation that can occur include rain, wet snow (snow or graupel that is
partially melted), or slush (snow or graupel that has almost completely melted).
A key physical process within the transition region is the melting of snow
particles. The rate of melting of an individual particle, in mass per unit time, is
proportional to the particle diameter. However, since the total mass of the particle is
proportional to the third power of diameter, large particles take considerably longer
to melt than small ones, under the same atmospheric conditions. Observationally,
the transition zone is often identifiable as the intersection of the radar “brightband”
(Fabry and Zawadzki 1995) with the ground, with melting snow particles being the
specific cause of enhanced radar returns (see previous footnote number 2).
The melting of snow occurs through several stages (Knight 1979; Fujiyoshi
1986). Water droplets initially appear within the snowflake lattice structure, produc-
ing wet snow; this continues until the particle’s liquid mass fraction is roughly 70%
(Mitra et al. 1990), when the lattice collapses and the particle is somewhat similar to
a drop but with ice within. Theriault et al. (2006) refer to such a particle as “slush”.
Eventually all the ice melts, yielding a raindrop. This multi-step process controls
the types of precipitation that can occur. However, within sub-saturated conditions,
the sublimation of snow is also a critical issue. Sublimation acts to cool the particle
and thereby reduce the melting rate. The exact nature of the precipitation within the
transition region then depends critically on these melting and sublimation processes.
The impact of melting and sublimation is certainly evident in observations of the
type of precipitation at the surface. For example, Matsuo et al. (1981) showed that
wet snow could reach the surface even if the temperature is C4ı C to 5ı C, as long
as the relative humidity is at or below 60%. In contrast, wet snow only reaches the
surface up to C1.5ı C to 2ı C if the atmosphere is near saturation. Lundquist et al.
(2008) found that C1.5ı C was the temperature at which approximately 50% of the
time the precipitation on a mountainside was rain and 50% it was snow, but they
also pointed out that there is considerable variation in this threshold.
7.6.2 Precipitation Types Within Transition Regions

over Sloping Terrain
The nature of the precipitation in the transition region evolves in a systematic

manner as one moves from the colder to the warmer side: dry snow; wet snow;
wet snow and slush; wet snow, slush and rain; slush and rain; and finally to rain
alone. This evolution follows directly from the melting behavior of snowflakes. The
higher the temperature, the greater the fractional melting.
The development of mixtures of precipitation at a given location occurs because
a spectrum of snowflake sizes is occurring (Gunn and Marshall 1958). The
different sizes melt at different rates and give rise to different precipitation types
simultaneously. For example, wet snow, slush and rain can occur simultaneously.
The smallest snowflakes have melted completely (rain); somewhat larger particles
have melted a great deal and have collapsed (slush); but the largest particles have
melted only slightly (wet snow). The exact thresholds between these types and their
combinations depend on the detailed nature of the melting and the morphology
of the particles. The particles also interact as they fall. Any collision between
particles of different types acts to maintain semi-melted particles and to reduce the
occurrence of all-liquid particles.
Transition regions can furthermore be an ideal location for extremely large
particles. There have been numerous reports in the literature of such particles being
many centimeters in diameter (Stewart 1992; Lawson et al. 1998). At temperatures
near 0ı C, aggregation to large snowflakes is enhanced by the “sticky” character of
the snow, and melting isn’t occurring or only slightly. As in the discussion above,
the presence of such particles would be expected to widen the transition region since
temperatures would need to be higher for complete melting.
On mountainsides, this transition of precipitation types will mimic that found
over flat terrain. The upper elevations will generally be on the cold side, and the
lower elevations on the warm side. The lowest level of the mountainside at which
snow or slush accumulates is typically referred to as the “snow level”, although this
term is also occasionally used for characterizing the lowest height of the seasonal
snowpack on a mountainside.
7.6.3 Impact of Precipitation Phase Changes

and Other Factors on Transition Regions
7.6.3.1 General Conditions
Phase changes of particles falling through the transition region can have a significant
impact on the environmental conditions. Melting and sublimation both extract
energy from the environment and this leads to atmospheric cooling. In the case
of melting, under saturated conditions, there is a tendency to produce an isothermal
layer at 0ı C (e.g., Findeisen 1940). Melting occurs down to this temperature but no
lower. One outcome is that, if rain is initially occurring at the surface, the steady
erosion of above-freezing temperatures due to melt cooling will tend to convert this
into snow. A critical issue is assessing the likelihood that the melting process can
drive the snow line down to the surface.
Assuming instantaneous melting, Atlas et al. (1969) developed an equation
giving the amount of melted snow (expressed here as liquid equivalent depth, dmelt )
needed to reduce temperatures in a layer down to 0ı C due to melt cooling. The
equation does not account for the impact of latent heat release from condensation,
nor of warm advection:

a cp H 2
dmelt D (7.1)
w Li 2
Here a and w are the densities of air within the isothermal layer and of liquid
water, respectively; cp is the specific heat of air; Li is the latent heat of fusion for
water; is the initial lapse rate, and H is the depth of the isothermal layer.
This equation illustrates the impact of melting on the atmosphere. Assuming for
example that ” is 5.6ıC km1 at 765 hPa, 1 mm equivalent of melted snow will
produce a 300 m deep 0ı C layer and 10 mm will produce a 1 km deep 0ı C layer.
Thus, realistic amounts of melting can have a significant impact on the depth of the
transition zone and resultant precipitation type at the surface.
Such cooling can significantly affect the location of the transition region,
whereby an area that is initially experiencing rain will eventually be subject to snow,
as first noted by Wexler et al. (1954). Stewart and McFarquhar (1987) extended
these calculations to estimate the speed of motion of the boundary that would be
associated with this alteration. They found, with reasonable precipitation rates of a
few mm h1 , the transition region would move towards the warmer air at speeds of
order of a few m s1 for several hours. The speed is slowly reduced since the initial
layer that must be cooled becomes progressively deeper and warmer and more time
is needed for melt cooling to drive the 0ı C level down to the surface. They also
found that the width of the transition region should, in general, become narrower as
time progressed. The leading (warm) edge would progressively slow as it encounters
warmer temperatures, whereas the trailing (cold) side would essentially move at the
same speed since the temperature reduction needed to produce a 0ı C layer does not
change across the transition region.
Only limited research has been conducted on what determines the depth of
the melting layer. Observations suggest that this vertical depth can vary from
less than 100 m to several 100 m. Such a variation is not surprising since the
depth of the melting layer required to completely melt snowflakes depends on
many factors such as air temperature, relative humidity, precipitation rate and
snowflake characteristics. For example, Matsuo and Sasyo (1981a) studied the
depth of the melting layer needed to completely melt snowflakes, considering
different sizes and densities of the snowflakes. They assumed a constant lapse rate of
6.5ı C km1 below the melting level. Assuming saturated environmental conditions,
snowflakes of different sizes can fall 100–600 m before melting completely. This
distance also depends critically on the snowflake density. For instance, the depth
needed to completely melt a snowflake can vary by a factor of up to 3 simply by
using different, yet realistic, densities. When the environmental air is saturated,
snowflakes will start melting when they reach air temperature of 0ı C. However,
if snowflakes fall within a sub-saturated environment the melting rate is decreased.
Matsuo and Sasyo (1981b) show a near-linear dependence of melting snowflakes on
the relative humidity and air temperature. For example, this relation indicates that
no melting will occur at temperature above C4ı C if the relative humidity is 50%
or lower. For numerical weather prediction, empirical techniques are used to assess
if snow will melt completely when falling through a melting layer (e.g. Bourgouin
2000). The technique is based on the depth and maximum temperature of the warm
layer.
Sublimation and evaporation also have an impact on the atmospheric moisture
content. In a sub-saturated environment, sublimation of snow and evaporation
of rain increase the amount of water vapor in the atmosphere and decrease the
temperature, driving the atmosphere toward saturation. In a saturated environment,
melt cooling can produce additional condensation in the form of cloud droplets
(Findeisen 1940; Donaldson and Stewart 1993; Woods et al. 2008b). If this occurs
at the surface, fog is produced, exacerbating low visibility already reduced by
precipitation in the transition region. The cloud droplets could, in principle, be
scavenged by the very precipitation particles that produced them via melting,
though the precise microphysical pathways in this particular situation have not been
studied.
7.6.3.2 Transition Regions and Orography
In orographic environments with upslope flow, the situation is more complicated

than over flat terrain. Studies have shown that often, the rain-snow transition layer
dips down to a lower altitude near the terrain than in the free atmosphere upstream,
by amounts ranging from 0 to 1,000 m, with typical amounts of a few hundred
meters (e.g., Lumb 1983a, b; Marwitz 1983; Medina et al. 2005; Houze and Medina
2005; Marigo et al. 2008). This was depicted schematically in Fig. 7.7, with the
0ı C isotherm bending downward as it intersects the terrain. There are at least four
potential physical reasons for this phenomenon:
1. Adiabatic cooling. As air rises upslope, it cools moist adiabatically above its
lifted condensation level, and if the lapse rate in the free atmosphere is moist
neutral, rising air near the terrain will be colder than air at the same height in the
free atmosphere upstream.
2. Larger particles. If the precipitation is more intense near the barrier due to
orographic lift, snow particles are likely to be larger, requiring more time to melt,
and a deeper transition layer near the barrier than upstream (Stewart et al. 1984).
There may be a positive feedback with this factor, in that the deepening melting
layer increases the depth over which partially melted particles can aggregate,
yielding larger particles that take longer to melt.
3. Stronger melt cooling. If the precipitation is more intense near the barrier due to
orographic lift, the melting rate, and hence melt cooling, are also more intense,
driving the melting layer down farther near the barrier than upstream (Atlas et al.
1969).
4. Differential depth of the mixing layer. Melt cooling produces a superadiabatic
lapse rate at the bottom of the isothermal layer, and therefore, the downward
penetration of the melting layer due to melt cooling is countered by vertical
mixing in the warm layer underneath (Findeisen 1940; Moore and Stewart 1985).
This effect is not accounted for in Atlas’ equation (7.1) above. The ability of
mixing to counteract the descending melting layer is weaker for a more stable
warm layer and for a shallower warm layer. Therefore, for the same initial height
of the melting level above sea level, and the same rate of melt cooling, the melting
layer over the higher terrain should be forced downward faster than over the
lower terrain upstream because the mixing layer is shallower.
Numbers 2, 3, and 4 above can be counteracted by horizontal winds, which
transport the diabatically modified air away from the orographic precipitation zone,
so calm or light-wind conditions near the surface enhance the lowering of the
melting layer.
Several of the above effects were examined in a systematic manner by Theriault
et al. (2008). They used the one dimensional kinematic cloud model of Theriault and
Stewart (2007) and Theriault et al. (2006), coupled with a microphysical scheme
that incorporates the various frozen or partially frozen precipitation types described
above. The initial temperature profile is comprised of a melting level 1 km above
the ground, with a surface temperature of 5ı C.
a b
Fig. 7.10 The evolution at the ground surface of (a) precipitation types and (b) temperature for
a 1-D simulation of stratiform precipitation using a microphysical scheme that predicts rain (R),
slush (SL), wet snow (WS), and dry snow (S) particle types. The transition region is located between
the grey dashed lines (Adapted from Theriault et al. 2008)
The time evolution of the surface precipitation types and temperature are shown
in Fig. 7.10. In this particular simulation it was assumed that snow falls from above
the melting level at a rate of 2 mm h1 within a saturated environment with respect to
water and with a vertical air velocity of 10 cm s1 . Initially, snow completely melted
as it passed through the 0ı C level, and significant rain first reached the surface at
25 min. From this point on, the various processes discussed above began to drive the
snow level downward. The first appearance of slush particles occurred at the surface
at 130 min. Rain decreased and slush increased until 180 min, when wet snow began
to appear. From 180 to 230 min, slush decreased and wet snow increased, until
finally at 230 min, wet snow suddenly transitioned to dry snow.
Several sensitivity experiments were run in which precipitation rate, vertical
velocity, and initial relative humidity were varied, and results of these are shown
in Fig. 7.11. While melt cooling was important in driving down the snow level,
adiabatic cooling contributed as well, as evidenced by the slower arrival of the snow
level at the ground with slower upward motion. Increased precipitation rate hastened
the lowering of the snow level through more melt cooling and a longer time required
for complete melting of the larger particles. Relative humidity also impacted the
lowering of the snow level. A sub-saturated environment reduced the condensation
by melt cooling and adiabatic ascent, and caused sublimation (or evaporation) of
falling precipitation. Both effects led to cooling, accelerating the lowering of the
snow level.
Another recent study by Minder and Durran (2010) used model simulations
of flow over an idealized barrier to specifically test the importance of the first
three factors listed above in lowering the melting layer near sloped terrain. They
found that, depending on various conditions, the melting layer lowered by up to
a
Vertical Air Velocity (cm/s)
20
15
10
5
0 60 120 180 240 300 360 420 480
Time (min)
b
Precipitation Rate (%)
20
15
10
5
0 30 60 90 120 150 180 210 240
Time (min)
c
Relative Humidity (%)
100
90
80
0 30 60 90 120 150 180 210 240
Time (min)
R SL WS S
Fig. 7.11 Duration of precipitation types in simulations like that shown in Fig. 7.10, but with
varied environmental conditions: (a) varying the value of the upslope wind, (b) varying the initial
precipitation rate and (c) varying the degree of saturation of the atmosphere. The control run
assumes an initial snowfall rate of 2 mm h1 , a vertical air velocity of 10 cm s1 and saturated
atmospheric conditions. The symbols are rain (R), slush (SL), wet snow (WS) and snow (S)
(Adapted from Theriault et al. 2008)
a few hundred meters against the terrain compared to far upwind. However, they
used a strong, frictionless cross-barrier flow which tended to immediately transport
diabatically cooled air up and over the barrier, limiting the time available for the
diabatic cooling to act. It is likely that a more stagnant air mass situated directly
against the upslope side of the barrier would allow diabatic cooling to be more
effective at lowering the melting layer near the barrier, and perhaps match some of
the more extreme observed cases where the lowering was as great as 1,000 m. This
stagnation could be accomplished through several mechanisms omitted in Minder
and Durran’s study, including cross-barrier shear, stronger windward blocking,
surface friction, and small-scale terrain variability.
7.6.4 Summary
The transition region is a critical aspect of many orographic storms. Precipitation

types evolve with time and across the transition region from dry snow to rain,
with wet snow and slush particles occurring in varying proportions in between.
What might seem at first to be a straightforward issue of assessing the height at
which snow, mixed precipitation, and rain occur, actually depends on a number of
thermodynamic, microphysical and dynamic factors that differ in important ways
over orography compared to the flat terrain upstream, often leading to a significantly
lower transition region in the mountains.
This section has addressed some of the fundamental issues associated with
the transition region along mountainsides. In actual events, there are many other
factors to consider, including the time evolving large-scale and regional forcings,
variations in the types of particles falling into the transition zone (habit and degree
of riming), and the generation of dynamic flows through interactions with orography
and precipitation (see for example, Marwitz 1987a, b; Marwitz and Toth 1993;
Heffernan and Marwitz 1996; Cox et al. 2005).
7.7 Snow Accumulation: Depth, Density, and Particle Types
A quantitative precipitation forecast (QPF) predicts the amount of precipitation,

expressed as a liquid-equivalent depth, that is expected to fall in a given period.
However, when that precipitation is expected to fall as snow, the forecast of actual
depth of new snow on the ground is also desired, since that is what is actually
experienced by drivers, road and airport maintenance crews, and participants in
winter recreational activities. To obtain a snow depth forecast, one must apply an
assumed “snow ratio” (or sometimes “snow to liquid ratio”) to the QPF to arrive at
a snow depth forecast. The snow ratio, R, is defined as the new snow depth, Dsnow ,
divided by the liquid equivalent depth of the same snow, Dliquid , and is inversely
proportional to the snow density: R D Dsnow /Dliquid D l /s , where l is the density
of liquid water, 1,000 kg m3 . A general forecasting rule of thumb is that s is
100 kg m3 (or R D 10, or 1 mm of liquid equivalent precipitation produces 1 cm
of snow), but studies have shown that s can range from 10 to 320 kg m3 (R can
range from 3.1 to 100) (Judson and Doesken 2000).
Snow density depends on a variety of factors both aloft and at the surface
(Roebber et al. 2003). The factors are summarized below, separated (as in Roebber
et al.) into in-cloud, sub-cloud, and ground-level factors:
In cloud:
• Snow particle type (habit and degree of riming). Different snow particle habit
types, which exhibit radically different physical structures, have varying amounts
of empty space within individual particles, and between particles within accumu-
lated snow, leading to different accumulation densities. Riming also tends to fill
in interstitial spaces without increasing the overall size of the particle, yielding
more dense accumulated snow. Several studies (discussed below) have confirmed
these effects.
Sub-cloud:
• Sublimation. As particles fall through sub-saturated layers, sublimation tends
to act most rapidly to deplete intricate protruding structures of snow crystals,
allowing them to arrange more compactly on the ground.
• Melting. Similarly, as particles fall through the 0ı C layer, melting also tends to
act first on the thinnest, protruding structures of the crystals, yielding particles
that can arrange more compactly on the ground.
• Fragmentation. Collisions in the air can cause particles to break into fragments
that are generally less complex than their parent crystals, allowing them to
arrange more compactly on the ground.
Ground-level:
• Wind fragmentation and compaction. Wind-driven collisions between particles
on the ground can significantly densify the snow, both by fragmentation, and by
forcing the colliding particles to arrange more compactly.
• Melting. As with melting aloft, further melting on the ground due to either warm
air or warm underlying ground leads to denser snow.
• Compaction by weight of overlying snow. Snow is known to gradually compact
under the weight of overlying snow via plastic deformation of the crystal
structure. This effect is relatively large when the snow temperature is close to
0ı C, but compaction rate decreases by a factor of 2 for each 6ı C below freezing,
according to Conway and Wilbour (1999)’s snow stability model. Compaction
studies such as Casson (2009) have demonstrated this process explicitly with
experimental apparatus and natural snow samples, whereas other studies such as
Roebber et al. (2003) and Ware et al. (2006) have demonstrated the effect by
inference, showing a statistically significant relationship between accumulated
snow density and the liquid equivalent depth of the 12-h accumulation (a measure
of its weight). One complication with this process is that it is a function both of
time and of overlying snow depth, which is also a function of time, so its effect
on density is dependent on the accumulation time interval. If accumulations were
measured every 1 h, for example, it is unlikely that compaction (or precipitation
rate) would have a noticeable effect on snow density.
• Ageing. Although much slower than the other processes discussed above, and
usually negligible within the first day or two after accumulation, vapor diffusion
within the accumulated snow can move water mass both locally from convex to
concave corners, and vertically through the snowpack in the presence of vertical
temperature gradients. These processes can lead to either increases or decreases
(i.e., formation of depth hoar) of density in the snowpack.
Roebber et al. (2003) also include the factors of particle size and aggregation,
and Dubé (2003) adds the factor of refreezing of partially melted snow. However,
the importance of these factors to snow density on the ground is not clear. The
effect of particle size on snow density has not been demonstrated, independent of
particle habit type. With regard to aggregation, all particles ultimately “aggregate”
rather violently upon colliding with the snow surface, so it is unclear that prior
aggregation aloft makes a difference in how densely the crystals arrange themselves
on the ground (Ware et al. 2006). Similarly, it is unclear if refrozen partially melted
snow yields a significantly different accumulation density than the same partially
melted snow would have yielded before being refrozen.
Various studies have attempted to relate the density of new snow to surface
information and station location alone. Judson and Doesken (2000), Roebber et al.
(2003), Wetzel and Coauthors (2004), and Byun et al. (2008) found that surface
temperature alone explained 0–52% of the variance in snow ratio. For a specific
location in the Wasatch Mountains, Alcott and Steenburgh (2010) were able to
increase the explained variance in snow ratio to as high as 68%, by using several
different predictor variables at different heights from a reanalysis data set, with the
key predictors being wind and temperature near crest level. But in general, surface
temperature offers limited predictive value, because snow density strongly depends
on particle habit type and degree of riming (Judson and Doesken 2000), and there is
limited correspondence between surface conditions and the conditions aloft where
the particle’s habit and degree of riming are determined. The important conditions
aloft include the temperature and humidity where the particles grew by deposition
(for habit type) and supercooled cloud water concentrations encountered during par-
ticle descent (for degree of riming). Power et al. (1964) took measurements during
portions of two winter seasons in Montreal, and verified that the density of new snow
is related to snow crystal forms, and is higher for rimed snow particles, though they
only considered the somewhat coarse categorization of “rimed” vs. “unrimed”.
Casson (2009) analyzed measurements of snow particle habit type, degree of
riming, and density of accumulated snow during two winter seasons of observations
in the Washington Cascade Mountains, and found a clear relationship between
particle type and density, as shown in Fig. 7.12a (note the y-axis is snow ratio instead
of density). As one moves from left to right, particle types shown vary from unrimed
or lightly rimed dendrites, to heavily rimed dendrites, to light or moderately rimed
cold-type crystals (sideplanes, assemblages of plates, etc.), to graupel-like snow,
and finally graupel. Although there is considerable scatter within each category, a
large portion of the variance in snow ratios is clearly tied to the habit type and
degree of riming. Also shown (Fig. 7.12b) is the relationship between snow density
and degree of riming, for both dendritic and non-dendritic particle types. Snowfalls
of both habit-type categories display a strong relationship between density and
degree of riming. These relationships provide motivation for the development of
Fig. 7.12 Accumulated snow ratio and density as a function of particle characteristics, determined
from ground observations in the Cascade Mountains during the winters of 2006–2007 and 2007–
2008. (a) Snow ratio as a function particle habits and degree of riming, at Snoqualmie Pass (large
gray circles) and Stevens Pass (black dots). (b) Snow density as a function of estimated percent of
mass of particles comprised of rime, for all dendritic and nondendritic particle types, with linear
fits shown. Whiskers show ˙1 standard deviation of the measurements for that rime mass fraction
(From Casson 2009. © Gerald Casson. Reprinted with permission)
microphysical schemes in NWP models that can predict particle habit type and
degree of riming as discussed in the next section, as such schemes would enhance
the capability to predict snow density directly from model output.
Because many of the factors that affect snow density have different climatologies
in different geographic regions, it is not surprising that snow density climatology
itself varies geographically. Even over flat terrain, variations are noticeable (Baxter
et al. 2005; Ware et al. 2006). Over complex terrain, variations can be dramatic.
For example, on the slopes of Mt. Olympus in Washington state, where moisture-
laden air from the Pacific Ocean is lifted by the steep topography, and low CCN
concentrations yield large, rimable droplets, the average density of new snow has
been estimated at 250 kg m3 (snow ratio D 4.0) (LaChapelle 1958). In contrast, at
Steamboat Springs, Colorado, where temperatures are generally colder, moisture is
less plentiful, and snow often falls in the form of unrimed or lightly rimed dendrites,
snow density averages around 72 kg m3 (snow ratio D 13.9), a difference of a factor
of 3.5 between the two locations.
Recent studies have devised several techniques for estimating snow ratio from
model forecast parameters. Dubé (2003) developed an algorithm that allows the
forecaster to assess the status of each of the factors affecting density listed above
based on forecast model output, and then assigns a snow ratio based on those
assessments. Roebber et al. (2007) trained a neural-network to predict broad snow
ratio categories, using observed snow ratios at 53 sites in the eastern U.S. and
Eta/NAM model forecasted lower-tropospheric parameters; similarly, Alcott and
Steenburgh (2010) trained a step-wise multiple linear regression procedure using
observed snow ratios at Alta Ski Resort in Utah and Eta/NAM forecasts. In
both cases, the forecasting methods improved snow depth forecasts considerably
compared to use of either a constant snow ratio or the standard National Weather
Service table that uses a simple surface temperature-based variation of snow
ratio. Many NWS offices now use a scheme recently developed by Cobb and
Waldstreicher (2005), which predicts snow ratio primarily from a determination
of overlap between the maximum precipitation growth zone and the temperature
zone for dendritic snow growth, based on forecast model-predicted temperatures
and upward motion. While this approach represents an improvement over the simple
surface temperature-based table, the scheme is known to perform less skillfully over
complex terrain, where terrain-influenced upward motion is not well represented by
the operational forecast models (Alcott and Steenburgh 2010).
7.8 Numerical Weather Prediction, Microphysical Schemes,

and Orographic Precipitation
The physical principles of cloud and precipitation microphysics, as described

in Sects. 7.2–7.6, enter into the forecast process in two important ways. First,
forecasters integrate their understanding of physical processes into the pattern
recognition techniques, conceptual models and heuristics that they apply on a
regular basis to assess what is currently happening and how the weather will evolve.
Second, these principles are incorporated into numerical weather prediction (NWP)
models, which serve as an indispensible guidance tool for the forecaster. Because the
models are imperfect representations of the actual full physics of the atmosphere, it
is important for forecasters to understand how models represent physical processes
and what the limitations and known deficiencies are, so that they can utilize the
model guidance from an informed position. Chapters 9, 10, and 11 of this book
reviews the dynamical and boundary layer aspects of representing orographic flows
in NWP models. This section specifically describes the representation of clouds and
precipitation within orographic storms.
It is important at the outset of this section, however, that we not raise expectations
too high for a grand enlightenment into biases and behaviors in model precipitation
physics that forecasters of orographic weather can watch for. As oversimplified as
model precipitation physics is, the pathways to precipitation production are incred-
ibly complex, and are interactive with (and sensitive to) each other and many other
processes in the model, such that it becomes very difficult to identify, understand,
and fix inaccurate behaviors in the model schemes. Those problems that have been
easily identifiable and explainable have, for the most part, been fixed already, and
remaining deficiencies may be largely unavoidable due to limitations of the bulk
approach to representing hydrometeor species, as well the representation of physical
processes that are fundamentally difficult to measure in the real atmosphere and
are highly complex. Having developed and used microphysical schemes extensively
in their own research and forecasting, the authors are continually surprised by the
complexity of interactions in the model, and the repeated experience of making
relatively simple changes in the model with expectations of a certain outcome,
only to have something completely different occur. Nevertheless, it is easier for
forecasters to develop a comfort level with the models as forecasting tools if they
are armed with a working knowledge of the principles employed in model physics
parameterizations, as opposed to perceiving the models as mysterious black boxes.
Since deep convection is uncommon during midlatitude winter orographic storms
as defined in Sect. 7.1, most of the precipitation generated by the models is in
response to grid-scale upward motion, and is therefore managed by grid-scale
microphysical schemes, rather than by subgrid-scale cumulus parameterization
schemes. Microphysical schemes keep track of a set of equations that predict the
quantity (both mass and number) of hydrometeor particles of different species at
every grid point. They include terms for initial formation of particles, growth by
condensation and deposition, loss by evaporation and sublimation, transfer of mass
among species due to particle collisions, and transport by both three-dimensional
advection and sedimentation. While the predicted variables and mass transfer
processes vary considerably between different schemes, a schematic of a typical
microphysical scheme is depicted in Fig. 7.13. The circles represent the predicted
hydrometeor mixing ratios, and the numerous arrows represent the pathways by
which water mass can be transferred from one hydrometeor species to another, via
the processes listed above. The remainder of this section describes the formulation
and limitations of microphysical schemes used in NWP models, with specific
attention to orographic precipitation.
Fig. 7.13 Schematic representation of the Reisner et al. (1998) bulk microphysical scheme,
as depicted in Colle and Zeng (2004a) (© American Meteorological Society. Reprinted with
permission.) Circles represent mixing ratios of predicted hydrometeor species (“c.w.”, “w.v.”, and
“c.i.” refer to cloud water, water vapor, and cloud ice, respectively). Arrows represent simulated
processes that transfer mass between species at each model time step. Processes are labeled
according to the coding convention used in the scheme. See Colle (2004a) for a list of the processes.
Short downward arrows on the three precipitating species (rain, snow and graupel) represent loss
or gain from fallout into or out of the model grid box
7.8.1 Types of Microphysical Schemes in NWP Models
7.8.1.1 Explicit Schemes
Explicit or bin microphysics schemes represent a particle size distribution with a set
of “bins” for each hydrometeor species, where each bin covers a distinct range of
particle mass. Using simple relations for mass as a function of diameter, each mass
bin correlates directly to particle diameters. To cover the range of all diameters
of water drops and ice crystals, researchers typically double the mass of each bin
compared to the preceding bin. Within each bin, a simple or complex particle
number distribution function is assumed and the mass mixing ratio and number
concentration of particles are predicted by a separate prognostic equation in the
model.
In order to cover the range of sizes of water drops in the atmosphere, the bin
scheme must realistically represent drops from approximately 0.1 m to nearly
6 mm. Therefore, even with mass doubling, the number of size bins for most
models is between 30 and 50. And, if one considers ice phase hydrometeors
like pristine ice, snow, and graupel, then the number of variables rapidly reaches
hundreds. Therefore, these explicit schemes are typically used only for parcel
models, 1-D column models, or idealized 2-D simulations. Research simulations
with bin microphysics in full three-dimensional case studies have been carried out
since the early 2000s, but often take days on large supercomputers to complete just
a few hours of forecast time.
7.8.1.2 Bulk Schemes
Due to the high computational cost of explicit microphysics schemes, a different

class of schemes was developed that did not explicitly represent particles of
different sizes. Such schemes are called bulk microphysical parameterizations
(BMP) because they predict the bulk properties of each hydrometeor species,
namely the mass of the entire population of particles of that species within a grid
cell. The shape of the particle size distribution (PSD) is assumed to adhere to a
mathematical form with one or more shape-defining parameters that can be set
constant, diagnosed, or predicted. Common forms are monodisperse (all particles
are the same size), exponential, or generalized gamma distributions. The latter is
expressed as
NO .D/ D N0 D exp.
D/ (7.2)
where NO .D/ is the number concentration of particles per unit volume per unit
diameter range, as a function of diameter, D; N0 is the intercept parameter; œ is
the slope parameter; and is the shape parameter. The PSDs represented by (7.2)
are most clearly depicted on a semi-log plot of NO .D/ versus D (Fig. 7.14). A
monodisperse PSD appears as a delta function on such a graph (not shown). When
is zero in (7.2), the gamma distribution reduces to the more simple exponential,
which appears as a straight line with negative slope on the semi-log plot, with N0 and
œ determining the intercept and slope of the line, respectively. Many studies have
shown the approximate exponential character of both rain (Marshall and Palmer
1948) and snow (Gunn and Marshall 1958) PSDs. However, it is also known that
for some species and under some conditions, PSDs are not simple exponential, in
which case can be set to a nonzero value to represent such PSDs as depicted in
Fig. 7.14.
Single moment bulk schemes predict only the mass mixing ratio of the hy-
drometeor species (proportional to the “third moment” of the PSD). The number
concentration is then diagnosed from the assumed shape of the PSD and one
other constant or deduced parameter such as y-intercept, slope, or characteristic
Fig. 7.14 The generalized

gamma distribution, for 1010
D 5 mm1 and
Number Concentration (m-4)

N0 D 3.0106 m4 , and
for three different values =-2
108
of (From Mitchell 1994.
=0
Society. Reprinted with 106
permission)
=+2
104
102
0 0.5 1 1.5 2
Diameter (mm)
diameters. When a BMP also explicitly predicts the number concentration of the
hydrometeor species, it is typically referred to as a two-moment scheme, because
the number concentration (proportional to the “zeroth moment” of the PSD), and
the mass mixing ratio (proportional to the “third moment” of the PSD) are both
explicitly predicted. As in the case of the single moment schemes, an assumed
functional form of the PSD must be assumed. The primary advantage of two-
moment schemes is their additional degrees of freedom to shift the distribution
toward the less numerous larger particles or toward the more numerous smaller
particles. Also, by using the properly determined mass-weighted and number-
weighted terminal fall speeds, the water and ice precipitate more realistically,
resulting in proper size sorting with relatively more large hydrometeors at low
altitudes and more small hydrometeors suspended above.
Some studies have added a third predicted moment in an effort to further increase
the flexibility in predicting the shape of the PSD and the ability to match observed
PSD evolutions. An example is the three-moment bulk scheme of Milbrandt and Yau
(2005), which predicts radar reflectivity factor (proportional to the sixth moment of
the PSD) as a third predicted variable for each hydrometeor type. They showed that
this scheme achieved a closer match to the observed PSD than a scheme in which
one of the PSD parameters was kept constant. Such schemes are computationally
expensive and not practical yet for operational use.
7.8.2 Review of Recurrent Problems with Bulk Schemes

in Orographic Environments
Colle et al. (1999, 2000, 2003, 2005a) and Colle and Mass (2000) consistently
showed a positive bias in forecasted precipitation above the windward slope and
(c) 4 km
Fig. 7.15 Bias scores ( 100%) at precipitation gauge sites for 4-km MM5 forecasts over the
Pacific Northwest during the 1997–1998 and 1998–1999 cool seasons. The 90% and 160% lines
are contoured with dashed and solid lines, respectively. Terrain is shaded for reference (From Colle
crest of major barriers for orographic winter storms in the northwestern and
northeastern United States, based on analyses of either single cases or seasonal
behavior of model forecasts using various versions of the MM5 and WRF models.
An example of such a bias in the MM5 in the Cascade Mountains can be seen in
Fig. 7.15, which depicts model-predicted precipitation over the course of two winter
seasons, as a percentage of observed, at precipitation sites in western Washington
state. Colle and Mass (2000) suggested that deficiencies in model microphysical
parameterizations were a primary cause, and that the most sophisticated BMP did
not generate the best precipitation distribution. Specifically, Colle and Mass (2000)
examined sensitivities of simulations of a Cascade Mountain winter flooding event
to snow fall speed and size distribution assumptions in the BMP, and found that
the windward-leeward precipitation distribution was particularly sensitive to both.
When compared with aircraft observations of ice particles aloft in a Cascade storm
studied during the IMPROVE project, Garvert et al. (2005) and Colle et al. (2005a)
found that model-predicted snow mixing ratios above the windward slope were
excessive. A similar study over the Wasatch Mountains in Utah (Colle et al. 2005b)
also found a windward precipitation bias, but only over the southern, wider part of
the barrier, not over the northern, narrower part, suggesting that wider barriers may
be more sensitive to errors in microphysical details aloft, consistent with the larger
snow sensitivities noted for wider barriers discussed in Sect. 7.3.2.
Many of the studies cited in the previous paragraph used progressively newer ver-
sions of microphysical schemes available in the MM5 and WRF mesoscale weather
prediction models. Recent experience with the Thompson et al. (2008) scheme in
the WRF model, which has evolved considerably from its roots in the Reisner et al.
(1998) scheme in the MM5 model, suggests that the windward precipitation bias
has been considerably reduced (B. Colle, personal communication). Key differences
which have contributed to the improvement include new representations of the
snow size distribution, more realistic particle mass and fall speed relationships, and
look-up tables that allow for more accurate collection processes. Another important
recent development has been the implementation of water mass-conserving positive-
definite advection schemes. Hahn and Mass (2009) showed that up to 20% of the
excessive orographic precipitation in their case study was due to the use of a non-
mass-conserving advection scheme. In the operational realm, the NWS continues
to use the relatively simple and cost-effective Ferrier et al. (2002) microphysics
scheme in its operational regional forecast model. Colle et al. (1999) demonstrated
a windward precipitation bias over the Cascade Range in the NWS’s Eta model
forecasts that ran operationally at that time with the Ferrier et al. microphysics
scheme, and the present authors are unaware if this behavior persists in the NWS’s
current North American Mesoscale (NAM) model forecasts, which also use the
Ferrier et al. scheme.
An additional area for potential improvement is in the interaction between the
relatively simplistic turbulent mixing scheme currently employed in models like
WRF and the microphysics scheme. Mesoscale models require a parameteriza-
tion for the vertical exchange of heat, moisture, and momentum by unresolved
turbulent motions. Most numerical models have separate parameterizations for the
microphysics and turbulent mixing and the latter rarely considers water processes
like condensation during mixing. Instead, water vapor and temperature are mixed
independently, which potentially leads to liquid water contents that are larger
than the value possible from adiabatic expansion and condensation of vapor
above its saturation point. Thus, the vertical mixing may be allowing excessive
accumulation of water content at lower levels in a model column and preventing
its proper mixing to higher altitudes, where it might subsequently be carried by
the flow over the terrain instead of contributing to the precipitation reaching the
ground.
7.8.3 Specific Uncertainties in Bulk Microphysical

Schemes, to Which Orographic Precipitation
Is Particularly Sensitive
7.8.3.1 Condensation and Warm Rain Processes
Nearly all bulk microphysical parameterizations are a collection of mini param-

eterizations that represent the various source and sink terms of the different
hydrometeor species. Perhaps one of the most pivotal sub-parameterizations is
that of “autoconversion” of cloud to rain, which attempts to represent the process
of cloud droplets colliding and coalescing into larger drops capable of attaining
precipitable fall speeds. There is no shortage of autoconversion treatments in the
literature (c.f. Gilmore and Straka 2008). Inherent in all autoconversion schemes
are various thresholds and/or mechanisms for deciding when to create rain, initially
from cloud droplets.
In nature, the creation of the first precipitation size particles depends heavily on
the activation of cloud condensation nuclei, which are essentially unknown in all
but the rarest circumstances during targeted field projects. Most models utilizing
BMPs do not explicitly predict the CCN for this reason. Instead, they activate a
pre-specified or variable number of droplets when the amount of water vapor in
the air exceeds saturation. When combining mass, number, and assumptions of the
size distribution, various mean sizes of drops can be computed with subsequent
assumptions determining whether sufficient drops reach precipitation size.
Most modern autoconversion schemes reflect the observed relationship that more
numerous but small droplets leads to little or no conversion to rain while fewer
large droplets leads relatively rapidly to rain. Once any rain is formed in a model,
the collection of small drops by the larger drops accelerates the rain production
significantly. Autoconversion itself acts as a seed to get more droplets collected
to produce rain in a model. Unfortunately, since so little is known about the
initial aerosol population that created the first droplets, the entire process of rain
production is fraught with potential errors.
Much of the advantage of high-resolution simulations to capture realistically
the terrain-induced updrafts is lost due to the lack of proper representation of
aerosol/CCN distributions and effects when determining onset of precipitation.
Specifically, simulations of orographic precipitation are directly impacted by the
method of autoconversion in shallow, moist flows impinging on the terrain. If a
model produces rain too early, it will fall out at lower elevations versus a model
that more slowly creates rain, which would simulate a higher starting elevation for
precipitation.
Not only does the autoconversion treatment affect the warm rain process, but also
the assumptions to convert cloud water to drizzle and rain also affects the ice in a
model. The various ice species all collide and interact with water drops. The earliest
BMPs used constant collision efficiencies between ice and water, but many newer
ones use variable efficiencies computed based on hydrometeor size. In general, the
smallest drops just follow the airflow around an ice particle and are not collected
due to the small inertia of the droplets. Relatively large cloud droplets and small
drizzle are more likely to be captured by a medium to large size snow or graupel
particle. This accretion or riming process is therefore peripherally related to the
initial aerosols and droplet spectra.
7.8.3.2 Ice Initiation
After decades of research, ice initiation remains one of the biggest challenges in
both observations and modeling. Determining how many ice crystals nucleate under
specific thermodynamic conditions is critical to subsequent properties of clouds and
balance of ice and liquid. If many tiny ice crystals are simulated, then an entire cloud
may glaciate since the lower vapor pressure of ice compared to water will produce a
vapor tendency away from liquid and towards the ice. Given that observations of ice
crystal number vary by approximately four orders of magnitude, with only a weak
relationship between temperature and number concentration, determining ice crystal
concentration in a model is an ongoing and vexing problem. As with cloud droplets,
ice usually initiates on a nucleus of foreign matter, perhaps most importantly on
mineral dust larger than 0.5–1.0 m. With strong updrafts, cloud droplets may
survive in liquid state to the homogeneous freezing point of approximately 38ı C
to 40ı C. Ice can also nucleate by freezing of deliquesced aerosols, especially at
cirrus cloud altitudes and temperatures.
Many bulk microphysics schemes continue to initiate ice at relatively high
temperatures (warmer than approximately 35ı C) and at or near ice supersaturation
via deposition of vapor directly into the ice phase, while observations generally
indicate that ice usually forms only after liquid droplets have been created. One
justification for continuing the practice of initiating ice below water saturation and
at temperatures warmer than 35ı C is due to sub-grid eddies that are not resolvable
by the model, which may include much higher saturations and colder temperatures
than what is resolved in the model grid volume.
7.8.3.3 Depositional Growth Rates
Once ice initiates in a model, determining its particle habit becomes the next
problem. Various ice shapes such as columns, plates, stellars, and dendrites grow by
vapor diffusion but each crystal type grows at different rates due to their different
shapes. If all ice crystals remained spherically shaped, then a population of crystals
would tend to grow uniformly. Instead, most ice crystals have one dominant and
one secondary growth axis and some ice crystals combine to form entirely different
growth regimes. Nearly all bulk schemes assume a vapor deposition growth rate
based on electrical capacitance analogies, representing the migration of vapor
toward an arbitrarily shaped particle in the same way that electrical charge migrates
toward a charged object. Some schemes assume variable capacitance depending

on crystal habit that most directly depends on temperature and ice supersaturation.
Other schemes assume constant capacitance or spheres. In a model with grid spacing
of a few km, a single crystal type is probably not applicable to the entire volume.
Observations of actual snow reaching the surface reveal that snow particles are
typically aggregates crystals of one or more habit types (Fig. 7.2), yielding different
shapes and values of capacitance, and it is not clear what the effect of an assumed
constant capacitance has on errors in depositional growth rate.
7.8.3.4 Particle Size Distributions
Identical to the assumption of water drops, the various forms of ice in bulk
microphysics schemes assume a simplified mathematical form with regard to the
number of crystals of specific sizes. Nearly all of the BMP codes in existence today
assume either a generalized gamma or exponential distribution. This assumption
appears rather robust in the case of snow, which has been shown for many years to be
well characterized by the exponential distribution. Field et al. (2005) measured snow
size distributions with aircraft probes and found a characteristic “shoulder” shaped
PSD, with relatively more small ice particles and fewer large snow particles than
the PSD represented by a purely exponential distribution, and this PSD has been
incorporated into the Thompson et al. (2004, 2008) BMP. However, there is still
some concern that a portion of the smallest ice crystals were produced by shattering
by the probes on the research aircraft as it flew through clouds.
7.8.3.5 Particle Type Issues
Mass and Fall Speed Versus Diameter Relationships
In conjunction with the size distribution assumption and the crystal habit assump-
tion, there are a variety of relationships between mass and fall speed of ice crystals.
In a large number of BMPs, the snow species is assumed to be composed of spheres
of relatively low density (0.1 g cm3 ). However, nearly all observations show a very
clear signal that density varies inversely with diameter, thus spherical shapes do not
approximate most snow. Setting the square of the diameter to be proportional to
mass retains the observed decrease in density as size increases.
Besides the mass relationship, BMPs require a relationship between particle fall
speed and size as well. In schemes with multiple snow species, each species needs
a corresponding fall speed. In theory, there exist relationships between crystal type,
drag, and fall speed for shapes like plates, needles, and columns, and aggregates
thereof (Mitchell 1996; Mitchell and Heymsfield 2005), that have been shown to
have good agreement with observationally derived power laws such as those shown
in Fig. 7.3. However, a challenge in utilizing such relationships is the ability to
diagnose or predict the snow particle type. A few research studies are beginning
to explore such capabilities (see below), but they are not specifically developed or
sufficiently tested for operational use.
Apportionment of Riming Growth Between Snow and Graupel
In BMPs with multiple ice categories, one species is typically graupel, which can
often look like a small Styrofoam pellet. Classical graupel forms when a small
snow crystal (a few 100 m in diameter) collects many cloud and/or drizzle size
water drops and freezes those drops onto the graupel surface. With enough riming,
graupel typically becomes a lump that is relatively spherical or conical. Its density
is generally higher than snow but still half or less than that of pure ice or hail.
Pinching a piece of graupel between your fingers gives the impression of pinching
a small piece of sponge.
In a BMPs, deciding when to convert rimed snow into graupel plays a crucial
role because graupel is typically two to four times denser and falls two to three
times faster than snow. Thus, apportioning between snow and graupel directly
affects the timing and amount of precipitation reaching the ground, especially in
orographic precipitation where there are often strong updrafts capable of sustained
liquid water production and therefore a source for continual riming of snow. In many
BMPs, the decision for creating graupel versus increasing snow mass is based on
simple thresholds for amount of snow and/or cloud water after which most snow
summarily becomes graupel. Eulerian models have no past history with amount of
rime collected by snow and therefore no physically correct method to transfer snow
mass into the graupel category. Instead, BMPs use ad hoc methods to decide and
perform mass transfer from snow to graupel. Furthermore, most BMPs also assume
that large water drops freeze into graupel because they lack a separate category for
a frozen drop or hail category. In this event, the BMP is almost certainly assigning
an incorrect density that is much lower than it should be.
Effects of Riming on Snow Particle Properties
Video disdrometer observations indicate that rimed snow particles fall approx-
imately 10–50% faster than unrimed snow particles of the same diameter, as
discussed in Sect. 7.2.1.4. As snow particles collect cloud droplets, their mass
and fall speed increase, while their maximum dimension increases much more
slowly, due to the droplets primarily filling in interstitial spaces in the particle (with
the exception of graupel). However, most BMPs allow all snow particles to grow
at constant density regardless of the source of growth (deposition or riming). If
the particles are actually growing significantly by riming (but not yet enough to
transition to graupel), the constant density approach will lead to a snow particle
population that is both too large in average size and also too slow falling. This
deficiency can be addressed by modulating the mass and fall speed versus diameter
relationships based on the growth history of the snow particles (discussed below).
7.8.4 Predicted Species
By the late 1990s, a large number of bulk microphysical schemes were in existence,
employed in both operational and research models. At the upper end of the range
of sophistication were five-class schemes with separate equations for cloud water,
rain, cloud ice, snow, and either graupel or hail. Recent studies have focused on
the complexity and variability of the frozen precipitation species, and have explored
the value of adding additional predictive equations to better describe this variability.
Woods et al. (2007) and Hashino and Tripoli (2007, 2008) added separate predictive
equations for different crystal habit types of ice particles. Morrison and Grabowski
(2008) and McCormick (2009) split the snow species into two categories, that
grown by deposition and that grown by riming, and carried separate prognostic
equations for each. Both of these efforts are aimed at accounting for the known large
variability of ice particle mass and fall speed for different habit types and degrees of
riming. McCormick (2009) found that the use of riming prediction tended to shift
windward precipitation farther up the windward slope due to more precipitation
being retained in the rimed snow category instead of being moved quickly into faster
falling graupel; and also to shift leeward precipitation back up the slope, due to
spillover snow having somewhat higher fall speeds characteristic of rimed particles.
The overall effect was to “squeeze” the precipitation closer to the crest from both
the windward and leeward sides.
Both habit and riming prediction are recently developed techniques, and it
remains to be seen whether they improve the accuracy of model forecasts, and
if those improvements are worth the added computational costs, compared to, for
example, predicting more moments of the particle size distributions (Milbrandt and
Yau 2005), including more accurate size-dependent riming (Saleeby and Cotton
2008), or using a simpler diagnostic approach to accounting for the effects of riming
on particle fall speeds and masses (Lin and Colle 2010).
7.8.5 Interactions with Other Model Processes
The microphysical parameterization is just one among many physical parameteriza-

tion schemes in use in mesoscale NWP models. Processes such as sub-grid cumulus
convection, planetary boundary layer physics, land surface processes, sub grid-
scale mixing in the horizontal and vertical directions, and radiative processes are
also represented by separate algorithms, and all of these processes interact with
precipitation physics to affect the final precipitation prediction at the ground.
Recent studies of orographic precipitation have shown that, while the choice
of microphysical scheme exerts the greatest influence on quantitative precipitation
forecasts, sensitivity to choices for the other physics schemes, particularly the
planetary boundary layer scheme and horizontal diffusion, are nearly as large (Zängl
2004; Jankov et al. 2007; Garvert et al. 2007). Additionally, even the treatment
of a process as basic as advection has recently been shown to be important in

prediction of winter orographic precipitation, with the use of a positive-definite
advection scheme for the hydrometeor species reducing precipitation amounts by
3–17% (Hahn and Mass 2009).
This chapter has summarized the current state of knowledge of the role that
cloud and precipitation microphysics plays in the development of winter-time
midlatitude orographic precipitation, in attempt to help “bridge the gap” between
what the research community has learned about this topic, and what forecasters
should know to engage in the forecast process from a more informed position.
Knowledge of microphysical processes in orographic precipitation enters into the
pattern recognition techniques, conceptual models and heuristics that the forecaster
applies on a regular basis to assess what is currently happening and how the
weather will evolve. It also helps the forecaster understand how model-generated
precipitation forecasts are arrived at, and what their inherent deficiencies might be.
Research in the realm of orographic precipitation and microphysics has been
particularly active in the last 10 years or so, spurred on by a number of field
projects and their associated observational phenomenological studies and numerical
modeling studies. Much has been learned about the dynamical processes by which
flow over orography induces upward motion, and how those processes interact with
cloud microphysics to determine the amount and distribution of precipitation across
orographic barriers.
However, in spite of the stated goals of these studies to fully quantify in space and
time the microphysical processes that determine the amount and spatial distribution
of precipitation, several important gaps in knowledge remain. We have not devel-
oped techniques, either through direct or remotely sensed observation, to measure
the mesoscale vertical velocity over an orographic barrier with sufficient accuracy
(within a few 10s of cm s1 ) to quantify the most basic input into the precipitation-
generating process, namely, the vapor supply rate for condensation and deposition.
We also have not directly quantified the growth processes of frozen particles by
vapor deposition and riming, a task that is essential to verifying the microphysical
pathways for growth seen in mesoscale NWP model forecasts. This pursuit is
hindered by fundamental experimental limitations in both the laboratory and the
real atmosphere. In the lab, growth of individual frozen particles can be simulated,
but reproducing realistic atmospheric conditions and interactions between large
populations of falling particles is logistically prohibitive. In the atmosphere, aircraft
can take measurements where realistic conditions and actual particle populations
exist, but they cruise through clouds at 100 m s1 , capturing only snapshots of
particle structures, masses, and numbers along a narrow horizontal path, rather than
following a population of falling particles and quantifying processes of interaction
and growth. Further limitations exist over orography, where rules of safe flight
prohibit operations within the lowest 1–2 km of the atmosphere, a region where
a large fraction of the growth of precipitation often occurs. Careful observation of
particles on the ground can yield more detail about particle types and structure, but
also does not quantify processes, is limited by confinement to the surface, and is
much less effective than aircraft at achieving large spatial sampling.
The field of in situ microphysical observations is in need of new technological
innovations that can quantify processes rather than particles. Remote sensing
holds some promise to fill this gap, with continued development in the areas of
polarimetric and dual-wavelength radar. The former helps distinguish particle types
and provides more accurate precipitation rate estimates, and the latter has shown
promise in mapping out the 3D cloud water field, a major step forward in cloud
observations. Even though these measurements do not directly quantify processes,
their ability to map out the full 3D population of particles and their characteristics
should make it easier to retrieve estimates of the processes that produce them.
As the cloud physics field moves forward with technological advances, new
theories and conceptual models, and improved microphysical schemes in NWP
models, it is imperative that the research community and operational forecasting
community continue to interact, because both provide necessary guidance and focus
to the other. By listening to the research community, the operational community can
exploit the latest findings to improve their forecast methods and skill; on the same
token, by listening to the operational community, the research community can focus
their efforts on problems that will translate to real forecast improvements.
Acknowledgements The authors would like to thank Fotini Katopodes Chow, Stephan
De Wekker, and Bradley J. Snyder for giving us the opportunity to contribute to this important
book, and for their hard work and patience in bringing it together. The first author is grateful to
John Locatelli and Art Rangno for very helpful discussions, and to Ken Libbrecht for permitting
us to use his exquisite snow crystal photographs. We would also like to thank two anonymous
reviewers for performing a careful scrutiny of a very long manuscript, and for their many helpful
comments. Much of the research that was reported on in this chapter was funded by the National
Science Foundation, with the first author being supported under grants 0242592 and 0634999.
The authors would also like to thank the Natural Sciences and Engineering Research Council of
Canada for its support.
References
Alcott, T. I., and W. J. Steenburgh, 2010: Snow-to-liquid ratio variability and prediction at a high-
elevation site in Utah’s Wasatch Mountains. Wea. Forecasting, 25, 323–337
Atlas, D., R. Tatehira, R.C. Srivastava, W. Marker and R.E. Carbone, 1969: Precipitation-induced
mesoscale wind perturbations in the melting layer. Quart. J. Roy. Meteor. Soc., 95, 544–560.
Auer, A.H., J.M. White, 1982: The combined role of kinematics, thermodynamics and cloud
physics associated with heavy snow episodes. J. Meteor. Soc. Japan, 60, 500–507.
Bader, M. J. and W. T. Roach. 1977. Orographic rainfall in warm sectors of depressions. Quart. J.
Roy. Meteor. Soc., 103, 269–280.
Bailey, M. and J. Hallett, 2004: Growth rates and habits of ice crystals between 20ı and 70ı C.
J. Atmos. Sci., 61, 514–544.
Bailey, M. P., and J. Hallett, 2009: A comprehensive habit diagram for atmospheric ice crystals:
confirmation from the laboratory, AIRS II, and other field studies. J. Atmos. Sci., 66,
2888–2899.
Banta, R.M., 1990: Atmospheric processes over complex terrain. Meteorological Monographs, 23,
W. Blumen, Ed., Amer. Meteor. Soc., 229–283.
Barstad, I., and R. B., Smith, 2005: Evaluation of an orographic precipitation model. J. Hydrome-
teor., 6, 85–99.
Barthazy, E. and R. Schefolda, 2006: Fall velocity of snow akes of different riming degree and
crystal types. Atmos. Research, 82, 391–398.
Baxter, M. A., C. E. Graves, and J. T. Moore, 2005: A climatology of snow-to-liquid ratio for the
contiguous United States. Wea. Forecasting, 20, 729–744.
Beard, K. V., and H. T. Ochs III, 1993: Warm-rain initiation: An overview of microphysical
mechanisms. J. Appl. Meteor., 32, 608–625.
Bell, G.D. and L. Bosart. 1988: Appalachian cold-air damming. Mon. Wea. Rev., 116, 137–161.
Bergeron, T., 1949: The problem of artificial control of rainfall on the globe. Part II, The coastal
orographic maxima of precipitation in autumn and winter. Tellus, 1, 15–32.
Bergeron, T., 1965: On the low-level redistribution of atmospheric water caused by orography.
Suppl. Proc. Int. Conf. Cloud Phys. Tokyo, pp. 96–100.
Borys, R. D., D. H. Lowenthal, and D. L. Mitchell, 2000: The relationships among cloud
microphysics, chemistry, and precipitation rate in cold mountain clouds. Atmos. Environ., 34,
2593–2602.
Borys, R. D., D. H. Lowenthal, S. A. Cohn, and W. O. J. Brown, 2003: Mountaintop and radar
measurements of anthropogenic aerosol effects on snow growth and snowfall rate. Geophys.
Res. Lett., 30, 1538.
Bougeault P., Coauthors, 2001: The MAP Special Observing Period. Bull. Amer. Meteor. Soc., 82,
433–462.
Bourgouin, P., 2000: A method to determine precipitation types. Wea. Forecasting, 15, 583–592.
Bousquet O., and B. F. Smull, 2003: Airflow and precipitation fields within deep Alpine valleys
observed by airborne Doppler radar. J. Appl. Meteor., 42, 1497–1513.
Bradley, M. M., 1985: The numerical simulation of orographic storms. Ph.D. dissertation,
University of Illinois, Urbana–Champaign, 261 pp.
Brewer, A. W., and H. P. Palmer, 1949: Condensation processes at low temperatures, and the
production of new sublimation nuclei by the splintering of ice. Nature, 164, 312–313.
Brown, S. R., 1970: Terminal velocities of ice crystals. Master’s thesis, Colorado State University.
Browning, K. A., F. F. Hill, and C. W. Pardoe, 1974: Structure and mechanism of precipitation and
the effect of orography in a wintertime warm sector. Quart. J. Roy. Meteor. Soc., 100, 309–330.
Browning, K. A., C. W. Pardoe, and F. F. Hill, 1975: The nature of orographic rain at wintertime
cold fronts. Quart. J. Roy. Meteor. Soc., 101, 333–352.
Bruintjes, R. T., T. L. Clark, and W. D. Hall, 1994: Interactions between topographic airflow and
Sci., 51, 48–67.
Byun, K.-Y., J. Yang, and T.-Y. Lee, 2008: A snow-ratio equation and its application to numerical
snowfall prediction. Wea. Forecasting, 23, 644–658.
Carruthers, D. J., and T. W. Choularton, 1983: A model of the feeder-seeder mechanism of
orographic rain including stratification and wind-drift effects. Quart. J. Roy. Meteor. Soc., 109,
575–588.
Casson, G. D., 2009: Evaluating the importance of crystal type on new snow instability. Master’s
thesis, University of Washington.
Choularton, T. W., and S. J. Perry, 1986: A model of the orographic enhancement of snowfall by
the seeder-feeder mechanism. Quart. J. Roy. Meteor. Soc., 112, 335–345.
Cobb, D. K., and J. S. Waldstreicher, 2005: A simple physically based snowfall algorithm.
Preprints, 21st Conf. on Weather Analysis and Forecasting/17th Conf. on Numerical Weather
Prediction, Washington, DC, Amer. Meteor. Soc., 2A.2. [Available online at http://ams.confex.
com/ams/pdfpapers/94815.pdf.]
Colle, B. A., 2004: Sensitivity of orographic precipitation to changing ambient conditions and
terrain geometries: An idealized modeling perspective. J. Atmos. Sci., 61, 588–606.
Colle, B. A., 2008: Two-dimensional idealized simulations of the impact of multiple windward
ridges on orographic precipitation. J. Atmos. Sci., 65, 509–523.
Colle B. A., and C. F. Mass, 2000: The 5–9 February 1996 flooding event over the Pacific
Northwest: Sensitivity studies and evaluation of the MM5 precipitation forecasts. Mon. Wea.
Rev., 128, 593–617.
Colle, B. A., and Y. Zeng, 2004a: Bulk microphysical sensitivities and pathways within the MM5
for orographic precipitation. Part I: The Sierra 1986 event. Mon. Wea. Rev., 132, 2780–2801.
Colle, B. A., and Y. Zeng, 2004b: Bulk microphysical sensitivities and pathways within the MM5
for orographic precipitation. Part II: Impact of different bulk schemes, barrier width, and
freezing level. Mon. Wea. Rev., 132, 2802–2815.
Colle B.A., and S. E., Yuter, 2007: The impact of coastal boundaries and small hills on the
precipitation distribution across southern Connecticut and Long Island, New York. Mon. Wea.
Rev., 135, 933–954.
Colle, B. A., K. J. Westrick, and C. F. Mass, 1999: Evaluation of the MM5 and Eta-10 precipitation
forecasts over the Pacific Northwest during the cool season. Wea. Forecasting., 14, 137–154.
Colle, B. A., C. F. Mass, and K. J. Westrick, 2000: MM5 precipitation verification over the Pacific
Northwest during the 1997–1999 cool seasons. Wea. Forecasting., 15, 730–744.
Colle, B. A., J. B. Olson, and J. S. Tongue, 2003: Multiseason verification of the MM5. Part I:
Comparison with the NCEP Eta Model over the central and eastern United States and impact
of MM5 resolution. Wea. Forecasting, 18, 431–457.
Colle B. A., M. Garvert, J. B. Wolfe, C. F. Mass, and C. P. Woods, 2005a: The 13–14 December
2001 IMPROVE-2 event. Part III: Simulated microphysical budgets and sensitivity studies.
J. Atmos. Sci., 62, 3535–3558.
Colle B. A., J. B. Wolfe, W. J. Steenburgh, D. E. Kingsmill, J. A. W. Cox, and J. C. Shafer, 2005b:
High-resolution simulations and microphysical validation of an orographic precipitation event
over the Wasatch Mountains during IPEX IOP3. Mon. Wea. Rev., 133, 2947–2971.
Colle, B. A., Y. Lin, S. Medina, and B. Smull, 2008: Orographic modification of convection and
flow kinematics by the Oregon coastal range and Cascades during IMPROVE-2. Mon. Wea.
Rev., 136, 3894–3916.
Conway, H., and C. Wilbour, 1999: Evolution of snow slope stability during storms. Cold Reg. Sci.
Technol., 30, 67–77.
Cooper, W. A., and C. P. R. Saunders, 1980: Winter storms over the San Juan Mountains. Part II:
Microphysical processes. J. Appl. Meteor., 19, 927–941.
Cotton, W. R., G. J. Tripoli, R. M. Rauber, and E. A. Mulvhill, 1986: Numerical simulation of the
effects of varying ice crystal nucleation rates and aggregation processes on orographic snowfall.
J. Climate Appl. Meteor., 25, 1658–1680.
Cox, J. A. W., W. J. Steenburgh, D. E. Kingsmill, J. C. Shafer, B. A. Colle, O. Bousquet, B. F.
Smull, and H. Cai, 2005: The kinematic structure of a Wasatch Mountain winter storm during
IPEX IOP3., Mon. Wea. Rev., 133, 521–542.
Demoz, B. B., R. Zhang, and R. L. Pitter, 1993: An analysis of Sierra Nevada winter orographic
storms: Ground-based ice-crystal observations. J. Appl. Met., 32, 1826–1836.
Dixon, R.W., L. Mosiman, B. Oberholzer, J. Staechelin, A. Waldvogel, and J.L. Collett, 1994: The
effect of riming on ion concentrations of winter precipitation. Part 1: A quantitative analysis of
field measurements. J. Geophys. Res., D6, 11, 517–11, 528.
Donaldson, N.R. and R.E. Stewart, 1993: Fog induced by mixed phase precipitation. Atmos. Res.,
29, 9–25
Doyle, J. D., 1997: The influence of mesoscale orography on a coastal jet and rainband. Mon. Wea.
Rev., 125, 1465–1488.
Dubé, I., 2003: From mm to cm : : : Study of snow/liquid water ratios in Quebec. Meteorological
Service of Canada, Quebec, QC, Canada. [Available online at http://www.meted.ucar.edu/
norlat/snowdensity/from mm to cm.pdf.].
Ferrier, B. S., Y. Jin, T. Black, E. Rogers, and G. DiMego, 2002: Implementation of a new grid-
scale cloud and precipitation scheme in the NCEP Eta model, Preprints, 15th Conference on
Numerical Weather Predication, San Antonio TX, Amer. Meteor., Soc., 280–283.
Fabry, F., and I. Zawadzki, 1995: Long-term radar observations of the melting layer of precipitation
and their interpretation. J. Atmos. Sci., 52, 838–851.
Field, P. R., R. J. Hogan, P. R. A. Brown, A. J. Illingworth, T. W. Choularton, and R. J. Cotton,
2005: Parameterization of ice-particle size distributions for mid-latitude stratiform cloud.
Field, P. R., A. J. Heymsfield, and A. Bansemer, 2006: A test of ice self-collection kernels using
aircraft data. J. Atmos. Sci., 63, 651–666.
Findeisen, W., 1940: The formation of the 0ı C isothermal layer and fractocumulus under
nimbostratus. Meteor. Z., 57, 49–54.
Fuhrer, O., and C. Schär, 2005: Embedded cellular convection in moist flow past topography.
J. Atmos. Sci., 62, 2810–2828.
Fujiyoshi, Y., 1986: Melting snowflakes. J. Atmos. Sci., 43, 307–311.
Garvert M. F., B. A. Colle, and C. F. Mass, 2005: The 13–14 December 2001 IMPROVE-2 event.
Part I: Synoptic and mesoscale evolution and comparison with a mesoscale model simulation.
J. Atmos. Sci., 62, 3474–3492.
Garvert, M. F., B. F. Smull, and C. F. Mass, 2007: Multiscale mountain waves influencing a major
orographic precipitation event. J. Atmos. Sci. 64, 711–737.
Gilmore, M. S., and J. M. Straka, 2008: The Berry and Reinhardt autoconversion parameterization:
A digest. J. Appl. Meteor. Clim., 47, 375–396.
Gocho, Y., 1978: Numerical experiment of orographic heavy rainfall due to a stratiform cloud.
J. Meteor. Soc. Japan, 56, 405–423.
Grabowski, W. W., 1989: On the influence of small-scale topography on precipitation. Quart.
J. Roy. Meteor. Soc., 115, 633–650.
Gunn, K. L. S., and J.S. Marshall, 1958: The distribution with size of aggregate snowflakes.
J. Meteor., 15, 452–461.
Hahn, R. S., and C. F. Mass, 2009: The impact of positive-definite moisture advection and low-level
moisture flux bias over orography. Mon. Wea. Rev., 137, 3055–3071.
Hallett, J., and S. C. Mossop, 1974: Production of secondary ice particles during the riming process.
Nature, 249, 26–28.
Hanesch, M., 1999: Fall velocity and shape of snowflakes. Ph.D. dissertation, Swiss Federal
Institute of Technology, 117 pp.
Harimaya, T., and M. Sato, 1989: Measurement of riming amount on snowflakes. J. Fac. Sci.,
Hokkaido Univ., Ser. 7, 8, 355–366.
Hashino, T. and G. Tripoli, 2007: The Spectral Ice Habit Prediction System (SHIPS). Part I: Model
description and simulation of vapor deposition process. J. Atmos. Sci., 64, 2210–2237.
Hashino, T. and G. Tripoli, 2008: The Spectral Ice Habit Prediction System (SHIPS). Part
II: Simulation of nucleation and depositional growth of polycrystals. J. Atmos. Sci., 65,
3071–3094.
Heffernan, E., and J. Marwitz. 1996: The Front Range blizzard of 1990. Part II: Melting effects in
a convective band. Mon. Wea. Rev., 124, 2469–2482.
Heggli, M. F., and D. W. Reynolds, 1985: Radiometric observations of supercooled liquid water
within a split front over the Sierra Nevada. J. Climate Appl. Meteor., 24, 1258–1261.
Heggli, M. F., and R. Rauber, 1988: The characteristics and evolution of supercooled water in
wintertime storms over the Sierra Nevada: A summary of microwave radiometric measurements
taken during the Sierra Cooperative Pilot Project. J. Appl. Meteor., 27, 989–1015.
Heggli, M. F., L. Vardiman, R. E. Stewart, and A. Huggins, 1983: Supercooled liquid water and
ice crystal distributions within Sierra Nevada winter storms. J. Appl. Meteor., 22, 1875–1886.
Heymsfield, A. J., and M. Kajikawa, 1987: An improved approach to calculating terminal velocities
of plate-like crystals and graupel. J. Atmos. Sci., 44, 1088–1099.
Hill, F. F., K. A. Browning, and M. J. Bader, 1981: Radar and raingauge observations of orographic
rain over South Wales. Quart. J. Roy. Meteor. Soc., 107, 643–670.
Hobbs, P. V., 1965: The aggregation of ice particles in clouds and fogs at low temperatures. J.
Atmos. Sci., 22, 296–300.
Hobbs, P. V., 1975: The nature of winter clouds and precipitation in the Cascade Mountains and
their modification by artificial seeding. Part I: Natural conditions. J. Appl. Meteor., 14, 783–
804.
Hobbs, P. V., and A. J. Alkezweeny, 1968: The fragmentation of freezing water droplets in free
fall. J. Atmos. Sci., 25, 881–888.
Hobbs, P. V., and D. Atkinson, 1976: The concentrations of ice particles in orographic clouds and
cyclonic storms over the Cascade Mountains. J. Atmos. Sci., 33, 1362–1374.
Hobbs, P. V., and A. L. Rangno, 1985: Ice particle concentrations in clouds. J. Atmos. Sci., 42,
2523–2549.
Hobbs, P. V., and A. L. Rangno, 1990: Rapid development of ice particle concentrations in small
polar maritime cumuliform clouds. J. Atmos. Sci., 47, 2710–2722.
Hobbs, P., R. Easter, and A. Fraser, 1973: A theoretical study of the flow of air and fallout of solid
precipitation over mountainous terrain: Part II. Microphysics. J. Atmos. Sci., 30, 813–823.
Houze, Jr., R. A., and S. Medina, 2005: Turbulence as a mechanism for orographic precipitation
enhancement. J. Atmos. Sci., 62, 3599–3623.
Houze, Jr., R. A., C. N. James, and S. Medina, 2001: Radar observations of precipitation and
airflow on the Mediterranean side of the Alps: Autumn 1998 and 1999. Q. J. R. Meteorol. Soc.,
127, 2537–2558.
Ikeda, K., R. M. Rasmussen, W. D. Hall, and G. Thompson, 2007: Observations of freezing drizzle
in extratropical cyclonic storms during IMPROVE-2. J. Atmos. Sci., 64, 3016–3043.
Jankov, I., P. J. Schultz, C. J. Anderson, and S. E. Koch, 2007: The impact of different physical
parameterizations and their interactions on cold season QPF in the American River basin.
J. Hydrometeor., 8, 1141–1151.
Jiang Q., and R. B. Smith, 2003: Cloud timescales and orographic precipitation. J. Atmos., Sci, 60,
1159–1172.
Jirak, I. L., and W. R. Cotton, 2006: Effect of air pollution on precipitation along the Front Range
of the Rocky Mountains. J. Appl. Meteor., 45, 236–245.
Judson, A., and N. Doesken, 2000: Density of freshly fallen snow in the central Rocky Mountains.
Kajikawa, M., 1976: Observation of falling motion of columnar snow crystals. J. Meteor. Soc.
Japan, 54, 276–283.
Kajikawa, M., 1989: Observation of the falling motion of early snowflakes. Part II: On the variation
of falling velocity. J. Meteor. Soc. Japan, 67, 731–737.
Kalina, M.F., and H. Puxbaum, 1994: A study of the influence of riming of ice crystals on snow
chemistry during different seasons in precipitating continental clouds, Atmos. Environ., 28,
3311–3328.
Kingsmill, D. E., P. J. Neiman, F. M. Ralph, and A. B. White, 2006: Synoptic and topographic
variability of northern California precipitation characteristics in landfalling winter storms
observed during CALJET. Mon. Wea. Rev., 134, 2072–2094.
Kirshbaum, D. J. and R. B. Smith, 2008: Temperature and moist-stability effects on midlatitude
orographic precipitation. Q. J. R. Meteorol. Soc., 134, 1183–1199.
Knight, C.A., 1979: Observations of the morphology of melting snow. J. Atmos. Sci., 36,
1123–1129.
Korolev, A., 2007: Limitations of the Wegener–Bergeron–Findeisen mechanism in the evolution
of mixed-phase clouds. J. Atmos. Sci., 64, 3372–3375.
Kunz, M. and C. Kottmeier, 2006: Orographic enhancement of precipitation over low mountain
ranges. Part II: Simulations of heavy precipitation events over southwest Germany. J. Appl.
Kusunoki, K., M. Murakami, M. Hashimoto, N. Orikasa, Y. Yamada, H. Mizuno, K. Hamazu, and
H. Watanabe, 2004: The characteristics and evolution of orographic snow clouds under weak
cold advection. Mon. Wea. Rev., 132, 174–191.
Kusunoki, K., M. Murakami, N. Orikasa, M. Hoshimoto, Y. Tanaka, Y. Yamada, H. Mizuno,

K. Hamazu, and H. Watanabe, 2005: Observations of quasi-stationary and shallow orographic
snow clouds: Spatial distributions of supercooled liquid water and snow particles. Mon. Wea.
Rev., 133, 743–751.
LaChapelle, E. R., 1958: Winter snow observation at Mt. Olympus. Proc. 26th Annual Meeting,
Bozeman, MT, Western Snow Conference, 59–63.
Lascaux F., E. Richard, and J.-P. Pinty. 2006: Numerical simulations of three different MAP IOPs
and the associated microphysical processes. Q. J. R. Meteorol. Soc. 132, 1907–1926.
Lawson, R. P., R. E. Stewart, and L. J. Angus, 1998: Observations and numerical simulations of
the origin and development of very large snowflakes. J. Atmos. Sci., 55, 3209–3229.
Lin, Y., and B. A. Colle, 2010: A new bulk microphysical scheme that includes varying degree of
riming and ice habit characteristics. J. Atmos. Sci., submitted.
Locatelli, J. D., and P. V. Hobbs, 1974: Fall speeds and masses of solid precipitation particles.
J. Geophys. Res., 79, 2185–2197.
Long, A. B., B. A. Campistron, and A. W. Huggins, 1990: Investigations of a winter mountain
storm in Utah. Part I: Synoptic analyses, mesoscale kinematics, and water release rates.
J. Atmos. Sci., 47, 1302–1322.
Lowenthal, D.H., R. D. Borys, W. Cotton, S. Saleeby, S. A. Cohn, and W. O. J. Brown, 2011:
The altitude of snow growth by riming and vapor deposition in mixed-phase orographic clouds.
Atmos. Environ., 45, 519–522.
Lumb, F.E., 1961: The problem of forecasting the downward penetration of snow. Meteor. Mag.,
90, 31–319.
Lumb, F.E., 1983a: Sharp snow/rain contrast-an explanation. Weather, 38, 71–73.
Lumb, F.E., 1983b: Snow on hills. Weather, 38, 114–115.
Lundquist, J., P. Neiman, B. Martner, A. White, D. Gottas and M. Ralph, 2008: Rain versus snow
in the Sierra Nevada, California: comparing Doppler profiling radar and surface observations
of the melting level. J. Hydrometeor., 9, 194–211.
Lynn, B., A. Khain, D. Rosenfeld, and W. L. Woodley, 2007: Effects of aerosols on precipitation
from orographic clouds. J. Geophys. Res., 112, D10225.
Magono, C., and C. W. Lee, 1966: Meteorological classification of natural snow crystals. J. Faculty
Sci., II, 321–335.
Marigo, G., T. Robert-Luciani and A. Crepaz, 2008: Snow level forecasting methods and
parameters: two practical examples on eastern Italian Alps., 13th Mtn. Meteor. Conf., Whistler,
Canada.
Marshall, J. S., and W.M. Palmer, 1948: The distribution of raindrops with size. J. Meteor., 5,
165–166.
Marwitz, J., 1980: Winter storms over the San Juan Mountains. Part I: Dynamical processes.
J. Appl. Meteor., 19, 913–926.
Marwitz, J., 1983: The kinematics of orographic airflow during Sierra storms. J. Atmos. Sci., 40,
1218–1227.
Marwitz, J. D., 1987a: Deep orographic storms over the Sierra Nevada. Part I: Thermodynamic
and kinematic structure. J. Atmos. Sci., 44, 159–173.
Marwitz, J. D., 1987b: Deep orographic storms over the Sierra Nevada. Part II: The precipitation
processes. J. Atmos. Sci., 44, 174–185
Marwitz, J., and J. Toth. 1993: The Front Range blizzard of 1990. Part I: Synoptic and mesoscale
structure. Mon. Wea. Rev., 121, 402–415.
Matsuo, T. and Y. Sasyo, 1981a: Melting of snowflakes below the freezing level in the atmosphere.
J. Meteorol. Soc. Jpn., 59, 10–24.
Matsuo, T. and Y. Sasyo, 1981b: Non-melting of snowflakes observed in subsaturated air below
freezing lever. J. Meteorol. Soc. Jpn., 59, 26–32.
Matsuo, T., Y. Sasyo and Y. Sato, 1981: Relationship between types of precipitation on the ground
and surface meteorological elements. J. Meteor. Soc. Japan, 59, 462–476.
McCormick, H. S., 2009: Observations and modeling of snow over the Washington Cascades.
Master’s thesis, University of Washington.
Medina, S. and R. Houze, 2003: Air motions and precipitation growth in alpine storms. Quart. J.
Roy. Meteor. Soc., 129, 345–371.
Medina, S., B. Smull, R. Houze and M. Steiner, 2005: Cross-barrier flow during orographic
precipitation events: Results from MAP and IMPROVE. J. Atmos. Sci., 62, 3850–3598.
Medina, S., E. Sukovich, and R. A. Houze, Jr., 2007: Vertical structures of precipitation in cyclones
crossing the Oregon Cascades. Mon. Wea. Rev., 135, 3565–3586.
Meyers, M. P., P. J. DeMott, and W. R. Cotton, 1992. New primary ice nucleation parameterizations
in an explicit cloud model. J. Appl. Met. 31, 708–721.
Milbrandt, J. A., and M. K. Yau, 2005: A multi-moment bulk microphysics parameterization. Part
II: A proposed three-moment closure and scheme description. J. Atmos. Sci., 62, 3065–3081.
Minder J.R., and D.R. Durran, 2010: Mesoscale controls on the climatology of the mountainside
snow line. J. Atmos. Sci., submitted.
Minder J.R., D.R. Durran, G.H. Roe, and A.M. Anders, 2008: The climatology of small-scale
orographic precipitation over the Olympic Mountains: Patterns and processes. Quart. J. Roy.
Meteor. Soc., 134, 817–839.
Mitchell, D. L., 1994: A model predicting the evolution of ice particle size spectra and radiative
properties of cirrus clouds. Part I: Microphysics. J. Atmos. Sci., 51, 797–816.
Mitchell D. L., 1996: Use of mass- and area-dimensional power laws for determining precipitation
particle terminal velocities. J. Atmos. Sci., 53, 1710–1723.
Mitchell D. L., and A. J. Heymsfield, 2005: Refinements in the treatment of ice particle terminal
velocities, highlighting aggregates. J. Atmos. Sci., 62, 1637–1644.
Mitchell, D. L., R. Zhang, and R. L. Pitter, 1990: Mass-dimensional relationships for ice particles
and the influence of riming on snowfall rates. J. Appl. Meteor., 29, 153–163.
Mitra, S.K., O. Vohl, M. Ahr and H. R. Pruppacher, 1990: A wind tunnel and theoretical study
of the melting behavior of atmospheric ice particles. Part IV: Experiment and theory for snow
flakes. J. Atmos. Sci., 47, 584–591.
Moore, G.W.K. and R.E. Stewart, 1985: The coupling between melting and convective air motions
in stratiform clouds. J. Geophys. Res., 90, 10659–10666.
Morrison H., and W.W. Grabowski, 2008: A novel approach for representing ice microphysics in
models: description and tests using a kinematic framework. J. Atmos. Sci., 65, 1528–1548
Muhlbauer, A. and U. Lohmann, 2008: Sensitivity studies of the role of aerosols in warm-phase
orographic precipitation in different dynamical flow regimes. J. Atmos. Sci., 65, 2522–2542.
Myers, V. A., 1962: Airflow on the windward side of a ridge. J. Geophys. Res., 67, 4267–4291.
Nakaya, U., 1954: Snow Crystals, Natural and Artificial. Harvard University Press, 510 pp.
Neiman P. J., F. M. Ralph, A. B. White, D. E. Kingsmill, and P. O. G. Persson, 2002: The statistical
relationship between upslope flow and rainfall in California’s coastal mountains: Observations
during CALJET. Mon. Wea. Rev., 130, 1468–1492.
Peterson, T. C., L. O. Grant, W. R. Cotton, and D. C. Rogers, 1991: The effect of decoupled low-
level flow on winter orographic clouds and precipitation in the Yampa River valley. J. Appl.
Meteor., 30, 368–386.
Power, B. A., P. W. Summers, and J. d’Avignon, 1964: Snow crystal forms and riming effects as
related to snowfall density and general storm conditions. J. Atmos. Sci., 21, 300–305.
Pruppacher, H. R., and J. D. Klett, 1997: Microphysics of Clouds and Precipitation. Kluwer
Academic, 954 pp.
Pujol, O., J.-F. Georgis, M. Chong, and F. Roux, 2005: Dynamics and microphysics of orographic
precipitation during MAP IOP 3. Q. J. R. Meteorol. Soc., 131, 2795–2819.
Ralph, F. M., and Coauthors. 1999: The California Land-falling Jets Experiment (CALJET):
Objectives and design of a coastal atmosphere–ocean observing system deployed during a
strong El Niño. Preprints, Third Symp. on Integrated Observing Systems, Dallas, TX, Amer.
Meteor. Soc., 78–81.
Ralph, F.M., P.J. Neiman, and R. Rotunno, 2005a: Dropsonde observations in low-level jets over
the northeastern Pacific Ocean from CALJET-1998 and PACJET-2001: Mean vertical-profile
and atmospheric-river characteristics. Mon. Wea. Rev., 133, 889–910.
Ralph, M. F., R.M. Rauber, B.F. Jewett, D.E. Kingsmill, P. Pisano, P. Pugner, R.M. Rasmussen,
D.W. Reynolds, T.W. Schlatter, R.E. Stewart, J.S. Waldstricher, 2005b: Improving short
term (0–48 hour) cool season quantitative precipitation forecasting: recommendations from
a USWRP workshop. Bull. Amer. Meteor. Soc., 86, 1619–1632.
Rangno, A. L., 2008: Fragmentation of freezing drops in shallow maritime frontal clouds. J. Atmos.
Sci., 65, 1455–1466.
Rangno, A. L., and P. V. Hobbs, 2005: Microstructures and precipitation development in cumulus
and small cumulonimbus clouds over the warm pool of the tropical Pacific Ocean. Quart. J.
Roy. Meteor. Soc., 131, 639–673.
Rasmussen, R. M., and Coauthors, 1992: Winter Icing and Storms Project (WISP). Bull. Amer.
Meteor. Soc., 73, 951–974.
Rasmussen, R. M., J. Vivekanandan, and J. Cole, 1999: The estimation of snowfall rate using
visibility. J. Appl. Meteor., 38, 1542–1563.
Rasmussen, R. M., I. Geresdi, G. Thompson, K. Manning, and E. Karplus, 2002: Freezing drizzle
formation in stably stratified layer clouds: The role of radiative cooling of cloud droplets and
cloud condensation and ice initiation. J. Atmos. Sci., 59, 837–860.
Rauber, R. M., 1987: Characteristics of cloud ice and precipitation during wintertime storms over
the mountains of northern Colorado. J. Climate Appl. Meteor., 26, 488–524.
Rauber, R., 1992: Microphysical structure and evolution of a central Sierra Nevada orographic
cloud system. J. Appl. Meteor., 31, 3–24.
Rauber, R., and L. O. Grant, 1986: The characteristics and distribution of cloud water over the
mountains of northern Colorado during wintertime storms. Part II: Spatial distribution and
microphysical characteristics. J. Climate Appl. Meteor., 25, 489–504.
Rauber, R. M. and L. O. Grant, 1987: Supercooled liquid water structure of a shallow orographic
cloud system in southern Utah. J. Climate Appl. Meteor., 26, 208–215.
Rauber, R., L. O. Grant, D. Feng, and J. B. Snider, 1986: The characteristics and distribution
of cloud water over the mountains of northern Colorado during wintertime storms. Part I:
Temporal variations. J. Climate Appl. Meteor., 25, 468–488.
Reinking, R. F., 1975: Formation of graupel. J. Appl. Meteor., 14, 745–754.
Reinking, R. F., 1979: The onset and early growth of snow crystals by accretion of droplets.
J. Atmos. Sci., 36, 870–881.
Reinking, R. F., J. B. Snider, and J. L. Coen, 2000: Influence of storm embedded orographic gravity
Reisner J., R. M. Rasmussen, and R. T. Bruintjes, 1998: Explicit forecasting of supercooled liquid
water in winter storms using the MM5 mesoscale model. Quart. J. Roy. Meteor. Soc., 124,
1071–1107.
Reynolds, D.W., and A.S. Dennis, 1986: A Review of the Sierra Cooperative Pilot Project. Bull.
Reynolds, D.W., and A. P. Kuciauskas, 1988: Remote and in situ observations of Sierra Nevada
winter mountain clouds: relationships between mesoscale structure, precipitation and liquid
water. J. Appl. Meteor., 27, 140–156.
Rhea, J. O., 1978: Orographic precipitation model for hydrometeorological use. Ph.D. dissertation,
Colorado State University, Atmospheric Science Paper 287, 198 pp.
Richard, E., N. Chaumerliac, J. F. Mahfouf and E. C. Nickerson, 1987: Numerical simulation
of orographic enhancement of rain with a mesoscale model. J. Climate Appl. Meteor., 26,
661–669.
Robichaud, A. J., and G. L. Austin, 1988: On the modelling of warm orographic rain by the seeder–
feeder mechanism. Q. J. R. Meteorol. Soc., 114, 967–988.
Roe, G. H., 2005: Orographic precipitation. Ann. Rev. Earth Planetary Sci. 33, 645–671.
Roe, G.H., and M. Baker, 2006: Microphysical and geometrical controls on the pattern of
orographic precipitation. J. Atmos. Sci., 63, 861–880
Roebber, P. J., S. L. Bruening, D. M. Schultz, and J. V. Cortinas, 2003: Improving snowfall
forecasting by diagnosing snow density. Wea. Forecasting, 18, 264–287.
Roebber, P. J., M. R. Butt, S. J. Reinke, and T. J. Grafenauer, 2007: Realtime forecasting of snowfall
using a neural network. Wea. Forecasting, 22, 676–684.
Rogers, R. R., and M. K. Yau, 1989: A Short Course in Cloud Physics. Pergamon, 293 pp.
Rosenfeld, D. and A. Givati, 2006: Evidence of orographic precipitation suppression by air
pollution–induced aerosols in the western United States. J. Appl. Meteor. Climatol., 45,
893–911.
Rotunno, R. and R. A. Houze, 2007: Lessons on orographic precipitation from the Mesoscale
Alpine Programme. Q. J. R. Meteorol. Soc., 133, 811–830.
Saleeby, S. M. and W. R. Cotton, 2008: A binned approach to cloud droplet riming implemented
in a bulk microphysics model. J. Appl. Meteor. Climatol., 47, 694–703.
Saleeby, S. M., W. R. Cotton, D. Lowenthal, R. D. Borys, and M. A. Wetzel, 2009: Influence of
cloud condensation nuclei on orographic snowfall. J. Appl. Meteor. & Clim., 48, 903–922.
Saleeby, S. M., W. R. Cotton, and J. D. Fuller, 2011: The cumulative impact of cloud droplet
nucleating aerosols on orographic snowfall in Colorado. J. Appl. Meteor. Climatol., 50,
604–625.
Sassen, K., A. W. Huggins, A. B. Long, J. B. Snider, and R. J. Meitin, 1990: Investigations
of a winter mountain storm in Utah. Part II: Mesoscale structure, supercooled liquid water
development, and precipitation processes. J. Atmos. Sci., 47, 1323–1349.
Sawyer J. S., 1956: The physical and dynamical problems of orographic rain. Weather, 11,
375–381.
Schultz, D.M., and Coauthors, 2002: Understanding Utah Winter Storms: The Intermountain
Precipitation Experiment. Bull. Amer. Meteor. Soc., 83, 189–210.
Sinclair, M.R, 1994: A diagnostic model for estimating orographic precipitation. J. Appl. Meteor.,
33, 1163–1175.
Sinclair M. R., D. S. Wratt, R. D. Henderson, and W. R. Gray, 1997: Factors affecting the
distribution and spillover of precipitation in the Southern Alps of New Zealand – A case study.
J. Appl. Meteor., 36, 428–442.
Smith, R. B., 1979: The influence of mountains on the atmosphere. Advances in Geophysics, 21,
Academic Press, 87–233.
Smith, R., 2003: A linear upslope time-delay model for orographic precipitation. J. Hydrol. 282,
2–9.
Smith R. B., I. Barstad and L. Bonneau, 2005: Orographic precipitation and Oregon’s climate
transition. J. Atmos. Sci., 62, 177–191.
Mountain winter storm cycle. Wea. Forecasting, 18, 1018–1036.
Steiner, M., O. Bousquet, R.A. Houze Jr., B.F. Smull, M., 2003: Air flow within major Alpine river
valleys under heavy rainfall. Q. J. R. Meteorol. Soc., 129, 411–431.
Stewart, R.E., 1992: Precipitation types in the transition region of winter storms. Bull. Amer.
Meteor. Soc., 73, 287–296.
Stewart, R.E. and G.M. McFarquhar, 1987: On the width and motion of a rain/snow boundary.
Water Res. Res., 23, 343–350.
Stewart, R.E., J.D. Marwitz, J.C. Pace, and R.E. Carbone, 1984: Characteristics through the
melting layer of stratiform clouds. J. Atmos. Sci., 41, 3227–3237.
Stickel, P. G., 1982: Observation of ice aggregation at temperatures near 50ı C. Proc. Conf. Cloud
Physics, Chicago, Amer. Meteor. Soc., 226–229.
Stoelinga, M., and Coauthors, 2003: Improvement of microphysical parameterization through
observational verification experiment. Bull. Amer. Meteor. Soc., 84, 1807–1826.
Storebo, P. B., 1976: Small scale topographic influences on precipitation. Tellus, 28, 45–59.
Theriault, J. and R.E. Stewart, 2007: On the effect of vertical air motion on winter precipitation
types. Natural Hazards and Earth System Sciences, 7, 231–242.
Theriault, J., R.E. Stewart and J.A. Mildebrandt, 2006: On the simulation of winter precipitation
types. J. Geoph. Res., 111, D18202.
Theriault, J., R.E. Stewart and W. Henson, 2008: The transition region along mountainsides. 13th
Mtn. Meteor. Conf., Whistler, Canada.
Thompson, G., R. M. Rasmussen, and K. Manning, 2004: Explicit forecasts of winter precipitation
using an improved bulk microphysics scheme. Part I: Description and sensitivity analysis. Mon.
Wea. Rev., 132, 519–542.
Thompson, G., P.R. Field, R. M. Rasmussen, and W.D. Hall, 2008: Explicit forecasts of winter
precipitation using an improved bulk microphysics scheme. Part II: Implementation of a new
snow parameterization. Mon. Wea. Rev. In press.
Wallace, J. M. and P. V. Hobbs, 2006: Atmospheric Science: An Introductory Survey. 2nd ed.
Academic Press, 483 pp.
Wang, P. K. and W. Ji, 2000: Collision efficiencies of ice crystals at low–intermediate Reynolds
numbers colliding with supercooled cloud droplets: A numerical study. J. Atmos. Sci., 57,
1001–1009.
Ware, E. C., D. M. Schultz, H. E. Brooks, P. J. Roebber, and S. L. Bruening, 2006: Improving
snowfall forecasting by accounting for the climatological variability of snow density. Wea.
Wetzel, M., and Coauthors, 2004: Mesoscale snowfall prediction and verification in mountainous
terrain. Wea. Forecasting, 19, 806–828.
Wexler, R., R.J. Reed and J. Honig, 1954: Atmospheric cooling by melting snow. Bull. Amer.
Meteor. Soc., 35, 48–51.
White, A. B., P. J. Neiman, F. M. Ralph, D. E. Kingsmill, and P. O. G. Persson, 2003:
Coastal orographic rainfall processes observed by radar during the California land-falling jets
experiment. J. Hydromet., 4, 264–282.
Woodcock, A. H., 1975: Anomalous orographic rains of Hawaii. Mon. Wea. Rev., 103, 334–343.
Woods, C.P., M.T. Stoelinga, J.D. Locatelli, and P.V. Hobbs, 2005: Microphysical processes and
synergistic interaction between frontal and orographic forcing of precipitation during the 13
December 2001 IMPROVE-2 event over the Oregon Cascades. J. Atmos. Sci., 62, 3493–3519.
Woods, C.P., M.T. Stoelinga, and J.D. Locatelli, 2007: The IMPROVE-1 storm of 1–2 February
2001. Part III: Sensitivity of a mesoscale model simulation to the representation of snow particle
types and testing of a bulk microphysical scheme with snow habit prediction. J. Atmos. Sci.,
64, 3927–3948.
Woods C. P., M. T. Stoelinga, J. D. Locatelli, and P. V. Hobbs, 2008a: Size spectra of snow particles
measured in wintertime precipitation in the Pacific Northwest. J. Atmos. Sci., 65, 189–205.
Woods, C. P., J. D. Locatelli, and M. T. Stoelinga, 2008b: The IMPROVE-1 storm of 1–2
February 2001. Part IV: Precipitation enhancement across the melting layer. J. Atmos. Sci., 65,
1087–1092.
Young, K. C., 1993: Microphysical processes in clouds, Oxford University Press, New York,
427 pp.
Yuter, S. E. and R. A. Houze Jr., 2003: Microphysical modes of precipitation growth determined by
S-band vertically pointing radar in orographic precipitation during MAP. Quart. J. Roy. Meteor.
Soc., 129, 455–476.
Zängl, G., 2004: Numerical simulations of the 12–13 August 2002 flooding event in eastern
Germany. Quart. J. Roy. Meteor. Soc., 130, 1921–1940.
Zängl, G., 2007. Small-scale variability of orographic precipitation in the Alps: Case-studies and
semi-idealized numerical simulations. Quart. J. Roy. Meteor. Soc., 133: 1701–1716.
Zängl, G., 2008: The temperature dependence of small-scale orographic precipitation enhance-
ment. Quart. J. Roy. Meteor. Soc., 134, 1167–1181.
Zikmunda, J., 1972: Fall velocities of spatial crystals and aggregates. J. Atmos. Sci., 29,
1511–1515.
Zikmunda, J., and G. Vali, 1972: Fall patterns and fall velocities of rimed crystals. J. Atmos. Sci.,
29, 1334–1347.
Chapter 8
Observational Techniques: Sampling
the Mountain Atmosphere
Robert M. Banta, C.M. Shun, Daniel C. Law, William Brown,

Roger F. Reinking, R. Michael Hardesty, Christoph J. Senff, W. Alan Brewer,
M.J. Post, and Lisa S. Darby
Abstract In this chapter several types of instruments are surveyed that have
contributed to increased understanding of the structure of atmospheric flows and
processes that drive these flows in complex terrain. After a brief example of the use
of in-situ measurements, those that are in contact with the air they are measuring,
a review of basic remote sensing principles serves as a background for a discussion
of several types of remote sensors. The review focuses on active remote sensing
systems, those that transmit and receive their own signal, including sodar, radar,
radar wind profilers, and lidar. Examples of the kind of data available from each
type of system allow an appreciation for how these systems contribute to advanced
understanding of atmospheric processes by providing data above the surface and
revealing the evolving vertical and horizontal structure of complex terrain flows. Ad-
vantages of deploying combinations of complementary instruments are described.
The roles of measurements and mesoscale numerical models are discussed in
a section where modeling studies have used high-resolution remote sensing and
R.M. Banta () • D.C. Law • R.F. Reinking • R.M. Hardesty • C.J. Senff • W.A. Brewer
• M.J. Post • L.S. Darby
NOAA Earth System Research Laboratory, 325 Broadway, Boulder, CO 80305, USA
e-mail: robert.banta@noaa.gov; Daniel.C.Law@noaa.gov; Roger.Reinking@noaa.gov;
Mike.Hardesty@noaa.gov; Christoph.Senff@noaa.gov; Alan.Brewer@noaa.gov;
M.J.Post@noaa.gov; lisa.darby@noaa.gov
C.M. Shun
Hong Kong Observatory, Kowloon, Hong Kong
e-mail: cmshun@hko.gov.hk
W. Brown
National Center for Atmospheric Research, Boulder, CO, USA
e-mail: wbrown@ucar.edu
C.J. Senff
Cooperative Institute for Research in the Environmental Sciences, Boulder, CO, USA
e-mail: Christoph.Senff@noaa.gov

410 R.M. Banta et al.
other advanced measurement systems to verify the models. Finally a brief section
describes turbulence measurement techniques by remote sensing that may be able
to provide profiles of turbulence quantities in mountain studies.
8.1 Introduction
Characterization of atmospheric structure and understanding of atmospheric pro-

cesses begin with observations and measurements. Obtaining measurement data
in mountainous terrain is an important but difficult challenge. In this chapter we
survey several measurement and analysis techniques that have proven valuable
in specifying the state of the atmosphere and studying atmospheric processes in
complex terrain.
Two broad categories of measurement system are “in-situ” and remote sensing.
In general, in-situ sensors are in direct contact with the fluid element or air “parcel”
being sampled. Remote-sensing systems use electromagnetic or acoustic (sound)
signals to probe atmospheric properties at some distance from the sensor. Human
touch (heat, cold, pressure, pain, etc.), taste, and smell could thus be regarded as in
situ, whereas sight and hearing exemplify remote sensing. In-situ sensors are often
tower mounted, but balloons, aircraft, kites, ships, automobiles or trucks, and other
conveyances have all been used to carry sensors to a region of the atmosphere to
be sampled. Remote sensors include the weather radar (and more recently Doppler
weather radar) made familiar by television news programs. Radiometers are the
most common instruments on satellites. Although we will not be discussing satellite
measurements in this chapter, ground-based and airborne radiometric and other
optical techniques can provide important data not only pertinent to atmospheric
radiation, but also indicative of atmospheric composition. Sodar has been an
important source of data above the Earth’s surface for decades, for such applications
as air quality, emergency response, and many others. More recently lidar and
Doppler lidar systems have exploited advances in laser and related technologies to
sense atmospheric particulates, wind fields, and concentrations of many atmospheric
gases, including water vapor and the pollutant ozone.
The purpose of atmospheric measurements in general is to provide information
on the state of the atmosphere in as much detail as required to address a need. This
information can be as simple as a temperature measurement in the morning to see
whether the daily forecast is on track, to decide about taking a coat to work, or an
animated sequence of weather radar scans to see whether precipitation is imminent.
Here we seek a more complete specification of the state of the atmosphere, which
might be useful for initializing the numerical weather prediction (NWP) models on
which the forecasts are based, or for understanding the meteorological processes that
have produced the current state (processes which the models should be simulating),
or for detailed studies to understand basic physical processes such as radiation,
microphysics, or turbulent exchange, where one of the goals could be improving
their representation in NWP models.
8 Observational Techniques: Sampling the Mountain Atmosphere 411
For these more exacting requirements, the goal of measurements is to determine

as quantitatively and completely as necessary the state of a region of the atmosphere.
Understanding how and why the atmosphere evolves requires knowledge of succes-
sive states of the atmospheric domain being studied. When the atmosphere in the
study domain is found to be steady, we wish to characterize the dominant balance
of processes maintaining the current state; when it is changing, we attempt to gain
insight into the imbalance of dominant processes that drive the evolution.
The process of inferring the atmospheric state from measurements involves
several steps. Instruments provide measured data, the data are analyzed, and the
analyses are interpreted into a conceptual picture of atmospheric behavior, which
can then be compared with existing analytical, numerical, or other conceptual
models. Sometimes the distinction between analysis and interpretation is blurred,
as when hand analyzing a surface pressure chart for isobars, pressure centers, and
fronts. Such analyses often subjectively consider conceptual models of cyclone and
anticyclone structure or large-scale terrain interactions during the analysis process.
At other times the analysis phase is objective, such as the application of statistical
techniques, and the results of these objective analyses must then be interpreted as a
separate step.
One of the key limitations of measurements is that the atmosphere is always
undersampled: there can never be enough measurements to completely specify the
mean and turbulent state of the atmosphere. This undersampling leads to ambiguity
in the interpretation. The goal of observational meteorology is to reduce this level
of ambiguity, ultimately to where the uncertainty is negligible with respect to the
requirements of the application for which the measurements are being made.
When considering sampling strategies for atmospheric systems, we can consider
two extremes, horizontally homogeneous and stationary. Horizontally homogeneous
atmospheres can be sampled using sensors and profiling instrumentation at one
fixed location. In this case the entire issue is the temporal evolution of the vertical
atmospheric structure, which by hypothesis can be measured at any location. The
important sampling objective is to obtain vertical profiles at fine enough time
resolution to be able to characterize changes to the atmospheric state. If we were
to hypothesize that the flow over the Great Plains of the central United States
was horizontally homogeneous, for example, the routine 0000 and 1200 UTC
rawinsonde profiles are insufficient to describe the diurnal patterns. Hourly profile
data may suffice for much of the daily cycle, but during the dawn and dusk
transitions even finer time resolution would be required to characterize the evolution
of the meteorological fields. At the other extreme are stationary atmospheric systems
or processes, such as some mountain waves, for which characterizing the spatial
structure is the objective. This requires sensors at many locations, mobile sensing
platforms, remote-sensing measurements capable of areal or volume scanning, or
combinations of these capabilities. Here the important sampling objectives are to
specify as completely as possible the spatial variability of the flow. For a purely
stationary system the observations do not need to be simultaneous.
The mountain atmosphere, of course, requires sampling of both the temporal
evolution and the spatial variability of flow systems, so it is necessary to exploit
09/15/03 + lidar + sodar

500
400
300
z (m)
200
100
0
1:30 2:30 3:30 4:30 5:30 6:30 7:30 8:30 9:30
Time (UTC)
Fig. 8.1 Ten-minute averaged profiles of high-resolution Doppler lidar wind speed (blue) and
Doppler sodar wind speed (red). Red dots show all available sodar data, and red pluses indicate
data having a high calculated confidence factor, which involves SNR, continuity, and other values.
Range of wind speeds within each time interval is 5–20 m s1 (Pichugina et al. 2008. © American
both measurement approaches, i.e., a mix of fixed sensors to monitor the temporal
changes (at many locations to also sample the spatial variability), with mobile
platforms to fill in the spatial gaps; additionally, remote sensing systems provide
broad coverage in time and space, and they provide data in inaccessible places.
Each type of instrument has its own set of strengths and limitations, so it is always
advisable to use arrays of as many different kinds of sampling system as possible,
although, of course, practical considerations become part of the decision-making
process; budgetary constraints, for example, force compromises that are generally a
necessary aspect of program planning.
Comparing measurements from two instruments that nominally measure the
same quantity can provide insight into the relative strengths and limitations of the
different measurement systems. Figure 8.1 shows hourly profiles of wind speed
during nocturnal low-level jet (LLJ) conditions at a site in southeastern Colorado,
in the U.S. Great Plains (Pichugina et al. 2008). The blue profiles came from
analysis of Doppler lidar scan data, and the red asterisks are from Doppler sodar.
The agreement below 150–200 m is good, but above the nose of the jet, differences
are apparent. Figure 8.2 also shows comparisons of wind profiles, both direction
and speed, from Doppler lidar scan data and radar wind profiler, this time aboard
a ship; differences are evident below 500 m (Wolfe et al. 2007). In both cases, the
differences in the profiles can be attributed to how each instrument samples the
atmosphere, and we will return to these sampling differences in later sections.
In this chapter we review several different sampling systems commonly used
to characterize the atmospheric state, including strengths and limitations of each
sampling method. Traditionally, tower-mounted, balloon-borne, and even aircraft-
borne in-situ sensors have been important sources of information on the mountain
atmosphere, but in recent decades, three types of remote sensor—radar, sodar, and
lidar—have led to significant advances in our understanding of mountain processes.
These instruments have been especially useful in providing information above the
Fig. 8.2 Wind-speed and direction profiles taken on shipboard in the Gulf of Maine. Black
pluses represent high-resolution Doppler lidar VAD profiles. Blue and red symbols are for radar
wind profiler data at 60- and 100-m resolution, as indicated (Wolfe et al. 2007. © American
Meteorological Society. Reprinted with permission). Rawinsonde profiles were also available for
some of the shipboard comparisons
earth’s surface, a traditionally difficult and expensive region to obtain atmospheric

data. The emphasis in this chapter will be on contributions from these three types
of instrument, but roles of other kinds of sensor, such as in-situ measurements and
radiometry, will also be briefly discussed.
The chapter is organized as follows. First we present an example of how in-situ
measurements including air chemistry can be used to characterize mountain flows.
In a later section, near the end of the chapter, this information will be supplemented
by remote sensing data to show how the interpretation can be enhanced by including
more comprehensive measurements. Second, a review of some basic remote-
sensing principles serves as a background for the discussions of the three types
of instrument. The following sections then describe the instrument types. Sodar, the
one type of instrument discussed that uses a sound pulse, has been a useful profiling
device for many decades. It is typically not scanned. We note here that the terms
radar and lidar can be viewed as categories, each including several different kinds
of instrument depending on the wavelengths and techniques employed to probe
the atmosphere. In these two sections, we give examples of several types of radar
or lidar, including the capabilities and how these capabilities have been used to
learn about atmospheric properties and processes. One type of radar, the radar wind
profiler, is employed differently enough from other types of radar that it is described
in its own brief section.
The descriptions of the types of remote-sensing system thus include a section
each on sodar, radar, radar wind profiler, and lidar, emphasizing how such sensors
have been employed to reveal atmospheric structure. These discussions are aimed
at an appreciation of some of the differences in the data obtainable by the
different sensors, which can affect the interpretation of the data analyses from these
instruments.
Data from different instruments often complement each other, and examples
where two instruments have been used together to take advantage of the strengths of
each and compensate for their limitations will be given. Examples are also included
where multiple instruments have been used in field campaigns to characterize
meteorological processes in mountainous settings, and the interpretation of these
processes takes advantage of the strengths of each sampling system. This summary
follows many reviews and books of the progress of measurement systems, for
example Lenschow (1986), Schwiesow (1986), Atlas (1990), Neff (1994), Wilczak
et al. (1996), Weitkamp (2005), Emeis et al. (2007), and in the last Book on Complex
Terrain, Neff (1990).
Two final sections address verification of mesoscale NWP models using high-
resolution remote sensing measurements, and examples of some remote-sensing
techniques that have been developed to measure turbulence profiles.
8.2 In-Situ Sensors
We begin with an example of the use of in-situ observations to study air flows
in mountainous terrain, to illustrate how a mixture of in-situ measurements can
give insight into the interpretation of relevant atmospheric processes. The example
is from the east–west Fraser River Valley 30 km east of Vancouver, British
Columbia, Canada, and 40 km from the coast. Air pollution transport processes
in this valley were also discussed in Chap. 5. The mountain range to the north of the
Fraser Valley is divided by numerous north–south valleys containing long, narrow
lakes, one of which—the valley of Pitt Lake—lies just to the north-northeast of a
surface measurement site located at H in Fig. 8.3. The case involves the diurnal
inflow-outflow cycle of polluted air from the main valley, passing into the Pitt Lake
valley during daytime and back out at night. In a later section we will supplement
the in-situ measurements described here with remote sensing measurements from
the same case, to further explore the important meteorological processes.
In-situ measurements taken just outside the mouth of Pitt Lake valley, are given
in Fig. 8.4, which shows time series of wind speed and direction sampled at 4 m
above ground at 15-min intervals, during evening transitions from mid afternoon
to 2 h after midnight. Shown are afternoon westerly winds shifting to stiff northerly
Fig. 8.3 Map showing Fraser River Valley east of Vancouver. Mountains north of the Fraser
Valley, including the Pitt Lake region, are shown by the 1,000 m contours. Location of Doppler
lidar indicated by L and surface chemistry and meteorological sensor site is at H (Banta et al.
1997a). M and N mark the southern end of the two mountain ridge lines that enclose the Pitt Lake
Valley. These show up as prominent terrain hits in Doppler lidar scans shown later in Sect. 8.8
flow between 6 and 7 p.m. (1800–1900 PST). The northerly flow lasts for 2–3 h, and
then becomes light and variable. Figure 8.3 shows the geographic context of these
measurements, which were taken at H. From the topographical setting we can infer
that the westerly afternoon winds are a combination of thermally forced upvalley
and sea-breeze flows along the Fraser Valley, probably modified by a component
of the larger-scale ambient winds. The northerly flow is consistent with an early-
evening, cool-air outflow (or downvalley flow as described in Chap. 2) from the Pitt
Lake Valley, whose slopes have gone into shadow during the late afternoon hours.
Measurements of trace atmospheric chemical species, mostly pollutants, can
significantly enhance insight into which meteorological processes are dominant. The
solid line in the top panel of Fig. 8.5 shows the behavior of ozone (O3 ) during the
evening transition of 3–4 August (as seen in Fig. 8.4, top panels). The northerly
flow after 1800 PST has lower O3 concentrations than the westerly flow it replaced,
and thus the northerly flow seems in some sense “cleaner” than the air in the Fraser
Fig. 8.4 Surface wind

measurements from site
marked H in Fig. 8.3. Shading
indicates periods of northerly
flow. Abscissa indicates hour
Pacific Standard Time (PST)
(Banta et al. 1997a)
Valley, which is blowing in from the direction of the Vancouver urban area. One
possibility is that this cleaner air could be pristine air from the Pitt Lake Valley.
In-situ measurements of other chemical species provide further insight into the
origin of the lower-O3 air mass. The upper panels of Fig. 8.5 show the O3 traces
for the nights of 3–4 and 4–5 August (solid lines) along with nitric oxide (NO), a
reactive gas emitted during combustion processes (including motor-vehicle traffic),
and the ratio NOX /NOY , which is a measure of the “photochemical age” of the
air. NOX , the sum of NO C NO2 , represents the concentration of freshly emitted
nitrogen oxides, and NOY , the sum of all nitrogen oxide species, including both
the freshly emitted (NOX ) species and those product species subsequently formed
by chemical reaction. In a freshly polluted air mass, most of the NOY consists of
NOX , so the ratio NOX /NOY is near unity. In air with aged pollutants, some of the
NOX has reacted to form product species, and therefore the ratio drops below 0.8
for an aged sample. NOY values in the northerly flow suggest that this is actually
polluted air despite the low O3 concentrations, and that the pollution in this air is
aged. Information provided by the carbon monoxide (CO) data (X’s in the lower
panels of Fig. 8.5) show CO concentrations exceeding 200 ppb even in the northerly
low-O3 air. This observation confirms that the northerly flow represents polluted
air—and not pristine air—coming from the direction of Pitt Lake.
Fig. 8.5 Left axis: mixing ratios of O3 (solid line) and NO (circles, multiplied by 10) at the site
marked H in Fig. 8.3 for the same times (PST) as in Fig. 8.4. Shaded bar indicates period of
northerly flow at site (Banta et al. 1997a)
This detailed example illustrates how a mixture of in-situ measurements—

here both meteorological and chemical—can give insight into the interpretation of
relevant atmospheric processes and the origins and fate of air masses. But questions
remain. All of these measurements were at one surface site, located 10 km outside
the valley where the northerly flow was presumed to have originated. Can one be
certain that this northerly flow actually came from the valley? And if so, does the
light-and-variable flow starting at 2100–2200 PST mean that the outflow only lasted
Table 8.1 In-situ measurements

Advantages Limitations
Excellent time series, temporal coverage Coverage limited to one point, or a line
possible (by advection)
Instrumentation often simpler Siting issues
Many measurables can be collocated Representativeness of measurement
Accessibility to instrumentation Interference by platform (including tower)
can be a concern
Procedures familiar –
Interpretation straightforward –
2–3 h each night? These questions will be further addressed in Sect. 8.8 of this
chapter using remote-sensing data.
Tower-mounted in-situ sensors give a picture of the local change (@/@t) of
measured quantities at a fixed height, which is important for documenting changes
in these quantities and the evolution of atmospheric processes. In-situ sensors are
also mounted on moving platforms to measure spatial structure. Well known to
meteorologists is the use of balloons to carry sensors aloft to profile the atmosphere.
An example from the previous case study is shown in Fig. 8.6, a tethered ozonesonde
launched from a small island in the middle of Pitt Lake. The wind profile from this
sounding shows a northerly (down-valley) 5 m s1 jet at 100 m above the surface,
with light-and-variable flow above. Below the jet maximum, the O3 profile shows
O3 decreasing to a minimum value at the surface. Taken with the surface in-situ
measurements, the sounding data suggest a picture where polluted air flows into the
valley during daytime (McKendry et al. 1997, 1998). At night this air flows back out,
but O3 and some other pollutants have been removed, especially near the surface.
The discussion in Sect. 8.8 of this chapter will further explore this hypothesis.
Overall, in-situ instrumentation mounted on towers is often the most inexpensive
way to obtain meteorological data. Arrays of such sites can be deployed to cover an
area of interest and document the spatial variability of the evolving surface flows. An
area where in-situ measurements are especially important is in the measurement of
precipitation and snowpack for hydrological applications. Blockage of radar signals
by terrain limits the ability to use radar to estimate precipitation intensity, as further
discussed in Sect. 8.6 of this chapter, so ground-based in-situ data, such as from rain
gauges, snow boards, and snow courses, are critical to predict runoff and flooding
potential in mountainous areas.
A comprehensive discussion of many factors involved in making accurate tower
measurements is presented in Kaimal (1986). Despite some limitations, such
as representativeness or the need for routine maintenance, these are in general
measurements that meteorologists have confidence in, so an important role for in
situ data is as a reference or “ground truth” for ground-based, airborne, and satellite
remote-sensing data as well as for NWP output. Some advantages and limitations of
in-situ sensing are listed in Table 8.1.
Fig. 8.6 Profiles of O3 , potential temperature, wind speed, and wind direction from a tethersonde
launched from a small island in Pitt Lake at 0124 PST (Banta et al. 1997a)
8.3 Remote Sensing Concepts
Before describing each type of instrument, we review some basic remote sensing
principles that apply to all of the remote sensing systems described.
8.3.1 Categories of Remote Sensor
The field of remote-sensing instrumentation, including but not limited to radar,

sodar, and lidar, has been divided into two broad categories, passive and active.
Passive sensors can be thought of as devices that “listen” for electromagnetic or
acoustic signals emitted by the environment or by other sources outside of the
instrument itself. The human remote senses, vision and hearing, are passive.
Perhaps the most common atmospheric passive instruments are radiometers.
They take advantage of two properties: matter emits electromagnetic radiation with
a wavelength œ dependence characteristic of its temperature and other properties,
such as composition, and gas molecules absorb (and re-emit) radiation in wave-
length bands characteristic of each chemical species. Some types of radiometer
sense wavelengths meant to characterize a source or source region, e.g., certain
radiometers on satellites and aircraft are used to sense properties of the Earth
(including ocean) surface. Microwave radiometers detect liquid water in clouds
and provide a path- or column-integrated value for mass of liquid-water (Hogg
et al. 1983; see Sect. 8.8 for an example). Temperature-profiling radiometers detect
multiple œ’s, each characteristic of a certain atmospheric layer (Westwater et al.
1990, Westwater 1997). Other types of radiometer obtain data from the intervening
atmosphere by detecting radiation from a well characterized source. A strong white
light source can be reflected off reflectors (mirrors) several kilometers away and
received back at a detector to sense the path-integrated absorption by a number
of atmospheric pollutants and other gases, using a technique called Differential
Optical Absorption Spectroscopy (DOAS) (e.g., Stutz et al. 2004). Solar Differential
Optical Absorption Spectroscopy (e.g., MAX-DOAS: Volkamer et al. 2009) and
Solar Occultation Flux (SOF: Mellqvist et al. 2010) use the sun as a source to
determine the absorption of natural and anthropogenic atmospheric gases. Another
optical technique, the Atmospheric Emitted Radiance Interferometer (AERI), has
demonstrated the capability to profile temperature, humidity, and other quantities
(e.g., Feltz et al. 2003). A technique for using global-positioning system (GPS)
signal has been successfully used to measure column-integrated water-vapor in
the atmosphere (Wolfe and Gutman 2000; Birkenheuer and Gutman 2005; Smith
et al. 2007). Thus, a variety of absorption-based techniques have been developed to
measure atmospheric temperature and trace constituents.
In active remote sensing the source of the signal is the instrument itself. As a
further stretch of the human analogy, one could regard the X-ray vision of some
fictional “superheroes” as an example of active remote sensing, assuming that the
source of the X-rays is within the superhero himself. As a more real-life example,
one could consider a hiker with a flashlight in the dark. If one thinks of the hiker-
flashlight as part of one “system,” the flashlight illuminates objects along the path.
If the hiker encounters smoke or fog, they are illuminated by the beam and detected
by the hiker’s vision, thus providing information about the atmosphere between the
hiker and the objects. A natural example of active remote sensing is echo-location
by flying bats.
Two categories of active remote sensor are continuous-wave (CW) and pulsed.
In both cases, the basic transmitted electromagnetic or acoustic signal s(t) is a wave
represented by a sine function. The three basic specifications for a sine function are
its frequency f, its amplitude A, and its phase ®.
s .t/ D A sin .2f t C ®0 / (8.1)
Frequency and wavelength are interchangeable, related by the speed of propa-

gation of the wave according to the relationship: c D œf or œ D c/f, where c is the
speed of light or sound, as appropriate. The phase ® is the position the wave along
the abscissa, whether at maximum positive, zero, maximum negative, or somewhere
in between, and ®0 is its initial position or displacement.
CW systems, consistent with the name, emit a continuous signal. For CW lidars
the received signal generally represents returns from a limited range of focus in
the atmosphere as specified by a range-dependent weighting function. The signal
varies as the atmosphere advects through the beam or as the pointing angle of
the instrument is changed. Other than knowledge that the atmospheric return is
maximized at a given distance from the instrument as described by the peak in
the weighting function, no explicit range information is obtained by this kind of
system. Range information can be obtained by mechanically changing the distance
at which the system is focused. Another approach used for CW radar to obtain
range information is to systematically vary in time, or “modulate,” the frequency of
the signal. Vertically pointing Frequency-Modulated, Continuous-Wave (FM/CW)
radars have produced spectacular images of the structure of the daytime and
nighttime boundary layer (BL) using this technique (see Sect. 8.6 of this chapter).
8.3.2 Pulsed Remote Sensors
Being able to determine atmospheric return signal as a function of range is

a tremendous advantage in studying the atmosphere, because of the ability to
document spatial structure, among many other reasons. A common approach is
to transmit a short pulse of energy as illustrated in Fig. 8.7 and then detect the
echo signals returning from the scattering target region, as a function of time delay.
This is the operating principle of pulsed remote sensors, including radar, sodar,
and lidar. The “—dar” in each system is an abbreviation for “detection and
ranging” (RAdio Detection And Ranging; SOund : : : ; LIght : : : ), implying that
the instrument not only detects the existence of a target, but also determines the
distance (range) to the target (Schwiesow 1986a). The distance d is calculated from
the time delay t, by d D ct, where c is the speed of propagation of the signal—either
the speed of light (3 108 m s1 ) or the speed of sound (3 102 m s1 ).
The signal pulse has to travel relatively great distances, scatter off miniscule
targets, then return to the instrument as a very faint echo. For the return to
be detectable, significant energy must be transmitted. For typical active remote-
sensing systems, the transmitter and receiver are the same hardware—the so-called
Fig. 8.7 Schematic of transmitted remote-sensing pulse. Horizontal axis is time, and within the
pulse, the wavelength of the transmitted signal is seen. Pulse length is variously defined as the time
interval from ½ or 1/e of the maximum amplitude of the pulse, here normalized to be ˙1
“monostatic” configuration. In this case the transmitter produces an “explosion”

of energy, comparatively speaking, and then a tiny fraction of a second later the
system must be able to listen for a small whisper of a return. But it takes time for
the hardware to settle down after a transmission, a time during which the instrument
cannot detect its signal. This time corresponds to a distance, a minimum range (aka
“dead zone”), which is an important system characteristic, especially for vertically
pointing systems (Fig. 8.8).
Ideally the transmitted pulse would turn on and off abruptly as a square pulse or
packet, but in reality a pulse has a shape more resembling Fig. 8.7, which also shows
the wavelength of the transmitted signal within the pulse. The pulse length is defined
to be the distance between the points where the signal amplitude drops to some
fraction of the peak power, such as ½ or 1/e. The pulse is sampled by accumulating
backscattered energy in discrete time bins after pulse transmission, or range gates.
The spatial, along-beam resolution of the measurement is determined by the pulse
length, as will be described in the next subsection. The three-dimensional volume
defined as the cross-beam, cross-sectional area of the pulse times the pulse length is
the pulse volume. Thus, an essential distinction between in-situ and remote sensing
is that in-situ sensors provide a point measurement, whereas a remote-sensing
measurement is distributed over a pulse volume, which is then averaged within each
range gate.
Fig. 8.8 (a) Vertical-slice Doppler-lidar scan taken at Colorado Springs airport, with mountains
to the west (right) and flow from right to left. Lidar is at (0,0), surrounded by a black semicircular
“dead zone,” which represents the minimum range detectable by this instrument. (b) Shallow 1ı
conical scan taken near Colorado Springs, Colorado, 1 April 1997; north is up, and west to the left.
Color scale indicates lidar radial velocities in m s1 . Lidar is located at the center of the plot, and
minimum range indicated by black circle surrounding lidar in center of plot. (c) 915 MHz profiler
wind profiles in the Great Salt Lake Valley, showing no wind-barb data in the “dead zone” below
120 m (shaded region). Wind barbs in knots
Part of the job of a remote sensing instrument system is to form and transmit the
outgoing signal. Properties of this transmission determine the kind of information
that can be obtained, including the strength of the signal, the duration of the signal
within a pulse, the angular width of the pulse, how often pulses are transmitted, and
the outgoing wavelength. These properties determine the maximum range of the
measurement, the spatial and temporal resolution of the data, and the nature of the
backscattered signal.
8.3.3 Characteristics of the Transmitted Signal
Important system characteristics of a remote sensing instrument are the transmitted

power, the pulse-repetition frequency (PRF), the pulse length, the beam width, and
its transmitted wavelength œ.
8.3.3.1 Basic System Properties
Increases in transmitted power improve system performance, most obviously max-

imum range. The tradeoffs are that more powerful transmitters are generally larger
and more expensive, and require more expensive supporting hardware. For example,
transmitting more power generates more heat, which requires more expensive heat
dissipation techniques and air conditioning.
The pulse repetition frequency (PRF) is the number of pulses or “shots”
generated per unit time. As will be discussed in Sect. 8.3.5 of this chapter, the returns
from individual shots tend to be noisy, so pulses are generally averaged together to
improve range and precision of the data. High pulse rates are an advantage for being
able to obtain accurate averaged data at short enough time intervals to be useful for
probing the atmosphere, but note that PRF cannot be increased arbitrarily without
consequence. If a new pulse is transmitted before the previous pulse has cleared
beyond the detectable range, backscatter from the previous pulse can show up as if
transmitted by the new pulse, leading to range ambiguities. The unambiguous range
for an active remote sensor is the distance corresponding to half the time between
pulses.
The pulse length, illustrated in Fig. 8.7, is the duration of the pulse times the
propagation speed. The range resolution of the measurement is determined by the
pulse length h. Two atmospheric targets separated by more than half the pulse length
(h/2) can be seen as separate entities by the instrument. The factor of ½ comes from
the geometric argument that the backscattered pulse folds back over itself in such
a way that when the trailing edge of the pulse is passing over a target, the return
signal from the leading edge will have originated from a distance h/2 farther out in
range, so that targets farther away than this will be seen as separate from the first
target. Note that in general, range-gate spacing, which may be adjustable for some
instruments, does not necessarily equate to range resolution. If range gates were
set smaller than the half-pulse length they would still not resolve (see as separate)
targets closer together in range than this. Often range gates will be selected to be
larger than the half-pulse length to average over more incoming signal and improve
the signal detectability, as will be described below, by sacrificing range resolution.
Thus, only for range-gates larger than h/2, the sampling volume and the range
resolution are determined by the range-gate spacing.
The angular width of the beam is important as a determinant of transverse
resolution in a scanning system, because it determines the transverse dimension
of the scattering volume in space over which the signal is averaged. One of the
uncertainties in a velocity measurement, for example, is the turbulent variability
of the atmospheric velocity fluctuations. A larger sampling volume averages over
more of such fluctuations, providing a better estimate of the mean Doppler velocity
in that volume. This is good for the mean estimates, but may be a limitation if one is
attempting to characterize the turbulent fluctuations or other atmospheric variations
that may be about the same size as or smaller than the sampling volume. For a given
œ the hardware that most affects the beam width is the antenna, so some aspects of
antennae will be discussed in a later section.
8.3.3.2 Wavelength œ
The wavelength is critical because it determines the nature of how the signal inter-
acts with the atmosphere, and thus precisely what the instrument “sees.” Possible
scattering targets for atmospheric remote sensors include rain drops, cloud droplets,
dust or aerosol particles, and atmospheric molecules. A storm radar transmitting
at œ D 10 cm would receive sufficient backscatter from raindrops to detect them,
but would not receive enough reflectivity from cloud droplets. Similarly, a cloud
radar operating at 8 mm could detect cloud droplets and rain drops, but not aerosol;
an infrared lidar transmitting at 10 m would detect aerosol particles but would not
detect molecular backscatter, and a lidar operating at 532 nm would detect backscat-
ter from both atmospheric molecules and aerosol particles. The scattering properties
of an instrument’s transmitted œ determine what the instrument can detect.
Remote sensing involves many tradeoffs, and one of them is in the selection of
œ’s. Shorter œ’s detect smaller particles as just illustrated, but the shorter-œ photons
have greater energy and greater interaction with the intervening medium, and thus
greater attenuation and less range, other factors (such as transmitted power) being
equal. In other words, in general, shorter-œ systems do not see out as far. Larger œ’s
detect larger particles or drops, and can detect signal at greater ranges, as a result of
lower-energy photons and reduced environmental interaction.
Equation for received power. For quantitative applications using radar or lidar
backscatter the relationship between the received power and the transmitted power
and other factors is expressed as the “radar equation” or the “lidar equation.”
A simplified form of the radar equation could be written:
Pr D CR 0 `i =r2 (8.2)

where Pr is the received power, CR 0 is a “radar constant” consisting mostly of

instrument-system properties (such as œ, transmitted power, beam cross-sectional
area, antenna gain), ` is a parameter between 0 and 1 representing attenuation losses
during the two-way propagation of the signal through the atmosphere, ¢ i is the cross-
sectional area of the individual scatterer as seen by the receiver, and r is the range
from the instrument to the target (after Rinehart 2010). A similar equation is used
to calculate lidar backscatter.
Scattering by droplets, particles, or molecules. One of the important aspects of
the interaction between a signal and the environment is the relationship between
œ, target size, and scattering properties. When radiation from an electromagnetic
signal pulse is intercepted by an individual point target, radiation is scattered out
of the direction of propagation of the radiation stream. The intercepted energy
is transferred to the target, which re-transmits the energy in all directions, but
not equally in all directions. For ordinary scattering the re-emission is at the
same œ as the incident signal (Battan 1973), referred to as elastic scattering
(Hardesty and Darby 2005). Other types of scattering will be described in Sect. 8.7
of this chapter. Of interest for monostatic remote-sensing systems is the energy
scattered back toward the receiver at an angle of 180ı from the direction of the
outgoing transmitted signal. It is this backscattered radiation that is detected by the
receiver.
One of the factors in (8.2) is the cross-sectional area intercepted by the targets
along the beam. Meteorological targets for radar (cloud water, precipitation, ice
particles) and lidar (aerosol and molecules) are not individual targets, but are
distributed through the pulse volume. The backscatter seen by the instrument is
the summation of the backscatter from all the individual targets encountered by the
signal pulse as it propagates outward from the transmitter.
Scattering is a complex function of target size, density, shape, and chemical
composition, so no simple relationship exists between actual backscattered radiation
and the geometric (physical) cross sectional area of targets. For convenience,
therefore, and presumably to provide some physical insight into the backscatter
problem, researchers have defined a fictitious cross-sectional area ¢ i , called the
backscatter cross section, defined in terms of the actual amount of energy scattered
back toward the sensor by the targets (e.g., Battan 1973). It is convenient at this point
to consider an individual scattering entity in a sampling volume within a range gate.
The subscript i is to indicate the cross section of an individual target, and later it
will refer to the ith target in a pulse volume. The area ¢ i is defined to be the circular
cross-sectional area a2 of a sphere of radius a that would scatter the same amount
of energy back toward the receiver as that actually backscattered by the real target—
if the scattering by the fictitious spherical target were the same in all directions, i.e.,
if the scattering were isotropic. As pointed out by Battan (1973), no atmospheric
targets are truly isotropic, even spherical ones, so strictly speaking ¢ i does not
represent anything physical, but it is useful because it does represent a quantity,
in units of area as required by the radar or lidar equation, that is proportional to the
amount of signal backscattered toward the sensor, for a given transmitted œ.
For some types of scattering targets, a straightforward relationship can be defined

between the backscatter cross section and physical variables. When the target
diameter is much smaller than the transmitted œ (Fig. 8.9, left inset diagram) the
relevant scattering process, referred to as Rayleigh scattering, can be expressed as
the relatively simple relationship:
ˇ ˇ ı
i D . 5 ˇK2 ˇ D6 /
4 (8.3)
where jK2 j is related to the composition-dependent index of refraction and D (D2a)

is the target diameter (Battan 1973). For a given target diameter, scattering is
inversely proportional to the fourth power of œ, meaning that shorter œ’s are much
more efficiently scattered than longer œ’s. In choosing a radar œ to investigate clouds
with tiny droplets (10–100 m), for example, shorter-œ radio waves of 1 cm or
less will be most effective; to use molecular backscatter, lidar with shorter œ’s in
the ultraviolet will be considerably more effective than longer infrared œ’s. On the
other hand, for a given instrument, where œ is then specified, backscattered energy is
proportional to the sixth power of target diameter. For a population of scatterers this
means that backscatter in the Rayleigh scattering regime is often dominated by a few
large targets in the scattering volume, and that, as the smaller targets become tinier,
the amount of energy backscattered rapidly becomes irrelevant even if they were
present in large numbers. For example, backscatter from atmospheric molecules is
negligible at longer infrared and microwave œ’s.
For targets whose diameter is about the same as or somewhat larger than the œ
(Fig. 8.9, right inset), the scattering interactions become more complex. Figure 8.9
shows the backscattered energy, expressed as ¢ i , for a sphere of radius a, normalized
by its geometric cross-sectional area. This nondimensional scattering quantity is
plotted as a function of nondimensional radius (expressed as circumference 2a)
normalized by œ. With a logarithmic scale along the ordinate, the simpler Rayleigh
regime appears linear for normalized diameters (circumferences) of less than 1, but
as diameters become greater than 1, the scattering cross section ratio becomes much
more complex, oscillating about a value of 1. Even though the “Mie-scattering”
equations describing the curve in Fig. 8.9 are valid for the entire range of the
abscissa, the linear portion of the curve for circumferences less than 1 is called the
Rayleigh regime, as we have already been doing, and the more complex interactions
for circumferences greater than one is often referred to as the Mie regime.
Because the meteorological targets for remote sensors are distributed through
the pulse volume, the concept of scattering cross section actually refers to the
summation of the scattering properties of the entire population of distributed targets
in the pulse volume—the number, size, and scattering effectiveness of the targets,
or ¢ D † ¢ i . It is necessary to adopt a beam-filling assumption, that the distributed
scatterers uniformly fill the scattering volume. We note that although strictly ¢ is
a fictitious quantity, calculations using actual size distributions can be reasonable
estimates of measured backscatter Pr . For example, Rinehart (2010) presents a
table (reproduced here as Table 8.2) with a hypothetical droplet size distribution
sorted into bins of 5 m width to illustrate the calculation procedure for the radar
0
log (σi • πa2)
Rayleigh regime
Mie regime
–1
–2
Rayleigh Mie
–3
–1 0 1
log c • a/λ
Fig. 8.9 Schematic representation of backscatter cross section of a spherical target normalized by
its circular cross sectional area, plotted as a function of radius of equivalent sphere normalized by
wavelength of signal (After Rinehart 2010 and Skolnik 1980). Inset: schematic representation of
relationship between particle size and signal wavelength for Rayleigh and Mie scattering regimes
reflectivity Z, represented as Z D ˙Di 6 . In each size bin it is assumed that the bin
is narrow enough that each scatterer in it can be assigned a diameter equal to the
midpoint of the bin without significant error, so the contribution of the reflectivity
in each bin can be represented as Zn D †Nj Dj 6 , where n represents the nth bin and
the summation over j represents all the droplets in the size range of the nth bin.
The contribution for each bin is summed over all n bins to arrive at a theoretical
reflectivity, as illustrated in Table 8.2. In cases where drop-size distributions have
been measured by aircraft, calculations such as this hypothetical one have yielded
reasonable values of radar reflectivity as measured by ground-based radars. Thus,
even though ¢ is not actually a physical cross-sectional area, for some applications
it can yield a reasonable estimate.
Clear-air signal: Bragg scattering. Another type of scattering that is important
for remote sensing occurs when a signal encounters sudden changes in the index
of refraction along its propagation path. For electromagnetic signals, index of
refraction variations that can produce such scattering generally consist mostly of
moisture gradients with a lesser contribution from temperature gradients, whereas
for sound signals, the major contribution to backscattering is from temperature
gradients, as further discussed in Sect. 8.4 of this chapter. Gradients strong enough
to induce backscatter can be produced by mean along-beam variations, as when air
masses change with height, but more commonly the primary scattering targets are
Table 8.2 Radar reflectivity

Diameter (m) Number/cm3 N D6 (mm6 m3 )
factor for small cumulus
cloud (From Rinehart 2010) 5 100 1.56 106
10 100 1.00 104
15 50 5.69 104
20 25 1.60 103
25 10 2.44 103
30 5 9.19 103
35 1 4.01 103
Total D 1.80 102
D>17.4 dBZ
turbulent fluctuations of moisture or temperature in the atmosphere. This type of

interaction with turbulence has been referred to as Bragg scattering (Gossard 1990).
According to the theory and supported by measurements (Gossard 1990), peak
Bragg backscatter comes from turbulent fluctuations of half the wavelength of the
signal (œ/2), so for a 10-cm radar, backscatter would come mostly from turbulent
eddy sizes near 5 cm; for a 4,500 Hz sodar (œ D 6.7 cm), from temperature fluc-
tuations near 3.3 cm; or from a 915 MHz radar wind profiler (œ D 32.8 cm), from
turbulent fluctuations of moisture with dimensions of 16.4 cm. Under conditions
when the scattering œ lies within a portion of the turbulence spectrum called the
inertial subrange—conditions that are often met in the atmosphere (see Sect. 8.10
for further discussion)—the turbulent backscatter is proportional to a quantity called
the structure function, which is a measure of the mean square difference between
the values of a variable at two locations in the atmosphere separated by a distance •.
For example for temperature T, the temperature structure function is defined as
CT 2 D < ŒT .x C ı/ T .x/ 2 > =ı 2=3 (Tatarskii 1971; Neff and Coulter 1986).
The structure function for electromagnetic signal CN 2 , which involves the humidity
and temperature structure functions and their covariance, is more complicated
(Chadwick and Gossard 1986). For such calculations the separation distance • is
taken to be equivalent to œ/2 of the instrument being used, the eddy size producing
maximum Bragg backscatter. This formula allows backscatter from a remote sensing
instrument to be directly compared with in-situ turbulence measurements measured
on a tower, and it further allows the backscatter to be expressed in terms of
meteorological turbulence variables.
For most atmospheric applications, therefore, the existence of backscattered
signal from this mechanism, which is essential to sodar and profiler operation, is
an indicator of the presence of sufficient finescale turbulence in the atmospheric
region being probed. For Doppler applications, it is assumed that these turbulent
fluctuations are carried by the mean wind, both horizontally and vertically.
Attenuation. The equation for received backscatter Pr (Eq. 8.2 or its lidar or sodar
counterpart) also contains factors for the decrease of intensity of return signal from
targets as the distance to the target increases. The decrease in backscattered signal
energy per unit area with distance, resulting from the expansion of the surface area
over which the signal is distributed, is proportional to 1/r2 , where r is the range
from the scattering volume back to the receiver. The other cause of diminishing
return signal is loss or attenuation of the signal by interaction with intervening
atmospheric constituents, which is here represented in a simplified manner by `,
a fraction between 0 and 1. The two components of attenuation are scattering,
which we have just discussed, and absorption. Scattering is often represented as
a scattering cross section ¢ C defined similarly to the backscattering cross section
just described, except accounting for energy scattered out of the beam in all
directions, rather than just backwards toward the instrument. Likewise, absorption
by atmospheric gas molecules or aerosol particles can be expressed as an absorption
cross section ¢ A . The two components are additive, so that the total attenuation ¢ T
can be written ¢ T D ¢ C C ¢ A . Attenuation is the total effect of such scattering and
absorption processes on signal loss as the pulse or beam propagates outward from
the transmitter and backwards toward the receiver, so the parameter ` must represent
all these effects.
8.3.4 Antenna Characteristics
An antenna is a device used to transmit/receive electromagnetic radiation to/from

free space. Although not strictly an antenna by this definition, the transducer used
in sodars to convert between electrical signal and acoustic radiation performs the
same function. Most antennas are reciprocal, which means they can both transmit
and receive signals at the same frequency. The main purpose of the antenna
during transmission is to concentrate the radiation in a preferred direction so as
to locate the target in space and to enhance its signal strength. Two important
antenna characteristics for our discussion are the antenna gain and its side-lobe
characteristics.
The antenna gain is a measure of its concentrating ability. An antenna with no
gain radiates equally in all directions, such as a candle, and is called an isotropic
radiator. Its radiation pattern is a sphere (area D 4 r2 ). An antenna with gain
concentrates the radiation in a preferred direction, as with a flashlight. Its gain is the
ratio of the antenna power (in the direction of its maximum) to that of an isotropic
radiator. It is often expressed in decibels, as will be described in the next section.
For example, an antenna that radiates uniformly over a hemisphere (area D 2 r2 )
would have a gain of 2–3 dB.
Figure 8.10 shows that the higher-gain antenna concentrates its radiation in
a smaller area called the main lobe. The figure also shows that some of the
radiation is emitted in directions other than the main lobe. These are called side
lobes and may allow signal returns from unintended targets (clutter) such as birds,
airplanes, ground objects (buildings, hills, trees, etc.), and ocean waves or “sea
clutter,” or admit signal from sources other than the remote sensor itself, from
angles very different from the nominal pointing angle of the antenna. Problems
0 dB
–10 dB
–20 dB
–30 dB
G = 7.4 dB G = 14.9 dB G = 25 dB
Fig. 8.10 Schematic antenna patterns for gains of (left) 7.4 dB (D 5.24), (middle) 14.9 dB (D 31),
and (right) 25 dB (D 316). As antenna gain increases, beam becomes more focused. Maximum
gain normalized to 1 in each example
Fig. 8.11 Sodar beamwidth and sidelobe patterns for acoustic frequencies of 1,000, 2,000, and
3,000 Hz for same acoustic antenna characteristics, scaled by the peak intensity occurring at
vertical (Neff 1975, 1990). Beam becomes narrower, gain increases at higher frequencies as
described in text
arise when sidelobe signal intercepts clutter targets with strong reflectivity, which
can overwhelm and obscure the faint return from the intended targets. Sidelobes
can be a significant problem for radar and sodar systems in mountainous or urban
topography, where such nearby hard-target obstacles are often present.
Antenna gain G is related to its collecting area A and the operating wavelength œ,
G D 4A=
2 . Figure 8.11 illustrates how sodar beam width (and therefore gain) and
sidelobe activity depend on system œ. When used to receive backscattered signals,
it is convenient to consider that a higher-gain antenna has a larger collecting area A,
which allows the detection of weaker signals. Also, as a practical matter, an antenna
of a certain gain is generally physically smaller for higher frequency (smaller œ)
operation than one for lower frequency (larger œ).
Much of the antenna designer’s job involves balancing the requirements of
operating wavelength, gain, side-lobe characteristics, physical size, and scanning
requirements. Recently, phased array antennas have proven useful for these designs.
A phased array combines the radiation from a large number of small antenna
transmitting and receiving elements positioned in an array. The phase and amplitude
of the signal radiated from each element may be controlled to electronically steer
the resultant combined beam and to shape the radiation pattern to minimize clutter
or interference.
8.3.5 Effects of Averaging
An active remote sensor transmits a signal at its characteristic œ, then “listens”

for a return at that œ (or at other œs of interest). But even with the transmitter
turned off, if the receiver is on, the system will still be “hearing” something.
This background noise may come from internal system electronics, or external
interference. For example, for some lidars, solar light scattered constantly into the
receiver by atmospheric molecules or by fine aerosol particles could be a source
of external noise. For radar wind profilers, radio transmissions near urban areas
can interfere with signal detectability. For sodar, acoustic noise from urban areas,
highways, airports, wind turbines, construction sites, or strong winds blowing over
the instrument enclosure or nearby objects can all be sources of external audio
interference.
The weak remote-sensing return signal is often dominated by the noise. In a
simplified manner, the signal-to-noise ratio (SNR) is a measure of the standard
deviation of the signal strength ¢ S compared with the standard deviation of the noise
¢ N (where in this context, ¢ indicates the customary statistical symbol for standard
deviation). In a uniform field of scatterers, SNR characteristically decreases with
range as 1/r2 , as previously discussed, because of the decrease in backscattered
signal per unit area. SNR is generally expressed in decibels (dB); if R is the ratio
¢ S /¢ N , SNR in dB is equal to 10 log R, where the logarithm is to the base 10. Thus,
if the ratio R is 1/100, log R D 2, and, SNR D 20 dB. For such a small SNR,
the likelihood of detecting the signal would seem to be very small. Many different
approaches exist to calculate SNR.
The solution to this detection problem is averaging. A return (Fig. 8.12c) consists
of the signal plus noise (see top panels of Fig. 8.12). Noise is a random process; the
average of the fluctuations in a white noise process, for example, approaches zero as
the number of independent samples that are summed (or “sample size”) N increases.
But the signal does not average to zero; instead, it averages to its mean value. Thus,
averaging over a long enough interval will expose the signal, as shown in Fig. 8.12.
In other words, averaging improves the detectability of the signal. The detectability
or sensitivity improves as N½ . The tradeoff is in the resolution of the averaged data
in time or space, as will be illustrated later.
Accurate, quantitative estimates of the backscattered energy are needed for
applications such as liquid-water contents or rainfall rates for radar, and aerosol
Fig. 8.12 Effects of averaging on (a) signal with (b) noise added (note difference in scale) for 1,
10, 100, and 1,000-sample running averages
20
15
10
0
dB
–5
–10
–15
–20
–25
–30
–30 –20 –10 0 10 20 30
Frequency Shift (Hz)
Fig. 8.13 Idealized example of a hypothetical Doppler spectrum, with returned power plotted as
a function of Doppler-shifted return-frequency offset, showing a peak at 10 Hz
concentration estimates for lidar. For such quantitative estimates the transmitted
power as well as system hardware characteristics, such as internal signal losses,
must be well characterized, i.e., the system must be well calibrated. Some other
aspects of using backscattered energy will be discussed with each type of system.
Doppler. For Doppler systems, it is only necessary to have enough signal (backscat-
tered energy) to obtain an estimate of the return frequency. This generally involves
estimating the signal energy returned to the receiver as a function of frequency, i.e.,
estimating the Doppler spectrum of the returned signal (Fig. 8.13), which can be
accomplished, e.g., by taking a Fourier transform of the time series of the return
signal. Radar and lidar use different approaches to determining Doppler velocities,
as described in the Appendix.
Because the returned frequency is independent of the amount of backscattered
energy (as long as enough signal energy is returned to make an estimation of its
frequency), accurate knowledge of the transmitted power and other system attributes
are not needed for an accurate estimate of the frequency. The return frequency is
used to calculate the Doppler velocity according to
Vdopp D
f =2 (8.4)
D 1=2 cf=f (8.5)

where Vdopp is the calculated velocity component along the beam, f is the Doppler
shift—the frequency difference between the transmitted and returned signal, and
œ is the wavelength of the transmitted signal, and the second version makes
use of œ D c/f. For example, assuming a radar operating frequency of 915 MHz
(œ D 32.76 cm), the Doppler shift of C10 Hz shown in Fig. 8.13 corresponds to a
target velocity of 1.638 m s1 toward the radar.
Stronger signal, or more averaging, improves the sensitivity or detectability of
the signal, such as allowing the velocity to be estimated to greater ranges. As an
example, Fig. 8.14 shows a Doppler lidar signal averaged over several different
time intervals. The plot illustrates the effect of increasing the number of Doppler
lidar shots or pulses averaged from 10 to 50 to 200. It is evident that increasing
the number of shots averaged means that these quantities can be measured with
confidence to greater ranges. For example, going from 10- to 50-shot averaging
represents a 5 decrease in time resolution, but range increases from 6.5 to 8.5 km,
a 30% enhancement. For applications where the time resolution is unimportant, or
during periods when conditions are changing very slowly, it would be worthwhile
to sacrifice the temporal resolution to obtain the extra range coverage—a good
example of how remote-sensing tradeoffs can be selected to fit the application. But
note that the increased averaging does not affect the value of the SNR or the Doppler
velocity at each range; the only changes result from slowly changing atmospheric
variations.
As a final note, the accuracy of a measurement consists of two parts, a systematic
offset or bias and the effect of the random noise fluctuations or precision. Increasing
the sample size N improves the precision of the measured mean value for the sample
and also provides an improved estimate of the bias. In the following sections, the
precision of a measurement will be described in many cases, but it is important to
keep in mind that the precision of a measurement, such as the standard deviation of
mean values calculated over a given averaging interval, depends on the averaging
time—an instrument can be made to seem to have more impressive performance by
averaging over longer intervals.
8.4 Sodar
Sodar is a profiling system that transmits a pulse of sound into the atmosphere. It
provides vertical profiles of acoustic backscatter and wind velocities in the lowest
tens to hundreds of meters, depending on the transmitted wavelength (frequency).
Typical transmitted frequencies are between 1,000 and 5,000 Hz, the lower frequen-
cies for instruments that provide vertical coverage up to several hundred meters, and
the higher frequencies for more compact systems having more limited coverage (up
to 100 or 200 m) but achieving much finer vertical resolution. The transmitted sound
pulse scatters off turbulence-scale fluctuations in the acoustic index of refraction
(described below), returning a weak signal back to the receiver. Technology for
generating and processing acoustic signals is well established, and the speed of
Fig. 8.14 Doppler lidar fixed-beam profile display of (top) signal-to-noise ratio and (bottom)
radial wind speed results for 10, 50, and 200 pulse averages. The change in averaging times
indicated by vertical lines can also be discerned as different tick spacings in the time axes of
the plot (transitions at 1:37 and 1:46) (Banta et al. 2007)
sound is relatively slow, especially compared with the speed of light (by a factor
of 106 ). Because of the availability of relatively inexpensive hardware components,
sodar can be an economical alternative to other active remote sensors. Sodar is a
very familiar instrument and many reviews have been written of sodar capabilities
and theory (e.g., Neff and Coulter 1986; Neff 1990; Coulter and Kallistratova
2004; Kallistratova and Coulter 2004). Because of the availability of these detailed
descriptions, this section is intended as an introduction to using sodar in complex
terrain.
Variations in the acoustic index of refraction that produce a backscattered signal
consist of gradients of temperature T occurring over distances of œ/2, typically 10
to a few tens of cm, as described by Tatarskii (1971) and reviewed, e.g., by Neff
and Coulter (1986) and Neff (1990). Such strong gradients can occur in the mean
temperature profile (Gossard et al. 1985; Balsley et al. 2003; Kouznetsov 2008),
but infrequently, so under more normal conditions the main source of backscatter
is turbulent temperature fluctuations, via Bragg scattering. Major source terms
in generating these fluctuations, as measured by the temperature variance T 0 2 ,
in a horizontally homogeneous atmosphere involve the turbulent vertical velocity
variance w0 2 and the stratification (lapse rate @™/@z):
@T 0 2 @ 0 2 @ @w0 0 @
/ w0 0 ; and / w0 2 (8.6)
@t @t @z @t @z
making use of the fact that at any given height, T0 and ™0 (the prime denoting a
departure from a horizontal or temporal mean value of T or ™) are nearly equivalent.
Acoustic backscatter signal is thus generated by velocity turbulence w0 2 in the
presence of a non-neutral lapse rate (@™/@z ¤ 0) both together, and if either is
zero, acoustic backscatter “targets” are not being produced. This is illustrated
schematically in Fig. 8.15, which shows how vertical-velocity fluctuations w0
interact with a stable (lower left) or an unstable (lower right) ™ profile to generate
temperature fluctuations T0 , because ™ is conserved during adiabatic upward or
downward displacement. In a neutral atmosphere (upper right) no T0 is generated.
This illustration provides an explanation for the behavior of the signal of the
sodar in Fig. 8.1. Below the nose of the LLJ strong shear produces turbulence in
the presence of stable stratification, in turn producing strong sodar backscatter. At
the nose of the jet, shear goes to 0 as does the generation of turbulent velocity
fluctuations, producing a minimum of backscatter, and thus low signal-to-noise
ratio, at that level. The result is greater uncertainty in the velocity estimates. Above
the nose of the jet, the stratification is often near neutral, meaning that even if
velocity turbulence is present, it may not produce very strong T0 signal. Figure 8.1
illustrates potential LLJ-related detection problems for sodar, but we note that
sensitive sodars can be used to study LLJs (Kallistratova 2008).
It is important to note that in a convective BL (unstable, lower right of Fig. 8.15),
warm T0 perturbations are generated near the surface but then are carried aloft in
updrafts by buoyancy through the depth of the BL. Thus, sodars can receive signal
Fig. 8.15 Schematic

representation of possibilities
for generating temperature
fluctuations for sodar signal.
Horizontal axis is potential
temperature and vertical axis
is height above ground. Red
line indicates environmental ™
profile, boxes represent fluid
elements or air “parcels,” and
arrows indicate vertical
displacement effects of
turbulence
through the BL depth even though the mean ™ profile appears to be near neutral.
Sodars have been used to track the depth of the convective BL, at least until the
signal near the top of the BL becomes too weak for the sodar to detect (Beyrich
1997; Coulter and Kallistratova 2004).
These arguments are true for acoustic backscatter (180ı), but for scattering angles
other than 180ı , acoustic index-of-refraction fluctuations are also comprised of
along-beam velocity fluctuations. Thus, if a receiving antenna is not the transmitting
antenna and is located at some distance from the transmitter (the “bistatic” configu-
ration), signal is obtained even in neutral conditions (upper right of Fig. 8.15) from
velocity turbulence. Although such systems are technologically more complicated
and more expensive, a few such systems have been demonstrated (Neff 1990;
Behrens et al. 2008).
Scattering contributes to the attenuation of the acoustic signal, but another factor
is the absorption of the sound wave by the intervening atmosphere. Water vapor
reduces these effects so reliable sodar data may be obtainable to greater heights
Table 8.3 Typical operating characteristics of sodar: examples; PRP is pulse repetition period
F (Hz) œ (cm) Range (m) Pulse length (m) r (m) PRP (s) Eff PRP (s)
1,000 30 40–2,000 40 20 12 1; 200
2,000 15 30–900 20 10 9 900
Mini 4,500 6:7 20–200 10 5 6 900
in a moist environment than in a drier environment, all other factors being equal.
A contributing factor to the height limitation shown in Fig. 8.1 was probably that
the measurements were taken in a dry region.
Table 8.3 shows operating characteristics of some sodar systems. Many sodars
transmit in the frequency range of 3,000–1,000 Hz, corresponding to a œ of
10–30 cm. These systems obtain signal to heights of several hundred meters with
resolutions of 10–30 m, starting at 50 m minimum range. Measurements are
generally averaged over several minutes, to achieve precisions of 1 m s1 or better.
System performance varies with œ depending on atmospheric conditions, so systems
that transmit multiple œ’s have been developed to optimize performance over a
wider range of conditions, for example, during daytime, when turbulent eddy sizes
are larger as well as at night when eddies are smaller.
Sodars operating at higher frequencies (4,500 Hz) cover the lower 100–200 m
or so of the BL at finer vertical resolution, as previously described. As shown in
Table 8.3, the minimum range of these instruments is 5–10 m, so they overlap
conveniently with the lowest BL region measured by towers. Because these systems
are very compact (antenna or “dish” sizes of <1 m), they have been referred to as
“mini-sodars” (Coulter and Kallistratova 2004).
Sodars measuring backscatter are normally aimed toward the zenith to obtain
vertical profiles. Doppler systems must measure the wind components in at least
three directions to calculate the total wind vector. If three beams are used, normally
two beams are tilted at some angle to the zenith (such as 18ı ) and aligned in
perpendicular horizontal directions to each other, to measure both components of
the horizontal wind. The third beam is pointed upward so the horizontal-wind
measurements can be corrected for w. Phased-array sodar transmitters can direct
beams in multiple directions (often five) to provide redundancy for the horizontal
wind estimation. Horizontal wind components are calculated assuming horizontal
homogeneity of the flow across the beam pattern. An issue in complex terrain is that
the beams are spatially separated, so that at sites where the flow is not horizontally
homogeneous, significant errors in the calculated horizontal winds can result.
As with some other remote sensors, an issue for sodar is the transmission and
detection of sidelobes (see Fig. 8.11). Mountainous terrain amplifies this problem
because of topographic obstacles, which may be forested, intercepting the sidelobe
pulse and returning a strong signal back to the sodar. For the wind profile, Doppler
processing may be dominated by a near-zero velocity signal from the topography,
or erroneous nonzero velocities from motions of nearby trees. These problems may
be addressable by data-processing software, but are also addressed by surrounding
the sodar transmitter by an enclosure that suppresses the transmission of sidelobe
Fig. 8.16 Sodar backscatter record from Great Salt Lake basin during morning hours (sunrise
1200 UTC) (top), with weak echoes in green and strong echoes in red. Inversion (arrow) rises
with surface heating (Courtesy R. Coulter). Sodar backscatter record from narrower valley during
morning transition (bottom). Stronger echoes are darker, and indicate inversion descent during
morning hours (Courtesy W. Neff)
signal out to the atmosphere. This enclosure also helps to shield the receiver from
ambient noise, which can interfere with the detection of the backscattered acoustic
signal.
The first routine sodar products were time-height cross sections of sodar
backscatter intensity, which have been used for more than five decades. Two types
of atmospheric structure were readily identifiable in the earliest such sodar data,
thermal plumes in the daytime convective BL and strong horizontal layering of the
nighttime stable BL. The layering evident in sodar data made the data useful for
studying a number of stable BL features, including waves, fronts and other density
currents, and LLJs (Neff 1986, 1990; Neff and King 1988; Coulter and Kallistratova
2004). Carefully calibrated sodars have also been used to determine several turbu-
lence quantities, including structure functions, turbulent dissipation, and turbulent
fluxes. According to Coulter and Kallistratova (2004) these techniques, “work well
only in skillful hands, and they demand further experimental examination.” An
overall conclusion of their review was that, “measurement of vertical profiles of
wind speed components is the main purpose of most contemporary sodars.”
Examples of sodar backscatter in mountain settings, showing obvious layering,
are shown in Fig. 8.16. The top panel shows a morning transition in a wide basin,
the Great Salt Lake basin in Utah in the western U.S. (Coulter et al. 2004). With
time running from left to right, the larger values of sodar backscatter (red) marking
the top of the shallow early-morning convective BL (layer marked by arrow) can be
seen to rise as this shallow unstable layer grows due to surface heating. The strong
Fig. 8.17 Doppler minisodar example from Colorado Front Range during passage of a shallow
Denver Cyclone. (a) Topographic map of north central Colorado (north up) shaded at 800-m
contour intervals, showing streamlines and position of Denver cyclone at 0830 UTC, 7 February
1991. Vertical arrows show cyclone path from north to south, and westerly arrow shows outflow
from a major canyon. (b) Contour map showing locations of Coal Creek Canyon and the two sodar
sites CC* and BR*. (c) Doppler minisodar wind profiles at site CC*, at mouth of canyon, showing
westerly flow for entire period. Time (UTC) runs right to left. (d) Wind profiles at site BR*,
showing passage of Denver cyclone (From Levinson and Banta 1995. © American Meteorological
sodar returns from the inversion at the top of the shallow developing convective BL
are a result of strong T0 generated, as this turbulent (large w0 2 ) BL grows upward into
the nocturnal inversion layer, where @™/@z is large. In contrast to such wide basins,
the different valley cross-sectional geometry of narrow valleys leads to a sinking of
the inversion layer at the top of the shallow convective layer during the morning
hours, as described in Chap. 2. This effect is shown in the bottom panel of Fig. 8.16
(layer indicated by arrow) for a narrow valley in northern California (Neff 1986).
Sodars having Doppler capability have also found widespread use in mountain-
ous terrain to provide wind profiles (Neff 1986; Neff and King 1988; Kallistratova
and Coulter 2004; Coulter et al. 2004; and as summarized by Neff 1990). Figure 8.17
Table 8.4 Sodar measurements

Inexpensive, deployable in arrays Coverage can be incomplete
Frequent profiles of wind and Interpretation of backscatter data often challenging
turbulence-related quantities
Layers visible Often may not be useful in noisy environments
Transportable Sound pulse annoying to neighbors
Limited data availability in strong winds
shows vertical nighttime profiles (right profiles) of the horizontal winds from two
mini-sodar sites outside the mouth of a canyon northwest of Denver, Colorado.
One site (CC) was located very near the mouth of the canyon, and the other site
(BR) was located 8 km farther east. On this night, a terrain-forced vortex called
a “Denver cyclone” formed just to the east of the mountains (Fig. 8.17a) and north
of the sodar sites (Levinson and Banta 1995). The vortex moved southward, passing
the sites between 0130 and 0330 LST (0830 and 1030 UTC). The sodar closest
to the mouth of the canyon (top) shows uninterrupted westerly canyon outflow all
through the night. The sodar 6 km farther east (bottom right) shows the westerly
outflow before 0800 UTC and after 1100 UTC, but this canyon drainage flow was
interrupted by windshifts due to the cyclone passage between 0830 and 1030 UTC.
The vortex flow at this location was very shallow, on the order of 50 m deep.
This example illustrates that because sodars are inexpensive, familiar, and relatively
straightforward to use, they can be deployed to multiple locations to give insight
into the horizontal variability of profile features. Here multiple sodars reveal the
strong spatial variability of the nighttime flow near the mountains—specifically that,
although the terrain-induced vortex came very close, its influence did not extend all
the way to the mountains.
Sodar backscatter and Doppler data provide an important source of data above the
surface. The technology is well developed, with a variety of sodar systems available
commercially. Intercomparison studies with in-situ sensors (Crescenti 1997; Coulter
and Kallistratova 2004) and other remote sensors (e.g., Pichugina et al. 2008) show
that mean wind estimates averaged over 5–10 min are reasonably accurate during
times of strong backscatter and high SNR. An issue for sodar, as for most remote
sensors, is availability of the data: are there some times when backscatter or SNR
are weak and data dropouts are apt to occur? A number of factors can limit vertical
coverage of sodar data, such as near-neutral conditions (including evening and
morning transitions), weak-turbulence conditions, dry conditions, and high-ambient
noise conditions. Some other limitations are that the rather loud signal is audible and
can be an annoyance to neighbors, and under strong wind conditions, the signal can
blow downstream fast enough to avoid being sensed back at the antenna, because
the propagation speed of sound is relatively slow. These and other advantages and
limitations of sodars are listed in Table 8.4.
8.5 Radar
Weather radar has been a critical component of storm detection, tracking, and
forecasting for many decades. It has been standard equipment in weather-service
and military forecast offices for more than half a century, and has undergone
many advances since its beginnings during World War II. As a research tool,
enhancements in technology and data processing techniques have led to rapid
advances in the understanding of thunderstorm structure and dynamics, and radar is
well known to the public because of its widespread exposure on television weather
reports and newscasts. Many introductory and advanced textbooks have appeared
on radar meteorology (e.g., Battan (1973), Rinehart 2010, Doviak and Zrnić 1993;
Bringi and Chandrasekar 2001), as well as several books. Mountainous topography
poses significant challenges for using radar to characterize clouds and storms, which
have been discussed in Chap. 1.
Weather radars operate in the microwave part of the electromagnetic spectrum
at wavelengths that range from 1 mm to 100 m (with corresponding frequencies
of 300 GHz down to 3 MHz) (see Table 8.5), all considerably longer than the
visible. At these œ’s, particularly below 10 cm, the scattering targets for radar are
primarily hydrometeors. The longer œ’s have been effectively exploited for radar
wind profilers using Bragg scatter from humidity and temperature fluctuations in
the atmosphere, presented in Sect. 8.6. As described in Chap. 7 on microphysics,
hydrometeors range in diameter from less than 10 m for cloud droplets or cloud
ice particles, those having negligible fall speeds through the air, to rain drops of
several mm diameter or hail, which can grow to several cm across. No single
radar can measure all cloud parameters, so selection of the radar will depend
on the specific measurements required, and the topographic and meteorological
circumstances for acquiring the intended measurements. For example, consistent
with their important severe-storm characterization and tracking roles, longer-œ
radars, such as the operational WSR-88D used by the U.S. National Weather
Service (formerly known as “NexRad”), are optimized for detecting the larger
particles and drops found in severe thunderstorms and squall lines. Their reduced
attenuation compared with shorter-œ instruments allows them to detect storms at
greater distances and to see through the far side of storms. On the other hand,
these radars are less useful for detecting non-precipitating clouds, because the
tiny cloud ice or droplets are effectively invisible to them as a result of the r6
dependence of Raleigh backscatter (Sect. 8.3.3.2). Weakly precipitating stratiform
and snow systems (which also contain smaller droplets and particles) will be difficult
to detect at any significant distance from the antenna. In mountainous terrain the
wide beam and sidelobes of these larger radars make it difficult to sample storms
close to topography. As described in Chap. 1, excessive ground clutter—returns
from stationary targets such as topography, buildings, clumps of trees—and beam
blockage by the terrain are critical limitations to the use of radar in the mountains.
Shallow storms that may produce copious precipitation can be completely obscured
by topography (White et al. 2003; Wetzel et al. 2004), or only the tops of deeper
444
Table 8.5 Operating characteristics of meteorological radar examples

“Band” œ (cm) F (GHz) Min range (m) Unambig max range (km) Peak Pwr PRF (Hz) Pulse length s (m) r (m)
Storm
WSR-88D S 10 2.8 200, 400 1,000 318–1,400 1.57 (470), 4.5 (1,350) 250, 1,000
TDWR C 5 5.6 48, 360 250 235–2,000 1.1(330) 150, 300
X 3 9 300 75 40 2,000 1 (300) 150
Cloud
Ka 0.86 34.86 150 15 200 10k 0.25 (70) 45–90
R.M. Banta et al.
storms may be detected. In such cases it is difficult to tell whether the hidden
low-level portion of the mountain storm is benign or menacing. Work in the Alps
has addressed how to optimize radar information obtained in mountain locations
(Germann and Joss 2004).
For characterizing, tracking, or studying weaker-precipitating systems or clouds,
meteorologists use shorter-œ radars. œ’s near 3 cm are effective in detecting
stratiform precipitation and snow, and vertically pointing radars with œ of a few mm,
for which cloud droplets/ice provide detectable Rayleigh backscatter, have been
used operationally to characterize cloud layers. Despite limited ranges compared
with the longer-œ radars, short-œ instruments often feature narrower beams and
weaker sidelobes, which make them more useful in mountainous settings, and they
are capable of finer range resolution (Table 8.5). Because of these properties and
the fact that the smaller fall speeds of snow and cloud particles with respect to
the surrounding air makes them better tracers of the flow, a research radar with
œ D 8 mm has been able to be used to study kinematics of air flow across mountain
barriers, as will be described later in this section.
8.5.1 Radar Measureables
Other important uses for radar besides storm tracking include quantitative pre-
cipitation estimation, studies of growth processes of tiny cloud droplets or ice
into large precipitating drops or ice particles (Chap. 7), and cloud and storm
dynamics (Chap. 6). For many of these applications it is desirable to relate the radar
backscatter, generally expressed as reflectivity Z, to the density of condensed water
or ice suspended in the cloud, i.e., the cloud’s “liquid water content” (LWC) or
“ice water content” (IWC). Unfortunately, Z is a complex function of LWC, size
distribution, particle shape, and particle composition or phase (liquid or ice), and Z
is only one piece of measurement information. A given measured value of Z could
be produced by many combinations of these variables, and in particular, the Z value
by itself cannot be associated with a unique LWC. To add to the complexity of the
problem, the most useful relationship is between Z and the rainfall rate at the surface
R. A number of physical processes affect precipitation between where it is measured
in the cloud and when it hits the ground, such as evaporation, melting, drop or
particle growth by microphysical interactions, and others. Nevertheless, driven by
applications such as the critical need for area-integrated rainfall in hydrology, a great
deal of research effort has been devoted to determining these Z–R relationships,
which vary with location, season, storm type, dominant microphysical processes,
and other environmental factors.
To help address these ambiguities in the Z-LWC (or Z–R) relationships, radar
design has been advanced to provide other measureables. These include frequency
of the return signal, from which the Doppler velocity is calculated, the polarization
characteristics of the return signal, and the use of multiple wavelengths.
Doppler velocity. The Doppler velocity is the radial velocity component of the
movement of scatterers along the radar beam, as described in Sect. 8.3.4 of this
chapter. For tiny cloud droplets or cloud ice this movement may closely approximate
the air movement (i.e., radial component of the wind), but precipitating raindrops
or ice by definition have a fall speed with respect to the air, and thus follow the
flow only with considerable inertia, i.e., often not very well. Nevertheless, Doppler
radar has been used very effectively to study storm structure and kinematics. In
research conditions the restriction of measuring only the radial-wind component
has been overcome by deploying two or more radars with overlapping coverage
of the scan volumes, from which the full three-dimensional motions can be
computed. Examples of such “dual-Doppler” or “multiple Doppler” studies in
mountainous terrain will be presented later in this section. Information on particle
size distributions can be obtained from vertically pointing Doppler radars. In this
mode the radar measures the variation in terminal velocities of the precipitating
drops/particles, adjusted for the air motion by also measuring the velocity of the
cloud droplets or cloud ice moving with the air, which often shows up as a separate
peak in the indicated velocity distribution.
Polarization. Polarization is a property of the transmitted and backscattered radia-
tion related to the orientation of the electric field (and the perpendicular magnetic
field) of the electromagnetic wave. The two basic types of polarization are linear
(horizontal and vertical senses) and circular (right-hand circular and left-hand
circular senses). Radars may transmit one sense of polarization (or may alternate
senses) and then listen for both senses in the return signal to determine whether the
scatterers have changed the polarization, or “depolarized,” any of the signal. The
amount of depolarization is related to the shape of the scatterers. For example, for
linear polarization, spherical scatterers (liquid cloud droplets or smaller raindrops)
tend to return equal amounts of signal in the vertical and horizontal channels,
whereas elongated particles, either flat or needlelike tend to fall with a horizontal
orientation and thus return more signal in the horizontal channel. For circular
polarization, spherical scatterers preserve the direction of rotation in the backscatter
and thus all the return signal is in one channel, whereas elongated particles
divert some energy into backscattered signal with the other rotation. An important
application for polarization is as an aid in discriminating between liquid-water- and
solid-phase (ice) hydrometeors. Spherical scatterers tend to be liquid droplets, and
elongated scatterers are most often ice, although at large reflectivities they may
represent large raindrops several mm in diameter that flatten as they fall (Rinehart
2010). A refinement of polarization capability is the use of elliptical polarization
of varying aspect ratios, which combines the advantages of the other two types
(Reinking et al. 2000).
The degree of depolarization is expressed as a depolarization ratio in decibels,
either the reflectivity depolarization ratio (or differential reflectivity) ZDR for linear,
or the circular depolarization ratio CDR for the circular. For linear polarization
and spherical scatterers the nearly equal distribution of reflectivity in the horizontal
channel ZH to that in the vertical channel ZV produces a ratio ZH /ZV of 1 and a ZDR
(D 10 log ZH /ZV ) of near 0. Elongated scatterers, for which ZH /ZV > 1, produce
positive values of ZDR . For circular polarization the CDR is defined as [10 log
ZP /ZO ], where ZP is the reflectivity in the parallel channel (same as transmitted) and
ZO is that in the orthogonal channel (opposite to transmitted). The case of circular
polarization is somewhat complicated in that the incident transmitted beam bounces
back from a population of spherical scatterers with no change in the direction of
rotation (e.g., clockwise) as seen by an observer at the antenna, but the direction
of propagation is reversed, so that right-hand circular polarization (e.g.) going out
becomes left-hand polarization coming back, and the polarization changes sign.
Thus spherical drops reverse the sign of the polarization but put little energy into
the opposite rotation direction, so the CDR is very large in magnitude but negative.
Flat or linearly oriented particles do generate polarization in the opposite channel
(appearing in the denominator) regardless of spatial orientation, leading to smaller
(negative) magnitudes of the CDR.
Polarization thus provides important pieces of information for interpreting the
content of clouds and storm systems. The depolarization ratios and other related
signal information, including the reflectivity Z itself and Doppler information, can
be combined into complex schemes to infer particle types and concentrations. Often
an independent measurement of the temperature profile showing the freezing level
is required for best results. One such procedure developed for NCAR’s 10-cm-œ
(S band) SPol radar is described by Vivekanandan et al. (1999). In this scheme, five
radar measureables (including reflectivity Z, ZDR , linear depolarization ratio, and
two others) plus the temperature (from a sounding) are used to classify data from
each range gate into 14 hydrometeor categories (and a 15th for insects, also visible
to the radar).
Multiple wavelengths. Because of the strong dependence of backscatter or reflec-
tivity on particle or droplet size, measurements at one œ give limited information
about the size distribution of the scatterers. But the ability to infer LWC depends
on whether the measured Z is from a large number of small droplets, a few large
droplets, or other possibilities, such as bimodal size distributions. Measurements
at more than one œ can address this problem, provided the separation in œ is large
enough. For layered clouds with reasonably uniform properties in the horizontal,
multi-wavelength datasets can be obtained using different instruments at the same
site each pointing vertically, which could be radars with different œ’s or radars and
lidars. Convective clouds and cloud systems, which are probed by scanning, are
spatially variable and often rapidly evolving, so a major challenge is to probe the
same volume of space by more than one œ. A scanning radar transmitting more
than one œ would address this problem, but considerable engineering obstacles
are associated with using one antenna to transmit and receive more than one œ,
because of the dependence of antenna properties on œ (see Sect. 3c). Despite these
challenges, multi-œ radars have been built and used to study storm structure.
Clear-air returns. An important class of applications involves radar returns when

clouds are absent, conditions that Gossard (1990) refers to as the “hydrometeor-
free atmosphere.” Such “clear-air” returns may have two sources, point targets,
such as insects or other floating material, and Bragg scattering from refractive-
index fluctuations, which are usually due to turbulence. The point-source biological
targets are warm-season continental targets that are not available during wintertime
conditions. Such floating debris has been shown to provide effective scattering
targets for studying the kinematics of forest-fire convection columns using radar
(Banta et al. 1992). Scanning Doppler radars have also proven valuable in the clear-
air BL, observing many features such as fine lines, density currents, convergence
zones, and many more. Ash particles and other debris in mountain forest-fire plumes
have also proven to be good radar targets (Banta et al. 1992). A comprehensive
review of these applications to boundary-layer research is given in Gossard (1990),
Kropfli (1986b), Kropfli (1990), and other applications have been described by
Fujita and McCarthy (1990).
An effective tool for using these effects to study BL structure is a profiling,
continuous-wave radar with frequency modulation (FM/CW radar) operating at
10 m, which provides ultra-fine resolution of 1.5 m up to heights of several
km. Doppler versions of this radar have been built.
8.5.2 Radar-Specific Issues
Radars, especially the longer-œ storm radars, are larger and higher powered, and
probe to greater distances, than most of the other remote sensing systems considered
here. As a result a few issues peculiar to radar arise.
Maximum range. The fact that radars, especially storm radars, are used routinely
to much greater distances than the other remote-sensing systems produces some
special issues for processing of radar data. One issue involves the time between
pulses Tr , which corresponds to a distance of travel Dr (Dr D c Tr ) of one pulse
before the radar stops “listening” for the return and transmits the next pulse. The
radar then listens for the next pulse. This distance, usually corresponding to the
maximum range displayed on radar images or scopes, is sometimes referred to as the
maximum unambiguous range. If a target existed beyond the distance Dr , however,
its echo would return to the radar during the time interval when the instrument is
trying to detect returns from the second pulse. The storm or other target would
then be displayed at an erroneous distance less than Dr , as if its echo came from
the second pulse. This usually is not much of a problem, because the 1/r2 decrease
in signal intensity makes signals beyond the pulse-repetition “distance” Dr much
weaker than those within that distance. But occasionally a very strong target beyond
this distance, such as a hailstorm, will produce such a strong echo that it will show
up prominently at the wrong distance from the radar in the display. Interesting
examples of such “range ambiguity” are given by Rinehart (2010) and Doviak and
Zrnić (1993), along with more detailed descriptions of these effects. Doviak and
Zrnić (1993) show an example where such a “second-trip” echo is superimposed on
a weaker echo, displayed at its correct range from the radar. A problem for Doppler
estimation in such a case is that the two signals are mixed together, since they arrived
at the same time, so it is difficult to estimate the mean frequency of the return signal
from either target, to estimate the mean Doppler wind speed for the corresponding
range gates.
An obvious solution to this range-ambiguity problem might be to increase the
time between pulses, thereby increasing the distance over which the radar listens for
each pulse before transmitting again. The problem for Doppler radar systems is the
sampling ambiguity for Doppler velocities, as discussed in the chapter Appendix,
which becomes worse for longer Tr . In other words, the requirement for short Tr (to
accurately measure the phase change from pulse to pulse for the Doppler velocity
calculation) competes with the desire for a longer Tr to reduce range ambiguity,
therefore requiring a tradeoff. Rinehart (2010) and Doviak and Zrnić (1993) point
out that the tradeoff inherent in this “Doppler dilemma” is more favorable for longer-
wavelength radars—a 10-cm radar does not have to compromise as much as a 1-cm
system. A number of technical approaches have been attempted to address these
issues (Doviak and Zrnić 1993), such as “encoding” each pulse with a characteristic
transmitted phase, so that the ranges to target can be properly sorted out and
displayed. Such processing is possible because the radar signal-generation system
produces a very pure frequency from pulse to pulse.
Siting. Radars are relatively large, high-powered systems, so locating appropriate
sites involves a number of considerations. The site must allow for radar scanning
that is sufficiently unobstructed so the desired cloud features can adequately be
observed, since terrain interception effectively blocks the signal and limits the
useful range. But some terrain can benefit radar data quality. In general the
ideal radar site from physical considerations would be at the center of a shallow
doughnut of slightly higher ground, a barren low ridge not far from the radar that
obscures interfering features such as towers, swaying, water-laden trees (leaves may
contain a high percentage of liquid water, making them good radar targets), or
moving vehicles. As with many other remote sensors, sidelobes can be a problem
especially in complex terrain, and such a site can reduce the impact of these
effects. In the mountains, these sites can be hard to find, so compromises are
almost always necessary. Ground-clutter maps can be obtained by scanning the radar
360ı in azimuth at minimal, and incrementally increased, elevations and identifying
interfering terrain, forests, or other stationary objects that always appear in the scans
even in the absence of precipitation.
For mountain locations, the ideal site meteorologically would be one that
experiences no significant wind or precipitation at the site, but offers a clear view of
both within scanning range. Near-ideal sites for observing weather over mountain
ranges are often found in an adjacent valley, basin, or plain. Strong winds can shift
antennas corrupting the measurements, and can damage antennas, so some climatic
estimate of extreme weather at the desired site is helpful. Snow that sticks to the
antenna can attenuate transmission and reception, and the weight can damage the
scanning machinery. Siting on high mountain slopes receiving considerable wet
snow or high winds usually severely limits acquisition of good measurements.
A mobile 3-cm “Doppler on Wheels” (DOW) radar developed for storm chasing
(Wurman et al. 1997), which avoids many of these siting problems, has been used
successfully in mountainous conditions during the Mesoscale Alpine Programme
(MAP).
Protective dome. NEXRAD, a permanent installation, employs a radar dome, or
radome, to protect the radar’s antenna from extreme weather. Radomes can be
expensive, and can collect rime ice or snow that attenuates the signal especially from
shorter wavelength radars. They are generally not used for ground-based research
radars, so other precautions must be taken to deal with weather at the site, which
sometimes must just be tolerated.
8.5.3 Scanning
The simplest data-acquisition mode for a radar is to point at a fixed angle, either
vertically or at a slant. Some radars, especially at the shortest and longest radar
œ’s, operate only in this way, but most others, especially precipitation and storm
radars, scan the antenna in azimuth and elevation. Scans in azimuth trace out
a cone or a sector of a cone in space, and thus reveal the horizontal structure
of clouds and storms, especially at lower elevation angles. Scans in elevation
produce vertical cross sections of data. Before versatile computer displays were
available, radar operator control panels had separate cathode-ray-tube scopes, one
dedicated to the azimuth scans and the other to the elevation scans, which were
called the “plan-position indicator” (PPI) and the “range-height indicator” (RHI),
respectively. Although these scopes no longer exist on modern radar consoles,
radar meteorologists still refer to azimuth scans as “PPIs” and elevation scans as
“RHIs.” The use of computers for radar control allows scan sequences to be pre-
programmed, and determining the most effective scanning strategies and appropriate
sequences is an important aspect of planning for field operations.
An important application of full 360ı azimuth scanning using Doppler velocity
data is in calculating vertical profiles of the horizontal wind speed and direction. The
profiles are obtained from radar conical scan data using the velocity-azimuth display
(VAD) technique, illustrated in Fig. 8.18. As the antenna sweeps through 360ı of
azimuth, each range gate describes a ring at a certain height in the atmosphere.
The radial velocities in this ring, plotted as a function of azimuth, form a sine wave,
the phase of the wave representing the direction of the mean wind and the amplitude
representing the mean wind speed (Lhermitte and Atlas 1961; Browning and Wexler
1968). Calculating these quantities for each range gate gives a profile of the mean
wind averaged over the ring.
Fig. 8.18 Schematic depiction of velocity-azimuth display (VAD) scan geometry. Elevation angle
of the conical scan is ’, height of range gate is h, and Vh is the magnitude of the horizontal
wind component. For a radar sensing precipitation, the fall speeds Vf of the precipitation may be
significant, and may need to be accounted for in the calculation (Wilson and Miller 1972)
An issue for VAD winds in mountainous terrain is whether the wind field is
horizontally homogeneous over the averaging ring. The degree of inhomogeneity
can be assessed by visual inspection of the plan view of the azimuth-scan plot,
or by looking at individual VAD plots if available. Experience with lidar VAD
data in complex terrain during research projects indicates that a modest degree of
inhomogeneity can be tolerated and still produce reasonable agreement with other
nearby measurements of the mean-wind profile.
Important applications for radar especially for research are in airborne opera-
tions, and these applications have been especially useful in mountainous terrain,
where weather systems show strong spatial variation. Several examples where
airborne radar has made important contributions to orographic precipitation studies
have been presented in Chap. 6 (Orographic Precipitation).
8.5.4 Examples
Radar is a familiar instrument to meteorologists, and an important operational tool

as described in the first chapter of this Book. Chapter 1 also describes some of the
limitations to using surveillance radars (such as the WSR-88D) near mountains,
including practical considerations, such as the wider separation between radar sites
in the mountainous Western U.S. than in other areas of the country. Examples of
the use of ground-based radars to investigate cold-season orographic storm systems
have also been given in Chap. 6 on orographic precipitation in the Oregon coastal
mountains, the Utah Rocky Mountains, and in hilly topography in New York State.
8.5.4.1 Cold-Season Orographic Precipitation
Precipitation in coastal mountains has been the subject of many field programs over
the past several decades in which storm radars play a prominent role, notably early
studies of Browning et al. (1974) and Hill et al. (1981) in Wales. Several such
programs have been carried out in the West Coast of the United States, which include
recent studies along the Oregon coast (Colle et al. 2008) and the Pacific Land-falling
Jets Experiment (PACJET: Neiman et al. 2008).
One recent series of studies in the coastal mountains of northern California north
of San Francisco illustrates the effects of the ambiguities between radar reflectivity
and rainfall rate (or cloud LWC). In this region precipitation from wintertime storm
systems hitting the coast, enhanced by orographic lifting, can produce significant
rainfall and valley flooding, but the NWS operational WSR-88Ds often have trouble
detecting the precipitation responsible (White et al. 2003), so that river forecasters
have difficulty anticipating the flooding. The California Land-falling Jets (CALJET)
project during January through March 1998 (Ralph et al. 2003) addressed this issue
by establishing research measurement sites, including one at a northern California
coastal location and a second 10 km inland at a coastal mountain site above the
town of Cazadero. Both sites had surface measurements of rainfall rate and other
meteorological variables. The mountain site also featured a 10-cm radar operating
in vertically pointing mode to obtain profiles of the radar reflectivity and vertical
motion of the raindrops, which can be nearly equated to their fall speeds. Although
operating at essentially the same œ as WSR-88Ds, this instrument, referred to as
SPROF (S-band, profiling radar), was specially engineered to have the sensitivity to
detect the weaker returns from smaller droplets.
During CALJET and subsequent projects along the West Coast the SPROF radar
observed different types of precipitation, which White et al. (2003) divided into
categories, two of which were bright-band (BB) precipitation and non-bright-band
(NBB) precipitation. The bright band is a radar reflectivity maximum in the vertical,
characteristic of stratiform precipitation, commonly observed near the freezing
level (Austin and Bemis 1950; Lhermitte and Atlas 1963). This peak occurs as
the low-reflectivity snow drifting down from aloft with small fall speeds begins to
melt (Fig. 8.19). The partially melted snowflakes in the melting layer exhibit high
reflectivity characteristic of water drops of the same size, but with low fall speeds
and therefore high particle-count densities characteristic of the snow, producing
much larger Z within the melting zone than above it. Fully melted raindrops fall
from the melting zone with much greater fall speeds, producing smaller drop
number densities and smaller reflectivities below the melting zone than in it, thus
resulting in the highest reflectivities within the melting zone (Fig. 8.19).
NBB rain was nonconvective (stratiform) rain that had no bright-band radar echo.
It was generally shallower than BB systems and more associated with orographic
upslope flow. A third “hybrid” category had characteristics of both other types.
Some storms were mostly BB or mostly NBB, but others had periods that alternated
between the two (Fig. 8.20).
Fig. 8.19 Contoured frequency-altitude-diagram plots of vertical velocity (left, m s1 downward
positive) and S-PROF radar reflectivity (right, dBZ) for winter 2003–2004 in the California coastal
mountains, for periods with bright-band (BB) rainfall (top panels) and for rainy periods with no
bright band (NBB). Heights were all normalized by bright-band heights, to the mean height for
the period of measurement (white line in upper right panel) (Martner et al. 2008. © American
The presence of a bright band implies the existence of active and effective cold-
cloud precipitation processes above the freezing level, which requires some depth
to the cloud, implying cold cloud tops. Strong SPROF reflectivities of 20–30 dBZ
in this upper layer (Figs. 8.19–8.20) support the existence of precipitation-sized
particles. During NBB events, the absence of a bright band and weak reflectivities
Fig. 8.20 Vertically pointing S-PROF reflectivity (dBZ, color scale at right) (Martner et al. 2008.
aloft suggest that, although ice particles may be present in the upper portions of
the clouds, they are not producing large falling snow or other precipitating ice to
any significant degree, and therefore the precipitation is mostly produced by warm-
rain (liquid-phase) collision-coalescence processes occurring near and/or below the
freezing level. Although NBB systems tend to be associated with lower rainfall
rates, they are also capable of producing the high rain rates (as measured at the
surface sites) that forecasters in the area associate with flooding potential.
White et al. (2003) and Martner et al. (2008) were able to find periods of each
type of rainfall when the rainfall rates R were the same. They compared such periods
for BB and NBB precipitation and found that the low-level SPROF reflectivities
were much smaller for the NBB precipitation, illustrating the non-uniqueness of the
relationship between R and reflectivity Z. White et al. hypothesized that this was
because BB precipitation was in the form of fewer large raindrops, which dominated
the strong reflectivities, whereas NBB rain was comprised of large numbers of
smaller drops, which produced much smaller reflectivities because of the r6 size
dependence on backscatter. This hypothesis was confirmed by surface drop-size
distributions measured during a subsequent field program in the winter of 2003/2004
(Martner et al. 2008). The impact of these findings on flood forecasting using the
WSR-88D network is two-fold, as summarized in Fig. 8.21. First, droplet sizes for
NBB rainfall, which occurs a large fraction of the time, are too small to be detected
by the operational radar. Second these rain systems are shallow and fall below the
lowest radar scanning angles. This latter problem is compounded by the way in
which the radar beam is bent or refracted upward in the lower atmosphere, even as
compared with visible light, as indicated by the upward curvature of the radar beam
in the figure.
The use of polarization to understand precipitation processes was investigated
during cold-season rain events in the European Alps during the Mesoscale Alpine
Fig. 8.21 (a) Composite winter-season profiles of (left) vertical velocity (m s1 ) and (right)
radar reflectivity dBZ for bright-band rain (solid lines) and non bright-band rain (dotted lines).
Altitudes for the dataset were normalized by bright-band height as in Fig. 8.19. (b) Conceptual
diagram of shallow NBB rain in the Coastal Mountains of California, and problems encountered by
operational WSR-88D radars in attempting to sample these clouds (White et al. 2003. © American
Programme (MAP). One of the cases described both in Chaps. 1 and 6 is the study
described by Medina and Houze (2003) and Rotunno and Houze (2007). Figure 6.4
from Chap. 6 contrasts a case of precipitation resulting from stable blocked flow
(left panels) with a case of precipitation from more unstable inflow (right panels),
which blows unblocked over the barrier, but drops the heaviest rainfall upstream of
the main Alpine barrier over the first topography encountered. These storms were
probed with NCAR’s SPol radar, which has Doppler and polarization capability.
The lower panels, which depict vertical cross sections of the Doppler velocity, show
that for the blocked case, the low-level wind speeds decrease as the flow approaches
the barrier, whereas in the unblocked case, a jetlike flow approaching the barrier
accelerates over the highest terrain, which reaches 3.5 km ASL. The precipitation
echoes (top panels) in the blocked case are more stratified, with a well-defined
bright band. In the unblocked case the neutral to slightly unstable approach flow
produces precipitation cells preferentially located over the first elevated topography
encountered by the moist airstream.
Polarization data from the SPol radar were used to infer particle type using the
previously mentioned scheme of Vivekanandan et al. (1999). In the case where
the low-level flow was blocked, the particle categorization indicated a layer of
intermittent dry snow in the large-scale ascending flow above the blocked layer,
and a layer of wet snow below the freezing level and bright band. This hydrometeor
configuration is consistent with gentle ascent over the major barrier, formation of
falling snow by ice and riming processes, and melting as the snow falls through
the freezing level. The unblocked case also exhibited layers of dry snow over
wet snow in the air ascending the large-scale barrier, but the cellular reflectivity
maximum over the first upstream orography had a graupel signature, suggesting
a convective nature to its precipitation, and indicating the importance of riming
processes in locally augmenting precipitation as well as the precipitation efficiency
in this location. The enhanced efficiency increases the removal of moisture from
the airstream, diminishing the amount of moisture available for falling out far-
ther downstream in the Alpine mountains. For accurate quantitative precipitation
predictions in the mountains, it is therefore important to represent this upstream
moisture removal mechanism in models used for this purpose. The radar analysis
algorithm thus produced a plausible distribution of precipitation types in agreement
with expectation.
A dataset obtained during a field campaign in the mountains of northern
New Hampshire proved useful for a radar operating at œ D 8 mm to investigate
downslope-flow phenomena during snowstorms. The project, the Mount Washing-
ton Icing Sensors Project (MWISP) in April 1999, was designed to test methods
of detecting aircraft icing conditions in wintertime cloud and snowstorm systems,
which may be transferable to operational situations. The structure and dynamics of
windstorms occurring on the lee slopes of mountain barriers are difficult to study
using longer-œ radar, because these radars preferentially sense larger particles that
do not follow the air flow very well; they have a wide beam and significant sidelobes
that prevent obtaining data near the topography; and downslope winds are often
Fig. 8.22 Vertical-slice scan from Mount Washington experiment (west to right) at 0003 UTC,
17 April 1999, showing (top) Doppler-radar radial velocity (color bar in m s1 ) and (bottom)
reflectivity (color bar in dBZ). Mount Washington Observatory (MWO) and radar site (CRB)
indicated (Martner et al. 2002. © American Meteorological Society. Reprinted with permission)
dry and devoid of hydrometeors. But radars of shorter-œ are capable of sensing
smaller particles, which are better tracers of the air flow, and can use antennae
more suited to mountain studies because of narrower beams and much reduced side
lobes.
Despite being to the lee of the barrier, the snow and ice particles in the airstream
were plentiful enough to provide a clear picture of the airflow, as shown in Fig. 8.23.
As also shown in Fig. 8.22, the flow had a two-layer structure, with westerly flow
aloft and easterly-component winds in the lowest km flowing over the ridge, so that
the radar observation site “CRB” was downwind with respect to the lower-layer
flow coming over the ridgetop, but upwind with respect to the 12 m s1 upper-
layer westerly flow (top panels). The lowest part of the westerly flow layer aloft
was strongly sheared, as evident in the high-reflectivity precipitation streamers in
the lower panel. The easterly-component flow over the ridgetops was laden with
Fig. 8.23 Same as Fig. 8.22, except for 0333 UTC, 17 April 1999 (Martner et al. 2002.
snow and ice particles from an upstream cell of high reflectivity, and snow may also
have been picked up from the surface by the wind. The snow particles remained
suspended in the flow to provide visual signal in the downslope flow, showing the
turbulent nature of this flow as it passed over the radar site.
By 3½ h later (Fig. 8.23) the structure of the snowstorm had evolved into an
upslope snowfall event in the westerly flow aloft, with a high-reflectivity band
sloping upward toward the peak (Martner et al. 2002). The snow falling from
this upper-level precipitation into the lower layer of easterly-component flow gave
continuous signal all the way to the surface. This continuous signal gave a very
clear picture of the layer of downslope flow from the east, and revealed that this
flow was in the form of a hydraulic jumplike feature (Fig. 8.23, top) similar to
those observed in laboratory flows and by Doppler lidar (see Sect. 8.7), including
flow reversals, or rotors, aloft at x D 2–3 km and at the surface westward from
x 1.5 km.
In another case study Reinking et al. (2003) investigated stable flow approaching
the barrier from the west that was blocked in the lowest few hundred meters. The
Fig. 8.24 Same as Fig. 8.22, except for 1515 UTC, 3 April 1999 (Reinking et al. 2003.
transition between the 13 m s1 winds 800 m above the valley floor and higher
and the stagnant (or reversed) flow in the blocked layer was a layer of strong shear.
Within this shear layer the radar scans revealed a striking wave structure to the flow
(Fig. 8.24), which Reinking et al. (2003) analyzed in detail.
8.5.4.2 Warm-Season Orographically Generated Moist Convection
Examples of the use of multiple Doppler radars to probe mountain thunderstorms

are rare in the literature, but some very nice illustrations come from the Colorado
mountains. Summertime thunderstorms originating in mountainous terrain are an
important source of growing-season moisture in the mountains and in regions
downwind, such as the U.S. Great Plains. Hail and other severe weather produced by
such thunderstorms have an adverse effect on crops and other human activity, and
lightning can start wildland fires. A detailed review of comprehensive studies of

deep moist mountain-generated convection in the Western U.S. was given in Banta
(1990). Results of recent field programs in the U.S. (Damiani et al. 2008) and Europe
(Kottmeier et al. 2008; Wulfmeyer et al. 2008) are becoming available.
Doppler radar has been an important tool for studying mountain thunderstorms.
An issue is whether significant differences exist in the structure and behavior of
mountain storms versus those studied over simpler topography. The studies of
Cotton et al. (1982) and Knupp and Cotton (1982a, b) used multiple-Doppler-
radar analyses of data taken in the Colorado mountains to study the initiation,
structure, and development of multiple thunderstorms that formed along a north–
south convergence line in a broad mountain basin (Fig. 8.25), and how the nature of
the convection changed as a moist Pacific front pushed south through the area.
A special problem for multiple Doppler computation of the three-dimensional
wind vector in the mountains is the boundary condition required for dual-Doppler
(and desirable even for multiple-Doppler) calculation of the vertical-velocity profile
w(z) using the continuity equation. Over flat terrain this integration begins by
assuming w D 0 at the surface, but this is not useful over sloped mountainous
topography, largely due to “unavailability of consistent multiple-Doppler data at
low levels.” Knupp and Cotton (1982a, 1987) addressed this by applying w D 0 at
the cloud echo top, assessing the resulting error to be <5 m s1 .
During early afternoon in the southern region of the basin, the thunderstorms
formed sequentially along the north–south convergence line, moved northward
with continuously evolving structure in the southerly ambient flow, and dissipated
(Fig. 8.25b). Later in the afternoon after the cold front passed through, however,
postfrontal conditions left a moister, more strongly sheared environment more
conducive to quasi-steady, propagating storm systems. As a result, as the front
passed through the southern portions of the basin, one of the southernmost cells
developed into such a large, quasi-steady storm, similar in structure to those
studied over the nearby Great Plains. The Doppler analysis (Fig. 8.26) showed
the ambient flow modified by being deflected around the updrafts and downdrafts
of the primary storm, diverting the direction and speed of propagation of smaller
nearby cells and providing clear examples of storm-storm interactions. Multiple
Doppler analysis of mountain thunderstorms thus indicated structure and behavior
similar to that observed over flatter terrain. The role of the topography was in
the localization and strength of initiation mechanisms (Banta 1990) and in the
channeling and “drainage” (downslope diversion) of gust fronts and other cold air
flows, as documented by surface mesonet data, which affected cell propagation and
initiation of subsequent cells (Cotton et al. 1982; Knupp and Cotton 1982a).
Other, larger-scale convective systems were also studied by radar, including the
life cycle of a mesoscale convective system featuring convective and stratiform
rain regions similar to those observed outside the mountains (Knupp and Cotton
1987), and a thunderstorm system that began in the mountains and propagated onto
the adjacent plains, where it spawned a tornado near Colorado Springs (Wetzel
et al. 1983).
8 Observational Techniques: Sampling the Mountain Atmosphere
Fig. 8.25 (a) Map of the mountain basin, South Park, Colorado, location of the measurement campaign. The three radar locations are indicated by black
triangles and colored yellow, and distance between the two southernmost mesonet sites is 27 km. (b) Radar reflectivity contours of storms in South Park plotted
with mesonet data for 2223 UTC (1523 local standard time), 19 July 1977. Solid reflectivity contours are 35 dBZ, and dotted contours, 50 dBZ. Five-minute
averaged potential temperature (K), water vapor mixing ratio (g kg1 ), and winds are plotted for each station. The three radar locations are again indicated by
black triangles and colored yellow (Cotton et al. 1982. © American Meteorological Society. Reprinted with permission)
461
Fig. 8.26 Conceptual model of the flow patterns within a mountain thunderstorm cell during its
intense quasi steady stage, as inferred from dual Doppler lidar analysis. Streamlines show storm-
relative airflow in each horizontal plane, and the ribbon arrows represent updrafts and downdrafts.
H’s and L’s denote regions of strong and weak flow, respectively, at the lower (l), middle (m), and
upper (u) levels. Heavy rain in the lower level is indicated by cross hatching (Knupp and Cotton
1982a. © American Meteorological Society. Reprinted with permission)
8.5.5 Discussion
It is difficult to overstate the significance of weather radar to modern-day me-

teorology, especially in light of the importance of water to human activity. The
use of storm or surveillance radar in the detection, tracking, and characterization
of severe storms is well known. Added capabilities, including the use of shorter
wavelengths, polarization, Doppler, the use of multiple wavelengths, and the use of
multiple Doppler radars have added significantly to the ability to understand and
characterize a much wider range of cloud and storm properties. This insight into
relevant microphysical and dynamical processes feeds into operational applications
in hydrology, aircraft icing, quantitative precipitation forecasting, and many others
in mountainous terrain.
Table 8.6 lists many of the advantages of radar as an atmospheric sampling
system, including the ability to detect storm systems at long range, its familiarity,
and its versatility to measure many different useful quantities. Some limitations
include the need for hydrometeors as scattering targets (except conditions where
insects are available), ground-clutter and sidelobe interference especially in complex
Table 8.6 Radar measurements

Distant detection of storm and precipitation Many types of radar application require
systems—long range hydrometeors for best performance
Senses clouds, precipitation Sidelobes, ground clutter often a problem for
interpretation in complex terrain
QPE—precipitation monitoring Angular resolution coarse for some applications
Scans fast Doppler velocity includes fall speeds of drops
Storm dynamics from Doppler In mountains, beam width, blockage may
prevent detecting lower parts of echo
Microphysical info from polarization, multi-œ –
Familiar—workhorse for operational –
meteorologists, hydrologists
Airborne—research and operational –
(e.g., hurricane)
Table 8.7 Operating characteristics of sample radar wind profilers

Frequency Altitude Typical peak
Type (MHz) œ (cm) range (km) Pulse length (m) r (m) power
UHF boundary 900–1,200 33–23 0.20–4 80–200 40–100 400 W–4 kW
layer
UHF 449 67 0.20–16.25 210, 255, 1,020 105–510 2–50 kW
tropospheric
VHF 50 600 1–20 300–2,000 5 kW–1 MW
terrain, and resolution which may be too coarse for some applications. The degree
to which they pose a problem depends on œ and of course the available resources to
study and implement advanced engineering solutions.
8.6 Radar Wind Profilers
Radar wind profilers (RWPs) are a class of radar that have a history of atmospheric
studies dating back several decades, and have been recently applied to studies of
mountain weather. A variety of different styles of wind profiler have been developed,
but all measure the wind profile in the atmosphere, as indicated in Table 8.7. Gener-
ally wind profilers are aimed upward to measure the wind profile of the atmosphere
directly overhead. They use longer wavelengths than scanning weather radar, which
allows them to detect echoes from the “clear-air” (such as turbulent fluctuations) in
clear and cloudy conditions, and many also detect precipitation overhead. The clear-
air returns are most often turbulent index-of-refraction fluctuations (Bragg scatter),
which tend to be dominated by moisture fluctuations as discussed in Sect. 8.3. RWP
technology is familiar to most meteorologists. It has been established and well char-
acterized for more than two decades and is used operationally for many applications.
8.6.1 Types of Profiler
Wind profilers come in several sizes (Table 8.7). The NWS “National Profiler
Network” (NPN), a network of 18 UHF tropospheric profilers distributed mainly
around the Great Plains, are examples of large powerful wind profilers, capable
of measuring up to 15 km altitude. More common are boundary-layer wind
profilers that provide profiles from 100 m up to 2–4 km, depending on signal
strength. At least a hundred such instruments operating at 915 MHz are scattered
around the U.S. and operated by a wide range of agencies such as the Federal
Aviation Agency (FAA), the California Air Resources Board, NOAA, NCAR, and
universities.
The most common wind measurement technique is known as “Doppler Beam
Swinging” (DBS), in which the radar points vertically and measures the Doppler
shifted frequency (giving vertical velocity), then tilts to some angle (for example,
20ı off vertical northward), measures the Doppler shift in that direction, then tilts in
an orthogonal direction (e.g. 20ı off vertical to the east). A typical “dwell time”
in each direction is 2 min, so the entire sequence takes 6 min to measure all
three wind components. Often five beams are used (e.g. to the south and west) to
provide some redundancy. The three Doppler shifts can then be put together with
trigonometry to derive the three-dimensional wind vector. The tilting is usually done
not by physically turning the radar, but using a flat phased-array antenna, which is
electronically steered by adjusting the phase of elements across the antenna (Cohn
et al. 2001).
One disadvantage of the DBS technique is that at significant heights above the
ground, the points in the atmosphere being measured on the tilted beams (the
“sampling volumes”) can be widely separated. For example, using beams having
a zenith angle of 20ı , the separation of the sampling volumes at 3 km altitude
is 1 km. The technique usually requires longer averaging times (typically 30–
60 min) of the beam-angle sequences to achieve ˙1 m s1 precision (Martner et al.
1993; Angevine and MacPherson 1995; Angevine et al. 1998; Baumann-Stanzer
2004). An inherent assumption is that over such long time periods, the motion of
the atmosphere over these widely separated volumes is similar. In mountainous
terrain the assumptions of homogeneity across the beam and stationarity may not be
valid under all conditions. Sidelobes can produce unpredicatable results when strong
scattering or reflecting targets are nearby to generate backscatter. As an example, the
discrepancies described in Sect. 8.1 (Fig. 8.2) between the shipborne profiler and
Doppler lidar wind-profile measurements were attributed to the reflection of profiler
sidelobe signal off the ship and nearby solid objects on deck and the subesquent
scattering by ocean waves, referred to as “sea clutter” (Wolfe et al. 2007). The sea
clutter often produced very noisy and erroneous returns in the lowest few hundred
meters of the wind profile, but reasonable values at higher altitudes of the profile,
when compared with simultaneous rawinsonde and Doppler-lidar profiles. These
kinds of effect can also be found in mountainous topography when terrain or other
obstacles are near the transmitter, so that careful siting and orientation of the radar
antenna may be required to minimize side-lobe interference with the desired radar
signal.
An alternative wind-measuring technique that has been demonstrated is the
spaced-antenna technique in which three or more separate but closely spaced
antennas are used. As atmospheric inhomogeneities move over the radar, the
separated antennas see their reflected signals at slightly different times, depending
on the antenna orientations to the wind. Since the antenna separation is known
(typically a few meters), the time differences can be used to derive the wind vector.
The advantage of this technique is that a wind measurement can be made very
quickly (typically in a minute or less). The disadvantage is that these radars are
more complex and require considerably more powerful transmitters.
Three radio-frequency bands are generally used by wind profilers (Table 8.7)—
as a rule, the longer the wavelength (i.e. lower the frequency), the higher into the
atmosphere the systems can probe. This is because the minimum scales of turbulent
eddies tend to increase in size with altitude, requiring longer wavelength radars to
detect the Bragg backscatter from the eddies. Longer-wavelength radars are larger
in physical size (for example 50 MHz systems capable of reaching 20 km may be
more than 100 m across) and also interfere with television broadcasts, so they are
generally operated only in remote locations.
A problem peculiar to RWPs is the detection of flocks of migrating birds, which
overwhelm the weaker Bragg scatter from which the winds are calculated. The
calculated “winds” thus represent bird motion rather than atmospheric flow. When
this interference occurs, it has been found to be a problem mostly above 1 km.
Methods for extracting the atmospheric signal have been devised (Merritt 1995;
Wilczak et al. 1995), but at times the interference is too strong even for these
techniques, and atmospheric wind data are unobtainable.
In addition to the winds, profilers also obtain profiles of the radar reflectivity. The
reflectivity reaches a maximum in the inversion at the top of the daytime convective
BL, and this peak provides a measurement of the BL depth. This topic is further
described in Sect. 8.10.
8.6.2 Temperature Profiles
A potentially useful extension of profiler capabilities is the measurement of

temperature profiles, obtained by adding an acoustic source to the transmission site,
to send a sound pulse up into the atmosphere. The acoustic frequency is chosen to
maximize Bragg scattering of the radar signal. The frequency range of the profiler
receiver is switched to sense the speed of sound, which is directly related to the
virtual temperature. Temperature profiles are generally available from 100 up to
as much as 1,000 m at height intervals of 50 m, using a 915 MHz profiler. An
example will be given in Sect. 8.8.
8.6.3 Profiler Applications
Wind measurements from these radars can be ingested into numerical models
in a similar manner to measurements from soundings. Studies have shown that
ingested measurements from the NPN profilers in the Great Plains can significantly
improve the forecasting of severe weather events in that region. Reflectivity
measurements from boundary-layer profilers are frequently used by air quality
agencies to determine daytime mixed-layer depths (thus pollution dilution) for air
quality forecasting.
The technology of radar wind profilers is mature enough that they have been
used for aviation applications. The FAA operates three wind profilers around Juneau
airport in Alaska. This airport is surrounded by steep terrain, and aircraft using the
airport experience considerable wind shear and turbulence. The profilers are used in
conjunction with surface sensors and numerical models to diagnose terrain-induced
flows that produce potentially hazardous landing or takeoff conditions. Profilers are
also a part of the wind-shear warning system at the Hong Kong International Airport
(Shun and Chan 2008), as described in Sect. 8.7.
Wind profilers also play key roles in mountain weather research programs.
A recent example is the T-REX (Terrain-Induced Rotors EXperiment) carried out
in the Sierra Nevada area in 2006. Three wind profilers (along with a wide range
of other instruments) were operated in the Owens Valley in this study of mountain
waves, rotors, and boundary layer dynamics. Figure 8.27 illustrates an example of
their observations during an easterly flow event over the Owens Valley. The three
wind profilers were operating in an east–west line across the north-south valley at
spacing of approximately 10 km; ISS2 was a DBS profiler on the western slope,
MAPR is a spaced-antenna profiler in the center of the valley, and MISS was a
mobile DBS profiler (mounted on a trailer) on the eastern slope.
The profilers observed the dominant easterly flow above the ridgetops. At lower
levels (below the 1-km altitude of the eastern ridgeline) the flow tended to be
channeled either southerly or northerly along the valley, however at times (circled
in red) westerly flow was observed across the valley, indicating a reverse-flow rotor
circulation.
On a larger scale, radar wind profilers were deployed to the Central Valley of
California during the Central California Ozone Study in summer of 2000. Analyses
of the diurnally varying flows into and out of the valley were analyzed and compared
with mesoscale model results by Zhong et al. (2003), Bao et al. (2008), and
Michelson and Bao (2008). Other examples of the use of profiler technology in
mountainous terrain include a year-long study of wind flow in opposing drainage
valleys in Colorado (one to the windward side of the Continental Divide and one im-
mediately opposite on the leeward side) by King (1997), the VTMX program (Doran
et al. 2002; Shaw et al. 2003; Banta et al. 2004); and others reported by Neff (1990).
Profilers can thus provide wind profiles to fill in the times between the twice-
daily rawinsondes. The technology is well established and has many advantages
(Table 8.8) including that wind profiles can be obtained in most weather conditions.
Fig. 8.27 Data from three profiler sites sited in an east–west line across the north south Owens
Valley during TREX. Wind speeds color coded (scale at right in m s1 ). Background synoptic
ridgetop flow was easterly on this day. Ellipses indicate regions of reversed rotor flow
Concerns for complex terrain operations include the separation of the sampling
volumes for the various wind components, which may violate the assumption of
horizontal uniformity of the flow in regions where the flow varies strongly in the
horizontal, and interference by sidelobes.
8.7 Lidar
Atmospheric lidar systems use ultraviolet (UV), visible, or infrared (IR) light at
wavelengths ranging from approximately 200 nm to 10 m. The different operating
wavelengths used by lidar are affected by very different atmospheric scattering
Table 8.8 Radar wind profiler measurements

Frequent wind profiles Minimum range, resolution coarse for some
Suitable for assimilation app’s (e.g., SBL, slope flows)
Reflectivity also useful Spatial separation (3 or 5 beam config)
Daytime mixed-layer depth available Susceptible to RF interference, especially urban
Many exist, can deploy in arrays, use Side-lobe effects
for trajectory calc’s
Unattended operation Migrating-bird contamination
All-weather –
and transmission properties, and as a result, “lidar” refers to many different types
of instrument. Scattering targets at the longer wavelengths are primarily aerosol
particles, but in the shorter visible and UV œ’s, scattering by molecules becomes
increasingly important. Since aerosol and molecules are universally present in the
atmosphere, the various types of lidar are important observational tools in clear air,
i.e., in the absence of cloud or in regions of weak turbulence. On the other hand,
light at lidar œ’s is attenuated by liquid water, so penetration of liquid-water clouds
or fog by lidar signals may often be limited to 200 m or so, depending on the
specific œ and cloud liquid-water contents involved (e.g., Tucker et al. 2009).
The use of optical œ’s, which are much shorter than the radio or sound waves
employed to probe the atmosphere, results in important differences between lidar
and other remote sensing instruments, such as weather radar. Instead of an antenna,
lidar uses a telescope to project and collect the signal. The resulting beam is
very narrow, sometimes referred to as a “pencil beam,” although wire- or filament
beam are probably more appropriate descriptors of the true aspect ratios involved.
Sidelobes are absent, for all practical purposes. The narrow beam and lack of
sidelobes are significant advantages for mountainous terrain studies, as clutter
(unwanted signals) due to stationary ground targets is almost a negligible effect
compared with radar or sodar. Lidars are generally used to probe shorter ranges
than radar, generally less than 30 km, so “second trip” echoes are not an issue for
PRFs typically used. In many Doppler-lidar systems the pulse-to-pulse frequency
can vary slightly, so to achieve desired velocity precisions on the order of 10 cm s1 ,
the exact outgoing pulse frequency is monitored for comparison to the return signal
(Post and Cupp 1990), which means that signals from each pulse must undergo some
processing as they arrive, before further averaging or other signal processing.
Because lidars operate at wavelengths in or close to the visible, eye safety is an
important consideration. Laser emission at visible wavelengths is also a concern to
aviation, as it could cause confusion to pilots especially near airports. One approach
has been to use non-eyesafe œ’s but direct the beam away from human activity
or employ a “spotter” to shut down transmissions when pedestrians, vehicles, or
aircraft come within range. Another approach is to avoid the œ’s that transmit
through the eye to the retina; these hazardous œ’s include near-UV, visible, and near-
IR œ’s less than 1.45 m. CO2 lasers, which generate an IR signal at 10 m, and
solid-state lasers at 2.02 m have been used successfully in eye-safe Doppler lidars
for many years (Post and Cupp 1990; Grund et al. 2001). More recent emphasis by
the telecommunications industry in developing electronics at eyesafe wavelengths
near 1.5 m have made off-the-shelf components available at reasonable cost,
sparking a rapid increase in cost-effective lidar systems, including some commercial
systems, at that wavelength. A third approach to eye safety is to use eye-sensitive
œ’s, but at low pulse power and/or an expanded beam cross section at very high
repetition rates. The low-power signal is very noisy, but useful signal is obtained
by averaging over many thousands of pulses, the so-called micropulse approach
(Spinhirne et al. 1995; Eloranta 2005).
The breadth of lidar operating œ’s provides a rich diversity of interactions
between the beam and the atmosphere that have been exploited for mountain
meteorological applications in many ways. Scattering by aerosol particles provides
information on the structure of aerosol layers, and polarization of the return
signal gives important information about particle shape. Absorption by specific
atmospheric constituents or frequency-shifted scattering can be used to remotely
profile concentrations of atmospheric species such as water vapor and ozone. The
frequency of the backscattered signal can be analyzed to yield information on
atmospheric motions using Doppler processing. These approaches are described in
greater detail below, following the discussion of Hardesty and Darby (2005). Books,
and other reviews about the use of lidar in the atmosphere are available (Measures
1984, Schwiesow 1986b, Weitkamp 2005; Hardesty and Darby 2005).
8.7.1 Types of Lidar
Scattering and transmission properties of light in the atmosphere are fundamental

to lidar measurements (Schwiesow 1986b). Scattering properties of individual
atmospheric particles and other targets, as well as populations of scatterers, were
described in Sect. 8.3.3.2. Size distributions of atmospheric aerosol often peak at
diameters on the order of 1 m. Such atmospheric aerosol particles may include
organic and inorganic particles, such as sulfates and nitrates, dust, soot, smoke, and
pollen, as well as liquid water and ice. For lidar systems operating at œ D 10 m,
scattering by such particles is in the Rayleigh regime, but for solid-state systems
transmitting signal at 2 m or less, these particles induce Mie scattering, consistent
with the discussion in Sect. 8.3 of this chapter. For most mountain boundary-
layer applications the primary interest has been in applications using aerosol
backscatter (as opposed to molecular). More generally lidar applications have also
used molecular scattering, many of which have involved determining temperature
and density profiles at high altitudes above the stratosphere (Behrendt 2005; Abo
2005). Lidars designed to exploit molecular scatter generally operate at short visible
or ultraviolet wavelengths, to take advantage of the strong increases in backscatter
at shorter œ’s (which go as œ4 for Rayleigh scatter (Sect. 8.3)), to maximize the
backscattered signal.
Some lidar systems designed to study quantitative aerosol properties in the

boundary layer and troposphere operate at œ 532 nm, where both aerosol and
molecular contributions are significant. In these cases, the aerosol backscatter is
determined by subtracting the molecular contribution from the total measured
backscatter; therefore, accurate estimates of molecular backscatter are required.
Techniques for separating aerosol from molecular backscatter using High Spectral
Resolution Lidar (HSRL) will be described in the next section (Eloranta 2005).
Rayleigh and Mie scattering processes are both elastic, i.e., the œ of the scattered
signal is not altered by the scattering process. Scattering from atmospheric atoms
and molecules can also produce, in addition, Raman scatter, in which energy is
exchanged between the scattered photon and the scattering species, changing the œ
of the scattered light. A fundamental characteristic of Raman scatter is this œ shift of
the scattered light from that of the incident light to a different œ by an increment that
is unique to the chemical species of the scattering molecule. As a result, the Raman-
scattered spectrum can be analyzed to identify the molecule (e.g., nitrogen or water
vapor), determine its number density, and estimate its energy state by comparing
the populations in different energy levels. The Raman scattering component of
backscattered light is typically weaker than the Rayleigh-scattered radiation by
a factor of 102 –103; however, use of high-quality optical filters to separate the
Raman-scattered light enables identification and monitoring of specific atmospheric
constituents such as nitrogen and water vapor (Wandinger 2005; Hardesty and
Darby 2005). As a result of the need to detect a weak signal, these systems often tend
to be large (physically). Raman techniques have been used successfully to profile
water vapor (Melfi et al. 1989; Goldsmith et al. 1998; Eichinger et al. 1999; Turner
et al. 2002; Whiteman 2003, Grzeschik et al. 2008), and the technique has also been
developed to measure temperature profiles.
Some of the laser light propagating toward and back from the scattering volume
may also be absorbed by specific atmospheric molecules and particles along
the propagation path. Gases and particles typically have distinctive absorption
characteristics as a function of œ. For example, most atmospheric gases absorb
only in certain discrete wavelength bands, whereas the absorption cross-sections
of aerosol particles usually vary slowly with wavelength. Absorption spectra of
atmospheric gases may be highly structured (Fig. 8.28), particularly for infrared œ’s
(e.g. water vapor, methane, and carbon dioxide), or they may be characterized by
rather broad absorption features, as is often the case for ultraviolet œ’s (e.g. ozone).
A technique known as differential-absorption lidar (DIAL) takes advantage of the
wavelength dependence of the absorption cross section to measure gaseous species
concentrations (Bösenberg 2005). In DIAL applications, usually two wavelengths
are transmitted, chosen such that the laser light at one wavelength is more strongly
absorbed by the species of interest than at the second wavelength (Fig. 8.28). The
difference in attenuation along the transmit/receive path is expressed as the ratio
of the two signals. From this ratio the concentration of the species of interest may
be calculated (Schotland 1974; Browell et al. 1985). An advantage of using this
ratio approach is that the lidar system quantities, such as the transmitted power
and the receiver efficiency, appear in both the numerator and denominator of the
Fig. 8.28 Absorption bands and DIAL: abscissa is wavelength (nm) in the UV and colored curves
show absorption cross section for O3 (red line) and two other gases (blue, green lines). Wavelength
range over which laser can be tuned is indicated by gray box, and three wavelengths that could be
selected for O3 DIAL measurements within this tuning range are indicated by black vertical lines,
which intersect the O3 band at three different absorption levels
ratio and thus cancel, making this approach “self calibrating,” assuming sufficiently
strong signal in both channels. Accurate quantitative data on species concentrations
can be obtained without accurate knowledge of transmitted power or other system
quantities.
One of the challenges to using the DIAL technique is to find an absorption
feature that is useable, given the atmospheric concentrations of the species to be
measured. If the absorption is too strong, the maximum range of the DIAL system
will be very limited, and if, on the other hand, the absorption is too weak, the
difference in the attenuation between the two DIAL wavelengths will not be enough
to make a meaningful measurement. Atmospheric gases that have been measured
using the DIAL technique include water vapor, ozone, sulfur dioxide, nitrogen
dioxide, ammonia, methane, carbon dioxide, mercury, and several volatile organic
compounds (VOCs). DIAL provides profiles of species concentrations that can be
used to form time-height plots. On mobile platforms it can provide measurements
of the three-dimensional distribution of atmospheric trace gases at high spatial and
temporal resolution (Sect. 8.7.4). This is especially desirable in mountainous areas
where a complicated flow field often results in a very heterogeneous distribution of
atmospheric gases.
Lidar systems are typically categorized according to the technique used to detect
the light returning from the atmosphere: coherent (heterodyne) and incoherent (or
direct) detection. Laser sources typically oscillate in multiple frequency modes

and thus transmit a range of œ’s about the nominal transmitted œ, the outgoing
signal photons having random phase. In lidars using incoherent detection, the
detectors receiving the backscattered signal accept œ’s spanning the bandwidth of
the transmitted range regardless of phase, and thus make optimum use of available
return energy. Lidar applications requiring measurement of the backscattered signal
power as the primary variable of interest, such as those measuring aerosol structure
or atmospheric constituent concentration by either DIAL or Raman configurations,
typically employ direct detection. In a direct detection receiver, the backscattered
signal is focused on a detector, which uses either analog detection or photon
counting to produce a detector output signal proportional to the incident signal
power. Because a narrow signal bandwidth is not necessary, direct detection systems
can employ very short pulses which provide excellent range resolution (1 m)
in measuring the backscattered energy. Direct detection instruments can operate
across the spectrum from the near infrared to the ultraviolet; the wavelength region
is chosen depending on whether aerosol or molecular backscatter is of primary
interest.
For coherent lidars, frequency purity of the transmitted pulse is critical, because
the high sensitivity of coherent lidars is achieved through precise spectrum anal-
ysis of narrow-band returns. A frequency-stable transmitted pulse is achieved by
“seeding” the laser cavity with light of the desired frequency from another laser,
which induces the source laser to release its energy in phase at that frequency,
i.e., as a coherent laser pulse (Post and Cupp 1990; Grund et al. 2001). The
backscattered signal received by lidars using coherent detection is optically mixed
with a continuous-wave signal, offset slightly in frequency from the transmitted
pulse, generated by a local-oscillator laser, and the resulting beat frequency is
sensed by the detectors. The Doppler frequency shift and Doppler velocity are
calculated from this measured frequency. Measurement of frequency is one of
the best developed capabilities in contemporary technology, so this part of the
measurement can be done to high precision. In coherent systems, backscattered
signals that are not phase-coherent or that are shifted outside the narrow acceptance
frequency bandwidth do not mix efficiently with the local oscillator and degrade
sensitivity, hence the need for a narrowband, frequency-stable transmitter. Also,
because the ability to accurately determine the peak of the Doppler spectrum (see
Fig. 8.13) is inversely proportional to signal bandwidth, Doppler instruments are
usually designed with longer pulses to decrease the bandwidth, and therefore they
typically have poorer range resolution (30–100 m) than intensity-measuring direct-
detection systems.
Direct detection can also be used to measure Doppler shift by adding an
interferometer before the detector, which acts as a spectrum analyzer. However,
practical optical-frequency analysis implementations of direct-detection Doppler
receivers are less precise than those obtainable using heterodyne techniques. So
although direct-detection lidars can measure winds from both aerosol and molecular
scatter, their reduced sensitivity and velocity precision means that in general
heterodyne instruments are preferable when sufficient aerosol is present to produce
a heterodyne signal. Direct detection Doppler systems are most applicable when
aerosol scattering is low and the return is dominated by the wideband molecular-
scattered signal.
Heterodyne and direct detection lidars are often complementary. Heterodyne
instruments produce precise wind measurements at moderate range resolution
when aerosol is present. Although aerosol-backscatter-intensity measurements from
heterodyne systems are also obtained, single-shot precision is poor due to the
interference and phase effects, which introduce the uncertainty into single-pulse
measurements that must be reduced by averaging. Heterodyne lidars are most
applicable for wind measurements in the boundary layer where aerosols are
abundant. Direct detection systems, on the other hand, can measure backscatter
signal intensity having both high precision (with sufficient signal) and high spatial
resolution in either aerosol- or molecular-scatter regimes, so they are generally
applied in aerosol, DIAL, and Raman lidars.
8.7.2 Backscatter Measurement Examples
The depth of the atmosphere to which lidars can obtain data varies depending on the
type of lidar and also depend on atmospheric conditions. Using lidar to study the
atmosphere requires appropriate scatterers to be present in the layer sampled. For the
shorter visible and UV œ’s molecular scattering produces return signal throughout
the atmosphere, but these shorter œ’s are subject to stronger attenuation effects.
For longer IR œ’s that depend on aerosol particles, the distribution of aerosol con-
centrations determines the depth of lidar coverage. Researchers using such aerosol
lidars are well aware that useful lidar signal and better overall lidar performance
extends to greater ranges in polluted, smoky, or other conditions of high aerosol
loading, as compared with cleaner, more pristine conditions in the atmosphere.
Aerosol concentrations in the troposphere generally decrease with height, with
highest concentrations in the ABL. Abrupt decreases in aerosol backscatter with
height have been used to determine the top of the daytime convective BL (e.g.,
Steyn et al. 1999; White et al. 1999; Hägeli et al. 2000; Tucker et al. 2009, 2010;
Emeis et al. 2008). Many solid-state lidars see the depth of the daytime convective
BL but not much deeper. On the other hand, NOAA’s more powerful CO2 lidar
TEACO2 (Post and Cupp 1990; see Table 8.9), operating at 10.59 m, typically
obtained signal to 3–5 km AGL, well above the depth of the BL (Figs. 8.8a and
8.29).
This lidar was used in a multiyear study of the vertical distribution of aerosol
backscatter, in which vertical profiles were measured several times per week for
more than 6 years. During this period the normal 3–5 km-deep aerosol layer was
abruptly increased in 1991 to span the entire troposphere and lower stratosphere
(Fig. 8.30a), after Mt. Pinatubo in the Philippine Islands erupted in June 1991 and
injected volcanic ash and sulfuric acid droplets deep into the atmosphere (Post et al.
1996). During this time of enhanced vertical coverage, the lidar, sited mostly in
Table 8.9 Lidar

Min Max Pulse Eff
range range length PRF
œ (m) (m) (km) (m) r (m) PRF (Hz) (Hz) References
Doppler
TeaCO2 10.59 1,200 20 900 450 10 3 Post and Cupp
(1990)
HRDL 2.02 200 5 30 30 200 2 Grund et al.
(2001)
2.02 150 10 50–100
1.5 1.5 <50 3 <20 18 20,000 1 Pearson et al.
(2009)
Backscatter
HSRL 0.532 50 30 15 7.5 4,000 2 Eloranta (2005)
Fig. 8.29 Vertical-slice Doppler-lidar scan taken at Colorado Springs airport, with mountains to
the west (right) and flow from right to left. Lidar is at (0,0), Doppler velocities in m s1 (color bar
across bottom)
Boulder, Colorado just to the lee of the Rocky Mountains, could be used to reveal
flow features near tropopause level. For example, Fig. 8.30b, a vertical cross section
of Doppler velocity, shows nearly complete coverage from the surface to more
than 20 km above ground, and reveals a series of overturning waves at 12–15 km
altitude having a wavelength of 6 km. Another example of a lidar-observed upper-
tropospheric overturning wave observed during this period was modeled by Clark
et al. (2000). The upper-tropospheric/lower-stratospheric aerosol concentrations
gradually subsided to more normal levels over 3 years, and Post et al. (1996)
noted an e-folding time constant for stratospheric aerosol loading of 500 days.
Backscatter from a population of aerosol particles is a function of the con-
centration, size distribution, shape, and composition (index of refraction) of the
aerosol as described in Sect. 8.3.3.2. In situations where aerosol properties are
not changing, backscatter variation thus depends on aerosol concentration, once
the range dependence has been accounted for. But one cannot necessarily rely on
aerosol properties to remain static; for example, aerosol pollutants can adsorb other
Fig. 8.30 Time-height cross section of weekly aerosol backscatter sampling for a 6-year period,
showing increase in depth when Mt. Pinatubo erupted in 1991 (top) (Post et al. 1997). Vertical-
slice scan taken during period of deep tropospheric/lower stratospheric loading (bottom), showing
wave pattern over mountains to west (right), as indicated by white bars. Peak speeds were 30 m
s1 . Range rings at 10 km, wind flow from right to left (Scan data courtesy of M.J. Post)
pollutants to change their size and composition. In an unstable convective BL, where
species are expected to be well mixed, lidar backscatter is often observed to increase
near the top of the BL, where the relative humidity of rising parcels of warm air
exceeds 80%. Some aerosol species, such as those containing sulfates, absorb water
vapor and swell in size in such moist conditions (the process of “deliquescence”),
affecting both the size distribution and composition. In cumulus-topped BL’s these
water-attracting aerosol particles enter cloud base as cloud-condensation nuclei
(CCN). An example of such an enhancement of backscatter in air entering cloud

base has been observed by Doppler lidar in a cumulus cloud capping a forest fire
convection column (“pyrocumulus”) (Banta et al. 1992). These examples illustrate
the limitations to assuming that aerosol concentrations are proportional to lidar
backscatter in a straightforward manner.
Enhanced backscatter from tracer materials released into the atmosphere has
been used in diffusion studies to track plumes by lidar. Figure 8.31 shows a
mountainous-terrain example where an oil-fog tracer plume was released from a
power-plant stack. The enhanced backscatter from the oil-fog droplets was used to
trace the path of the plume and its rate of spread (Eberhard et al. 1987). It is of
interest that in this case, the plume was not strongly channeled by the valley in
which the power plant was located.
The addition of polarization to backscatter measurements provides important
information on the shape of the scatterers (Sassen 2005). As for radar, a signal
pulse having one polarization is transmitted, then the receiver detects the return
energy versus range for both polarizations, from which a depolarization ratio is
calculated. Depolarization ratios near zero again indicate spherical scatterers, and
larger depolarization ratios generally indicate elongated, non-spherical scatterers.
Depolarization can discriminate between different types of aerosol, for example tro-
pospheric and stratospheric haze often have near-zero depolarization ratios, whereas
ratios during dust events, for example Asian or Saharan dust, are significantly
greater than zero. In clouds, near-zero ratios near cloud base and rapid attenuation
with height are generally associated with the presence of liquid-water droplets,
whereas larger ratios would mean ice crystals or snow.
An example of data from an unattended, low-energy, high-pulse-rate lidar in the
Arctic is shown in Fig. 8.32. The existence of depolarized returns to 5 km height
between 0900 and 1700 UTC indicates a stratiform ice cloud. Although lidar signal
can penetrate only a short distance into a water cloud, ice particles are very effective
scatterers, and lidar has long been used to study cirrus clouds (Sassen 2005; Grund
et al. 2001). Before 0800 or after 1700 in Fig. 8.32, the strong attenuation of the
signal above cloud base at 800 m and the existence of near-zero depolarization
ratios (spherical shapes) at cloud base is a signature of liquid-water droplets in
the cloud at those times, despite freezing Arctic temperatures. The presence of
liquid water rather than ice in such clouds has a very strong effect on the radiative
properties of the cloud.
An important œ for aerosol studies has been 532 nm, because of hardware
availability and its proximity to the size of many aerosol populations (<1 m). As
mentioned previously, molecular scattering at this œ must be accounted for to do any
quantitative work. A technique for calculating the backscatter due solely to aerosol
is high-spectral-resolution lidar (HSRL), which uses the frequency information
(Doppler spectrum) to separate the backscatter due to molecular activity from
that due to aerosol, since each has a distinctive, characteristic distribution as a
function of frequency (Eloranta 2005). This technique with polarization has been
used successfully to discriminate between types of cloud layers (liquid or ice) and
types of aerosol.
Fig. 8.31 Map showing location of early-morning oil fog release experiment at Tracy power Plant
in Truckee River Valley, 7 August 1984, with terrain contours at 100-m intervals. Solid and dashed
lines to north of lidar trace the centroids of the plumes for six time periods between 0330 and
0500 PDT. Lidar cross-section azimuths indicated by thin solid lines. Vertical cross sections of oil
fog distribution of concentrations within plume (bottom two panels). Vertically and horizontally
integrated plume concentration distributions plotted along horizontal and vertical axes (Eberhard
00:28 02:36 05:00 07:28 09:56 12:24 14:52 17:20 19:48 22:12
Fig. 8.32 Time-height plot of lidar depolarization-ratio. Water clouds have near-zero ratios,
whereas ice clouds have of 0.2–0.3 or more (Intrieri et al. 2002. © American Meteorological
Determination of aerosol size distributions by remote sensing would be very

useful for many applications. A common approach is to assume a size distribution
and use the remotely sensed data to estimate the parameters of the distribution,
for example, Post et al. (1996, 1997) proposed and verified the use of a log-
normal distribution, which has three parameters (number density, median radius,
distribution width). In theory, it is possible to retrieve as many parameters of aerosol
size distribution as one has independent and perfect remote-sensing measurements.
The independent measurements can be, for instance, lidar backscatter measurements
at different wavelengths and/or different polarizations, different angles of scatter,
measurements of attenuation (scattering and/or absorption), or measurements of
emission.
In practice, however, backscatter measurements at wavelengths that differ by
less than a factor of two are often not independent. Moreover, all real-world
measurements are imperfect: they contain noise and errors in calibration. These
practical matters severely limit the accuracy of the retrievals of aerosol size
distribution parameters, even for aerosol distributions that are well-behaved (known
shape and refractive index). E.g., Post (1996) has shown that for optimistically low
3% measurement errors, perfect calibrations, widely-space backscatter wavelengths
(0.355, 1.06, and 10.6 m), and ideal aerosol particles, retrieval errors of mode
radius and distribution width exceed 30% for mode radii between 0.2 and 0.9 m,
and widths between 1.3 and 1.9 m. For less well-behaved aerosol populations or
larger measurement errors, the accuracy of the retrievals decreases further. Thus
the use of remotely sensed data to estimate aerosol size distributions remains a
challenge.
Fig. 8.33 Cross-valley vertical-slice scan of lidar aerosol-backscatter from Owens Valley during
TREX. Sequence of scans indicates rotation as shown in schematic (lower panel) (De Wekker and
Mayor 2009. © American Meteorological Society. Reprinted with permission)
However, when the application is calculating the radiative transfer properties of

aerosol layers (for example in a model), remote-sensing retrievals using a multi-
parameter approach have one advantage, even for irregular and inhomogeneous
aerosol populations. The advantage is that the estimated (retrieved) parameters
are the same ones needed for radiative-transfer calculations. The remotely sensed
signals are themselves radiative in nature; thus the information they provide is a
direct indication of the radiative properties of the region (e.g., layer) being probed.
The calculations will be accurate over the range of parameters spanned by the
remote-sensing retrieval. Thus lidar remote sensing data provide a method for
directly verifying model calculations of radiative effects of aerosol layers, whether
or not the actual aerosol size distribution is well represented by the assumed
distributions used for the retrievals and the models.
The spatial distribution of aerosol in the atmosphere is not uniform, but tends
to appear “lumpy” in images. These “lumps” may remain intact for periods of
time, and have been used to track or characterize atmospheric flow features. They
have also been used to calculate wind speeds (Pirronen and Eloranta 1995; Mayor
and Eloranta 2001). Figure 8.33 is one of a series of vertical cross sections
from the Owens Valley in California during the TREX campaign, showing the
Sierra Nevada to the left (west) and the westerly mean winds blowing over the
mountains from left to right. The arrows indicate the rotation in a lee-side rotor
as inferred from animations of this aerosol-backscatter field (De Wekker and Mayor
2009).
8.7.3 Doppler Lidar Examples
For a coherent lidar signal, the other kind of information besides backscattered
energy is the returned frequency, from which the radial or Doppler velocity is
calculated. Many different approaches to using the Doppler signal have been tried,
as characterized in reviews by Werner (2005) and Hardesty and Darby (2005).
As with all atmospheric backscatter lidars, Doppler systems can employ either
coherent (heterodyne) or incoherent (direct) detection techniques. Direct-detection
systems use frequency-sensitive/dependent filters or interferometers to determine
the backscattered frequencies. One advantage that short-œ (e.g., 532 nm or smaller)
direct-detection systems have is they can use molecular backscatter, which can
provide velocity signal in regions of low aerosol concentration, such as the upper
troposphere, making them useful for satellite applications. But the molecular
backscatter signal is much weaker than aerosol returns, and the molecular Doppler
spectrum is much wider and noisier than that of aerosol. For applications near the
surface, such as mountain flows, wind-speed measurements of acceptable precision
from direct detection systems require large transmitted energies, large-aperture
receiving telescopes, and averaging times that are often too long for useful velocity
estimates. Therefore, for the current state of technology, coherent lidar systems
are the preferred systems for Doppler wind sensing in the lower troposphere
(Werner 2005).
As with other remote sensors (such as radar), and regardless of which detection
technique is employed, Doppler lidars can be continuous wave (CW) or pulsed.
CW lidars were used decades ago to investigate peak wind speeds in waterspouts
(Schwiesow et al. 1981). For CW systems, no explicit range information is obtained
by the backscattered signal, but the system optics are focused, such that they
receive most of the return signal from a specific range interval, which increases
(quadratically) as the focus is changed to be farther from the instrument. Some of
these systems are used in staring mode, others are scanned. One method of obtaining
wind profiles is by tilting the transmitter at a fixed elevation angle and scanning
a full 360ı in azimuth, then using the VAD technique to calculate the direction
and speed of the wind. The horizontal wind measured by such a CW system is
for the vertical layer where the beam is focused, and a profile is obtained by
(mechanically) changing the focus to a different range covering a different altitude
spread. Some systems currently available allow for five separate focus heights up
to 300 m, which are each scanned separately, so a complete five-level profile
takes several minutes to complete. An intrinsic issue with these systems is that
a hard target at some distance from the instrument, such as a cloud layer, can
produce a backscattered signal that overwhelms the signal in the focused region
and contaminates the Doppler-wind calculation (Courtney et al. 2008).
Despite the many Doppler-lidar approaches available, pulsed Doppler lidar is
what one is referring to, when “the term ‘Doppler lidar’ is used in connection with
wind measurements,” according to Werner (2005). Even pulsed Doppler lidar has
different approaches to forming the pulse. The “standard” approach of building
up and releasing pulse energy into the atmosphere has been described by Post
and Cupp (1990) and Grund et al. (2001). A second approach uses a CW laser,
and electronically “chops” the beam into weak pulses, which are released into an
amplifier. The amplified pulse is then released into the atmosphere. This is referred
to as a master-oscillator, power amplifier (MOPA) configuration (Brewer et al. 1998;
Pearson et al. 2009), also used in radar. Several types of Doppler lidar described in
this survey are becoming available commercially. It is not clear at this point which
technologies will be most advantageous for probing the atmospheric boundary layer,
especially in complex terrain, or under what conditions one may have an advantage
over the others. In the remainder of this section we will consider results from pulsed
Doppler lidars.
Scanning pulsed Doppler lidars have been used in many mountain-meteorology
studies to show the spatial and temporal variability of terrain-generated flows
(Nieman et al. 1988). As with radar, azimuth scans at fixed elevation angles
generate a cone of data. At low elevations, the shallow cones are nearly horizontal.
One example of strong horizontal variability has been shown in Fig. 8.8b, where
westerly-wind outflows from the mountains onto the plains near Colorado Springs,
Colorado, were channeled by upstream gaps and valleys in the mountain barrier
to the west (here red, yellow indicate flow toward the lidar). Between the outflow
plumes, topographic eddies of reversed flow can be seen to the lee of the mountains
(green regions on the left side of the figure).
Another example (Fig. 8.34) illustrates the variability of nocturnal flows in the
basin of the Great Salt Lake, Utah, and the interactions among flows of different
length and time scales during VTMX. Figure 8.34 shows the Doppler velocity
distribution from a nearly horizontal conical lidar scan at 0.5ı elevation angle.
A southerly, thermally-forced LLJ is a major feature at the scale of the entire
basin. At the largest scales, the formation and onset of this jet are controlled by
the synoptic-scale pressure gradient across the basin. At smaller than basin scale,
canyon outflows from the mountain range to the east were found to be sensitive to
the existence and timing of this basin-scale LLJ (Banta et al. 2004). These mesoscale
flow interactions had a significant effect on the dispersion of tracer material released
in the northeast part of the basin near Salt Lake City (Darby et al. 2006), further
described later.
The vertical structure of mountain flows is of great interest for applications and
NWP verification. Scans in elevation angle at fixed azimuth, which produce a verti-
cal slice or cross section of data, reveal vertical structure, as already demonstrated in
Figs. 8.8a and 8.29. Such vertical-slice scans performed perpendicular to mountain
barriers from downwind sites have revealed a variety of ground-level and elevated
rotors and hydraulic jumplike structures, such as those shown in Fig. 8.35. Rotors
are indicated as regions of reversed flow (opposite sign) from the background winds.
Doppler lidar was an important research tool for investigating the structure of valley
and gap flows during the Mesoscale Alpine Programme (MAP) in the European
Alps during fall 1999 (Drobinski et al. 2003; Flamant et al. 2002; Mayr et al. 2004;
Gohm et al. 2004; Weissmann et al. 2004; Durran et al. 2003; Rucker et al. 2008),
and the recent Terrain-Induced Rotors Experiment (TREX: Grubišić et al. 2008).
Fig. 8.34 Lidar conical sector scan at 0.5ı elevation of Great Salt Lake basin (north up, west to
left), showing formation of southerly down-basin jet complicated by outflows from canyons to the
east. Color scale is Doppler velocity in m s1 (Banta et al. 2004. © American Meteorological
Three-dimensional volume scans can be obtained by performing repeated az-

imuth sector scans but incrementing the elevation between scans (or, alternatively,
performing elevation sector scans at incremented azimuths). Volume scans can be
used to study both the vertical and horizontal structure of the flow, provided the
flows of interest are reasonably stationary over the volume scan period. Examples
where along-valley volume scans were analyzed to reveal the cross-valley structure
of nocturnal down-valley flows are shown in Fig. 8.36 for a modest valley (Brush
Creek, Colorado: bottom panel) and a very deep canyon (the Grand Canyon: top
8 Observational Techniques: Sampling the Mountain Atmosphere
Fig. 8.35 Examples of lidar vertical-slice scans through rotors. Upper and middle left: color plot (top) and accompanying vector plot of east–west vertical-slice
scan through a Front Range windstorm in Boulder, Colorado (Banta et al. 1990; Clark et al. 1994), Top right: mountain wave rotor structure in Owens Valley,
California during TREX (Courtesy Dr. Andreas Dörnbrack). Bottom panels: elevated rotors as seen by east–west Doppler lidar scans over Colorado Springs,
483
Colorado
Fig. 8.36 Cross sections derived from 3-D volume scans (top) in the Grand Canyon: map (left)
shows location of lidar on canyon rim at N in lower right of map; scans were performed to the north,
and right panel shows a vertical cross section across the canyon at night, with blue colors indicating
cold air flowing toward the lidar in the lower portion of the canyon and reds indicating flow away
from the lidar in the upper portions of the canyon (Banta et al. 1999. © American Meteorological
Society. Reprinted with permission). Cross section of nocturnal flow in Brush Creek in western
Colorado (bottom), with yellows indicating down-valley flow going away from the lidar (Post and
Neff 1986. © American Meteorological Society. Reprinted with permission)
panels). Both show an asymmetric core or jet of stronger flow within the valleys
above the valley bottoms. Volume scans have also been used to investigate the
structure of valley outflows near the Front Range of Colorado (Banta et al. 1995,
1996; Levinson and Banta 1995; Darby et al. 1999); in the Lower Fraser River
Valley near Vancouver BC, Canada (Banta et al. 1997a); and the Brush Creek dataset
was used to calculate along-valley mass fluxes as a function of distance along the
valley axis (Neff 1990). As with Doppler radar, dual-Doppler lidar datasets have
been used to determine the three-dimensional vector wind field of complicated flows
in mountainous terrain (Hill et al. 2010).
Partly because of their technological complexity, Doppler lidar systems have

been mostly research tools. But the availability of solid-state lasers and other
components at reasonable cost have led to more reliable systems that can be operated
unattended for extended periods and commercialized, and this in turn has led to
the possibility of using Doppler lidars to address operational problems. One of
the most significant and ambitious operational applications to date is the use of
Doppler lidar at the Hong Kong International Airport (HKIA) by the Hong Kong
Observatory (HKO) to detect potentially hazardous wind shear and turbulence for
aircraft approaching and taking off from this airport (Shun and Lau 2002; Shun et al.
2003, 2004; Banta et al. 2007; Shun and Chan 2008).
8.7.3.1 Operational Lidar at Hong Kong International Airport
A number of different types of flow that affect aviation have been identified by
lidar at HKIA. Frontal structures, such as microbursts, gust fronts, and sea-breeze
fronts (Fig. 8.37d), have a well-defined coherent vertical and horizontal structure in
the lowest several hundred meters of the atmosphere. On the other hand, phenomena
producing the terrain-induced wind shear, which is of significant concern for aircraft
on approach or departure, tend to have high spatial and temporal variability and
come in various forms. For example, Fig. 8.37a reveals a mountain wake area
with lower wind speed and a different wind direction from the background flow,
caused by the complex terrain of Lantau Island south of the airport, when large-scale
easterly winds blow in the springtime. Under certain conditions in spring, multiple
wind jets could be observed to emerge from the various gaps of Lantau Island
(Figs. 8.37b and 8.38a), some bringing föhn-like temperature increases of several
degrees (Shun et al. 2003; Banta et al. 2007). These gap-outflow jets can penetrate
into the background easterly flow, producing sudden decreases (then subsequent
increases) in the headwind component experienced by aircraft landing at the airport,
(e.g., Fig. 8.39). In other stably-stratified conditions, lee waves (Fig. 8.37c), reverse
flows, and hydraulic jumplike features (Figs. 8.37e, f and 8.38b) are observed by the
lidar.
To detect the highly changeable winds encountered by the aircraft along the
glide paths under terrain-induced wind-shear conditions (Figs. 8.39 and 8.40), HKO
devised an innovative glide path scan (GPScan) strategy for the lidar (Shun and Chan
2008), in which the lidar is scanned in azimuth and elevation simultaneously in a
coordinated manner, so that the beam is always aimed at the mean glide-slope path
(Fig. 8.41). The purpose of the GPScan procedure is to derive the along-glidepath
profiles of the headwind, and hence the wind shear encountered by aircraft along
the glide paths, as illustrated in Fig. 8.42. An automated Lidar Windshear Alerting
System (LIWAS) was developed based on these GPScans. The system was able
to detect 76% of the significant wind shear events (headwind change of 15 kt or
more) reported by pilots over the most-used approach corridor (07 L arrival) of
HKIA under clear-air conditions. One such event captured by the lidar is shown in
Fig. 8.42.
Fig. 8.37 Wind shear flow patterns observed by the lidar at HKIA. (a) Mountain wake observed
in the 1ı conical scan of the lidar. (b) Accelerated gap flow (labeled “J”) observed in the 1ı
conical scan. (c) Mountain wave observed in the 4.5ı conical scan. (d) As a contrast, westerly sea
breeze against the background easterly winds as observed in lidar horizontal azimuth scan (at 0ı
elevation), showing less spatial variability compared with terrain-induced wind shear. (e) Hydraulic
jump of cross-mountain airflow observed in the vertical-slice scan of the lidar at 95ı azimuth [see
(f)]. (f) Reverse flow of a recirculation vortex underneath the hydraulic jump in (e) in the 1ı conical
lidar scan (Shun and Chan 2008. © American Meteorological Society. Reprinted with permission)
Velocity (m/s)
a 10
2003 / 02 / 21 18 : 12 : 43 UTC 2003 / 02 / 21 18 : 14 : 09 UTC
23
20
8
17
6 14
11
4
8
2 5
Range (km)
2
0
−2
−2 −5
−8
−4
−11
−6 −14
−17
−8
−20
−10 −23
Velocity (m/s)
2003 / 02 / 21 10 : 55 : 30 UTC
b 5
23
20
17
4 14
11
8
3 5
Height (km)
−2
2 −5
−8
−11
1 −14
−17
−20
0
−23
0 5 10
S Rwy Range (km)
Fig. 8.38 (a) Hong Kong International Airport—gap flows producing wind shear across aircraft
glide paths. (b) Vertical cross section through gap flow, showing windstorm-like (jumplike)
structure (Banta et al. 2007)
Fig. 8.39 Vertical-slice scan approximately along glide path, showing spatial variability of
headwinds along glide path. In this case, surface easterly winds prevailed in the background and
affected the eastern side of the airport (as indicated by the winds blowing towards the lidar [green
colors]), but disturbed flow with different wind directions affected the western side of the airport
(as indicated by the winds blowing towards as well as blowing away (brown colors) from the lidar)
(Banta et al. 2007)
Fig. 8.40 Successive 1ı conical sector scans illustrating the temporal variability of terrain-
induced wind shear when southwesterly winds blew across the complex terrain of Lantau Island.
The cool (warm) colors represent winds toward (away) from the lidar (see scale at the bottom). The
arrows indicate the movement of the wind shear features marked by circles within the subsequent
4 min (Shun and Chan 2008. © American Meteorological Society. Reprinted with permission)
To ensure lidar-data availability for operational wind shear alerting and to better
resolve the headwind component along the glide paths, two lidars of the same
model have been deployed at each of the two parallel runways of HKIA since
October 2006. The launch of LIWAS at HKIA in December 2005, after 3 years
of experimentation and operational trial, signifies that the anemometer-based and
Fig. 8.41 Schematic diagram of a GPScan of the lidar (top). The scanning areas of GPScans
over the final 3 n mi along the arrival runway corridors of HKIA (bottom) (Shun and Chan 2008.
radar-based wind shear detection technologies deployed worldwide in the twentieth

century have been further advanced by the addition of Doppler lidar—a step closer
to all-weather coverage (Shun and Chan 2008). The routine use of automated
scanning sequences and the ability to remotely program and change scans and
scan sequences, as accomplished here, are important capabilities for the further
advancement of Doppler lidar as a tool for operational and research applications.
Fig. 8.42 Headwind profiles along the glide paths constructed from the GPScans of the lidar.
Significant wind shear detected by the automatic algorithm of LIWAS, highlighted in red, was
corroborated by a pilot report at around the same time (Shun and Chan 2008. © American
8.7.3.2 Time Dependence and Calculated Quantities
A major use of surface-based Doppler lidar scan data has thus been to show the
spatial variability and structure of mountain flows for research and operational
applications. By repeating the scans, the evolution of these structures can also
be studied. For example, Clark et al. (1994) showed that sequences of vertical-
slice scans could be used to identify and track gusts as they formed and moved
downstream in a windstorm, Darby et al. (1999) tracked the movement of a
mesoscale cold front just downwind of the Rocky Mountains, Darby et al. (2002)
showed the development of sea-breeze structure as affected by coastal mountains in
California, and Fig. 8.40 shows gust propagation near HKIA.
In addition to the image data, numerical scan data can be further analyzed
for high-resolution profile information or other quantitative information that can
be extracted from the scan. Conical scans analyzed using the VAD technique
(Sect. 8.5.3) produce vertical profiles of the horizontal winds to high precision (Hall
et al. 1984). Sequences of such profiles have been used in complex-terrain studies
to investigate flow evolution (Banta et al. 1993; Clark et al. 1994; Levinson and
Banta 1995). Figure 8.43 shows an example of Doppler-lidar VAD profiles on two
nights taken from a site in the middle of the basin of the Great Salt Lake during
VTMX, showing differences in the evolution of the southerly down-basin low-level
jet (See Fig. 8.34). On the first night (top), the flow was strong and southerly all
afternoon and night, but on the second night the afternoon flow began as a daytime
up-basin (northerly) flow that changed after 1900 local time to the nocturnal down-
basin jet. Although the flow in midbasin was very similar at the times of tracer
releases after midnight (shaded) in the Salt Lake City urban area, differences in the
early-evening development of local slope-flow and canyon-outflow characteristics

between the nights led to dramatic differences in the dispersion of the tracer after
midnight (Darby et al. 2006). On the first night local-flow development near the
foothills was insignificant, and the tracer was rapidly transported away by the strong
southerly LLJ flow. On the second night local flow development was strong, pushing
the basin-scale southerly flow to the west, replacing it with the lighter and more
disorganized slope and canyon outflows, and resulting in high tracer concentrations
lingering near the city (Darby et al. 2006). VAD vertical profiles of the horizontal
wind at high vertical resolution in the lower few hundred meters of the atmosphere
can thus be very useful in documenting flow evolution. Compact Doppler-lidar wind
profiling systems using similar techniques to provide high-resolution wind profiles
in the lowest 1000–2000 m of the BL have become commercially available.
Mean wind-speed profiles can also be calculated from vertical-slice scans
oriented along the mean wind direction. This technique is most useful when the di-
rectional shear of the wind with height is small. Time-height cross sections derived
from such profiles revealed two scales of the sea breeze in California, an early-
morning, shallow sea breeze and an afternoon deep layer of onshore flow (Intrieri
et al. 2002; Banta et al. 1993, Banta 1995). Numerical experiments on the effects
of the land-sea contrast and the two ranges of mountains in this region showed that
the heated mountainous inland topography had a strong role in the generation of
deeper afternoon onshore flow (Darby et al. 2002a). Another example of this kind
of analysis is given in Fig. 8.44, which shows the evolution of westerly air flow over
the Front Range of the Rocky Mountains near Colorado Springs, and the generation
of reversed flow (light shading prior to cold front) due to rotor activity.
Conditions of significant directional wind shear with height can be addressed
using two Doppler lidars. Vertical-slice scans are coordinated between lidars so that
the scans intersect at multiple locations, where both horizontal wind components
can be determined as a function of height, each vertical line of intersection being
referred to as a “virtual tower” (Calhoun et al. 2006).
Other quantitative data such as maximum speeds or dimensions of flow features
have been extracted from lidar data. Time series of values at points in the
measurement domain are also useful. For example, time series of along-canyon wind
velocities near the bottom of the Grand Canyon and at a level just below the rim of
the canyon showed that the flow was vertically mixed at times, but at other times,
under slack synoptic forcing, flow in the bottom of the canyon exhibited a diurnal
thermally forced up-canyon/down-canyon flow pattern (Banta et al. 1999). A similar
time series of winds at 90 m above ground was used during VTMX to document the
evolution of the nocturnal LLJ in the Great Salt Lake basin (Banta et al. 2004).
Surface-based Doppler lidar data are useful for revealing the local spatial
variability and flow structure on scales smaller than the scan diameter. But these
examples have shown that the passage of longer duration, and therefore larger-scale,
features can be studied using sequences of profiles or other time-series information
extracted from the scan data. To study the spatial structure of features larger than
lidar scans, lidars have been mounted on mobile platforms. Airborne platforms in
particular have proven very useful for lidar applications over mountains.
Fig. 8.43 Vertical profiles of the horizontal wind derived from Doppler-lidar 5ı elevation-angle
conical scans. Full barbs represent 5 m s1 wind speeds, half barbs, 2.5 m s1 . Shaded bars
indicate times of downtown SF6 tracer releases. (a) 25/26 October 2000 (VTMX IOP-10);
(b) 17/18 October 2000 (VTMX IOP-7) (From Darby et al. 2006. © American Meteorological
Fig. 8.44 Colorado Springs time-height cross sections of the westerly wind component between
0000 and 1230 UTC 1 Apr 1997, showing considerable temporal variability in the mountain wave
flow aloft. Solid contours represent westerly component flow at 2 m s1 contour intervals, and
dashed contours represent easterly component flow. Westerly flow greater than 8 m s1 has been
shaded dark gray. All easterly component flow (rotors before 1000 UTC and postfrontal flow
after 1000 UTC) has been lightly shaded. The zero contour is boldface. These profiles represent
conditions 12–14 km west of the lidar, nearly adjacent to the foothills (Darby and Poulos 2006.
8.7.4 Airborne Applications
Because of the strong spatial variability of flows in mountainous terrain, lidars

operating from airborne platforms are well suited to addressing flow evolution and
atmospheric transport issues in these regions. Aerosol generated in the polluted Los
Angeles basin and in the valley boundary layer near Vancouver, B.C., has been used
in airborne lidar images to trace vertical and horizontal transport due to mountain
flows in those regions (Wakimoto and McElroy 1986; Hoff et al. 1997; Hayden et al.
1997). Airborne High Spectral Resolution Lidar has been used to study the structure
of aerosol layers in the lower atmosphere (Hair et al. 2008). The depth of aerosol
layers marking the tops of air masses being transported over mountain barriers have
been used in studies of downslope wind dynamics and the applicability of hydraulic
theory (Flamant et al. 2002; Gohm and Mayr 2005). Other investigations of terrain-
related flows using airborne Doppler lidars have been described for example by
Reitebuch et al. (2003), Bilbro et al. (1984), Rothermel et al. (1998), and Tollerud
et al. (2008).
Fig. 8.45 Vertical curtains of airborne DIAL measurements of O3 along the aircraft flight track
over Front Range foothills, superimposed on a Google Earth map of the region near Denver,
Colorado. Color scale indicates O3 concentrations in ppb
DIAL systems sensing ozone, used to study pollution transport over simple
topography (Alvarez et al. 1998; Senff et al. 1998; Banta et al. 1998, 2005), have
also been used in complex terrain (e.g., Langford et al. 2010). As an example
of ozone measurements using such a DIAL system in complex terrain, Fig. 8.45
shows the ozone distribution on the afternoon of 31 July 2008 over the Rocky
Mountains to the west of the urban corridor extending from Denver, Colorado,
northward to the Wyoming border. These measurements were collected with the
NOAA/ESRL airborne ozone lidar (Alvarez et al. 2011) flying at approximately
16,500 ft (5,030 m) ASL. On that day, southeasterly winds pushed polluted air from
the Denver urban area into the Foothills and over the Continental Divide, resulting in
high ozone concentrations in Rocky Mountain National Park and west of the Divide.
Note that under these dry afternoon convective conditions, the transport carried
the ozone directly over the barrier and was not strongly channeled by topographic
features.
The kind of information available from airborne aerosol or DIAL lidars is very
useful for showing the integrated effects of BL flow on atmospheric transport, and
thus should also be very useful for verifying numerical model output (Sect. 8.9). It is
important to emphasize that the impact of studies using this type of data is far greater
when coordinated with arrays of ground-based instrumentation to more completely
characterize meteorological conditions and reduce ambiguities in the interpretation
of the airborne remote-sensing data.
Table 8.10 Advantages/limitations

Finely detailed flow near terrain (narrow Vertical range limited by aerosol
beam, no sidelobes)
Clear air sampling Research lidar systems require expert technicians
High temporal and spatial resolution Limited cloud penetration; attenuation by fog, rain
Scans a volume of space –
Detailed model evaluation –
Aerosol—BL height, cloud base –
Versatile –
Some unattended lidars available –
(backscatter and Doppler)
8.7.5 Discussion
The term lidar encompasses many different types of instrument—differences arising

from the wavelength employed and the atmospheric interaction exploited. In its
various forms, lidar is able to address many of the issues related to mountain
meteorology, including spatial variability, measurements close to topography, and
documentation of flow evolution. Advantages of lidar listed in Table 8.10 include
that the scattering targets, including aerosol particles, are widely distributed in the
ABL even in the absence of clouds, turbulence, or gradients of temperature or water
vapor. Limitations include the need for line-of sight, the complexity of the systems,
and the inability to penetrate very deeply into cloud and fog.
Scanning systems such as lidar are versatile, providing direct structural informa-
tion about flow features smaller than the scan diameter via images of the scans. But
they are also extremely valuable for providing high-precision profile information
and other quantitative data, available as a function of time, obtained by analyzing
scan data in ways to demonstrate the evolution of flow features and processes on
larger scales. The ability to measure and map out details of the flow to a degree
previously unimaginable means that processes can be documented and understood,
and NWP models can (and should) be verified, to a much finer standard than thought
possible prior to the availability of this kind of data. Impacts on numerical modeling
verification will be further discussed in Sect. 8.9.
8.8 Instrument Combinations
Capabilities of individual instruments or instrument types can give insight into the
current state of the atmosphere and the processes determining that state. A theme
of this chapter has been that each instrument type has strengths and limitations.
It is often advantageous therefore to deploy instrumentation in combinations that
optimize each system’s contributions based on its strengths, while having other
1600
Tv (C)
1500
14.7
1400 13.5
1300 12.3
1200 11.1
1100 9.9
1000 8.7
Height (m agl)
900 7.5
800 6.3
5.2
700
4.0
600
2.8
500
1.6
400
0.4
300
–0.8
200 –2.0
100 –3.2
0 22 20 18 16 14 12 10 8 6 4 2 0
29-Mar-1997 Time (UTC) 28-Mar-1997
PAF
0 5 10 50 knots
Fig. 8.46 Example of 915-MHz profiler wind profiles with RASS temperatures superimposed in
color (Data from Colorado Springs airport)
measurements available to compensate for its limitations. One type of advantage

to deploying instrument combinations is to extend coverage and address the limita-
tions, for example, using a 915 MHz profiler along with a minisodar (Coulter et al.
1999). The sodar provides high-resolution profiles below the 100-m minimum
range of the profiler and a quality check on the lowest profiler range gates. The
profiler extends the vertical coverage of the measurements up to 2–4 km. Similarly
for the profiler at sea previously discussed, a co-located Doppler lidar provides VAD
winds in the lowest several 100 m’s to compensate for the ship-borne profiler sea-
clutter limitation. Here profiles calculated from the lidar scan data also provide high
temporal and spatial resolution at low levels where finer resolution and precision are
needed.
Another type of advantage is to provide complementary information, for example
a profiler with RASS is an example of an instrument combination for measuring
wind and temperature profiles nearly simultaneously (Fig. 8.46). Wind-sensing
instrumentation provides the kinematic fields, but it is also important to document
the thermodynamic fields to understand the role of stability and other dynamic
processes in determining how the flow evolves and to verify NWP models.
Another example is the use of lidar and radar to study interactions between clouds
and the subcloud layer, where the lidar gives data in the subcloud layer but not very
deeply into the liquid-water clouds, and the radar sees in-cloud and precipitation
Fig. 8.47 Doppler lidar vertical slices of radial velocity through thunderstorm gust front (top
panels) (Darby et al. 2002. © American Meteorological Society. Reprinted with permission),
moving left to right. Positive values (red) indicate flow away from lidar. Rapidly moving cold
front exhibiting classic density current structure as it passes (right to left) through the Alpine Wipp
Valley south of Innsbruck, Austria (after Darby et al. 2000 and Gohm et al. 2010) (bottom panel).
Here positive values (red) indicate flow toward the lidar
processes, but little below cloud base. A dramatic example of this kind of analysis
was the Arctic study presented by Intrieri et al. described in Sect. 8.7. The lidar gave
accurate estimates of cloud base, whereas the radar gave the best determination
of cloud top, providing an optimum determination of cloud depth and content as
needed for radiation calculations (Fig. 8.32).
8.8.1 Density-Current Structure
Advantages of combinations may be illustrated by considering pairs, but real insight

comes from deployments of multiple instrument types. A mixture of sensors can
provide important insight into the structure and mixing properties of flow features.
For example, Darby et al. (2002b) used a Doppler lidar and a ground-based O3
lidar along with tower in-situ measurements at the same site to investigate vertical
mixing effects of a rapidly moving small-scale cold front, a thunderstorm gust
front (Fig. 8.47, upper panels). Time series (Fig. 8.48) of temperature (fifth panel),
surface-measured O3 at 10 m (top panel, blue lines), and DIAL-measured O3 at
Afternoon - Corrective Mixing Turbulent Formation of RL GF TW LLJ Wind shift Downward mixing
Intense
Photochemistry dominating mixing mixing Formation of RL Elevated turb.
a 120
22 June 1999
100
O³, ppbv
80
60 1035 m AGL
405 n AGL
10 m AGL
40
b
1.2
Sigma_w (cm/s)
1.0
0.8
0.6
0.4
0.2
c 2
Vetical velocity (m/s)
-1
-2
d 500
400
Height (M AGL)
Varience (mˆ2/sˆ2)
300
0 0.50 1.00
200
100
0
e 34
Temp (c)
32
30
28
15
26
f 360
Wind dir. (deg)
10 m AGL
270
180
90
0
15 16 17 18 19 20 21
Time (CST)
Fig. 8.48 Time series of (a) O3 concentrations (ppb) at 10, 405, and 1,035 m AGL; (b) standard
deviation of the vertical wind speed ¢ w at 8 m AGL (cm s1 ); (c) w (m s1 ) at 8 m; (d) variance of
the lidar-measured horizontal wind (m2 s2 ); (e) temperature at 8 m (C); and (f) wind direction at
10 m (degrees true north). Abbreviations used in time line at top: RL residual layer, GF gust front,
TW turbulent wake region, and LLJ low-level jet (Darby et al. 2002b)
altitudes of 405 and 1,035 m (green and red lines, top panel) show a typical
late-afternoon decrease in temperature and O3 after 1600 CST. Surface O3 then
continues to decrease, but upper-level O3 levels off. The gust front passed the site
just after 1800 CST, indicated by an abrupt shift from southeasterly to northwesterly
winds (bottom panel)—and marked by the dotted red lines through all panels. The
postfrontal flow is much more turbulent than the prefrontal flow, as indicated by
the tower sonic anemometer measurements of the vertical-velocity w variance ¢ w 2
and w itself (second and third panels). This enhanced turbulence extends through a
deep layer as shown in the time-height cross section of the along-wind variance ¢ u 2
in the fourth panel, measured by the Doppler lidar. This deep layer of turbulence
remixes the O3 profile, as the O3 at all three levels (including the 10-m tower
level) becomes equal by 1830 CST (top panel). After 1930 CST the more normal
nighttime pattern is reestablished, as the winds return to a southerly direction,
and the near-surface declines in O3 indicate decoupling from the upper-level O3
values at 1,035 m. Conceptual and numerical models of such density currents have
hypothesized a region of strong turbulent mixing behind the head of the density
current (e.g., Droegemeier and Wilhelmson 1987). Using remote sensing along with
surface-based meteorological and turbulence sensors, it was possible to verify the
existence of the strong turbulence, its occurrences through a deep layer, and its
effects in mixing a trace species (O3 ) through the deep layer. This study took place in
relatively homogeneous terrain, but a similar Doppler-lidar-based study of a rapidly
moving cold front through an Alpine valley near Innsbruck, Austria (Darby et al.
2000; Gohm et al. 2010) revealed virtually the same vertical structure as found here
(Fig. 8.47, lower panel).
8.8.2 Valley Outflows
In the case study presented in Sect. 8.2, data from multiple in-situ sensors, including
surface meteorological sensors, chemistry measurements, and an ozonesonde, were
used to study a northerly flow that routinely began just after sunset at a site in the
Fraser River Valley, the main east–west river valley east of Vancouver, BC, Canada
(Banta et al. 1997a). This flow came from the direction of a tributary valley to the
north of the Fraser Valley. As reproduced in Fig. 8.49, the O3 trace (solid line)
showed reductions in O3 concentrations during the period of the northerly flow
(indicated by the shaded bar at top), indicating that the northerly flow was “cleaner”
of O3 than the main-valley air mass it replaced. This air could have been more
pristine air from the tributary valley, the Pitt Lake Valley, but the existence of other
pollutants meant that this was previously polluted air.
A Doppler lidar, located to the south of the surface site at H in Fig. 8.3 and south
of the Pitt Lake Valley opening, scanned the flow structure in this region. The lidar
data showed the flow in this region to be complex and continually evolving through
the night (Banta et al. 1997a), but afternoon low-elevation conical scans (Fig. 8.50a)
showed polluted air from the main valley flowing into the Pitt Lake Valley during
Fig. 8.49 Same as upper panel of Fig. 8.5, showing O3 decrease at the onset of the Pitt Valley
outflow (Banta et al. 1997a)
daytime hours. At night it clearly measured northerly-component flow extending out

of the Pitt Lake Valley opening and over the surface measurement site for much of
the time (Fig. 8.50b). Even when the surface winds were light and variable later in
the evening, Doppler-lidar data showed the outflow jet aloft persisting much of the
night.
A schematic view of the possibilities is shown in Fig. 8.51, which shows the
polluted air going into the valley during daytime (top panel) and the two options
for nocturnal outflow. The removal of O3 but not CO occurs through dry deposition
to the vegetated surfaces. The middle panel shows the establishment of downvalley
flow through the valley, which could either bring clean air into the valley or could
promote deposition along the valley floor. But it has already been established that
the low-O3 air is not clean, unpolluted air, and furthermore, the floor of the Pitt
Lake Valley is a water surface, to which deposition of ozone is negligibly slow.
The other possibility (lowest panel) is that the deposition occurs in nocturnal cold-
air slope flows draining down the valley sidewalls (see Chap. 2). The downslope
flows converge midvalley then flow out of the valley. The combination of surface
meteorology, surface chemistry, and scanning Doppler lidar measurements allow us
to conclude that it is the latter mechanism that is responsible for the O3 removal.
8.8.3 Precipitation-Enhancement Effects of Mountain Waves
A final example of an effective use of multiple instrumentation systems is a study of

the influence of mountain waves on precipitation in central Arizona. In this region
water is critical, so that identifying times of seeding potential—when wintertime
cloud systems could be seeded to produce even small increases in the amount of
precipitation reaching the ground, to increase streamflow and fill reservoirs—is a
critical application of meteorological principles. Counteracting potential benefits is
Fig. 8.50 Doppler lidar conical scans at 2ı elevation of radial velocity. a) Afternoon scan showing
flow into Pitt Lake Valley to north (top). b) Nighttime scan showing flow coming out of Pitt Lake
valley (bottom). Lidar L is at center of each plot and range rings are at 5 km. North is up, surface
measurement site indicated by H, and mountain ridges indicated in Fig. 8.3 are denoted M and N.
Negative velocities (cold colors) indicate flow toward lidar and positive velocities (warm colors)
flow away. Fraser and Pitt Rivers are outlined in blue (Banta et al. 1997a)
Fig. 8.51 Conceptual view of flow going into valley during daytime (top panel), and two
possibilities for nocturnal outflow (bottom two panels)
the possibility of hazardous flash flooding events, which could be made worse by
inappropriate cloud seeding. These effects were studied during a wintertime project
described by Reinking et al. (2000) and Klimowski et al. (1998), which took place
in early March of 1995. In this study a variety of instrumentation platforms were
deployed to a region of central Arizona, where the topography consists of a series
of three nearly-parallel, northwest-southeast-oriented ridges spaced 50 km apart.
Southwesterly flow in moisture-bearing storm systems blew perpendicular to these
ridges, so that when this flow is statically stable (which is typical), the upstream
ridges produce wave activity that extends downstream (Fig. 8.52). Ascending air
in the wave updrafts enhances liquid water concentrations and thus precipitation
over the downstream topography Bruintjes et al. (1994). If much of this liquid is
supercooled, a potential for beneficial cloud seeding to increase precipitation at the
ground exists.
These possibilities were explored using a combination of instrumentation that
featured a scanning microwave radiometer (SMR) and the 8.66-mm-wavelength
radar already described in Sect. 8.5 which measured backscatter reflectivity Z,
Doppler velocities, a signal-to-noise variable, and the elliptical depolarization ratio
(EDR). The EDR, similar to the CDR described in Sect. 8.5, allows discrimination
Fig. 8.52 Schematic north–south cross section showing central Arizona topography and location
of wave structures. a) w contours in m s1 , b) isentropes at 1 K intervals (Reinking et al. 2000.
between spherical liquid-water droplets and ice crystals, which are non-spherical.
Rawinsonde, a surface mesonet, and a raingauge network also provided critical
measurements, and supplementary data from a Doppler sodar, a high-spectral-
resolution lidar (HSRL), and an instrumented aircraft proved useful in the analysis.
The basic precipitation systems were warm-sector synoptic-scale cloud bands
moving over the topography, where orographic lifting and wave processes modified
and amplified them. Radiometer (SMR) data clearly highlighted the passage of
the cloud bands as significant increases in total-column water vapor (precipitable
water Pw ) and integrated liquid water. The SMR also performed horizon to horizon
elevation scans, which took about 5 min to complete, showing the evolution of
the wave structure and intensity through each event by documenting the horizontal
variation of the vertically integrated moisture fields. The 8-mm radar gave a much
more detailed view of the wave structure (using scans analogous to those shown
in Figs. 8.22–8.24), and also the presence of liquid versus ice via polarization, the
height of the freezing level via the bright band, and the cloud-level winds using VAD
scanning and processing.
These measurements showed that the lee-wave updrafts generated significant
(supercooled) liquid water and also some ice particles, but not enough ice for
precipitation of any amount. Precipitating ice did fall out the bases of the upslope
orographic (as opposed to wave) clouds though. Radar scans showed that the moist
upslope flow produced an orographic cloud on the upwind side of the barriers, which
set up a seeder-feeder configuration (see Chap. 6) as ice precipitation from the wave
clouds fell through the orographic cloud collecting cloud water. A conclusion of the
study was that it is likely that the seeder-feeder mechanism produced considerable
precipitation that would not otherwise have reached the ground from either cloud
type alone. The SMR and radar were able to determine precipitation periods when
wave augmentation was present and when not. Precipitation from periods of lee-
wave activity accounted for 75–85% of the total, including during the period
preceding a flash-flood event that occurred during the project.
Monitoring of spatial structure and temporal changes of the moisture, wind, and
other fields by remote sensing systems thus revealed the critical role of wave-
induced updrafts in producing enough condensate for precipitation to form in
these Arizona cloud systems. The amounts of supercooled liquid water in these
clouds and the low precipitation efficiencies, which were also calculated from the
measurements, suggest favorable seeding opportunities, especially early in the storm
cycles. Flash flooding was observed to occur when an existing snowpack became
saturated, so a forecasting signature in the radar data was identified as: high values
of liquid water and a low freezing level (bright band just above the height of the
topography) in the presence of snowpack on the slopes of the ridges.
These have been a few examples to illustrate how deployment of a variety of
instrument types can significantly increase the ability to characterize and understand
flows and other meteorological processes occurring in a region of the atmosphere.
Such instrumentation chosen to address the problem of interest can produce
a more complete, comprehensive picture of features, structure, and processes
occurring in the atmosphere, against which NWP models can and should be
validated. Of course, producing such a comprehensive picture assumes that not
only that the instrument platforms are deployed to the same area, but also that an
integrated analysis of the combined dataset is performed. Many other examples
could be cited from other complex-terrain field projects and subsequent analyses
that have occurred over the past decade, for example the Mesoscale Alpine
Programme (MAP: Bougeault et al. 2001; Mayr et al. 2004), the Terrain-induced
Rotor Experiment (TREX: Grubišić et al. 2008), the Cumulus, Photogrammetric,
and Doppler Observations Experiment of 2006 (CUPIDO: Damiani et al. 2008),
the Convective and Orographically Induced Precipitation Study (COPS: Wulfmeyer
et al. 2008; Kottmeier et al. 2008), and others.
8.9 Measurements and NWP in Complex Terrain
Meteorological applications and research have become increasingly reliant on

numerical weather prediction (NWP) models. Observationalists will point out
that such models are imperfect analogs of the atmosphere. The finite-difference
approximations ignore higher-order terms that may be important particularly in
areas of interest (strong spatial or temporal gradients), may produce resolution-
dependent results, and are often excessively diffusive. Many key physical processes
are poorly represented or parameterized, leading to unreliable results. Modelers
counter that measurements provide incomplete coverage, instrument sites may be
unrepresentative or poorly exposed especially in complex terrain, and instruments
are subject to measurement errors that are not always well characterized. Models
provide information about key quantities that are difficult to measure, such as
horizontal pressure or temperature gradients, and about atmospheric conditions

between measurement sites that are consistent with the equations of motion. Since
both viewpoints are essentially correct, this is a situation where each viewpoint
would profit from the other. For example, the question of how well the “virtual
meteorology” generated by models relates to the real atmosphere can only be
answered by observations (Banta 2010).
NWP modeling is often an important tool for interpreting phenomena in mea-
sured datasets. Conversely, measurement studies make important contributions
to numerical-model development by providing data for model validation and
verification studies. The other significant use of measured data (besides the obvious
need for initial data fields, and boundary conditions for retrospective model runs)
is in targeted process studies to improve parameterizations, which will not be
emphasized in this section despite its importance. Verification of model output in
mountainous terrain often consists of more or less qualitatively determining whether
observed features are reasonably predicted, but more quantitative procedures, such
as interpolating measured data to an analysis grid (or alternatively, interpolating
model data to the location of the measurements) and calculating rms differences
between analyzed and modeled values, have also been developed in attempts to
evaluate model performance.
The availability of detailed remote-sensing datasets as part of field-campaign
deployments has made it possible to verify model output in much greater detail
than if such data were unavailable. A few examples illustrate. Drobinski et al.
(2003) included a ground-based Doppler lidar and a scintillometer in addition to
conventional instrumentation to compare with mesoscale simulations of downslope
windstorm (foehn) structure in the complicated topography of the central Alps. The
study by Darby et al. (2002a) at the central California coast (Sect. 8.7) included
sensitivity studies on the roles of land-sea contrast, the coastal mountain range, and
the inland Sierra Nevada in reproducing the diurnal cycle of the wind profile. The
use of time-height cross sections of westerly-component flow derived from Doppler-
lidar wind profiles, compared with similar analyses from model runs, showed that all
three features are necessary for agreement between model and observations. Some
other studies have included intercomparisons between NWP output and detailed
remote-sensing results, including studies in the Colorado Front Range (Poulos and
Bossert 1995; Poulos et al. 2000; 2007) and in the Central Valley of California
(Zhong et al. 2004; Bao et al. 2008; Michelson and Bao 2008). In southern France
Drobinski et al. (2005) studied northerly Mistral flow through the Rhone River
Valley using an airborne Doppler lidar to supplement arrays of more conventional
instrumentation. This study, echoing a general result of all these studies, found
that model analyses from an MM5 run were capable of good agreement with
measurements in some regions of the simulation domain, but in others the model
was unable to reproduce key dynamic features (such as flow reversals) or important
small-scale variability.
A few studies have used the high resolution of Doppler-lidar scan data to directly
compare with mesoscale model output. Clark et al. (1994) performed detailed
comparisons of high-resolution, three-dimensional simulations of a downslope
windstorm in Boulder, Colorado with measurements that included a Doppler lidar.

They concluded that, “the model prediction was successful in : : : that the model
predicted a windstorm when one did in fact occur.” The model also predicted some
other dynamic-relevant aspects that were observed by the lidar, including the gross
structure of the hydraulic jumplike flow to the windstorm, a turning of the wind from
westerly to northwesterly with height, and a 10–15-min periodicity to the surface
gusts. But other details were poorly simulated, including the location and timing of
the windstorm event itself, and the location and temporal/spatial variability of the
jump feature.
In the complex topography near Colorado Springs, Colorado, Darby and Poulos
(2006) used Doppler lidar and other instrumentation to study the life cycle of a lee-
wave rotor event. This project was aimed at the applied objective of understanding
terrain-generated low-level clear-air turbulence (CAT) encountered by aircraft on
approach to the airport. Unique to this study were simultaneous research aircraft
measurements of moderate terrain-induced CAT and lidar vertical-slice scans
showing the atmospheric structure responsible for the turbulence (middle panel,
Fig. 8.53). East–west lidar cross sections such as Fig. 8.53a were compared with
model-generated cross sections for the same time and location. Many of the flow
features, such as some rotor flow reversals were captured by the model, but other
important features were completely missed. Of those flow features that appeared
in the simulation, many were displaced in space or time, elongated or otherwise
distorted, or exhibiting erroneous magnitudes (e.g., Fig. 8.54). The lidar scan data
were interpolated to the model grid for calculation of rms differences. The rms
differences were much larger above 1 km AGL than below, attributed in part to the
lower resolution aloft due to the stretched grid with height. The simulation would
thus again be considered successful in the qualitative sense that it produced lee-
wave rotor structure when one was observed, suggesting that at least some of the
important physical processes were adequately represented by the model. But for
applications requiring accurate locations, timing, or magnitudes of variables, such
as peak wind speeds, these detailed comparisons point to a need for improvement in
the model.
As described in Sect. 8.7, a wealth of both conventional and remote-sensing
data was available during the month-long Vertical Transport and Mixing (VTMX)
campaign in the mountainous basin of the Great Salt Lake, Utah (October 2000).
The instrumentation used for these studies included a surface mesonet, radar wind
profilers at several locations, rawinsondes, and a Doppler lidar located near the
middle of the basin. This array of instruments was used in two mesoscale-modeling
studies which made comprehensive comparisons between measurements and model
output: Zhong and Fast (2003) made detailed temperature, moisture, and wind
comparisons with variable fields generated by three different models (RAMS, MM5,
and Meso Eta), and Fast and Darby (2004) made direct, detailed (including point-
by-point) comparisons of model output with scan data from the Doppler lidar.
The VTMX campaign was designed to study the complex nocturnal flows in the
basin. Figure 8.34 shows an example of this complexity for a synoptically quiet
Fig. 8.53 East–west cross sections of Doppler velocity taken by Doppler lidar at Colorado Springs
airport. Cross sections are for a) 0120 UTC, b) 0611 UTC, and c) 0938 UTC on 1 April 1997. West
and mountains to right, basic ridgetop wind flow from right to left. Star in panel b indicates location
where King Air research aircraft doing missed approaches encountered moderate turbulence
(Darby and Poulos 2006. © American Meteorological Society. Reprinted with permission)
Fig. 8.54 Comparison of the east–west wind component from an east–west cross section of
RAMS grid 4 (top) and the lidar east–west wind component obtained from an east–west range–
height scan (bottom). The RAMS cross section has been cropped to match the lidar’s range. West
is to the right. Note that lidar data within 2-km range of the lidar have been deleted because the
elevation angle of the beam is too steep to calculate a horizontal wind in this region. Solid contours
represent the westerly component flow and dashed contours represent the easterly component flow
and contours are every 4 m s1 . The bold line is the zero contour. Westerly wind speeds greater
than 16 m s1 have heavy shading and the easterly component flow has light shading. (Darby and
Poulos 2006. © American Meteorological Society. Reprinted with permission)
night, after the thermally forced midbasin winds, which had blown from a northerly
(upbasin) direction during the daytime, shifted to a southerly downbasin LLJ at
night. Adding to the complexity are the cold-air outflow jets emanating from the
canyons in the mountain range to the east. The interaction of these canyon outflows
with the southerly jet was a convergence line where upward motions were focused.
The types of analysis performed by Fast and Darby (2004) included comparisons
of individual lidar scans with simulated output at different times during each case-
study night, time-height cross sections of wind speed, statistics (mean, standard
deviation) of radial winds averaged over a horizontal region (box) as a function of
hour of the night, the location of the basin convergence zone (see Fig. 8.34), and
point-by point regression analyses of model versus measured radial-wind velocities
for each scan, for each of the eight study nights, and for the entire sample of 1.3
million data/model-value pairs. Fast and Darby also presented comparison statistics
for radar wind profiler data at two locations in the basin with model output, which
supplemented visual comparisons of 30-min profiles of profiler versus model for
individual nights, that had been presented by Zhong and Fast (2003). In addition to
the profiler comparisons, Zhong and Fast (2003) also presented model comparisons
with rawinsonde profiles and surface mesonet data, including turbulence and flux
measurements.
The availability of such high-resolution validation data allowed both studies
to conclude overall that the models runs “were able to qualitatively reproduce
the general features in boundary-layer structure and evolution,” but they found
“relatively large errors associated with the timing, dimensions, and magnitude”
of the simulated features (Zhong and Fast 2003), consistent with results already
described.
Some specific areas proved challenging for the models. Using the Doppler-lidar
scan data, Fast and Darby (2004) found that correlations were weak when the flow
exhibited strong spatial variability that was poorly represented in the simulations
and during transition periods when the onsets of wind shifts were mistimed by
the model. LLJ speeds were often much lower than observed (by several m s1 ),
the height of the maximum speed was often off, and examples of 90ı directional
errors—and even near 180ı where observed flow reversals were unpredicted—were
noted. Lulls in wind speed observed in the middle of some of the study nights
were not predicted. The authors noted that larger errors were found at levels above
ridgetops than below, “indicating that the interactions of the valley and ambient
winds near the height of the surrounding mountains were not simulated as well as
the flow within the valley.”
An important finding of the Zhong and Fast (2003) study was that despite
significant differences among the three models and model configurations used in
their study, the “types of forecast errors were surprisingly similar,” suggesting that
the results were general and related to the state of the art in model parameterizations,
or perhaps even more fundamental limitations of finite-difference modeling. As with
previous studies the three models were found to “qualitatively reproduce general
features of the BL structure and evolution,” such as the diurnal character, but they
noted major discrepancies in timing, location, and magnitude of flow features and
predicted variables in all three. In addition to these errors in the representation of
flows, the largest errors appeared aloft between 1 and 2 km above the basin floor,
even for the case of strong synoptic forcing, again indicating, “that the dynamic and
thermodynamic effects of the mountains on the ambient flow and/or the regional
pressure gradients were not simulated adequately” (Zhong and Fast 2003, p. 1314).
All three models exhibited a significant cold bias at all levels below ridgetops.
A comparison of the near-surface static stability or ™ gradient, using the ™ difference
between 15 and 150 m as a metric, showed that all three models produced weaker
nocturnal inversions than observed, and the agreement was especially bad for the
strong midbasin inversion that formed on nights with weak synoptic winds. The cold
bias and near-surface stability errors can be viewed as an indication of the models’
inability to correctly represent all relevant dynamic interactions, especially near
the surface, for such thermally forced circulations, for which the primary forcing
is by surface interactions. A further result of the Zhong-Fast intercomparisons is
that these model biases were situation dependent, differing for example between
strong-wind and weak-wind nights, and of course between night and day. This
result indicates that a simple universal bias or linear-regression adjustment is not

apt to correct this problem. The existence of such situation-dependent biases further
indicates that the practice of ensembling may not necessarily produce results that
converge to a solution that relates to future states of the real atmosphere.
To further investigate the predicted discrepancies in near-surface inversion
strength and the cold biases, Zhong and Fast (2003) looked at factors that could
contribute to weak inversions in models, including vertical resolution, excessive
vertical diffusion, and the surface energy budget. Doubling the vertical resolution in
the lowest kilometer produced a small improvement in the temperature profile, but
errors were still large. Model diffusion depends on the turbulence parameterization
employed, the TKE magnitude predicted, and the value of the eddy diffusivities
KH and KM . Two different turbulence schemes were substituted for the original
(Gayno-Seaman) scheme with little difference, indicating that current state-of-the-
art schemes all perform about the same with respect to this problem. TKE values
and accuracy varied from site to site, but overall the TKE values were too large
in two of the models, which would generate excessive mixing. Consistent with the
TKE, the simulated KH values were also considerably too large, likewise leading to
excessive mixing.
In the surface energy budget, the magnitudes of the simulated surface heat fluxes
were too large both day and night, so excessive downward nocturnal heat fluxes
could contribute to the cold biases at night. Daytime solar radiation fluxes were
well simulated, but the net radiation was significantly underpredicted, although one
of the models (Meso Eta) significantly overpredicted both, despite the cold bias. At
night the longwave radiation loss was overpredicted by all three models, consistent
with the cold bias.
Thus, the cold bias in most cases was associated with errors in the radiative
fluxes. Simulated nocturnal inversion weakness was associated with excessive
mixing. The issue of excessive diffusion in models is not straightforward to address.
A component of the diffusion is due to finite-difference numerical operators, and this
may exceed the magnitude of atmospheric diffusion. A certain amount of explicit
diffusion is generally required for numerical stability of finite-difference models.
Simply setting the diffusion coefficients or equivalent to small values is not an option
without significant alterations of other modeled processes, and in regions where
numerical diffusion exceeds atmospheric values, it may not be possible to make the
modeled diffusion small enough. The question of how this affects the model results
is an important one.
The studies that have used high-resolution measurements to compare with
mesoscale NWP output as summarized in this section demonstrate that one can find
“model friendly” periods or days or events, with reasonable agreement between
measurements (or selected measurements) and NWP output, which one could tout
as model successes. Alternatively, one can also find examples where agreement
was very poor. To be useful for operations and many other applications, consistent
performance is desirable, or at least to know when the model prediction is good and
when not.
The availability of comprehensive measurements at high time and space res-

olution makes it possible to validate model output to a much higher standard
than previously. In spite of this availability, relatively few studies have used such
observational resources to test model capability, such as the examples given here.
Consistent findings from these studies include that the gross features of the flow
were often represented in a qualitative or sometimes semi-quantitative sense. The
models often had difficulties with:
• Timing of appearance or onset of phenomena
• Location, dimensions, or sometimes even existence of atmospheric flow features
• Quantitative aspects of the flows and processes
• The largest discrepancies were often found at or above the ridgetops, rather than
near the surface or within valleys, indicating that the models are not capturing
very well the interactions among the flow within the valley or low areas, the flow
at and above ridgetops, and the topography.
From a modeling or research point of view the overall ability of models
to represent gross aspects of atmospheric phenomena in a qualitative or semi-
quantitative sense is significant and encouraging, indicating that many or most
of the relevant physical processes—and in some situations all of the dominant
processes—may be adequately simulated. Indeed, the demonstrated usefulness of
current NWP models as tools or guidance in formulating operational forecasts
must be viewed as a triumph. For some applications the discrepancies may be of
little consequence. Many other current applications or operations, however, require
very accurate quantitative predictions of certain quantities, for example freeze
warnings, air quality and emergency response (which require accurate predictions
of concentrations of atmospheric contaminant species as a functions of space and
time), and wind energy (which requires very precise predictions of wind speed).
These applications (esp. wind energy) and others may also require accurate timing
and location predictions for changes in atmospheric conditions. The fact that wind
direction can be off by 90ı , or even 180ı is some conditions, in complex terrain
should be of concern for emergency response.
For applications such as these, it is critical for users to understand the limitations
of model predictions (Banta 2010). It is the job of the research community to
characterize these limitations, as well as the strengths, of NWP models for the
users to understand how best to use the information provided by the models.
This should be accomplished through careful, comprehensive detailed comparisons
between model output and the best available measurement datasets. Of course the
ultimate goal of measurement programs in this context should be the improvement
of model skill and performance. Detailed measurement campaigns, such as those
described here, give indication of model limitations/deficiencies, but may not in
general provide guidance on how to improve the models. Even more comprehensive
and carefully planned measurement programs will be required to sort out the
representation of the complex interactions of processes on all relevant length and
time scales that must be enhanced to make NWP models better.
8.10 Turbulence
The measurement of turbulence and turbulent fluxes in mountainous terrain, as in

the atmosphere in general, is still an important frontier. Besides the crucial role
of turbulent diffusion in transport and dispersion applications, turbulent mixing
processes are a crucial unknown in the understanding of many mountain dynamics
issues. For example, BL processes exhibiting a diurnal cycle are driven by the
heating and cooling cycle, which starts at the earth’s surface and is transferred
upward into the atmosphere by turbulent mixing. The rate of upward diffusion or
transport, which varies from day through night, is important to the dynamics of
the flow systems and the ability to simulate them. (The major exceptions to surface
forcing of the diurnal cycle here are direct radiative absorption by clouds, aerosol, or
other atmospheric species, which can directly heat the atmosphere, but are generally
less important drivers of diurnal wind systems.)
This becomes especially a problem for NWP modeling, where all other terms
in the dynamic momentum equations are explicitly calculated, but the turbulent
mixing term must be filled in with an approximation or “parameterization,” which
attempts to represent the physical processes of mixing, but which may be non-
physically based. Despite this critical need for turbulent-flux measurements, few
exist in mountainous settings, and most of those are at the surface. Therefore,
this section will present a brief discussion of some approaches that have been
attempted over more homogeneous terrain, with the prospect that some of these
techniques may prove useful to address complex-terrain dynamic issues in the
future. Before discussing these remote-sensing approaches, some examples of
tower-based turbulence measurements, some in complex terrain, will be discussed.
Near the surface, tower-mounted in-situ measurements of turbulence quantities
have been made in mountainous settings. Surface heat fluxes have been measured
using eddy-correlation methods in studies of the surface energy budget (Whiteman
1990 and Chap. 2). An important consideration in making such measurements is that
they require a sufficiently long averaging period during which the turbulence can be
considered stationary, which is generally more than 15 min for variance (second-
order) quantities such as TKE but much longer for higher-order quantities or for
covariances of variables that are not highly correlated (Lenschow et al. 1994). The
problem in complex terrain, of course, is the proliferation of locally forced and
obstacle-related flows, whose interplay produces nonstationary behavior in tower-
measured time series.
The stable boundary layer (SBL) is especially difficult. Mahrt and Vickers (2006)
studied nocturnal stable conditions using tower-measured profiles of turbulence and
flux quantities in a broad mountain basin and compared them with similar profiles
obtained over flat terrain. They used special processing and analysis techniques to
separate the purely turbulent motions from larger-scale (lower-frequency) variations
(Vickers and Mahrt 2003; 2006). Under light wind conditions they found a shallow,
weak, turbulent BL in both cases, but over the flatter terrain, turbulence levels
became vanishingly small above the BL, whereas in the mountain-basin profiles,
turbulence magnitudes remained larger above the BL. Mahrt and Vickers (2006)
attributed the stronger turbulence above the mountain BL as compared with the flat-
land BL to advection of upstream-induced disturbances to the measurement site.
Thus, turbulence measurements can be made with care using tower-mounted in-
situ instrumentation in mountainous terrain, but the interpretation of profile data
becomes complicated by the likelihood of horizontal advection effects. In rugged
terrain, spatial representativeness of the measurement also becomes an important
issue.
A number of remote-sensing techniques for profiling turbulence quantities have
been attempted. The hope for the future is that some of these techniques will prove
reliable enough to be useful in providing accurate estimates of the magnitude of
turbulence quantities above the surface, their profile shapes, and the evolution of
these profiles through diurnal or large-scale changes. The need for accurate, reliable
flux measurements is compounded by the fact that the term in the dynamic equations
is the flux divergence, i.e., the vertical gradient of the fluxes @F/@z, where F is the
vertical flux of some quantity. In the absence of such data, it is unknown whether
current NWP simulations are modeling them properly or how to better represent
them in the models.
At present the only methods for continuously measuring turbulent fluxes
with height in complex terrain are with tall towers instrumented with 3-D sonic
anemometers (Poulos and Burns 2003). We can classify the remote-sensing
techniques as intrinsic pulse- or averaging-volume measurements of turbulence
quantities, and extrinsic procedures, where pulse-volume averaged values of wind
speed, for example, are used as data points to calculate statistics or other quantities.
These procedures include time-series measurements, scanning techniques, and
proxy measurements. Some of these techniques have been described in reviews by
Englebart et al. (2007) and Emeis et al. (2007).
Intrinsic turbulence measurements include the backscatter from sodar or radar
and the second moment or “Doppler spread” of the Doppler spectrum (e.g.,
Fig. 8.13). The backscatter can be directly related to useful turbulence quantities
(structure functions) via the Bragg-scattering relationships (Sect. 8.3.3.2) if the
half wavelength of the signal transmitted by the instrument lies within the inertial
subrange of the turbulence, illustrated in Fig. 8.55. This condition is routinely
met in deep boundary layers during daytime convective conditions over rela-
tively flat topography. But in the nocturnal SBL or in rugged terrain, the pulse
volume often includes turbulent eddies in the production or energy-containing
range, which precludes a generalized quantitative relationship between backscatter
and atmospheric turbulence quantities. Another caveat in using these techniques
especially in complex terrain (as well as at night) is the assumption that the pulse
volume is uniformly filled with scatterers. Mean gradients (e.g., shear) or other
variations across or within the pulse volume will enhance deviation quantities within
the volume and contribute to the variances, structure functions, etc., even though
they are measurement artifacts other than turbulence.
Sodar backscatter or radar reflectivity is an intrinsic measure of the intensity of
the small-scale turbulence within the pulse volume. Such small-scale turbulence is
Fig. 8.55 Turbulence kinetic energy spectrum, showing spectral density E versus wavenumber k
both on logarithmic scales (Wyngaard 1986. © American Meteorological Society. Reprinted with
permission)
Z a b c d
zi
Θ U B Cn2
Fig. 8.56 Schematic typical daytime convective boundary layer profiles of a) potential tempera-
ture, b) wind speed, c) aerosol backscatter as measured by lidar, and d) reflectivity as measured
by radar wind profiler, showing maximum reflectivity at top of convective layer (Banta and White
2003)
strong within and weak above the BL, so that profiles of sodar or profiler backscatter
have been used extensively to determine CBL depth (e.g., Coulter 1979; Beyrich
1997; White et al. 1999; Seibert et al. 2000; Banta and White 2003; Emeis et al.
2008). Because of strong turbulence within the BL and frequent strong gradients of
humidity and temperature at the top of the BL, backscatter often peaks at the top
of the BL, as illustrated in Fig. 8.56d. When present, this peak is taken to be the top
of the BL; when weak or absent, the BL is taken to be the top of the layer of strong
turbulence.
Extrinsic, calculated turbulence quantities generally use the pulse-volume data
as if they were point measurements, attributed to some location within the sampling
volume, such as the centroid. Some approaches are to use these values as time
series, to use scan-averaging techniques, and to use measured values as proxy

measurements to obtain other variables.
Straightforward applications of extrinsic processing are exemplified by time-
series analyses of vertical velocity w obtained at each range gate when remote
sensors are operated pointing vertically. The instruments can then be thought
of as equivalent to very tall towers that measure only w (Martner, B., personal
communication). Vertical velocity and vertical velocity variances w0 2 have been
measured using Doppler radar and lidar (Lenschow etal. 1994; Tucker
et al. 2009).
Other quantities include third and fourth moments w 0 3
and w0 4 (Lenschow
et al. 2000), coherence, structure functions, spectra (Lothon et al. 2006), and fluxes
or covariances using collocated remote sensors (Senff et al. 1994, 1996). Profiles
appeared to be consistent with expectation, but no independent verification data
were available for these studies.
Scanning techniques have been tested using both azimuth and elevation scanning.
The VAD technique uses data from conical scans to profile the mean horizontal
wind by fitting a sine wave to the velocity variation for each range gate as a
function of azimuth, as described in Sect. 8.3. However, by removing the mean sine
wave from the data, the residual velocities can be further analyzed for turbulence
information. At certain specified elevation angles, analysis of the fluctuations can
produce profiles of turbulence kinetic energy (TKE), momentum flux, and the
vertical flux of TKE (triple-correlation term) (Wilson 1970; Kropfli 1986a). This
technique has been used with Doppler radar (Kropfli 1986; Frisch 1991; Frisch
et al. 1992) and Doppler lidar (Eberhard et al. 1987; Banta et al. 1997b) to calculate
profiles of these quantities.
In another approach, elevation (vertical) scans pointed along the mean-wind
direction have been analyzed by binning the measurements in horizontally oriented
vertically stacked bins, and calculating the mean wind and variance in each bin
(Banta et al. 2002, 2006), forming profiles of the mean-wind speed and streamwise
variance. Results of this technique showed good agreement with tower-mounted
sonic anemometer data in an extensive series of regression analyses (Pichugina et al.
2008). An important consideration in attempting to use these scanning techniques
(VAD or vertical-slice scan procedures) in mountainous terrain is the need for
horizontal homogeneity across the scan ring- or volume-averaging region being
measured. Careful siting or scan orientation may be required to use these techniques
successfully in mountainous terrain.
A method for calculating TKE dissipation " from Doppler-lidar radial velocity
data has been described by Frehlich and Cornman (2002). The method uses velocity
differences at different range-gate separations down to the range resolution of the
instrument, and expressions valid in the inertial subrange of the turbulence. Since
the current state of the art in Doppler lidar has minimum resolution down to 30 m,
this technique would be most useful for deep daytime convective boundary layers,
where the range-gate spacing would be within the inertial subrange.
As an example of proxy measurements, a formula for heat flux can be reduced
to an expression relating the third moment of w to heat flux, so that theoretically,
vertically pointing velocity measurements can be used to infer heat flux profiles
(Wyngaard and Cote 1971; Gal-Chen et al. 1982)

g 0 0 1 @p 0 " @ 1 03
w D w0 C C w (8.7)
0 0 @z 3 @z 2
by ignoring the pressure term. This expression assumes horizontal homogeneity, so

may be difficult to apply in mountain settings.
To date these techniques must be regarded mostly as in the demonstration stage,
as pointed out for sodar by Coulter and Kallistratova (2004). Verification data, when
available, have been mostly limited to individual profiles or data points, so the
generality and usefulness of these techniques has not been established. A critical
but difficult research need, therefore, is to develop and verify procedures for the
measurement of turbulence quantities above the Earth’s surface. A step in that
direction is the recent study by Pichugina et al. (2008), who verified velocity
variance profiles calculated from Doppler lidar scan data, taken during 11 nighttime
study periods of 8–12 h each, against data from a 120-m tower.
8.11 Conclusions
Traditionally in science, careful, well conceived measurements and observations

lead to insight and understanding of physical processes. Enhanced understanding
leads to improved conceptual, analytical, and numerical models. Currently in the
atmospheric boundary layer many physical processes are not well characterized or
understood, because of a need for high-quality measurement datasets, especially
measurements above the surface and in complex terrain. Recent attention has been
directed at the idea of matching measurement-system capabilities to the meteoro-
logical features and phenomena that need to be characterized, a “phenomenological
approach to observational requirements,” according to the 2009 National Research
Council (NRC) report (National Research Council, 2009).
In this chapter we have emphasized the need for high-quality, comprehensive
datasets to characterize as completely as possible the structure and evolution
of meteorological fields over a domain, to gain insight into the kinematics and
dynamics driving the changes in those fields. These fields are needed not only
to advance basic understanding, but also to assess the capabilities today’s NWP
models, to determine why at times they do not reliably reproduce the behavior of
the atmosphere, and to correct shortcomings in model physics and other modeling
practices. Current technology in many cases offers choices of sensing systems. For
example, scanning, pulsed Doppler lidar provides the high spatial and temporal
resolution and high precision in the lowest several hundred meters of the atmosphere
needed for research and many applications, and as a bonus, provides scan image
data often useful in interpreting the profile data. For applications requiring data
through a deeper layer, or during cloudy or foggy conditions, where high resolution
and precision are less important, radar wind profilers represent a well characterized
option, and sodar provides a useful, inexpensive alternative that provides profile
data, even in fog or cloud conditions. In this example, choice of an appropriate
wind-profile measurement system should depend on how well the strengths versus
limitations of each type of instrument match with the phenomena to be measured
and the prevailing conditions as they affect instrument performance.
The notion of choice among instrument systems implies that a decision must
be made to avoid redundancy and added expense. But the discussions of strengths
versus limitations of measurement systems forces a reconsideration of what con-
stitutes redundancy, especially if all-weather information is required as for many
operational applications. As in the previous example, a Doppler lidar may provide
the best profile data most of the time, but to provide the needed information to deeper
levels or during fog or low-cloud conditions, it may be necessary to have radar wind
profiler and/or sodar available. Thus the design of an all-weather, high-resolution
profiling site may require deployment of all three types of profiling system, despite
the apparent redundancy in wind measurement systems. This is the situation in the
complex terrain surrounding the Hong Kong International Airport, where profiler,
sodars, a surface mesonet, and a Terminal Doppler Weather Radar are all used along
with the two Doppler lidars to fill different observational niches.
This chapter has emphasized wind-profiling technologies, which are especially
important at smaller scales that often characterize complex-terrain and mountain
flows. As described in the 2009 NRC report, the continued development and use
of temperature and trace species, especially water vapor, are very important to
accompany the wind-profiling capability. Datasets comprised of high-resolution
measurements of all these quantities over a sufficiently large domain to cover all
the important interacting scales of motion will be great assets in advancing the state
of the art in mountain and complex-terrain meteorology.
Acknowledgments Portions of the work in this chapter were performed under the NOAA
Air Quality Program. The authors thank Drs. Gregory S. Poulos, Stephan De Wekker and
two anonymous reviewers for helpful suggestions leading to significant improvements to the
manuscript, and to Debra Dailey-Fisher for her expert figure preparation.
Appendix
To begin to understand Doppler processing it is necessary to probe deeper into

sine-wave behavior. Recall that the three basic specifications for a sine function
(Fig. 8.57, top panel) are its frequency f, its amplitude A, and its phase ®:
s .t/ D A sin .2f t C ®/

Fig. 8.57 Sine-wave signal S(x)

as a function of time t (top).
Outgoing signal at an instant
of time as a function of
distance x, reflecting off an
object at distance D1 . Return
signal indicated by dashed
curve (middle). Same as
middle panel, except
reflecting object is at distance 0 t
D2 (bottom)
0 x
D1
1
0 x
D2
2
and that the frequency and wavelength are interchangeable, related by the speed of
propagation of the wave according to the relationship: c D œf or œ D c/f. Radar and
lidar have fundamentally different approaches to the Doppler determination.
The radar case is illustrated in Fig. 8.57. First consider for convenience a
continuous-wave (CW) signal, and for now consider this signal at an instant of time
as a function of the distance x from a source (Fig. 8.57, middle panel), instead of
as a function of t. Also for convenience, consider the instant when the CW signal
has initial phase ®0 D 0 at its source, i.e., where x D 0—the source would probably
be the instrument antenna. Then, let the signal intercept a target at distance D1 and
return to the source location. Note that where the signal intercepts D1 , it is most
often at a different phase from the phase at the source ®0 , because in general D1 is
not at an integral multiple of œ from the source. The signal “reflected” back to x D 0
has also returned at a ® different from ®0 . If the target is not moving at D1 along x,
we can change perspective and look at the “reflected” signal received back at x D 0
as a function of time. This transmitted source signal is a temporally varying sine
wave of frequency f, as in the top panel of Fig. 8.57. The received signal back at the
source position x D 0 as a function of time will also be the sine wave synchronized
(at frequency f ) with the transmitted signal, but with a constant phase offset ®1 .
(The mean particle/droplet motion produces a relatively tiny change in the frequency
of the backscattered signal, but at radar wavelengths this frequency shift is not used
to calculate Doppler velocities.)
If the target is located at a different distance D2 , then the phase intercept at D2
will be changed (Fig. 8.57, bottom panel) and the phase offset received back at
the source position will also have shifted, i.e., ®2 will be different from ®1 .
Where this argument becomes related to Doppler processing for radar, is if the target
moves from D1 to D2 in a time of t, then the speed of movement is related to the
rate of change of the phase offset, expressed as [(®1 ®2 )/t]. This leads to
a derivative expression d(®)/dt, which is proportional to the speed of movement
of the target (see Doviak and Zrnić 1993, Eq. 3.30; Battan 1973, Eq. 8.2; Rinehart
2010, p. 99). At Doppler radar œ’s, this phase-shift approach works for atmospheric
returns, even though the scatterers are not a solid target as portrayed here, because
(1) the scatterers in each pulse volume as a whole have not moved more than a
distance equivalent to œ in the time interval between pulses, and (2) the distributed
scatterers are not significantly rearranged in the time interval between shots in such
a way that the phase information within each pulse volume is distorted. The latter
could be stated as, for scatterers seen at radar œ’s, the atmospheric decorrelation
time due to finescale turbulence motions is larger than the pulse repetition time of
the measurements. Doppler radars therefore actually measure the phase shift from
pulse to pulse, which has units of frequency, but is not really a measurement of the
frequency of the signal. Therefore this method has sometimes been referred to as
“pseudo-Doppler” processing.
For the much smaller œ’s of lidar these conditions are not met. The distributed
aerosol-particle population does move a distance of more than the lidar œ in the
time between successive pulses, and turbulent motions do rearrange the particles
with respect to each other in such a way that the phase information in each range
gate is too scrambled to be useful for velocity calculations. Lidar systems perform
a more traditional Doppler approach, by directly measuring the frequency of the
return signal and comparing it to the outgoing frequency, calculating the Doppler
shift as in (8.5).
References
Abo, M., 2005: Resonance scattering lidar. Ch. 11 in Lidar: Range-Resolved optical Remote
Sensing of the Atmosphere, ed. C. Weitkamp. Springer Science C Business Media, Inc., New
York, pp. 307–324.
Alvarez II, R. J., C.J. Senff, R.M. Hardesty, D.D. Parrish, W.T. Luke, T.B. Watson, P.H. Daum,
and N. Gillani, 1998: Comparisons of airborne lidar measurements of ozone with airborne
in situ measurements during the 1995 Southern Oxidants Study. J. Geophys. Res., 103,
31,155–31,171.
Alvarez II, R. J., C. J. Senff, A.O. Langford, A.M. Weickmann, D.C. Law, J.L. Machol, D.A.
Merritt, R.D. Marchbanks, S.P. Sandberg, W.A. Brewer, R.M. Hardesty, R.M. Banta, 2011:
Development and application of a compact, tunable, solid-state airborne ozone lidar system for
boundary layer profiling. J. Atmos. Oceanic Technol., 28, 1258–1272.
Angevine, W.M., and J.I. MacPherson, 1995: Comparison of wind profiler and aircraft wind
measurements at Chebogue Point, Nova Scotia. J. Atmos. Ocean. Technol., 12, 421–426.
, P.S. Bakwin, and K.J. Davis, 1998: Wind profiler and RASS measurements compared with
measurements from a 450-m-tall tower. J. Atmos. Ocean. Technol., 15, 818–825.
Atlas, D., 1990: Radar in meteorology: Battan Memorial and 40th Anniversary Radar Meteorology
Conference, D. Atlas, ed. Amer. Meteor. Soc., Boston, 806 pp.
Austin, P.M., and A.C. Bemis, 1950: A quantitative study of the “bright band” in radar precipitation
echoes. J. Meteor., 7, 145–151.
Balsley, B.B., R.G. Frehlich, M.L.J.Y. Meillier, and A. Muschinski, 2003: Extreme gradients in
the nocturnal boundary layer: Structure, evolution, and potential causes. J. Atmos. Sci., 60,
2496–2508.
Banta R.M., L.D. Olivier, and J.M. Intrieri, 1990: Doppler lidar observations of the 9 January 1989
severe downslope windstorm in Boulder, Colorado. Preprints, 5th Conf. on Mountain Meteor.,
25–29 June, Boulder CO, pp. 68–69.
, 1990: The role of mountain flows in making clouds. Ch. 9 in Atmospheric Processes over
Complex Terrain, ed. William Blumen, Meteorological Monogr., 23 (No. 45), 229–283.
, L.D. Olivier, E.T. Holloway, R.A. Kropfli, B.W. Bartram, R.E. Cupp, and M.J. Post, 1992:
Smoke column observations from two forest fires using Doppler lidar and Doppler radar. J.
Appl. Meteor., 31, 1328–1349.
, L.D. Olivier, and D.H. Levinson, 1993: Evolution of the Monterey Bay sea breeze layer as
observed by pulsed Doppler lidar. J. Atmos. Sci., 50 (24), 3959–3982.
, 1995: Sea breezes shallow and deep on the California coast. Mon. Wea. Rev., 123, 3614–3622.
, L.D. Olivier, W.D. Neff, D.H. Levinson, and D. Ruffieux, 1995: Influence of canyon-induced
flows on flow and dispersion over adjacent plains. Theoretical and Appl. Climatology, 52,
27–42.
, , P.H. Gudiksen, and R. Lange, 1996: Implications of small-scale flow features to
modeling dispersion over complex terrain. J. Appl. Meteor., 35, 330–342.
, P.B. Shepson, J.W. Bottenheim, K.G. Anlauf, H.A. Wiebe, A.J. Gallant, T. Biesenthal, L.D.
Olivier, C.-J. Zhu, I.G. McKendry, and D.G. Steyn, 1997a: Nocturnal cleansing flows in a
tributary valley. Atmos. Environ., 31, 2147–2162.
, B.W. Orr, C.J. Grund, D.H. Levinson, A.S. Frisch, and S.D. Mayor, 1997b: Estimation of
TKE and momentum flux from Doppler lidar scans during LIFT. Preprints, 12th Symp. on
Boundary Layers and Turbulence, Amer. Meteor. Soc., 11–12.
, C.J. Senff, A.B. White, M. Trainer, R.T. McNider, R.J. Valente, S.D. Mayor, R.J. Alvarez,
R.M. Hardesty, D.D. Parrish, and F.C. Fehsenfeld, 1998: Daytime buildup and nighttime
transport of urban ozone in the boundary layer during a stagnation episode. J. Geophys. Res.,
103, 22,519–22,544.
, L.S. Darby, P. Kaufmann, D.H. Levinson, and C.-J. Zhu, 1999: Wind flow patterns in the
Grand Canyon as revealed by Doppler lidar. J. Appl. Meteor., 38, 1069–1083.
, R.K. Newsom, J.K. Lundquist, Y.L. Pichugina, R.L. Coulter, and L. Mahrt, 2002: Nocturnal
low-level jet characteristics over Kansas during CASES-99. Bound.-Layer Meteor., 105,
221–252.
, and A.B. White, 2003: Mixing-height differences between land-use types: Dependence on
wind speed. J. Geophys. Res., 108, (D10), 4321.
, L.S. Darby, J.D. Fast, J.O. Pinto, C.D. Whiteman, W.J. Shaw, and B.D. Orr, 2004: Nocturnal
low-level jet in a mountain basin complex. Part I: Evolution and implications to other flow
features. J. Appl. Meteor., 43, 1348–1365.
, C.J. Senff, T.B. Ryerson, J. Nielsen-Gammon, L.S. Darby, R.J. Alvarez, S.P. Sandberg,
E.J. Williams, and M. Trainer, 2005: A bad air day in Houston. Bull. Amer. Meteor. Soc., 86,
657–669.
, Y.L. Pichugina, and W.A. Brewer, 2006: Turbulent velocity-variance profiles in the stable
boundary layer generated by a nocturnal low-level jet. J. Atmos. Sci., 63, 2700–2719.
, L.S. Darby, W.A. Brewer, R.M. Hardesty, C.M. Shun, C.M. Cheng, B.L. Choy, O.S.M. Lee,
S.T. Chan, and S.Y. Lau, 2007: Detection and diagnosis of wind shear and turbulence using
Doppler lidar at Hong Kong International Airport. Final Report: NOAA/ESRL and Hong Kong
Observatory Report, 130 pp.
, 2010: High-resolution verification of mesoscale models using remote-sensing measurements.
5th Int’l Symposium on Computational Wind Engineering (CWE2010), Chapel Hill NC, 23–27
May 2010.
Bao, J.-W., S.A. Michelson, P.O.G. Persson, I.V. Djalalova, and J.M. Wilczak, 2008: Observed and
WRF-simulated low-level winds in a high-ozone episode during the Central California Ozone
Study. J. Appl. Meteor. Climatol., 47, 2372–2394.
Battan, L.J., 1973: Radar observation of the Atmosphere. University of Chicago Press, Chicago,
324 pp.
Baumann-Stanzer, K., 2004: The UHF wind profiler at Vienna airport—data quality control and
comparisons to rawinsonde data. Meteor. Atmos. Phys., 85, 165–174.
Behrens, P., S. Bradley, and S. von Hunerbein, 2008: A scanning bi-static SODAR. 14th Intl.
Symp. for the Advancement of Boundary Layer Remote Sensing, IOP Conf. Series: Earth and
Environmental Science, 6 pp.
Behrendt, A., 2005: Temperature measurements with lidar. Ch. 10 in Lidar: Range-Resolved
optical Remote Sensing of the Atmosphere, ed. C. Weitkamp. Springer Science C Business
Media, Inc., New York, pp. 273–306.
Beyrich, F., 1997: Mixing height estimation from sodar data—A critical discussion. Atmos.
Environ., 31, 3941–3953.
Bilbro, J., G. Fichtl, D. Fitzjarrald, M. Krause, and R. Lee, 1984: Airborne Doppler lidar wind field
measurements. Bull. Amer. Meteor. Soc., 65, 348–359.
Birkenheuer, D. and S. Gutman, 2005: A comparison of GOES moisture-derived product and GPS-
IPW data during IHOP 2002. J. Atmos. Ocean. Technol., 22, 1840–1847.
Bösenberg, J., 2005: Differential-absorption lidar for water-vapor and temperature profiling. Ch. 8
in Lidar: Range-Resolved optical Remote Sensing of the Atmosphere, ed. C. Weitkamp.
Springer Science CBusiness Media, Inc., New York, pp. 214–239.
Bougeault, P. and Coauthors, 2001: The MAP Special Observing Period. Bull. Amer. Meteor. Soc.,
82, 433–462.
Brewer, W.A., V. Wulfmeyer, R.M. Hardesty, and B.J. Rye, 1998: Combined wind and water-vapor
measurements using the NOAA mini-MOPA Doppler lidar, Proceedings, 19th International
Laser Radar Conference, Annapolis, MD, July 6–10.
Bringi, V.N., and B. Chandrasekar, 2001: Polarimetric Doppler weather radar. Principles and
applications. Cambridge Univ. Press, xxivC636 pp pp.
Browell, E. V., S. Ismail, and S. T. Shipley, 1985: Ultraviolet DIAL measurements of O3 profiles
in regions of spatially inhomogeneous aerosols, Appl. Opt.,24, 2827–2836.
Browning, K.A. and Wexler, R., 1968: The determination of kinematic properties of a wind field
using Doppler radar. J. Appl. Meteor. 7, 105–113.
Browning, K.A., F.F. Hill, and C.W. Pardoe, 1974: Structure and mechanism of precipitation and
effect of orography in a wintertime warm sector. Q. J. Royal Meteor. Soc, 100, 309–330.
Bruintjes, R.T., T.L. Clark, and W.D. Hall, 1994: Interactions between topographic airflow and
Sci., 51, 48–67.
Calhoun, R. R, Heap, M. Princevac, J. Sommer, H. Fernando, and D. Ligon, 2006: Virtual towers
using coherent Doppler lidar during the Joint Urban 2003 Dispersion Experiment. J. Appl.
Chadwick, R.B., and E.E. Gossard, 1986: Radar probing and measurement of the planetary bound-
ary layer: Part I. Scattering from refractive index irregularities. In Probing the Atmospheric
Boundary Layer. D.H. Lenschow, ed. Amer. Meteor. Soc., pp. 163–182.
Clark, T.L., W.D. Hall, and R.M. Banta, 1994: Two- and three-dimensional simulations of the
9 January 1989 windstorm: Comparison with observations. J. Atmos. Sci., 51, 2317–2343.
, , et al., 2000: Origins of aircraft-damaging clear-air turbulence during the 9 December

1992 Colorado downslope windstorm: Numerical simulations and comparison with observa-
tions. J. Atmos. Sci., 57, 1105–1131.
Cohn, S.A., W.O.J. Brown, C.L. Martin, M.E. Susedik, G. Maclean, and D.B. Parsons, 2001:
Clear air boundary layer spaced antenna wind measurement with the Multiple Antenna Profiler
(MAPR), Annales Geophysicae, 19(8), 845–854, 2001.
Colle B. A., Lin Y., Medina S., Smull B., 2008: Orographic modification of convection and flow
kinematics by the Oregon coastal range and Cascades during IMPROVE-2. Mon. Wea. Rev.,
136, 3894–3916.
Cotton, W.R., R.L. George, and K.R. Knupp, 1982: An intense, quasi-steady thunderstorm over
mountainous terrain. Part I: Evolution of the storm-initiating mesoscale circulation. J. Atmos.
Sci., 39, 328–342.
Coulter, R.L., 1979: A comparison of three methods for measuring mixing-layer height. J. Appl.
Meteor., 18, 1495–1499.
, G. Klazura, B.M. Lescht, T.J. Martin, J.D. Shannon, D.L. Sisterton, and M.L. Wesley, 1999:
The Argonne boundary layer experiments facility: Using minisodars to complement a wind
profiler network. Meteor. Atmos. Phys., 71, 53–59.
, M.S Pekour, and T.J. Martin, 2004: Elevated stratified layers observed with SODAR during
VTMX. Meteor. Atmos. Phys., 85, 115–123.
, and M. Kallistratova, 2004: Two decades of progress in SODAR techniques: A review of 11
ISARS proceedings. Meteor. Atmos. Phys., 85, 3–19.
Courtney, M., R. Wagner, and P. Lindelöw, 2008: Testing and comparison of lidars for profile and
turbulence measurements in wind energy. 14th Intl. Symp. for the Advancement of Boundary
Layer Remote Sensing, IOP Conf. Series: Earth and Environmental Science, 14 pp.
Crescenti, G.H., 1997: A look back on two decades of Doppler sodar comparison studies. Bull.
Damiani, R., and Coauthors, 2008: Cumulus, Photogrammetric, In-situ, and Doppler Observations:
The CuPIDO 2006 experiment. Bull. Amer. Meteor. Soc., 89, 57–73.
Darby, L.S., W.D. Neff, and R.M. Banta, 1999: Multiscale analysis of a meso-“ frontal passage in
the complex terrain of the Colorado Front Range. Mon Wea. Rev., 127, 2062–2081.
, A. Gohm, L.B. Nance, S. Gabersek, R.M. Banta, and S. Sandberg, 2000: Foehn flow in the
Austrian Alps interrupted by a cold front passage, Part I. Ninth Conf. on Mountain Meteor.,
7–11 August 2000, Aspen, Colorado, 79–82.
, R.M. Banta, and R.A. Pielke, Sr.: 2002a: Comparisons between mesoscale model terrain
sensitivity and Doppler lidar measurements of the sea breeze at Monterey Bay. Mon. Wea. Rev.,
130, 2813–2838.
, R. M. Banta, W.A. Brewer, W.D. Neff, R.D. Marchbanks, B.J. McCarty, C. J. Senff,
A.B. White, W.M. Angevine, E.J. Williams, 2002b: Vertical variations in O3 concentrations
before and after a gust front passage. J. Geophys. Res., 107 (D13), 4321.
, K.J. Allwine, and R.M. Banta, 2006: Nocturnal low-level jet in a mountain basin complex.
Part II: Transport and diffusion of tracer under stable conditions. J. Appl. Meteor., 45, 740–753.
, and G.S. Poulos, 2006: The evolution of lee wave/rotor activity in the lee of Pike’s Peak
under the influence of a cold frontal passage: Implications for aircraft safety. Mon. Wea. Rev.,
134, 2857–2876.
De Wekker, S. F. J. and S. D. Mayor, 2009: Observations of atmospheric structure and dynamics in
Meteor. Clim., 48, 1483–1499.
Doran, J.C., J.D. Fast, and J. Horel, 2002: The VTMX 2000 campaign. Bull. Amer. Meteor. Soc.,
83, 537–551.
Doviak, R.J., and D.S. Drnić, 1993: Doppler radar and weather observations, 2nd ed. Academic
Press, San Diego, 526 pp.
Drobinski, P., C. Heaberli, E. Richard, M. Lothon, A.M. Dabas, P.H. Flamant, M. Furger, and
R. Steinacker, 2003: Scale interaction processes during the MAP IOP-12 south Foehn event in
the Rhine Valley. Quart. J. Royal Meteor. Soc., 129, 729–753.
, S. Bastin, V. Guenard, J.-L.Caccia, A.M. Dabas, P. Delville, A. Protat, O. Reitebuch, and

C. Werner, 2005: Summer mistral at the exit of the Rhone Valley. Quart. J. Royal Meteor. Soc.,
131, 353–375.
Droegemeier, K.K. and R.B. Wilhelmson, 1987: Numerical simulation of thunderstorm outflow
dynamics. Part I, Outflow sensitivity experiments and turbulence dynamics. J. Atmos. Sci., 44,
1180–1210.
Durran, D.R., T. Maric, R.M. Banta, L.S. Darby, and R.M. Hardesty, 2003: A comparison of ground
based Doppler lidar and airborne in situ wind observations above complex terrain. Quart. J.
Roy. Meteor Soc., 129, 693–713.
Eberhard, W.L., G.T. McNice, and S.W. Troxel, 1987: Lidar sensing of plume dispersion: analysis
methods and product quality for light-scattering tracer particles. J. Atmos. Ocean. Technol., 4,
674–689.
Eichinger, W.E., et al. 1999: The development of a scanning Raman water vapor lidar for boundary
layer and tropospheric observations. J. Atmos. Ocean. Technol., 16, 1753–1766.
Eloranta, E.E., 2005: High spectral resolution lidar. Ch. 5 in Lidar: Range-Resolved optical Remote
Sensing of the Atmosphere, ed. C. Weitkamp. Springer Science C Business Media, Inc., New
York, pp. 143–163.
Emeis, S., M. Harris, and R.M. Banta, 2007: Boundary-layer anemometry by optical remote
sensing for wind energy applications. Meteor. Zeitschrift, 16, 337–347.
, K. Schäfer, and C. Münkel, 2008: Surface-based remote sensing of the mixing-layer height—
A review. Meteor. Zeitschrift, 17, 621–630.
Engelbart, D.A.M., M. Kallistratova, and R. Kouznetsov, 2007: Determination of the turbulent
fluxes of heat and momentum in the ABL by ground-based remote-sensing techniques (a
Review). Meteor. Z., 16, 325–335.
Fast, J.D., and L.S. Darby, 2004: An evaluation of mesoscale model predictions of down-valley and
canyon flows and their consequences using Doppler lidar measurements during VTMX 2000.
J. Appl. Meteor., 43, 420–436.
Feltz, W.F., H.B. Howell, R.O. Knuteson, H.M. Woolf, and H E. Revercomb. 2003: Near
continuous profiling of temperature, moisture, and atmospheric stability using the Atmospheric
Emitted Radiance Interferometer (AERI). J. Appl. Meteor., 42, 584–597.
Frehlich, R., and L. Cornman, 2002: Estimating spatial velocity statistics with coherent Doppler
lidar. J. Atmos. Ocean. Technol., 19, 355–366.
Frisch, A.S., 1991: On the measurement of second moments of turbulent velocity with a single
Doppler radar over non-homogeneous terrain. Bound.-Layer Meteor., 54, 29–39.
, B.W. Orr, and B.E. Martner, 1992: Doppler radar observations of the development of a
boundary-layer nocturnal jet. Mon. Wea. Rev., 120, 3–16.
Flamant, C., P. Drobinsky, L. Nance, R. Banta, L. Darby, J. Dusek, M. Hardesty, J. Pelon, and E.
Richard, 2002: Gap flow in an Alpine valley during a shallow south foehn event: Observations,
numerical simulations and hydraulic analog. Quart. J. Roy. Meteor Soc., 128, 1173–1210.
Fujita, T., and J. McCarthy, 1990: The applications of weather radar to aviation meteorology.
Ch. 31a in Radar in meteorology: Battan Memorial and 40th Anniversary Radar Meteorology
Conference, D. Atlas, ed. Amer. Meteor. Soc., Boston, pp. 657–681.
Gal-Chen, T., M. Xu, and W.L. Eberhard, 1982: Estimations of atmospheric planetary boundary
layer fluxes and other turbulence parameters from Doppler lidar data. J. Geophys. Res., 97,
18,409–18,423.
Germann, U. and J. Joss, 2004: Operational measurement of precipitation in mountainous terrain.
In P.Meischner, editor, Weather Radar: Principles and Advanced Applications,vol XVII of
Physics of Earth and Space Environment, chap. 2, 52–77, Springer Verlag.
Gohm, A., G. Zängl, and G.J. Mayr, 2004: South foehn in the Wipp Valley on 24 October 1999
(MAP IOP 10): Verification of high-resolution numerical simulations with observations. Mon.
Wea. Rev., 132, 78–102.
, and G.J. Mayr, 2005: Numerical and observational case-study of a deep Adriatic bora. Quart.
J. Roy. Meteor Soc., 131, 1363–1392.
, , L.S. Darby, and R.M. Banta, 2010: Evolution and structure of a cold front in an Alpine
valley as revealed by a Doppler lidar. Quart. J. Royal Meteor. Soc., Part B, 136, 962–977.
Goldsmith, J.E.M., F.H. Blair, S.E. Bisson, and D.D. Turner, 1998: Turn-key Raman lidar for
profiling atmospheric water vapor, clouds, and aerosols. Appl. Opt., 37, 4979–4990.
Gossard, E.E., J.E. Gaynor, R.J. Zamora, and W.D. Neff, 1985: Finestructure of elevated stable
layers observed by sounder and in situ tower sensors. J. Atmos. Sci., 42, 2156–2169.
Gossard, E.E., 1990: Radar research on the atmospheric boundary layer. Ch. 27a in Radar in
meteorology: Battan Memorial and 40th Anniversary Radar Meteorology Conference, D. Atlas,
ed., Amer. Meteor. Soc., Boston, pp. 477–527.
Gossard, E.E., R.G. Strauch, and R.R. Rogers, 1990: Evolution of dropsize distributions in liquid
precipitation observed by ground-based Doppler radar. J. Atmos. Ocean. Technol., 7, 815–828.
Grubišić, V. and Coauthors, 2008: The Terrain-Induced Rotor Experiment: A field campaign
overview including observational highlights. Bull. Amer. Meteor. Soc., 89, 1513–1533.
Grund, C.J., R.M. Banta, J.L. George, J.N. Howell, M.J. Post, R.A. Richter, and A.M. Weickmann,
2001: High-resolution Doppler lidar for boundary layer and cloud research. J. Atmos. Ocean.
Technol., 18, 376–393.
Grzeschik, M., et al. 2008: Four-dimensional variational data analysis of water vapor Raman lidar
data and their impact on mesoscale forecasts. J. Atmos. Ocean. Technol., 25, 1437–1453.
Hägeli, P., D.G. Steyn, and K.B. Strawbridge, 2000: Spatial and temporal variability of mixed-layer
depth and entrainment zone thickness. Bound.-Layer Meteor., 97, 47–71.
Hall, F.F., R.M. Huffaker, R.M. Hardesty, M.E. Jackson, T.R. Lawrence, M.J. Post, R.A. Richter,
and B.F. Weber, 1984: Wind measurement accuracy of the NOAA pulsed infrared Doppler
lidar. Appl. Optics, 23, 2503–2506.
Hardesty, R.M., and L.S. Darby, 2005: Ground-based and airborne lidar. Encyclopedia of Hydro-
logic Sciences, Malcolm G. Anderson, Ed., Wiley, pp. 697–712.
Hair, J.W., C.A. Hostetler, A.L. Cook, D.B. Harper, R.A. Ferarre, T.L. Mack, W. Welch,
L.M. Isquierdo, and F.E. Hovis 2008: Airborne High Spectral Resolution Lidar for profiling
aerosol optical properties. Appl. Opt., 47, 6734–6752.
Hayden, K.L., et al. 1997: The vertical chemical and meteorological structure of the boundary layer
in the Lower Fraser Valley during Pacific ’93. Atmos. Environ., 31, 2089–2105.
Hill, F.F., K.A. Browning, and M.J. Bader, 1981: Radar and rain-gauge observations of orographic
rain over South-Wales. Q. J. Roy. Meteor. Soc, 107, 643–670.
Hill, M., R. Calhoun, H.J.S. Fernando, A. Wieser, A. Dörnbrack, M. Weissmann, G. Mayr, and
R. Newsom, 2010: Coplanar Doppler lidar retrieval of rotors from T-REX.
Hogg, D.C., F.O. Guiraud, J.B. Snider, M.T. Decker, and E.R. Westwater, 1983: A steerable dual-
channel microwave radiometer for measurement of water vapor and liquid in the troposphere.
J. Climate and Appl. Meteor., 22, 789–806.
Hoff, R.M., M. Harwood, A. Sheppard, F. Froude, J.B. Martin, and W. Strapp, 1997: Use of
airborne lidar to determine aerosol sources and movement in the Lower Fraser Valley (LFV),
BC. Atmos. Environ., 31, 2123–216234.
Intrieri, J.M., C.G. Little, W.J. Shaw, R.M. Banta, P.A. Durkee, and R.M. Hardesty, 1990: The
Land/Sea Breeze Experiment (LASBEX). Bull. Amer. Meteor. Soc., 71, 656–664.
Intrieri, J.M., M.D. Shupe, T. Uttal, and B.J. McCarty, 2002: An annual cycle of Arctic cloud
characteristics observed by radar and lidar at SHEBA. J. Geophys. Res. 107.
Kaimal, J.C., 1986: Flux and profile measurements from towers in the boundary layer. In Probing
the Atmospheric Boundary Layer. D.H. Lenschow, ed. Amer. Meteor. Soc., pp. 19–28.
Kallistratova, M.A., and R.L. Coulter, 2004: Application of SODARs in the study and monitoring
of the environment. Meteor. Atmos. Phys., 85, 21–37.
Kallistratova, M.A., 2008: Backscattering and reflection of acoustic waves in the stable atmo-
spheric boundary layer. IOP Conf. Ser., Earth Environ. Sci.,1,012001 (012014 pp.).
King, C.W., 1997: A climatology of thermally forced circulations in oppositely oriented airsheds
along the Continental Divide in Colorado. Ph.D. Dissertation, Univ. of Colorado; NOAA Tech.
Memo. ERL ETL-283, 152 pp.
Klimowski, B.A., et al. 1998: The 1995 Arizona Program: Toward a better understanding of winter
storm precipitation development in mountainous terrain. Bull. Amer. Meteor. Soc., 79, 799–813.
Kottmeier, C., and Coauthors, 2008: Mechanisms initiating deep convection over complex terrain
during COPS. Meteor. Zeitschrift, 17, 931–948.
Kouznetsov, R.D., 2008: Quantitative estimate of the role of temperature gradients in sodar signal
formation. IOP Conf. Ser., Earth Environ. Sci.,1,012039.
Kropfli, R.A., 1990: The atmospheric boundary layer: Panel report. Ch. 27b in Radar in
meteorology: Battan Memorial and 40th Anniversary Radar Meteorology Conference, D. Atlas,
ed. Amer. Meteor. Soc., Boston, pp. 528–533.
Knupp, K.R., and W.R. Cotton, 1982a: An intense, quasi-steady thunderstorm over mountainous
terrain. Part II: Doppler radar observations of the storm morphological structure. J. Atmos. Sci.,
39, 343–358.
, and , 1982b: An intense, quasi-steady thunderstorm over mountainous terrain. Part III:
Doppler radar observations of the turbulence structure. J. Atmos. Sci., 39, 343–358.
, and , 1987: Internal structure of a small mesoscale system. Mon. Wea. Rev., 123,
2029–2050.
Kropfli, R.A., 1986a: Single Doppler radar measurements of turbulence profiles in the convective
boundary layer. J. Atmos. Ocean. Technol., 3, 305–314.
, 1986b: Radar probing and measurement of the planetary boundary layer: Part II. Scattering
from refractive particulates. In Probing the Atmospheric Boundary Layer. D.H. Lenschow, ed.
Amer. Meteor. Soc., pp. 183–199.
Lenschow, D.H., 1986: Probing the atmospheric boundary layer, D.H. Lenschow, ed. Amer.
Meteor. Soc., Boston, 269 pp.
Langford, A. O., C. J. Senff, R. J. Alvarez, II, R. M. Banta, and R. M. Hardesty, 2010: Long-range
transport of ozone from the Los Angeles Basin: A case study. Geophys. Res. Lett., 37, L06807.
Lenschow, D.H., J. Mann, and L. Kristensen, 1994: How long is long enough when measuring
fluxes and other turbulence statistics? J. Atmos. Ocean. Technol., 11, 661–673.
, V.O. Wulfmeyer, and C.J. Senff, 2000: Measuring second-through-fourth-order moments in
noisy data. J. Atmos. Ocean. Technol., 17, 1330–1347.
Levinson, D.H., and R.M. Banta, 1995: Observations of a terrain-forced mesoscale vortex and
canyon drainage flows along the Front Range of the Colorado Rockies. Mon. Wea. Rev., 123,
2029–2050.
Lhermitte, R., and D. Atlas, 1961: Precipitation motion by pulse Doppler radar. Preprints, 9th
Radar Meteorology Conf., Kansas City, Amer. Meteor. Soc., 218–223.
, and , 1963: Doppler fall speed and particle growth in stratiform precipitation. Preprints,
10th Radar Meteorology Conf., Washington DC, Amer. Meteor. Soc., 297–302.
Lothon, M., D.H. Lenschow, and S.D. Mayor, 2006: Coherence and scale of vertical velocity in the
convective boundary layer from a Doppler lidar. Bound.-Layer Meteor., 121, 521–536.
Mahrt, L., and D. Vickers, 2006: Extremely weak mixing in stable conditions. Bound.-Layer
Meteor., 119, 19–39.
Martner, B.E., and Coauthors, 1993: An evaluation of wind profiler, RASS, and microwave
radiometer performance. Bull. Amer. Meteor. Soc., 74, 599–613.
, R.F. Reinking, and R.M. Banta, 2002: Radar observations of downslope flow at Mount
Washington. Preprints, 10th Conf. on Mountain Meteorology, Park City, UT, AMS, 412–415.
, S.E. Yuter, A.B. White, S.Y. Matrosov, D.E. Kingsmill, and F.M. Ralph, 2008: Raindrop size
distributions and rain characteristics in California coastal rainfall for periods with and without
a radar bright band. J. Hydromet. 9, 408–425.
Mayor, S.D., and E.W. Eloranta, 2001: Two-dimensional vector wind fields from Volume Imaging
Lidar data. J. Appl. Meteor., 40, 1331–1345.
Mayr, G.J., L. Armi, S. Arnold, R.M. Banta, L.S. Darby, D. Durran, C. Flamant, S. Gabersek, A.
Gohm, R. Mayr, S. Mobbs, L.B. Nance, I. Vergeiner, J. Vergeiner, and C.D. Whiteman, 2004:
Gap flow measurements during the Mesoscale Alpine Programme. Meteor. Atmos. Phys., 86,
99–119.
McKendry, I.G., and Coauthors 1997: Elevated ozone layers and vertical down-mixing over the
Lower Fraser Valley, BC. Atmos. Environ.,31, 2135–2146.
, D.G. Steyn, R. M. Banta, W. Strapp, K. Anlauf, and J. Pottier, 1998: Daytime photochemical
pollutant transport over a tributary valley lake in southwestern British Columbia. J. Appl.
Meteor., 37, 393–404.
Measures, R.M., 1984: Laser Remote Sensing—Fundamentals and Applications. Wiley-
Interscience, New York.
Medina, S., and R.A. Houze, 2003: Air motions and precipitation growth in Alpine storms. Quart.
J. Roy. Meteor. Soc.129, 345–371.
Melfi, S.H., D.N. Whiteman, and R.A. Ferrare, 1989: Observation of atmospheric fronts using
Raman lidar moisture measurements. J. Appl. Meteor., 28, 789–806.
Mellqvist, J., J. Samuelsson, J. Johansson, C. Rivera, B. Lefer, S. Alvarez, and J. Jolly, 2010:
Measurements of industrial emissions of alkenes in Texas using the Solar Occultation Flux
method, J. Geophys. Res., 115, D00F17.
Merritt, D.A., 1995: A statistical averaging method for wind profiler Doppler spectra. J. Atmos.
Ocean. Technol., 12, 985–995.
Michelson, S.A., and J.-W. Bao, 2008: Sensitivity of low-level winds simulated by the WRF model
in California’s Central Valley to uncertainties in the large-scale forcing and soil initialization.
J. Appl. Meteor. Climatol., 47, 3131–3149.
National Research Council, 2009: Observing Weather and Climate from the Ground Up: A
Nationwide Network of Networks. National Academies Press, 234 pp.
Neff, W.D., 1975: Quantitative evaluation of acoustic echoes from the planetary boundary layer.
NOAA Tech. Report ERL 322-WPL, Boulder, 38 pp.
, 1986: On the use of sodars to study stably stratified flow influenced by terrain. Atmos. Res.,
20, 279–308.
, 1990: Remote sensing of atmospheric processes over complex terrain. Ch. 8 in Atmospheric
Processes over Complex Terrain, ed. William Blumen, Meteorological Monogr.,23 (No. 45),
173–228.
, 1994: Mesoscale air quality studies with meteorological remote sensing systems. Int. J.
Remote Sensing, 15, 393–426.
, and R.L. Coulter, 1986: Acoustic remote sensing. In Probing the Atmospheric Boundary
Layer. D.H. Lenschow, ed. Amer. Meteor. Soc., pp. 201–235.
, and C.W. King, 1988: Observations of complex terrain flows using acoustic sounders:
Drainage flow structure and evolution. Bound.-Layer Meteor., 43, 15–41.
Neiman, P.J., R.M. Hardesty, M.A. Shapiro, and R.E. Cupp, 1988: Doppler lidar observations of a
downslope windstorm. Mon. Wea. Rev., 116, 2265–2275.
, F.M Ralph, A.B. White, J.D. Lundquist, and M.D. Dettinger, 2008: Meteorological charac-
teristics and overland precipitation impacts of atmospheric rivers affecting the west coast of
North America based on eight years of SSM/I satellite observations. J. Hydrometeor., 9, 22–47.
Pichugina, Y.L., R.M. Banta, B. Jonkman, N.D. Kelley, R.K. Newsom, S.C. Tucker, and W.A.
Brewer, 2008: Horizontal-velocity and variance measurements in the stable boundary layer
using Doppler-lidar: Sensitivity to averaging procedures. J. Atmos. Ocean. Technol., 25,
1307–1327.
Pearson, G., F. Davies, and C. Collier, 2009: An analysis of the performance of the UFAM pulsed
Doppler lidar for observing the boundary layer. J. Atmos. Ocean. Technol., 26, 240–250.
Pirronen, A.K., and E.W. Eloranta, 1995: Accuracy analysis of wind profiles calculated from
volume imaging lidar data. J. Geophys. Res., 100, 25,559–25,567.
Post, M.J., and R.E. Cupp, 1990: Optimizing a pulsed Doppler lidar. Appl. Optics, 29, 4145–4158.
Post, M.J., 1996: A graphical technique for retrieving size distribution parameters from multiple
measurements: Visualization and error analysis. J. Atmos. Ocean. Technol., 13, 863–873.
, and W.D. Neff, 1986: Doppler lidar measurements of winds in a narrow mountain valley.
, C.J. Grund, A.M. Weickmann, K.R. Healey, and R.J. Willis, 1996: Comparison of Mount
Pinatubo and El Chichon volcanic events: Lidar observations at 10.6 and 0.69 m. J. Geophys.
Res., 101, 3929–3940.
, C.J. Grund, D. Wang, and T. Deschler, 1997: Evolution of Mount Pinatubo’s aerosol size
distributions over the continental United States: Two-wavelength lidar retrievals and in situ
measurements. J. Geophys. Res., 102, 13,535–13,542.
Poulos, G.S., J.E. Bossert, T.B. McKee, and R.A. Pielke, 2000: The interaction of katabatic flow
and mountain waves. Part I: Observations and idealized simulations. J. Atmos. Sci., 57, 1919-
1936.
Poulos, G.S., and S.P. Burns, 2003: An evaluation of bulk Ri-based surface layer flux formulas for
stable and very stable conditions with intermittent turbulence. J. Atmos. Sci., 60, 2523-2537.
——, 2007: The interaction of katabatic flow and mountain waves. Part II: Case study analysis and
conceptual model. J. Atmos. Sci., 64, 1857-1879.
Poulos, G.S., and J.E. Bossert, 1995: An observational and prognostic numerical investigation of
complex terrain dispersion. J. Appl. Meteorol., 34, 650–669.
Ralph, F.M., and Coauthors, 2003: The impact of a prominent rain shadow on flooding in
California’s Santa Cruz Mountains: A CALJET case study and sensitivity to the ENSO cycle.
J. Hydrometeorol., 4, 1243–1264.
Reinking, R.F., J.B. Snider, and J.L. Coen, 2000: Influences of storm-embedded orographic gravity
, A.S. Frisch, B.W. Orr, D.L. Korn, L.R. Bissonnette, and G. Roy, 2003: Remote sensing
observations of effects of mountain blocking on travelling gravity-shear waves and associated
clouds. Bound.-Layer Meteor., 109, 255–284.
Reitebuch, O., and Coauthors, 2003: Determination of airflow across the Alpine ridge by a
combination of airborne Doppler lidar, routine radiosounding and numerical simulation. Q.
J. Royal Meteor. Soc., 129, 715–727.
Rinehart, R.E., 2010: Radar for Meteorologists. 5th Ed., Rinehart Publications, Nevada, MO, USA,
482 pp.
Rothermel, J. and Coauthors, 1998: The Multi-Center Airborne Coherent Atmospheric Wind
Sensor. Bull. Amer. Meteor. Soc., 79, 581–599.
Rotunno, R. and R.A. Houze, 2007: Lessons on orographic precipitation from the Mesoscale
Alpine Programme. Quart. J. Roy. Meteor. Soc.133, 811–830.
Rucker, M., R.M. Banta, and D.G. Steyn, 2008: Along-valley structure of daytime thermally driven
flows in the Wipp Valley. J. Appl. Meteor., 47, 733–751.
Sassen, K., 2005: Polarization in lidar. Ch. 2 in Lidar: Range-Resolved optical Remote Sensing
of the Atmosphere, ed., C. Weitkamp. Springer Science CBusiness Media, Inc., New York,
pp. 19–42.
Schwiesow, R.L., R.E. Cupp, P.C. Sinclair, and R.F. Abbey 1981: Waterspout velocity measure-
ments by airborne lidar. J. Appl. Meteor., 20, 341–348.
Schwiesow, R.L., 1986: A comparative overview of active remote sensing techniques. In Probing
the Atmospheric Boundary Layer. D.H. Lenschow, ed. Amer. Meteor. Soc., pp. 5–18.
Schwiesow, R.L., 1986a: A comparative overview of active remote-sensing techniques. In: Probing
the Atmospheric Boundary Layer. D.H. Lenschow, ed., Amer. Meteor. Soc., pp. 129–138.
Schwiesow, R.L., 1986b: Lidar measurement of boundary-layer variables. In: Probing the Atmo-
spheric Boundary Layer. D.H. Lenschow, ed., Amer. Meteor. Soc., pp. 139–162.
Schotland, R. M., 1974: Errors in the lidar measurement of atmospheric gases by differential
absorption. J. Appl. Meteorol., 13, 71–77.
Seibert, P., F. Beyrich, S.-E. Gryning, S. Joffe, A. Rasmussen, and P. Tericer, 2000: Review and
intercomparison of operational methods for the determination of the mixing height. Atmos.
Environ., s34, 1001–1027.
Senff, C.J., J. Bösenberg, and G. Peters, 1994: Measurement of water-vapor flux profiles in the
convective boundary layer with lidar and radar-RASS. J. Atmos. Ocean. Technol., 11, 85–93.
, , , and T. Schaberl, 1996: Remote sensing of turbulent ozone fluxes and the ozone
budget in the convective boundary layer with DIAL and Radar-RASS: A case study. Beitr. Phys.
Atmos., 69, 161–176.
, and Coauthors, 1998: Airborne lidar characterization of power plant plumes during the 1995
Southern Oxidants Study. J. Geophys. Res., 103, 31,173–31,189.
Shaw, W.J., L.S. Darby, and R.M. Banta 2003: A comparison of winds measured by a 915 MHz
wind profiling radar and a Doppler lidar. Preprints, 12th Symp. on Meteor. Obs. and Inst., Long
Beach CA, Amer. Meteor. Soc., P1.8.
Shun, C.M., and S.Y. Lau, 2002: Implementation of a Doppler Light Detection and Ranging
(LIDAR) system for the Hong Kong International Airport. Preprints, 10th Conf on Aviation,
Range, and Aerospace Meteorology, Portland OR, Amer. Meteor. Soc., 8.3.
, C.M. Cheng, and O.S.M. Lee, 2003: LIDAR observations of terrain-induced flow and its
application in airport wind-shear monitoring. Proc. Int. Conf on Alpine Meteor. (ICAM) and
Mesoscale Alpine Programme (MAP), Brig, Switzerland, 6.6.
, S.Y. Lau, C.M. Cheng, and H.Y. Chiu, 2004: LIDAR observations of windshear induced by
mountain lee waves. Preprints, 11th Conf on Mountain Meteorology and MAP Meeting 2004,
Mount Washington Valley, NH, Amer. Meteor. Soc., 11.2.
, and P.W. Chan, 2008: Applications of an infrared Doppler lidar in detection of wind shear.
J. Atmos. Ocean. Technol., 25, 637–655.
Skolnik, M.I., 1980: Introduction to Radar Systems. New York, McGraw-Hill Book Co., 581 pp.
Smith, T.L., S.G. Benjamin, S.I. Gutman, and S.R. Sahm, 2007: Forecast impact from assimilation
of GPS-IPW observations into the Rapid Update Cycle. Mon. Wea. Rev., 135, 2914–2930.
Spinhirne, J.D., J.A.R. Rall, and V.S. Scott, 1995: Compact eye-safe lidar system. Rev. Laser Eng.,
23, 112–118.
Steyn, D.G., M. Baldi, and R.M. Hoff, 1999: The detection of mixed layer depth and entrainment
zone thickness from lidar backscatter profiles. J. Atmos. Ocean. Technol., 16, 953–959.
Stutz, J., B. Alicke, R. Ackermann, A. Geyer, A. White, and E. Williams, 2004: Vertical profiles of
NO3, N2O5, O3, and NOx in the nocturnal boundary layer: 1. Observations during the Texas
Air Quality Study 2000. J. Geophys. Res., 109.
Tatarskii, V.I., 1971: Effects of the Turbulent Atmosphere on Wave Propagation. Keter Press,
Jerusalem.
Tollerud, E.I., and Coauthors, 2008: Mesoscale moisture transport by the low-level jet during the
IHOP field experiment. Mon. Wea. Rev., 136, 3781–3795.
Tucker, S.C., and Coauthors, 2009: Doppler lidar estimation of mixing height using turbulence,
shear, and aerosol profiles. J. Atmos. Ocean. Technol., 26, 673–688.
, R.M. Banta, A.O. Langford, C.J. Senff, W.A. Brewer, E.J. Williams, B.M. Lerner, H.D.
Osthoff, and R.M. Hardesty, 2010: Relationships of coastal nocturnal boundary layer winds
and turbulence to Houston ozone concentrations during TexAQS 2006, J. Geophys. Res., 115,
D10304.
Turner, D., R. Ferrare, L.H. Brassuer, W. Feltz, and T. Tooman, 2002: Automated retrievals of
water vapor and aerosol profiles from an operational Raman lidar. J. Atmos. Ocean. Technol.,
19, 37–50.
Vickers, D., and L. Mahrt, 2003: The cospectral gap and turbulent flux calculations. J. Atmos.
Ocean. Technol., 20, 660–672.
——, 2006: A solution for flux contamination by mesoscale motions with very weak turbulence.
Vivekanandan, D.S., Zrnic, S.M. Ellis, R. Oye, A.V. Ryzhkov, and J. Straka, 1999: Cloud
microphysics retrieval using S-band dual polarization radar measurements. Bull. Amer. Meteor.
Soc., 80, 381–388.
Volkamer, R.M., S. Coburn, B. Dix, R. Sinreich, 2009: MAX-DOAS observations from ground,
ship, and research aircraft: maximizing signal-to-noise to measure “weak” absorbers. Proc. of
SPIE Vol. 7462 “Ultraviolet and Visible Ground- and Space-based Measurements, Trace Gases,
Aerosols and Effects VI” (Eds. J.R. Herman, W. Gao) 746203.
Wandinger, U., 2005: Raman lidar. Ch. 9 in Lidar: Range-Resolved optical Remote Sensing of
the Atmosphere, ed. C. Weitkamp. Springer Science C Business Media, Inc., New York, pp.
241–272.
Wakimoto, R.M. and J.L. McElroy, 1986: Lidar observations of elevated pollution layers over Los
Angeles. J. Appl. Meteor., 25, 1583–1599.
Weissmann, M.D., G.J. Mayr, R.M. Banta, and A. Gohm, 2004: Observations of the temporal
evolution and spatial structure of the gap flow in the Wipp Valley on 2 and 3 October 1999.
Mon. Wea. Rev., 132, 2684–2697.
Weitkamp, C., 2005: Lidar: Range-Resolved optical Remote Sensing of the Atmosphere. Springer
Science C Business Media, Inc., New York, 455 pp.
Werner, C., 2005: Doppler wind lidar. Ch.12 in Lidar: Range-Resolved optical Remote Sensing
of the Atmosphere, ed. C. Weitkamp. Springer Science CBusiness Media, Inc., New York, pp.
325–354.
Westwater, E.R., J.B. Snider, and M.J. Falls, 1990: Ground-based radiometric observations
of atmospheric emission and attenuation at 20.6, 31.65, and 90.0 GHz: a comparison of
measurements and theory. Antennas and Propagation, IEEE Transactions, 38, 1569–1580.
Westwater, E., 1997: Remote sensing of tropospheric temperature and moisture by integrated
observing systems. Bull. Amer. Meteor. Soc., 78, 1991–2006.
Wetzel, P.J., W.R. Cotton, and R.L. McAnelley, 1983: A long-lived mesoscale convective complex.
Part II: Evolution and structure of the mature complex. Mon. Wea. Rev., 111, 1919–1937.
Wetzel, M., M. Meyers, R. Borys, R. McAnelly, W. Cotton, A. Rossi, P. Frisbie, D.Nadler,
D. Lowenthal, S. Cohn, and W. Brown, 2004: Mesoscale snowfall prediction and verification
in mountainous terrain. Wea. Forecasting, 19, 806–828.
White, A.B., C.J. Senff, and R.M. Banta, 1999: A comparison of mixing depths observed by
ground-based wind profilers and an airborne lidar. J. Atmos. Ocean. Technol., 16, 584–590.
White, A.B., P.J. Neiman, F.M. Ralph, D.E. Kingsmill, and P.O.G. Persson, 2003: Coastal
orographic rainfall processes observed by radar during the California Land-Falling Jets
Experiment. J. Hydrometeor., 4, 264–282.
Whiteman, C.D., 1990: Observations of thermally developed wind systems in mountainous terrain.
Ch. 2 in Atmospheric Processes over Complex Terrain, ed. William Blumen, Meteorological
Monogr.,23 (No. 45), 5–42.
Whiteman, D.N., 2003: Examination of the traditional Raman lidar technique. II: Evaluating the
ratios for water vapor and aerosols. Appl. Opt., 42, 2593–2608.
Wilczak, J.M., and Coauthors, 1995: Contamination of wind profiler data by migrating birds:
Characteristics of corrupted data and potential solutions. J. Atmos. Ocean. Technol., 12,
449–467.
, E.E. Gossard, W.D. Neff, and W.L. Eberhard, 1996: Ground-based remote sensing of the
atmospheric boundary layer: 25 years of progress. Bound.-Layer Meteor., 78, 321–349.
Wilson, D.A., and L.J. Miller, 1972: Atmospheric motion by Doppler radar. Ch. 13 in Remote
Sensing of the Troposphere, Ed. V.E. Derr, U.S. Dept. of Commerce Report.
, 1970: Doppler radar studies of boundary layer wind profiles and turbulence in snow
conditions. Proc. 14th Conf. on Radar Meteor., Tucson, Amer. Meteor. Soc., 191–196.
Wolfe, D., C. Fairall, M. Ratterree, W.A. Brewer, J. Intrieri, C. Senff, B. McCarty, S. Tucker, D.
Law, A. White, and D. White, 2007: Shipboard multi-sensor wind profiles from NEAQS 2004:
Radar wind profiler, high-resolution Doppler lidar, GPS rawinsonde. J. Geophys. Res., Vol. 112.
Wolfe, D.E. and S.I. Gutman, 2000: Development of the NOAA/ERL ground-based GPS water
vapor demonstration network: Design and initial results. J. Atmos. Ocean. Technol., 17,
426–440.
Wulfmeyer, V., et al., 2008: The convective and orographically induced precipitation study. Bull.
Amer. Meteor. Soc., 89, 1477–1486.
Wulfmeyer, V., and Coauthors, 2008: The convective and orographically induced precipitation
study: A research and development project of the World Weather Research Program for
improving quantitative precipitation forecasting in low-mountain regions. Bull. Amer. Meteor.
Soc, 89, 1477–1486.
Wurman, J., J. Straka, E. Rasmussen, M. Randall, and A. Zahrai, 1997: Design and deployment of a
portable, pencil-beam, pulsed, 3-cm Doppler radar. J. Atmos. Ocean. Technol., 14, 1502–1512.
Wyngaard, J.C., and O.R. Cote, 1971: Budgets of turbulent kinetic energy and temperature variance
in atmospheric surface layer. J. Atmos. Sci., 28, 190-&.
Wyngaard, J.C., 1986: Measurement physics. In Probing the Atmospheric Boundary Layer. D.H.
Lenschow, ed. Amer. Meteor. Soc., pp. 5–18.
Zhong, S.Y., and J. Fast, 2003: An evaluation of the MM5, RAMS, and Meso-Eta models at
subkilometer resolution using VTMX field campaign data in the Salt Lake Valley. Mon. Wea.
Rev., 131, 1301–1322.
Zhong, S.Y., C.D. Whiteman, and X.D. Bian, 2004: Diurnal evolution of three-dimensional wind
and temperature structure in California’s Central Valley. J. Appl. Meteorol., 43, 1679–1699.
Chapter 9
Mesoscale Modeling over Complex Terrain:
Numerical and Predictability Perspectives
James D. Doyle, Craig C. Epifanio, Anders Persson, Patrick A. Reinecke,

and Günther Zängl
Abstract This chapter presents an overview of numerical modeling techniques and

methods that are relevant for the simulation and prediction of processes and phe-
nomena over complex terrain. The formative years of numerical weather prediction
provide an important reference perspective on the current state of the science for
prediction of terrain-forced flows using modern complex modeling systems. New
numerical methods and challenges for numerical weather prediction applications
in complex terrain include basic issues that range from the formulation of model
vertical coordinates for sloping terrain to the need to use consistent formulations
to represent the key forcing contributions in the equations of motion. An overview
of methods introduced to improve numerical simulations over steep terrain due to
the pressure gradient representation and numerical filtering is provided. Advantages
of higher-order accuracy methods for the prediction of mountain waves at high
resolution are discussed as well.
The recent development of mesoscale ensemble techniques such as the Ensemble
Kalman Filter (EnKF), data assimilation, and mesoscale adjoint models have
allowed researchers to investigate the mesoscale predictability of terrain-forced
flows. Examples of application of these tools used to explore the limits and
characteristics of mesoscale predictability and results from recent predictability
studies are discussed.
J.D. Doyle () • P.A. Reinecke

Marine Meteorology Division, Naval Research Laboratory, 7 Grace Hopper Avenue,
Monterey, CA 93943-5502, USA
e-mail: james.doyle@nrlmry.navy.mil
C.C. Epifanio
Texas A&M University, College Station, Texas, USA
A. Persson
UK MetOffice, Exeter, United Kingdom
G. Zängl
Deutscher Wetterdienst, Offenbach, Germany

532 J.D. Doyle et al.
In spite of the significant advances that have been made over the past several
decades in the various components of numerical modeling systems, a number of
outstanding challenges remain for the simulation and prediction of terrain-forced
flows, which are highlighted in this chapter. Some of these challenges include:
the fidelity of numerical methods, new methods amenable to variable resolution
mesoscale and global models, applications over steep terrain and interactions with
the land surface, interactions between numerical methods and physical process
parameterizations, and predictability of atmospheric flows and processes over
complex terrain.
9.1 Historical Perspective: Bjerknes’ Dream

of the Newtonian Weather Forecast (1903)
When weather forecasting along scientific-mathematic principles, numerical

weather prediction (NWP), was seriously discussed for the first time more than
100 years ago, one of the main concerns was the problem of representing the wind
and weather in complex terrain. As often happens in science, the triggering factor
for the scientific discussion was quite trivial. In this case it was the unusually
stormy Scandinavian summer of 1903, spurring a debate the following autumn
regarding the possibility of improving weather prediction. At the time, Sweden was
the only country in northwestern Europe without a special storm-warning system
for its coasts. This motivated Nils Ekholm, the leading Swedish meteorologist,
to propose a system of upper-air measurements, obtained by kite or balloon, to
provide observational indications of approaching storms (see Kutzbach 1979 for a
biography of Ekholm and contemporaries in Europe).
The weather forecast office was governed by the Royal Swedish Academy of
Science, which found that there were no established scientific rules or physical laws
used in weather forecasting at the time, in contrast to the Newtonian-based forecasts
of the tides, the moon and the planets.
One of Ekholm’s friends, the Norwegian Vilhelm Bjerknes (Fig. 9.1), pro-
fessor in physics at Stockholm Högskola (later renamed Stockholm University),
believed that observational information was insufficient to produce reliable weather
predictions. Bjerknes suggested that new and improved methods of forecasting
needed to be developed and applied to utilize the observations. He agreed with the
Swedish Academy that Newtonian and other physical laws, which formed the basis
for celestial predictions, should ideally also form the basis for scientific weather
forecasts.
At the Stockholm Physics Association meeting on Saturday 24 October 1903
Bjerknes presented a lecture “A rational method for weather forecasting”. The
speech was given extensive coverage in contemporary Swedish newspapers such
as “Stockholm Dagblad” and “Dagens Nyheter” (see Friedman 1989 for further
references). From newspaper accounts we can reconstruct Bjerknes’ speech (trans-
lation from Swedish by Norman Phillips):
9 Mesoscale Modeling over Complex Terrain: Numerical and Predictability... 533
Fig. 9.1 Vilhelm Bjerknes

(1862–1951) (From Shapiro
and Grønås 1999.
Society. Reprinted with
permission)
[The] weather is determined by the wind or air motion together with the density, pressure,
temperature, and humidity of the air, which corresponds mathematically to seven variables.
With the help of hydrodynamics and thermodynamics, together with knowledge of the
heat radiation from the sun and outgoing radiation from the earth, seven equations can
be formulated for determination of the seven variables.
But a pure analytic solution would exceed the capability of the times. In an early
vision of the “butterfly effect” Bjerknes commented that one could not take into
consideration every little eddy on the street corners in Stockholm. The problem
could, according to Bjerknes, be divided into two, one dynamic and one physical.
We first considered that temperature and humidity are unchanged for a short time and
determine changes in the wind, density, and pressure through graphical construction. With
help of that result and the heat that is exchanged, changes in temperature and humidity
would then be determined and continue so with the values for these. Topographic charts
must be used so that attention can be paid to the elevation of the land, and not, as is usual,
‘reduced to sea-level’.
Bjerknes stressed that the method he suggested did not differ essentially from the
method usually used for weather forecasting.
The difference is only that we must use all knowledge in a systematic way, including the
knowledge of mechanics and physics that have been gathered for us by investigators since
the time of Galileo and Newton.
Bjerknes’ lecture was, according to the newspapers, followed by a “lively

discussion” where the problem of the earth’s topography became the centerpiece
of the discussion. Two meteorologists in the audience, Professor H.E. Hamberg and
Dr. Westman suspected that friction and the effect of the “unevenness of the earth’s
surface” might be impossible to account for in the calculations. Bjerknes replied that
this would certainly be the case with an analytic treatment of the problem. But in the
graphical method he had suggested, irregularities introduced no difficulty, because
one could account for them with topographic maps and consider friction through
pure empirical investigations.
The Scandinavian Peninsula is characterized by a 1–2 km high mountain ridge

which not only shapes its own weather but also affects traveling cyclones arriving
from the North Atlantic. Nils Ekholm showed with examples how this topography
had a strong effect on the formation and movement of cyclones and pointed out that
Bjerknes’ method, if it could be practically applied, should without doubt lead to
a very significant step forward. From this, said Ekholm, followed the need to pay
more attention to unevenness in the earth’s surface than was generally done.
Over the Christmas and New Years holidays 1903–1904 Bjerknes seems to
have pondered the problems about Newtonian weather prediction, because on 7–8
and 10 January 1904 he published a long text in Norway’s leading newspaper
“Aftenposten” on weather forecasts and “the possibilities to improve upon them”
and shortly afterwards a scholarly article about weather forecasting “considered
from the standpoints of mechanics and physics” (Bjerknes 1904). The Newtonian
challenge was repeated when Bjerknes gave his inaugural lecture in Leipzig in 1913:
The problem of accurate pre-calculation that was solved for astronomy centuries ago must
now be attacked in all earnest for meteorology : : : It can take years to bore a tunnel through
a mountain. Many a worker may not live to see the day of the breakthrough. Nevertheless,
that will not prevent others from riding through the tunnel later at speed of an express train.
(Bjerknes 1914; Friedman 1989)
Perhaps it is no coincidence that the impact of Bjerknes’ vision would have its
most profound influence among Newton’s fellow countrymen. On 27 May 1910
Bjerknes, invited by the Royal Meteorological Society, delivered a lecture at Uni-
versity College in London about “rational methods” in weather forecasting. Almost
a third of the speech dealt with air motions in mountain ranges (Bjerknes 1910). An
English meteorologist of Italian descent Leon Bonacina wrote enthusiastic reviews
in “Quarterly Journal” and “Symonds Meteorological Magazine” of Bjerknes’
ideas as they appeared in foreign journals. The Head of the Met Office, Napier
Shaw, a good friend of Bjerknes, published a meteorological pastiche of Newton’s
“Principia” called “Principia Atmospherica” (Shaw 1914) and encouraged the
physicist and mathematician, L. F. Richardson, to realize Bjerknes’ Newtonian
dream (Ashford 1985; Persson 2000).
Bjerknes had made clear that “for the complete solution” of the meteorological
forecast problem it would be necessary at certain periods of investigation to extend
the observations to the higher atmospheric levels by “kites and registering balloons.”
On 20 May 1910, a week before Bjerknes’ London speech, an international under-
taking along these lines had in fact been launched over Europe. The purpose was
not weather forecasting, but to investigate the possible effects on the atmospheric
flow when the earth on that day passed through the tail of Halley’s famous comet.
No effect was found, but the large amount of upper-air observations was in a few
years time published and came to good use for Richardson’s calculations, which
he conducted while serving as an ambulance driver on the Western Front in WWI
(Richardson 1922; Ashford 1985).
Richardson’s 1922 calculations were a complete failure and at the time were
seen as proof that weather forecasting along “Newtonian” mathematical lines was
not feasible. Not only were his computed pressure tendencies wrong by an order
of a magnitude or two, a global simulated pressure pattern was forecast to move

westward, instead of eastward. Furthermore, the whole endeavor was entirely
impractical since it would demand the constant around the clock work of 64,000
mathematically trained clerks to provide real-time forecasts. There can be little
doubt that the failure of his trial forecast persuaded most meteorologists to ignore
his work (Lynch 1999, p. 72). Today we know that the excessive tendencies were
due to induced gravity waves that would have dampened out by themselves and
would not harm the forecasts beyond 24 h (Kalnay 2003, pp. 5–6). Richardson’s
large scale pressure patterns were actually planetary Rossby waves, which indeed
are supposed to move retrogressively westward (Lynch 1992). So in essence there
was really nothing meteorologically wrong with Richardson’s calculations. As often
is the case, correct mathematical equations were not properly understood because
of misconceptions and erroneous expectations.1 The computational problems were
solved after WWII, with the emergence of the computer within the world of science.
One would perhaps imagine that meteorologists would be quite eager to use these
new “mathematical machines”, but in some ways the opposite was the case. The
contemporary leaders of the meteorological community, including Jule Charney and
Carl-Gustaf Rossby, failed to imagine any tangible approaches of the computer,
even for a rather non-controversial project of short-range weather forecasting
(Lindzen et al. 1990). However, in August 1946, two non-meteorologists, Vladimir
Zworykin from the Radio Corporation of America and Johann von Neumann from
the Institute for Advanced Study at Princeton University, convened a meeting at
Princeton University to address the potential application of computers for weather
prediction. While Richardson’s failed experiment hung like dark clouds over some
participants and others regarded NWP as an inspiring, yet unattainable goal (Harper
2008), the two organizers were more optimistic about the potential of the computer.
They shared an underlying dream of using a computer to create a global atmospheric
numerical model and then using the model to explore the feasibility of weather
modification for defense purposes (Kwas 2001). After the 1946 meeting the idea
gradually emerged that atmospheric predictions may be possible with a simplified
set equations based on the “quasi-geostrophic” nature of the atmosphere; however,
even such a simplification exceeded the capability of the early computers. Only
when the atmospheric flow was stripped down to almost its simplest form by
assuming purely barotropic conditions was it possible, in 1950, to run the first
successful numerical forecasts on the then state-of-the-art ENIAC machine.
There was optimism after the ENIAC runs. Five-day-a-week operational NWP
started in Sweden already in autumn 1954 (Persson 2005a). It turned out to be
much more complicated to extend the simple one-layer barotropic model to two-
or three-level baroclinic models and at the end of the 1950s there were again
serious doubts if NWP was possible at all (Persson 2005c). The introduction of
orography or “complex terrain” in these quasi-geostrophic models had not been
high on the agenda. Norman Phillips’ early general circulation model, which was
1
An interesting discussion of erroneous interpretation of dynamical equations in relation to flow
over mountainous terrain is provided by Holton (1993).
a milestone in numerical weather prediction, for example did not include the
effects of mountains (Phillips 1956; see Lewis 1998 for the historical background).
In the early 1950s there was some controversy about the effects of the earth’s
surface on the atmospheric flow. While Charney and Eliassen (1949) and Bolin
(1950) suggested that large-scale orography, primarily the Rocky Mountains and
the Himalayas, could generate the stationary planetary waves, Smagorinsky (1953)
showed that non-uniform heating could also produce such a flow (see Smith (1979)
for a historical overview).
But it was not until the 1960s that computers and mathematical insights into finite
difference calculations had reached a level where Bjerknes’ vision could be realized.
Early primitive equation (PE) experiments by Hinkelmann (1959) and Charney
(1955) and Smagorinsky’s two-layer general circulation model (Smagorinsky 1965)
set the foundation for more sophisticated approaches. By this time L.F. Richardson’s
vision was recognized as an inspiring blueprint for modern NWP since the
numerical approach provided a foundation for modern techniques. Dover’s re-issue
of his book in 1965 became a large “hit” and when in 1966 NMC (today’s NCEP)
launched their first weather forecast, based on a six-level PE-model; Bjerknes’
vision was realized.
However, although the introduction of the PE-models was a “paradigm shift” it
was not everywhere seen as a step forward; some leading meteorologists regarded
it as just a matter of engineering, little more than a demonstration that realistic
equations were being used and applied properly (Lorenz 1960, p. 418). It would not,
they argued, teach us anything about the atmosphere, and in particular not about the
flow over or near mountains. This attitude sometimes led to a conservative attitude
towards PE-models, which hampered the development of NWP in some countries
(Persson 2005b).
Lorenz (1960) argued that it is through the omission of specific physical pro-
cesses contained within the equations, such as friction or the effects of mountains,
that we may acquire new insight concerning their relative importance. In other
words, much can be learned about the atmosphere through the use of incomplete
or approximate atmospheric models. But such a numerical laboratory must be based
on skillful simulations of atmosphere, such as possible through primitive equation
models.
The representation of topography in models, something which had not been
discussed in Richardson’s book, had been treated in a very basic way in early NWP
models. The first numerical simulations of flow over two-dimensional terrain by
Foldvik and Wurtele (1967) and Eliassen and Rekustad (1971) captured the basic
properties of mountain waves and showed the promise of numerical approaches
for the future studies of flow over complex terrain. The success of the PE-
model from the 1970s onwards naturally led to even more realistic and accurate
methods to include physical processes, in particular when nonhydrostatic equations
were introduced. More than 100 years after the idea of NWP was born and the
fundamentals of atmospheric flow over complex terrain was first discussed, the
influence of terrain on the atmosphere still remains a basic challenge for today’s
meteorologists.
9.2 Basic Numerical Methods for the Prediction

of Atmospheric Flows over Complex Terrain
Sections 9.2 and 9.3 provide an overview of basic concepts in NWP, including
discretization methods, stability limitations and basic finite-differencing methods.
9.2.1 Problem Description
The starting point for NWP is the basic laws governing atmospheric dynamics: New-
ton’s second law of motion, the first law of thermodynamics, and the conservation
of mass. To keep things simple, we adopt a Cartesian coordinate system2 and let our
primary thermodynamic variables be pressure and potential temperature. To a good
approximation, the basic laws are then
@u 1 @p
C u ru D C f v C Sx
@t @x
@v 1 @p
C u rv D f u C Sy
@t @y
@w 1 @p
C u rw D g C Sz
@t @z
@
C u r D Q
@t
@
C r .u/ D 0 (9.1)
@t
where u D (u,v,w) is the wind vector, is the potential temperature, p is the pressure,
is the density, and f is the Coriolis parameter, which is a function of the north-
south coordinate y.
In NWP, the system (9.1) is applied in a filtered or coarse-grained sense, meaning
that only disturbances larger than some pre-defined spatial scale are explicitly
considered. The effects of smaller-scale phenomena must then be estimated from
this larger-scale model state, a problem referred to as parameterization. Some
examples of parameterized processes in NWP include small-scale turbulent fluxes,
cloud microphysics and latent heating, small-scale cumulus convection (for large-
scale and climate models) and radiative transfer processes (among many others).
In (9.1), these parameterized effects are represented by Sx , Sy , Sz and Q. The
remaining non-parameterized terms are called the dynamical core of the model.
2
Most practical forecasting models are formulated in some version of spherical coordinates. For
derivation of the basic laws in spherical coordinates, see Gill (1982, Sect. 4.12) or Holton (2004,
Chap. 2). For discussion of the numerical complexities introduced by spherical coordinates, see
Williamson and Laprise (1998) or Randall (2000).
Next Forecast Cycle
Data Initial Dynamical Model Post-

Assimilation State Core Forecast processing
Observations Physical
Parameterizations
Fig. 9.2 Schematic information flow in a simple NWP system, including data assimilation,
dynamical core and physical parameterization components
In addition to the basic equations, a full specification of the problem requires

boundary conditions and an initial state. At the upper boundary, the main goal is to
remove upward propagating disturbances from the domain without reflection, either
through damping layers or through wave propagation conditions. Further discussion
is given in Durran (1999, Chap. 8). For lateral boundaries (if any), the fields are
often simply specified, using data from a global model or from an overlapping
larger-scale model run. Conditions at the lower boundary must account for heat,
momentum and moisture transfers between the atmosphere and the ground. These
surface exchanges occur at small spatial scales and are thus typically represented
through parameterized fluxes. Finally, the initial state for the model is defined
by optimally combining new observational data with the previous model forecast
for the time in question (from the prior forecast cycle) to give a best guess of
the atmospheric state. This problem of optimally combining data and prior model
forecasts is referred to as data assimilation.
Figure 9.2 shows a schematic view of a simple deterministic NWP system,
including data assimilation, physical parameterization and dynamical core compo-
nents. The present chapter focuses mainly on the dynamical core, as well as on
predictability issues related to the system as a whole. Discussions of model param-
eterization components can be found in Chap. 10. Numerical weather prediction
applications and weather forecasting in complex terrain is addressed in Chap. 11.
9.2.2 Basics of Spatial Discretization
To be stored on a computer, the continuous field variables in (9.1) must be broken

into finite sets of data values – that is, the fields must be discretized. Methods
for discretizing the fields fall into two general categories: grid-point methods and
series-expansion methods.
In grid-point methods, the fields are defined on a finite set of discrete grid
points dispersed appropriately about the calculation domain. A simple example
is shown in Fig. 9.3a, in which a one-dimensional (1D) function f(x) is sampled
at nine grid points in x. Grid-point methods in turn come in two varieties: finite-
difference methods and finite-volume methods. With finite-difference methods the
Fig. 9.3 Methods of spatial discretization for a 1D function. (a) Finite-difference. Tick marks
along the x-axis indicate grid points, with the sampled data values at the grid points shown by x’s.
(b) Spectral. Dotted lines show three Fourier components for the 1D interval, while the solid line
shows the sum of the three components. (c) Finite-element. Shown are the linear chapeau functions
(dashed) defined by (9.3), along with the sum of the functions (solid)
solution is defined solely at the grid points, with no assumptions made about the
function values in between. Spatial derivatives are approximated using appropriate
sum and difference formulae, as described further in Sect. 9.3. In finite-volume
methods, the function values on the grid are actually averages of the function over
an associated grid cell volume. The spatial derivatives in (9.1) are then replaced
by flux calculations across the grid cell surfaces, which in turn implies an assumed
function structure at points in between the grid. Further discussion can be found in
Ferziger and Perić (2002) or Durran (1999, Chaps. 2 and 5).
In series-expansion methods, the discretization is accomplished not by defining
the function at discrete grid points, but rather by projecting the function onto a finite
set of pre-defined basis functions. As an example, consider the 1D Fourier expansion
N
X
2nc x 2nc x
f .x/ an cos C bn sin (9.2)
nD0
L L
where L is the length of the solution domain and c is the mathematical constant.
Since the basis functions cos(2n c x/L) and sin(2n c x/L) for this problem are always
the same, the solution is specified completely by storing the appropriate coefficient
amplitudes an and bn . The spatial derivatives of f are then evaluated analytically
using (9.2). Of course, the approximation is that, on the computer, the series must
be truncated at finite N, rather than letting N ! 1.
Series-expansion methods are grouped into two broad categories based on the
type of basis functions employed. If the basis functions are orthogonal [as with the
sin and cos in (9.2)], then the method is called a spectral method. A given basis
function in the spectral method extends throughout the entire domain, as illustrated
by the Fourier basis functions in Fig. 9.3b. By contrast, if the basis functions are
nonzero only over a small portion of the model domain (termed an element), then
the method is called a finite-element method. As an example, consider a 1D finite-
element expansion using piecewise linear basis functions, as defined by
( jxxn j
X
N
1 jx xn j x
x ;
f .x/ an n .x/ where n .x/ D (9.3)
nD1 0; otherwise
where xn D (n1)x and where the width of the associated elements is 2x. An
example for this expansion is shown in Fig. 9.3c.
9.2.3 Time Differencing
To save memory, a model will typically store the solution only at a few model
time levels at any given point in the simulation. This restricted nature of the time
discretization means that series expansion methods like that in (9.2) and (9.3) are not
possible for the time coordinate. Thus, even if the spatial discretization is spectral
or finite-element, the associated time discretization will be finite-difference.
Time differencing methods can be grouped into two basic categories: explicit
methods and implicit methods. To illustrate, suppose the system (9.1) is written
symbolically as
@s
D F .s/
@t
where s is the discretized solution vector and where F includes all terms in (9.1),
apart from the Eulerian time derivatives. Suppose further that the goal is to advance
the solution from time level n to time level n C 1, and that the solution is stored in
memory for levels n, n1, : : : , nk (where k is typically 0, 1 or 2). If the method
for advancing to n C 1 relies solely on F(sn ), F(sn1 ), : : : , F(snk ), then the method
is explicit. If the method also involves F(snC1 ), then it is implicit.
Since s at levels n, n1, : : : , nk is already known, the calculation of a single
explicit time step is relatively straightforward. However, explicit schemes also have
a fundamental limitation in that the size of the time step cannot be too large (as
described in Sect. 9.2.4), meaning that many time steps are needed to achieve a
given forecast time. Implicit methods are often free of this time step limitation.3
3
Implicit methods are still limited by the smoothness of the solution. In the NWP context, this
means that the time step should be short enough that the weather and parameterizations at the
beginning of each step are still relevant at the end of the step.
However, the presence of the F(snC1 ) term in an implicit method converts each
time step into a complex algebraic problem, which adds significant computational
overhead per step. Which of the two approaches is ultimately more efficient depends
on the problem to be solved and on the way in which the two schemes are
implemented.
A popular compromise between implicit and explicit differencing is to use
implicit schemes only for the most restrictive terms in F, while the remaining
terms are treated explicitly. Such methods are called semi-implicit schemes. In
practice, the terms treated implicitly in a semi-implicit scheme are usually those
associated with rapidly propagating waves, as discussed in the following two
subsections.
9.2.4 Fast Modes and Stability Limits
One of the most difficult aspects of NWP is the wide range of disturbance timescales
involved. On the fast end of the scale are acoustic waves, with periods ranging from
fractions of a second to several tens of seconds (depending on the wavelength). On
the slow side are the baroclinic and Rossby modes (several days to weeks), while
gravity waves and convective motions are somewhere in between (tens of minutes
to hours).
For the purpose of synoptic-scale day-to-day weather prediction, the baroclinic
and Rossby modes are generally the most important (i.e., have the largest relative
amplitude). These modes are slow, and a model integration could thus in principle
use relatively large time steps (tens of minutes, say) while still maintaining
reasonable accuracy. Unfortunately, for explicit time differencing schemes (as
defined above) the time step is constrained not by accuracy considerations, but
rather by numerical stability constraints for the fast acoustic waves. In particular,
the time steps required for numerical stability of the fast modes are often orders
of magnitude smaller than those needed for accuracy in the slower modes. This of
course compromises the model efficiency.
Not surprisingly, a large number of methods have been developed to overcome
this limitation. One approach is to use implicit (or semi-implicit) time differencing,
which has the effect of artificially slowing (and thus stabilizing) the fast acoustic
modes, thus allowing larger time steps. Another approach is to use explicit
differencing schemes but to integrate with two different time steps: a small step for
terms associated with acoustic modes for stability, and a longer step for everything
else to improve efficiency. These two approaches can also be combined, as in the
popular split-explicit, vertically implicit schemes of Klemp and Wilhelmson (1978)
and Wicker and Skamarock (2002). Still another approach is to modify the basic
equations of motion so as to eliminate sound waves altogether, while only slightly
altering the slower modes. Popular examples are the anelastic equations of Ogura
and Phillips (1962) and the straightforward hydrostatic approximation (as described
below).
Each of these methods has its advantages and disadvantages, and the reader is
referred to Durran (1999, Chap. 7) for further discussion.
9.2.5 Hydrostatic and Nonhydrostatic Models
In most cases, NWP models use much larger horizontal grid spacing (or the
equivalent for series-expansion methods) than vertical spacing, meaning that shorter
wavelengths can be represented in the vertical. Since the frequency of acoustic
waves varies inversely with wavelength, the highest frequency modes in the system
are then the vertically propagating sound waves.
High frequency, vertically propagating acoustic modes impose a severe con-
straint on the time step when using explicit differencing schemes (as described
above). A popular method to overcome this constraint is to use the hydrostatic
approximation, in which the vertical momentum equation in (9.1) is replaced by the
hydrostatic balance. The hydrostatic balance completely eliminates vertical sound
waves,4 which allows much larger time steps to be used than would otherwise be
possible. The approximation also tends to streamline model numerics somewhat,
meaning that hydrostatic models tend to be more efficient than nonhydrostatic codes
(for the same t).
The main drawback to the hydrostatic framework is accuracy – specifically,
the approximation applies only to atmospheric disturbances that have much larger
horizontal spatial scales than vertical scales. For the large-scale weather pre-
diction that has traditionally been the domain of NWP, this assumption is well
satisfied; however, for smaller-scale regional prediction the approximation begins
to break down. Practical experience (e.g., Janjić et al. 2001) suggests that for
length scales larger than roughly 10 km, the hydrostatic constraint is sufficiently
accurate, while for shorter length scales the errors in the approximation become
noticeable.
The push toward finer spatial grids and smaller-scale weather prediction has
meant that most prediction centers have recently transitioned toward nonhydrostatic
models, at least for regional forecast purposes. Using semi-implicit time differenc-
ing (or other means) to stabilize the vertical sound waves, these nonhydrostatic
models can be made almost as efficient as their hydrostatic counterparts.
4
The hydrostatic approximation also eliminates sound waves in the horizontal. However, the
hydrostatic system does still support a similar mode – the Lamb wave – that propagates horizontally
at the speed of sound. The time step constraint due to horizontal propagation is thus the same as
that for the nonhydrostatic problem.
9.3 Finite Differencing
At present, most regional forecasting models are formulated as finite-difference or

finite volume models. The present section thus gives a brief overview of basic finite-
differencing concepts.
9.3.1 An Example Problem
To ground our discussion, we consider a simplified analog to (9.1): specifically, the

one-dimensional advection equation
@ @
CU D0 (9.4)
@t @x
where U is a constant. In the finite-difference method, is stored at a set of discrete
grid points in x, say at x1 , x2 , : : : , xN . To keep things simple, we assume for the
moment that these N grid points are evenly spaced in x (as in Fig. 9.3a), with the
distance between any two neighboring points denoted by x.
Given on the grid, the basic solution method for (9.4) is roughly as follows.
For each time level in memory, the spatial derivative @ /@x is estimated for each
grid point using the data on the grid (as discussed below). This estimate of @ /@x
in turn gives information about @ /@t, and the resulting time tendencies are used to
advance the solution forward in time. If the method is explicit, then each grid point
can be updated in time independently. If the method is implicit then the calculation
is more involved (see description in Sect. 9.2.3), but the basic conceptual idea is
similar.
9.3.2 Truncation Error and Order of Accuracy
Finite-differencing refers to a method for approximating the derivatives of a function

using a gridded data set. To see how this works, recall that for a continuous
function the derivative measures the slope at a particular point. To approximate the
derivative using grid point values, we might then try the average slope between
two neighboring points (as illustrated in Fig. 9.4a); for instance, to estimate the
derivative at point xi , we might try
ˇ
@ ˇˇ
i C1 i
(9.5)
@x ˇxi x
where i is shorthand for (xi ), so that iC1 is (xi C x).

∂ψ ∂ψ
a ∂x b ∂x
ψi+1 − ψi ψi+1 − ψi−1

Δx 2Δx
xi−1 xi xi+1 xi−1 xi xi+1
Fig. 9.4 Approximating derivatives by finite differences. Shown are the continuous function (x)
(solid black line), the tangent line to (x) at the point xi (solid magenta), and the approximate
tangent line as estimated by the difference approximations (blue dashed). (a) The one-sided
difference defined by (9.5). (b) The centered difference defined by (9.8)
To explore this approximation further, suppose we expand (xi C x) as a

Taylor series in x. The result is
ˇ ˇ ˇ

i C1 i @ ˇˇ 1 @2 ˇˇ 1 @3 ˇˇ
D C x C .x/2 C : : : (9.6)
x @x ˇxi 2 @x 2 ˇxi 6 @x 3 ˇxi
showing that as long as x is sufficiently small, our average slope approximation

does in fact approximate the derivative. For small x we focus mainly on the leading
error term in (9.6), and as a shorthand notation we write
ˇ
i C1 i @ ˇˇ
D C O.x/ (9.7)
x @x ˇxi
where the O(x) notation means that the leading error term is proportional to x.
The full error expression in (9.6) is referred to as the truncation error for the method,
while the leading power of x in the truncation error is the order of accuracy for
the scheme. The scheme defined by (9.5–9.7) is thus first-order accurate.
To derive more accurate approximations, we need to use a wider range of grid
points. For instance, taking points to the left and right of xi leads to the second-order
scheme
ˇ ˇ
i C1 i 1 @ ˇˇ 1 @3 ˇˇ
D C .x/2 C : : : (9.8)
2x @x ˇxi 6 @x 3 ˇxi
(as illustrated in Fig. 9.4b) while a fourth-order method looks like

ˇ ˇ
4
i C1 i 1 1 i C2 i 2 @ ˇˇ 1 @5 ˇˇ
D .x/4 C : : : (9.9)
3 2x 3 4x @x ˇxi 30 @x 5 ˇxi
If x is sufficiently small, then these schemes of higher order will have smaller
truncation errors.
A scheme that uses an equal number of points to the left and right of xi [as
in (9.8) and (9.9)] is called a centered-differencing scheme, while a method using
more points on one side or the other [as in (9.7)] is a one-sided method. In practice,
most one-sided schemes are unstable unless the side with more grid points is the
upstream side. A one-sided method is thus also an upstream differencing scheme.
9.3.3 Numerical Dispersion, Dissipation and Filtering
To gain insight into what these various differencing methods imply, suppose we
again consider (9.4), but this time we let the spatial derivative be replaced by a finite-
differencing method while the time derivative remains continuous. For instance,
using a first-order upstream scheme we have (for U > 0)
@ i i 1 @ @ U @2 U @3
CU D CU 2
x C .x/2 C : : : D 0
@t x @t @x 2 @x 6 @x 3
(9.10)
which reinforces the notion that the difference between (9.4) and (9.10) is de-
termined entirely by the truncation error. Assuming a Fourier mode of the form
O exp .ikx i!t/ then leads to the associated discrete dispersion relation
Uk Uk
¨ D Uk i .kx/ .kx/2 C : : : (9.11)
2 6
where k is the (real) wavenumber and ! is the complex frequency.
Inspection of (9.11) shows that the different terms in the truncation error have
distinct effects. The truncation terms with odd-order derivatives of [such as the
(x)2 term in (9.10)] lead to errors in the real part of (9.11), which in turn implies an
error in the phase speed of the mode. Such phase-speed errors are called numerical
dispersion errors. By contrast, terms associated with even-order derivatives produce
errors in the imaginary part, which in turn leads to a damping effect called numerical
dissipation. In both cases, the error is proportional to (kx)m , where m is the power
of x in the associated truncation error term.
These dispersion and dissipation concepts point to an important distinction
between upstream and centered differencing schemes. Specifically, as in (9.11),
upstream schemes always produce both dispersion and dissipation errors together.
By contrast, expanding out the errors in (9.8) or (9.9) shows that centered schemes
produce only dispersion, with no dissipation.
This lack of dissipation would appear to give a distinct advantage to centered
schemes, but in practice this is not necessarily the case. The reason is that all
differencing schemes perform poorly at short wavelengths (or large kx), and
having some k-dependent dissipation to remove these short wavelengths is typically
advantageous. In fact, models that use centered differencing methods will usually
Fig. 9.5 Phase-speed error

for second-order (dot-dashed)
and fourth-order (solid)
centered differencing 80
schemes as applied to the
advection equation (9.4) as a
function of the number of
% Error
grid points per wavelength.
Thin dotted line shows the
5% error threshold 40
0
2 6 10 14
Number of points per wavelength
add an extra even-order derivative (or numerical filtering) term to the basic
equations, which then mimics the effects of numerical dissipation. A discussion
of numerical filtering issues in steep terrain can be found in Sect. 9.8.
9.3.4 Grid Resolution and Accuracy
Roughly speaking, the concept of grid resolution refers to the number of grid points
spanning a given spatial feature, with a larger number of points implying better
resolution. Suppose we take the feature of interest to be a single wavelength, so
that the resolution is measured by N D 2 c /kx. From Sect. 9.3.3, we know that
for any given wavelength the spatial differencing errors depend at leading order on
(kx)m , where m is the power of x in the leading-order truncation terms. Since
kx D 2 c /N, this differencing error is then a function of two independent factors:
(1) the order of accuracy of the scheme (which determines m); and (2) the grid
resolution (which determines kx).
The roles of the accuracy order and grid resolution in determining the overall
error in a calculation are illustrated in Fig. 9.5. Shown in the figure is the phase-
speed error produced by two schemes: the second-order centered scheme defined
by (9.8) and the centered fourth-order scheme described by (9.9). In both cases the
errors for sufficiently good grid resolution (large N) are essentially negligible, while
in the poor resolution limit (specifically for N D 2) the errors for both schemes reach
100%. In between, the errors vary with N, with the error for any given resolution
being smaller for the fourth-order scheme.
The effects of grid resolution and accuracy order in simulating hydrostatic and
nonhydrostatic mountain waves are described in Sect. 9.4.
a b
z y
x x
Fig. 9.6 (a) Schematic of a vertically stretched grid, allowing higher grid resolution near the
ground. (b) A horizontally nested grid, allowing higher resolution over a limited area
9.3.5 Grid Stretching and Nesting
If the disturbance has a single lengthscale, then a uniformly spaced grid is a

reasonable approach. However, in most practical cases the lengthscales of interest
vary from place to place, and in this type of problem a uniform grid becomes
impractical. As a typical example from NWP, consider the atmospheric boundary
layer, where the wind and temperature gradients in the vertical are often much larger
(implying shorter lengthscales) than in the free atmosphere above. Resolving these
short lengthscales requires small z, but setting z this small everywhere would be
very inefficient.
The need to resolve multiple lengthscales has led to several methods for using
variable grid spacing. In the case of the boundary layer, the most common approach
is to use vertical grid stretching, in which z has a small value near the ground but
then increases with height. An example of a stretched grid is shown in Fig. 9.6a. In
general, the changes in the grid spacing from one level to the next must be relatively
small, since rapid changes can lead to reflections in the poorly resolved modes. The
grid spacing change between adjacent levels is thus typically no more than 10–20%.
Stretched grids can also be used in the horizontal, but in NWP a more common
approach is to use horizontal grid nesting. In nesting, a grid with smaller grid
spacing (the fine grid) is embedded within a larger grid at lower resolution (the
coarse grid), as illustrated in Fig. 9.6b. The boundary conditions at the sides of the
nested grid are taken from the coarse-grid calculation. If apart from these boundary
conditions the fine grid is essentially decoupled from the coarse grid, then the
simulation is said to be one-way nested. If, on the other hand, the fine grid results
feed back to the coarse grid (typically through averaging of some sort), then the
grids are two-way nested.
9.3.6 Staggering
In many cases, the grid resolution can be enhanced by using staggered grids, in
which the various disturbance variables in (9.1) are stored at different points in
a b c
Δx
u, v u, v v
ui−1 pi− 1 ui pi+ 1 ui+1
2 2
p, θ u p, θ u
u, v u, v v
Fig. 9.7 Grid staggering. (a) 1D staggered grid configuration in which the grid points for pressure
(solid dots) and velocity (x’s) are staggered by half grid intervals. (b) The B-grid configuration, in
which p and are stored at the dots, and u and v at the x’s. (c) The C-grid, with p and again stored
at the dots, u at the x’s, and v at the plus signs. In all cases, the distance between two neighboring
grid points of the same type is one grid interval (x or y)
space. As an example, consider the grid configuration shown in Fig. 9.7a, in which
u is stored at the points xi D x1 C (i1)x, while p is stored at the staggered points
xi C 1 D x1 C .i 1=2/x. To compute @p/@x at the u points, we might then use
2
ˇ ˇ
pi C 1 pi 1 @p ˇˇ 1 @3 p ˇˇ
2 2
D C .x/2 C : : : (9.12)
x @x ˇxi 24 @x 3 ˇxi
which is a centered second-order difference similar to (9.8), except evaluated over

x rather than 2x. As implied by the errors in (9.8) and (9.12), the staggering
essentially doubles the resolution (or halves the grid spacing) as compared to the
unstaggered case.
Unfortunately, this improved resolution in some terms often implies a cost in
other terms. To see how this works, consider the horizontal C-grid staggering shown
in Fig. 9.7c. As with (9.12), the staggering of pressure and velocity on the C-grid
allows more accurate pressure-gradient computations, along with the divergence
computation in (9.1). On the other hand, the Coriolis and advection terms on the
C-grid require four-way averaging of u and v, which then degrades the accuracy of
these terms. This type of give and take between terms means that no one staggering
is optimal for all applications – the best choice of staggering is usually a matter of
context.
Widely used staggerings in NWP include the C-grid staggering in Fig. 9.7c and
the B-grid in Fig. 9.7b. Analysis and experience suggest that convection and internal
waves are more accurate on the C-grid, while the B-grid is more accurate at large
scales, because no averaging is needed for the Coriolis terms.
9.4 Impact of Order of Accuracy on Mountain Waves
When simulating flow over topography special attention must be given to the treat-
ment of the shortest resolvable waves represented on the numerical mesh. Recent
work has explored the minimum horizontal resolution needed to simulate mountain
waves in both hydrostatic and nonhydrostatic regimes. Davies and Brown (2001)
conducted a series of numerical simulations and concluded that for hydrostatic flow,
features resolved by 10x converged to the true solution; however, appreciable skill
was found for 6x-wide features and qualitative agreement occurred for 4x-wide
features. In contrast, Reinecke and Durran (2009a) demonstrated that hydrostatic
features resolved by four grid points and simulated with second-order advective
schemes attenuated the standing mountain wave amplitude by more than 35%.
Furthermore, they found that for such coarse resolution the morphology of the
mountain wave was significantly distorted compared to the continuous solution with
significant wave energy propagating upstream of the barrier.
For shorter, nonhydrostatic horizontal wavelengths Reinecke and Durran (2009a)
showed that when using second-order accurate finite differences to model non-
hydrostatic flow over a supposedly well resolved 8x-wide mountain, significant
over-amplification can occur in standing mountain waves. As will be seen below,
the vertical velocity in these marginally resolved nonhydrostatic waves can be
as much as 30–40% stronger than the corresponding continuous solutions. This
overamplification has been observed in NWP mountain wave simulations over the
European Alps (Doyle and Jiang 2006) as well as the Oregon Cascades (Garvert
et al. 2007). One possible consequence of mountain-wave overamplification is an
increased tendency for the predicted waves to break down and generate severe
downslope winds (Peltier and Clark 1979). Furthermore, nonhydrostatic modes
with more intense upward vertical motions may lead to an increase in orographic
precipitation (Colle 2008).
To demonstrate the potential for marginally resolved nonhydrostatic waves
to be overamplified in numerical solutions, we consider exact discrete solutions
for steady, linear Boussinesq flow over two-dimensional topography. Following
Reinecke and Durran (2009a) we will consider the case when (9.1) are discretized
on an Arakawa C grid (Mesinger and Arakawa 1976; see Sect. 9.3.6; Fig. 9.7c)
with horizontal grid spacing x and vertical grid spacing z. We will consider
the behavior of predominately nonhydrostatic mountain waves over a topographic
profile specified as
(
x 4 ˇ x ˇ
h0
1 C cos 4a ; if ˇ 4a ˇ 1I
h.x/ D 16
(9.13)
0; otherwise;
which is similar to the familiar witch-of-Agnesi profile except that its elevation
drops to zero at a distance of 4a from the ridge crest. Nonhydrostatic solutions in
which ı D Na/U D 1.8 (corresponding to dimensional parameters of NBV D 0.01 s1 ,
U D 25 m s1 , and a D 4.5 km, where NBV is the Brunt Väisällä frequency, U is
2nd 4th
a b c
z
x x x
Fig. 9.8 The non-dimensional vertical velocity for discrete linear nonhydrostatic (ı D 1:8)
Boussinesq flow over an isolated ridge with (a) second-, and (b) fourth-order horizontal advection
schemes as well as the (c) continuous solution. For (a) and (b) the horizontal grid spacing is
x 0 D 0:67 resulting in eight grid points across the ridge. The contour interval is 0:125U h0 =a
and the zero contour is omitted (From Reinecke and Durran 2009a. © American Meteorological
the horizontal velocity, and a is the half width) will be considered. The horizontal
advective terms will be computed with either second- or fourth-order accurate
finite differences. The horizontal grid spacing is x D 2/3a, which corresponds to
supposedly well-resolved eight grid points across the barrier. At the same time,
z D 0:3U =NBV , implying very good vertical resolution of the mountain waves.
The normalized vertical velocity forced by the topographic profile (9.13) is
shown for second- and fourth-order solutions as well as the continuous solution
in Fig. 9.8. The horizontal axis is normalized by the mountain half-width so that
x’ D x/a, and the vertical axis is normalized by the vertical wavelength of a steady,
two-dimensional hydrostatic mountain wave, z’ D zNBV /(2 c U). The continuous
solution (Fig. 9.8c) shows a classic nonhydrostatic gravity wave with substantial
dispersion of the wave packets and energy propagating downstream (Durran 1986).
Comparing the continuous solution to the discrete solution computed with second-
order-accurate finite differences shows a surprising difference in both the amplitude
and the propagation character of the mountain wave. The maximum amplitude of the
vertical velocity is nearly 30% larger in the second-order solution compared to the
continuous solution. Furthermore, the second-order solution is seen to be relatively
non-dispersive and contains a significant amount of wave energy directly over the
ridge crest.
In contrast, the discrete fourth-order accurate solution (Fig. 9.8b) compares more
favorably to the continuous solution. In this case, the maximum vertical velocity
is only 10% larger than the continuous solution with a notable dispersion of the
wave packets. As in the continuous solution, substantial amounts of energy are
propagating downstream in the fourth-order solution. Increasing to a sixth-order
accurate solution (not shown) reduces the amplitude error even further to 3%. As
will be shown next, the over-amplification apparent in Fig. 9.8a is a direct result
of the second-order-accurate scheme being unable to adequately represent waves

forced by the 8x topography.
The errors observed in the simulated mountain wave can be understood by
comparing the discrete group velocity vector for the second- and fourth-order
numerical schemes with the continuous group velocity vector. In the continuous
system the horizontal and vertical components of the group-velocity vector cg are
NBV l 2 NBV kl
cgx D U 3
and cgz D 3
(9.14)
.k 2 C l 2 / 2 .k 2 C l 2 / 2
where k is the horizontal wave number and l is the vertical wave number. While
(9.14) describes both steady and non-steady motions, our attention is restricted to
steady flow by requiring l 2 D .NBV =U /2 k 2 . Figure 9.9 shows cgx and cgz
for steady flow (solid lines) as functions of the normalized horizontal wavelength
ı 0 D
x NBV =.2c U / D NBV =.U k/ . In context of the nonhydrostatic mountain
wave, the range of ı 0 presented in Fig. 9.9 contains most of the power forced by
the ı D 1:8-wide mountain shown in Fig. 9.8. For example, Fourier analysis of
the vertical velocity field at the surface shows the dominant horizontal wavelength
is ı 0 D 1:55 (corresponding to
x 8x). For decreasing ı 0 the influence of
nonhydrostatic motions becomes important as evident by the increasing amount of
downstream propagating wave energy (Fig. 9.9a). Near the nonhydrostatic cut-off
of ı 0 D 1, cgz drops rapidly to 0 and cgx increases to U; at these short horizontal
scales the majority of wave energy is propagating downstream.
Now consider the effect of discretization on the group-velocity vector. In an
analogous manner to the continuous system, the discrete version of cg can be
derived by differentiating the discrete dispersion relationship with respect to k and
l (Reinecke and Durran (2009a)). The steady-state horizontal and vertical discrete
group velocities are plotted in Fig. 9.9a, b for the second- and fourth-order numerical
schemes. The number of grid-points-per-wavelength indicated along the top axis
of both panels corresponds to the numerical resolution of the waves forced by the
topography in Fig. 9.8.
As evident in Fig. 9.9, the second-order scheme is unable to accurately approx-
imate the correct nonhydrostatic group velocities for any value of ı 0 between 0.5
and roughly 2.5. The downstream component of the group velocity is significantly
reduced compared to the continuous system and there is a non-trivial vertical
component of the group velocity for wavelengths which would be evanescent
in the continuous system (ı 0 < 1). These errors in the group velocity lead to an
accumulation of wave energy over the topography resulting in over-amplification
of the vertically propagating wave (Fig. 9.8a). In contrast, at the same wavelengths
the fourth-order scheme more faithfully represents the downstream propagation of
cg resulting in a more realistic nonhydrostatic wave (Fig. 9.8b).
a 4Δx 8Δx 12Δx 16Δx
0.8U
0.4U
cgx
−0.4U
−0.8U
b 4Δx 8Δx 12Δx 16Δx
0.8U
0.6U
cgz
0.4U
0.2U
δ
Fig. 9.9 The discrete group velocities in the (a) horizontal and (b) vertical directions as a
function of the normalized horizontal wavelength. The top label indicates the number of points
resolving ı 0 (From Reinecke and Durran 2009a. © American Meteorological Society. Reprinted
with permission)
9.5 Vertical Coordinates
The atmospheric primitive equations (9.1) are typically integrated numerically in

a transformed vertical coordinate represented by (x, y, z, t) that replaces the
geometric height. In the past, relatively simple coordinate transforms have been
proposed using the hydrostatic pressure (e.g., Phillips 1957), potential temperature
(e.g., Eliassen and Raustein 1968), and a hybrid of the two coordinates (e.g., Bleck
1978; Benjamin 1989; Hsu and Arakawa 1990) in order to define . For example,
consider that (x, y, z, t) is a new vertical coordinate and has properties such that
is monotonic with respect to height. The gradient operator,r , can be defined
along the coordinate surface through application of the chain rule. For example,
the horizontal pressure gradient in the transformed coordinate is defined as
@p
rz p C r z D r p (9.15)
@z
The second term on the lefthand side in (9.15) is referred to as a terrain metric term.
There are a number of possible choices for the vertical coordinate . The
choice of pressure as the transformed variable simplifies the form of the continuity
equation in the hydrostatic system (e.g., Durran 1999), however the possibility
of constant pressure surfaces intersecting the lower boundary makes it difficult
to utilize. To address this, Phillips (1957) suggested using a normalized vertical
coordinate D p/ps , where ps is the surface pressure. The D 1 surface coincides
with the lower boundary terrain and relaxes slowly to a flat coordinate surface at the
upper boundary where D 0. In nonhydrostatic systems, a similar terrain-following
coordinate devised by Gal-Chen and Somerville (1975) is often used:
zt .z h/
D ; (9.16)
zt h
where zt is the depth of the model computational domain, z is the physical height
(AMSL), and h D h(x,y) is the terrain elevation. Gal-Chen and Somerville highlight
two necessary conditions for well-posed terrain-following coordinates. The first is
that the terrain transformation must map the irregular computational domain to a
rectangular domain (x, y, ), where the boundary conditions are imposed such that
(x, y, h(x,y)) D 0 and (x, y, zt ) D zt . The second condition is that the transformation
must be invertible, which requires the inverse transformation, z D z(x, y, ), to exist,
implying that (z) is a strictly monotonic function of height. An example of terrain-
following coordinate surfaces near the surface for the Gal-Chen and Somerville
transformation is shown in Fig. 9.10a for multiple mesoscale mountains following
Schär et al. (2002).
It follows that the transformed set of 2D inviscid nonhydrostatic and nonrotating
Boussinesq equations can then be written as
2 ˇˇ ˇˇ 3
ˇ ˇ
@u @ ˇ @u ˇ @ ref ˇ @ ref ˇ
C u ˇˇ C ! ˇˇ C cp 0 4 ˇ C x ˇ 5 D 0;
@t @x @ x @x ˇ @ ˇ
x
(9.17)
ˇ ˇ ˇ
@w @w ˇˇ @w ˇˇ @ ref ˇˇ ref
Cu C! C cp 0 z ˇ g D 0; (9.18)
@t @x ˇ @ ˇx @ ˇ ref
x
zt
2
a
zt
4
0
zt
2
b
zt
4
0
zt
2
c
zt
4
0
x/ a
Fig. 9.10 Vertical coordinate surfaces based on the (a) Gal-Chen and Somerville vertical coordi-
nate transformation, (b) hybrid coordinate, (c) SLEVE coordinate using a scale-dependent decay
of terrain features. The vertical axis is normalized by the total domain depth (Modified from Schär
ˇ ˇ
@ @ ˇˇ @ ˇˇ
Cu ˇ C! D 0; (9.19)
@t @x @ ˇx
ˇ ˇ
@ 0 @ 0 ˇˇ @ ref C 0 ˇˇ
Cu C! ˇ
@t @x ˇ @ ˇ
x
ˇ ˇ ˇ ! 0
R @u ˇ ˇ ˇ
C ref C 0 @u
ˇ C x ˇ C z @w ˇ ref C d D 0;
cv @x ˇ @ ˇx @ ˇx dt
(9.20)
d
!D D x u C z w; (9.21)
dt
where u is the horizontal wind, w is the vertical wind, ™ is the potential temperature,
and D .p =p0 /R=cp is the Exner function. The hydrostatically balanced reference
potential temperature and Exner function are given by ™ref and ref while the
perturbation Exner function is . Ideally, the reference state would be identical
to the mean state; however, in NWP this choice is not straightforward since the
mean state can be a function of space and time. In this transformed equation set,
the horizontal derivatives are computed along the transformed coordinate surfaces
through application of the chain rule (9.15) and the definition of the vertical
coordinate transformation.
In the Gal-Chen and Somerville vertical coordinate system, the imprint of the
underlying terrain decays linearly with height. A second type of approach was
proposed by Simmons and Burridge (1981) that makes use of a hybrid pressure-
based coordinate, often applied to hydrostatic models (Fig. 9.10b). These hybrid —-p
coordinate approaches have proven to be especially valuable as the vertical extent
of models has been extended into the middle atmosphere. The basic premise allows
for a smooth transition from terrain-following levels near the surface to isobaric
levels in the upper troposphere or lower stratosphere (Fels et al. 1980; Simmons
and Burridge (1981); Simmons and Strüfing 1983; Eckermann 2008). The quasi-
horizontal isobaric surfaces in the upper portion of the model minimize numerical
errors that may arise due to strongly sloped coordinate surfaces (e.g., see Sects. 9.7
and 9.8).
A terrain-following hybrid vertical coordinate has been proposed for nonhy-
drostatic models by Schär et al. (2002). In contrast to the —-p hybrid coordinate
system, this smooth level vertical (SLEVE) coordinate provides smooth and quasi-
horizontal coordinates at middle and upper levels. The SLEVE coordinate involves
a splitting of the model topography into a large-scale (filtered) part and a small-
scale deviation, with the characteristic that the small-scale topographic features
decay faster with height than the larger-scale terrain components. A comparison
of an example set of vertical coordinate surfaces for the conventional vertical
coordinate system, hybrid coordinate, and the SLEVE coordinate is shown in
Figs. 9.10a–c. Schär et al. (2002) found the SLEVE coordinate to mitigate truncation
errors introduced by the coordinate transformations in the presence of complex
topography. Zängl (2003) extended the SLEVE approach to pressure-based —-
coordinates and noted that the accuracy of horizontal advection is substantially
improved due to the reduced slope of the coordinate surfaces. Furthermore, the
pressure-based type of the SLEVE coordinate has been generalized by Zängl (2007)
to provide an adaptive functionality. The vertical position of the coordinate surfaces
in this system is made time dependent through the use of a prognostic equation.
Isentropic or hybrid isentropic coordinates have been applied to a wide range
of meteorological applications using limited area and global circulation models
(e.g., Benjamin 1989; Hsu and Arakawa 1990; Bleck and Benjamin 1993; Johnson
et al. 2000, 2002). The advantages of modeling in isentropic coordinates have been
addressed in the literature in detail (e.g., Hsu and Arakawa 1990; Webster et al.
1999; Arakawa 2000). Some of the advantages of isentropic coordinates include
(1) increased vertical resolution near fronts, inversions, and the tropopause, (2)
the isentropic surfaces act as material surfaces for adiabatic processes, which can
be particularly advantageous for high-altitude transport, (3) the pressure gradient
takes a form that is irrotational when the curl is computed along the coordinate
surfaces, which implies a potential reduction in related discretization errors, and (4)
may provide improved conservation properties including that of Ertel’s potential
vorticity. Some disadvantages of modeling in isentropic coordinates include a lack
of vertical resolution in unstable boundary layer regions and the intersection of
model surfaces with the surface of the earth. These disadvantages are largely
ameliorated by using a hybrid isentropic coordinate system in which the boundary
layer is represented by a terrain-following coordinate (Bleck 1978). Zängl (2007)
demonstrated that the adaptive vertical coordinate discussed above can emulate a
hybrid isentropic system. The hybrid isentropic coordinates were found to improve
the representation of the tropopause because of the enhanced vertical resolution
in its vicinity; however, the hybrid isentropic coordinates are sub-optimal for
representing the dynamics of breaking gravity waves because of inadequate vertical
resolution due to the near neutral vertical stratification in the vicinity of the
overturning waves.
In the vicinity of steep topography, Mesinger and Janjić (1985) found significant
errors in computing the horizontal pressure gradient force in models that make
use of terrain-following vertical coordinates (e.g., see Sect. 9.7). Step coordinate-
type approaches have been introduced to mitigate truncation errors that arise in the
vicinity of steep terrain. In step coordinate models, the topography is represented as
discrete steps with the tops of the step layer coincident with the nearly horizontal
coordinate surfaces. Step coordinates have been applied operationally in limited-
area hydrostatic and nonhydrostatic models, such as the previous generation Eta
Model at the National Center for Environmental Prediction (NCEP) (Mesinger
1984; Mesinger et al. 1988; Janjić 1990; Gallus and Rancic 1996) and have been
shown to improve the performance of terrain-forced flows on the synoptic scale
(e.g., Black 1994). However, the step vertical coordinate may generate spurious
gravity wave activity due to flow separation downstream of the mountain that
introduces an artificial vorticity source at the corner of each step (Gallus and Klemp
2000). The differences between analytic solutions and step coordinate models of
flow over two-dimensional mountains are significant unless the vertical grid spacing
is very small compared to the height of the mountain. On the other hand, terrain-
following vertical coordinates have been shown to produce accurate solutions
provided the vertical grid interval is small compared to the gravity wave vertical
wavelength.
9.6 Consistency of Terrain Transform Metric Terms
In this section the necessity for the consistent treatment of the terrain metric terms
in finite difference approximations to (9.1) will be discussed. Klemp et al. (2003)
(here after KSF2003) demonstrated that inconsistencies in the evaluation of the
terrain metric terms can have a significant negative impact on the quality of steady
mountain wave solutions. The source of this inconsistency and a brief demonstration
of its impact on steady two-dimensional Boussinesq flow will be discussed in this
section.
Following KSF2003 we assume small amplitude perturbations around a basic
state of uniform wind speed U, static stability NBV , and a hydrostatically balanced
reference state Exner function ref . In idealized models the reference state Exner
function is typically specified to be the mean state, ;N however, for NWP such a
choice is not straightforward since the mean state can vary both temporally and
spatially. In this case the mean state Exner function is expressed as N D ref C , Q
where Q is a residual. The total Exner function is then D ref C Q C 0 . For
simplicity, we assume that Q is time independent and varies only in z. Substituting
ref D Q C 0 into (9.17) and applying the chain-rule to the terms involving ,Q
the steady linearized zonal momentum equation can be written as
ˇ " ˇ #
@u0 ˇˇ @ 0 ˇˇ x x
U C c Q
C c Q
D 0; (9.22)
@x ˇ @x ˇ
p 0 z p 0 z
z z
" ˇ #
@b 0 ˇˇ x 2 x
U NBV
2
C U NBV C N 2 w D 0; (9.23)
@x ˇ z z
where b 0 D g 0 =0 and NBV 2

D g =0 d N =d z is the Brunt-Väisälä frequency. In
(9.22) and (9.23) x D @ =@x , z D @ =@z , and x = z is referred to as the metric
term. Note that without terrain the metric term vanishes and (9.22–9.23) are identical
to the zonal momentum and thermodynamic equations in the Cartesian coordinate
system.
The first bracketed term in (9.22) represents the contribution of the horizontal
pressure gradient along surfaces of constant , while the second bracketed term
is the contribution perpendicular to surfaces of constant . Similarly, the first and
second bracketed terms in (9.23) are associated with the advection of buoyancy
along surfaces of constant and perpendicular to surfaces of constant , respec-
tively. In order for (9.22) and (9.23) to be consistent with the zonal momentum
and thermodynamic equations in a Cartesian coordinate system, the metric terms
must cancel. For these terms to cancel, the order of accuracy of the finite difference
must be the same between the first and second bracketed terms in (9.22–9.23). For
example, if the horizontal pressure gradient is computed with second-order finite
differences, then the metric term in (9.22) must also be computed with a second-
order finite difference. Likewise, in (9.23), if the horizontal advection of buoyancy
is computed with fourth-order finite difference then the horizontal metric term must
be computed in the same way.
If the metric terms in (9.22) and (9.23) are not differenced in an identical
manner then the inconsistencies will result in a spurious forcing of the gravity
wave solution. As shown in KSF2003 the forcing will be the most significant at
short horizontal wavelengths near the nonhydrostatic cutoff for vertical propagation,
x D 2U =NBV . Additionally, the forcing in (9.22) is nearly constant with height
whereas the forcing associated with (9.23) linearly decreases with height, resulting
in the maximum spurious forcing near the surface with weaker forcing aloft [see
KSF2003, Eqs. 33 and 34].
To demonstrate the impact of these numerical errors we consider steady two-
dimensional Boussinesq flow over a multi-scale ridge specified as
(
x 4 2 x ˇxˇ
h0
1 C cos 4a cos 4
; if ˇ 4a ˇ 1I
h.x/ D 16
(9.24)
0; otherwise:
This topographic profile is shown schematically in Fig. 9.10 and is similar to

that used by KSF2003, except that the small-scale ridge variations are enveloped by
a cos4 function instead of a Gaussian function. As in KSF2003, the broad mountain
scale is specified as a D 5 km while the half-width of the small-scale ridges is
D 1 km. Choosing values of the wind speed and Brunt-Väisälä frequency to be

U D 10 m s1 and NBV D 0:01 s1 implies that the non-dimensional mountain half
widths for the broad- and small-scale ridges are NBV a =U D 5 and NBV
=U D 1,
respectively. Therefore, the forced waves are in the nonhydrostatic regime with
substantial forcing near the cutoff for vertical propagation. To further simplify the
problem we set Q D 0 in (9.22) and only consider the role of the terrain metric
inconsistencies with respect to the advection of buoyancy in (9.23).
Figure 9.11a shows the vertical velocity of the continuous steady-state solution
for flow over terrain defined by (9.24). The horizontal scale is normalized by the
broad mountain half-width x 0 D x =a and the vertical scale is normalized by the
vertical wavelength of a two-dimensional steady hydrostatic mountain wave z0 D
NBV z =U . Evident in the figure is several evanescent waves superimposed on a
broader scale vertically propagating gravity wave. These short evanescent waves are
forced by the small-scale ridges and decay rapidly with height, whereas the broader
vertically propagating wave is forced by the envelope of the topographic profile.
Now consider what happens if the horizontal derivatives are approximated with
finite differences with various orders of accuracy (see KSF2003). Figure 9.11b
shows the discrete numerical solution for flow over the same ridge but with the
horizontal advective terms computed with a fourth-order accurate finite difference
approximation and the horizontal metric term computed with a second-order
accurate finite difference approximation. The horizontal grid spacing is x D
0:125
which should generate a nearly converged numerical solution; however,
inspection of the vertical velocity field reveals significant distortion in the vertically
propagating portion of the mountain wave with an approximate wavelength of half
a b c
Fig. 9.11 The non-dimensional vertical velocity for flow over multi-scale topography in the
transformed vertical coordinate system. The topographic profile is given by (9.24) with a D 5 km
and œ D 1 km. Plotted is the (a) continuous solution, (b) discrete solution with fourth-order
differencing of the advective terms and second-order differencing of the terrain metric term, and
(c) fourth-order differencing for both the advective terms and the terrain-metric terms. The contour
interval is 0.125 Uh0 /a and the zero contour is omitted (Modified from Klemp et al. 2003)
that of the vertical velocity field. The distortion decreases linearly with height and is
a direct result of the spurious forcing associated with discrepancies in the truncation
errors for the metric term and the buoyancy advection term. The solution presented
in Fig. 9.11c is nearly identical to the solution in Fig. 9.11b except that the horizontal
metric is computed with a consistent fourth-order finite difference approximation.
Visual inspection of the continuous solution (Fig. 9.11a) and the numerical solution
computed with consistent numerics (Fig. 9.11c) reveals nearly identical solutions.
9.7 Pressure Gradient Representation in Terrain-Following

Coordinates
As discussed in the previous section, the transformation from Cartesian into terrain-
following coordinates involves a splitting of any horizontal gradient into two terms,
that is, the horizontal gradient along the (in general) sloping coordinate surface plus
a metric term involving the local vertical gradient of the field times the slope of the
coordinate surface. In the case of the pressure gradient term, (9.15) can be written as

1 1 @p @
rj p D rj p rj z : (9.25)
z @ @z
Even with consistent horizontal differencing, this transformation is a potentially

important source of discretization errors, as the two terms may not properly cancel
each other for a field with zero horizontal gradients but nonzero vertical gradients.
Pressure is the most critical variable in this aspect because the vertical pressure
gradient typically is several orders of magnitude larger than the horizontal gradient.
An inadequate representation of the pressure gradient may induce spurious air

motion in mountainous topography, compromising the model’s ability to capture,
for example, wintertime cold-air pools and fog in valleys or basins. Careful
consideration of the numerical representation of pressure gradients is therefore
highly important.
Not surprisingly, pressure gradient errors are generally considered to be one of
the most important disadvantages of terrain-following coordinates. Much effort has
been spent on reducing the pressure-gradient error in hydrostatic sigma-coordinate
models, for which the problem had already been realized since the late 1960s
(Corby et al. 1972; Gary 1973; Mesinger 1982; Mahrer 1984; Mihailović and
Janjić 1986; Janjić 1989). In hydrostatic models, the crucial issue is to obtain
hydrostatic consistency between the surface pressure and the geopotential field,
which needs to be determined by vertically integrating the hydrostatic equation. This
is particularly difficult in the presence of changing vertical temperature gradients,
such as those near inversion tops or at the tropopause. Additional problems arise
in spectral models because of the convergence properties of Fourier transforms,
being such that isolated mountains may induce global errors in the pressure gradient
calculation (Janjić 1989). In nonhydrostatic models, the problem can be somewhat
alleviated by introducing a hydrostatically balanced reference atmosphere and
using the deviation of the pressure p from its reference value p0 , the perturbation
pressure p’, as a prognostic variable (e.g. Grell et al. 1995). This way, the order-of-
magnitude discrepancy between horizontal and vertical gradients is greatly reduced.
Introducing this splitting into (9.25) and assuming rj p0 D @p 0
rj z, which is
@z 0 @

0
automatically fulfilled for p0 D p0 (z), yields the expression rj p @p
1
@ @z
rj ,
z
which is formally the same except that full pressure is replaced by its perturbation.
For simplicity, we restrict our further
ˇ analysis to the one-dimensional problem
0ˇ ˇ
@p0 @z ˇ
without the density factor, i.e. @p@x ˇ @z @x
. Discretizing this expression on an

unstaggered grid with centered differences, one obtains
ˇ
@p ˇˇ p 0 i C1;k p 0 i 1;k p 0 i;kC1 p 0 i;k1 zi C1;k zi 1;k
ˇ ; (9.26)
@x i;k 2x zi;kC1 zi;k1 2x
whereas a C-grid staggering yields

ˇ
@p ˇˇ p 0 i C1;k p 0 i;k

@x ˇi C1=2;k x
.p 0 i C1;kC1 C p 0 i;kC1 / .p 0 i C1;k1 C p 0 i;k1 / zi C1;k zi;k
(9.27)
.zi C1;kC1 C zi;kC1 / .zi C1;k1 C zi;k1 / x
when assigning the velocity points to half indices. Inspecting these expressions
suggests that the C-grid staggering is advantageous because on a C grid, the vertical
gradient is evaluated at the same horizontal positions as the horizontal gradient,
whereas different grid points enter into this calculation on an unstaggered grid.
This reduces the potential for inconsistencies related to curved (convex or concave)
orography. An issue that cannot be addressed with C-grid staggering is related to
the ratio between the slope steepness and the vertical layer spacing. In general,
the height distance between adjacent grid points differs from the vertical layer
0
spacing, so the vertical positions at which the @p @z
@z and @x terms are evaluated do
not match each other. This is particularly problematic near the surface, where the
vertical model layer spacing is usually quite dense (in order to properly represent
PBL processes) and in fact may be an order of magnitude smaller than the height
difference between neighboring grid points.
Inspecting (9.27) reveals that discretization errors trivially vanish if p’ is constant
with height, and also for equally spaced model layers if p0 varies linearly with z.
For a C-grid discretization with equally spaced model layers, a Taylor expansion
3 0
reveals that the leading pressure gradient error is proportional to @@zp3 . To quantify
the contributions to this leading error term, consider the linearized ideal gas equation
p0 C p 0 D .0 C 0 /R.T0 C T 0 / and take its second derivative with respect to z.
Assuming a hydrostatically balanced reference state with T0 .z/ D T00 C .z z00 /
and neglecting quadratic perturbation quantities, one obtains
@2 p 0 @2 T 0 @T 0 d 0 d 2 0 @0 @2 0
2
D 2
R0 C 2 R C T 0 R 2 C 2 R C 2 RT0 ;
@z @z @z dz dz @z @z
0 p0
which, using @p@z
D g0 , d 0
dz
D .gCR
RT0
/0
and 0 RT0
p0
RT02
T 0, yields after a
few lines of algebra
@3 p 0 3p0 g gQ gO 0 p0 g @T 0 p0 g @2 T 0 gQ
gO
g
D 4
T 2 3 .gO C 2g/
Q C 3 3 p0 I (9.28)
@z3 3
R T0 R T0 @z RT02 @z2 R T0
gQ D g C R , gO D g C 2R . Assuming T0 D 285 K, ” D 6.5103 K m1 ,

0 2 0
p0 D 20 hPa, T0 D 20 K, @T@z D 5 K km1 and @@zT2 D 5 K km2 yields values of
about 1.8 108 Pa m3 , 5.1 108 Pa m3 , 2.1 107 Pa m3 and 2.6 109 Pa m3 ,
respectively, for the terms on the r.h.s of (9.28), which suggests that for suitably
chosen sea-level reference parameters, the largest pressure gradient errors can be
expected in and around inversion layers where the lapse rate deviates strongly from
the reference and/or changes rapidly with height.
To illustrate these calculations, idealized numerical experiments have been
conducted with the MM5 (Grell et al. 1995). They use one domain with 101 101
grid points, a horizontal grid spacing of 1 km, 50 layers in the vertical with a model
top at 100 hPa, and dry physics. In the domain center, a Gaussian mountain with a
peak height of 2,500 m and an e-folding width of 5 km is placed. The simulations
are initialized with a hydrostatically balanced atmosphere at rest that ideally should
remain so all the time. The reference atmosphere assumes a sea-level temperature
of 275 K and a vertical temperature gradient of ddlnT p D 48 K, corresponding
to ddTz 6:2 K km1 in the lower troposphere. Three idealized temperature
profiles are considered, starting from a sea-level temperature of 273.15 K. Their
Brunt-Väisälä-frequencies are NBV 2 D 104 s2 (4 104 s2 ) below (above) a

reference pressure of 250 hPa (T1), NBV 2 D 2 104 s2 everywhere (T2), and
NBV 2 D 104 s2 except for an inversion layer between 900 and 825 hPa where
NBV 2 D 103 s2 (T3). All experiments use the truly horizontal temperature dif-
fusion scheme implemented by Zängl (2002) because this is needed to suppress
the development of small-scale numerical noise in an atmosphere at rest (see Zängl
2002, 2003; Zängl et al. 2004). Unless mentioned otherwise, the MM5 version of the
SLEVE coordinate (Schär et al. 2002; Zängl 2003) is used, which ensures that the
coordinate surfaces at 250 hPa (the tropopause level of T1) are virtually horizontal.
The results after 6 h of simulation are displayed in Fig. 9.12. For profile T1, only
weak circulations with horizontal speed maxima of about 0.1 m s1 appear over the
slopes of the mountain (Fig. 9.12a). This is related to the fact that the low-level lapse
rate of profile T1 is rather close to that of the reference atmosphere. The tropopause
does not induce notable disturbances due to the use of the SLEVE coordinate.
With an ordinary ¢-coordinate, however, the numerical errors around the tropopause
exceed those at low levels by an order of magnitude (Fig. 9.12b). Profile T2, which
deviates more notably from the reference lapse rate in the troposphere, entails wind
speed maxima of about 0.5 m s1 (Fig. 9.12c; note the different ordinate range).
However, the by far strongest circulation appears when a marked inversion is present
in the lower troposphere (profile T3, Fig. 9.12d). The wind speed reaches maxima
of 2.5 m s1 for the setup chosen here, which may constitute a substantial error
under weak-wind anticyclonic conditions. In fact, the type of spurious circulation
appearing for profile T3 is believed to be one of the primary reasons for the
difficulties of terrain-following-coordinate models in maintaining cold pools and
fog layers in valleys or basins.
9.8 Numerical Diffusion and Filtering over Steep Terrain
There are basically two fundamental reasons behind the need for numerical diffusion
or filtering, one physical and one numerical. The physical one is the energy cascade
towards small scales, where kinetic energy is ultimately dissipated into internal
(thermal) energy. It can be most simply illustrated by the 1D nonlinear advection
equation, @u@t C u @x D 0. Assuming a sinusoidal structure in the velocity field,
@u
u D sin kx, one obtains u @x @u

D k sin kx cos kx D 12 k sin 2kx. Thus, nonlinear
advection induces higher wave numbers (or shorter wavelengths) of motion than
originally present in a field. In a numerical grid point model, this interferes with the
fact that the shortest resolvable wavelength is 2x, corresponding to kmax D x
.
Using trigonometric relations, it can be shown that energy transports to wave
numbers k > kmax are aliased into 2kmax k in a numerical grid point model,
which implies that the energy would accumulate near k D kmax unless applying a
proper means of dissipation, leading inevitably to numerical instability.
Besides physically based diffusion schemes, which in many models are applied
only in vertical direction, this can be achieved either with an explicit numerical
a b
c d
Fig. 9.12 Idealized simulations with a mountain in an initially resting atmosphere. Results are
shown at t D 6 h for temperature profiles (a) T1, (b) T1 with standard —-coordinate, (c) T2, (d)
T3 (see text for explanation). Black lines denote potential temperature (contour interval 2 K),
pink (brown) lines upward (downward) vertical wind speed (contours starting at 1 cm s-1 with a
multiplicative interval of 2), and colors denote horizontal wind speed (see shading key)
diffusion term or with a numerical scheme that implicitly damps small-scale

structures at a rate proportional to the local Courant number (e.g. upwind-biased ad-
vection discretizations). The discussion about the preferable method is still ongoing
among model developers, as one might argue that an explicit diffusion term allows
better control over diffusion than an implicit one, but the so-called background
diffusion coefficient that is independent from the local flow properties needs to
be quite large in practice. Moreover, explicit numerical diffusion entails the risk
of systematic errors over mountainous topography. Numerical diffusion is usually
realized as a fourth-order hyperdiffusion term, which in discretized form reads
@4 . i C2 C i 2 / 4. i C1 C i 1 / C6 i
4

@x x 4
for an arbitrary variable . Compared to second-order diffusion used to describe
turbulent mixing processes, the fourth-order diffusion is much more scale-selective,
which allows for imposing enough damping on scales near 2x to ensure numerical
stability without dissipating too much energy on larger scales well resolved in the
numerical model. A disadvantage is that applying the —-coordinate transformation
to fourth-order diffusion leads to a large number of metric terms, which would be
numerically expensive and prone to substantial discretization errors. Thus, fourth-
order diffusion is usually computed along the terrain-following coordinate surfaces
without accounting for metric terms.
Experience shows that this issue is relatively unproblematic for momentum
because the height-dependence of wind speed is – at least at low levels – more
related to the height above ground than to absolute height (because the frictional
approximation to the turbulent drag effect follows the surface), and momentum
diffusion does not induce spurious motion in an atmosphere at rest. On the other
hand, remarkably large systematic errors may arise for variables having a strong
vertical stratification, most notably temperature and moisture. The problem is
illustrated in Fig. 9.13a for temperature in a narrow valley. Computing diffusion
at the edge of the valley bottom (marked with a diamond) involves two points at
the valley bottom and two along the slope, the latter having lower temperatures
due to the assumed vertical gradient. Using actual temperature as diffused variable
therefore would cool the valley bottom. More generally, as diffusion always
attempts to smooth differences along the surface where it is applied, an isothermal
temperature profile will tend to form in the valley in the absence of other forcing
mechanisms. Likewise, the atmosphere tends to be dried in valleys and moistened
over mountains by the moisture diffusion because the mixing ratio (or specific
humidity) usually decreases rapidly with height.
To reduce the systematic errors related to numerical diffusion, a variety of
modifications have been proposed in the literature. The most common one is to
subtract the reference temperature profile before computing temperature diffusion,
which brings the lapse rate that the diffusion attempts to establish in narrow valleys
close to climatological values (e.g. Grell et al. 1995). However, as illustrated below,
the diurnal temperature cycle in narrow valleys may still be compromised. This
problem may be reduced by using a slope-dependent diffusion coefficient that
approaches zero for steeply sloping coordinate surfaces (e.g. Doms and Schättler
2002). A more accurate way is to compute diffusion truly horizontally by using
vertically interpolated values on the external grid points (Ballard and Golding
1991; Li and Atkinson 1999; Zängl 2002). This method, which is illustrated in
Fig. 9.13b (triangles), will be referred to as z-diffusion in the following. While z-
diffusion is straightforward to implement at sufficient vertical distance from the
ground, a special treatment is obviously required when the data points needed for
its computation would intersect the ground. Extrapolating the temperature profile
below the ground constitutes an uncontrollable source of errors and has never
been used to the authors’ knowledge. A very simple method is to switch back to
ordinary diffusion along model levels where the ground would be intersected by
z-diffusion (Ballard and Golding 1991). However, unless combined with a slope-
dependent diffusion coefficient, this largely retains the related numerical errors in
narrow valleys. A somewhat more advanced method is to use one-sided z-diffusion
a b
Fig. 9.13 Schematic illustration of diffusion calculation in an idealized valley for (a) uniform
temperature decrease with height, (b) a low-level inversion (as indicated by the labeling on the right
axis). Diamonds indicate the grid point for which diffusion is to be computed, crosses the external
grid points of the diffusion stencil, and dashed lines in (b) symbolize terrain-following coordinate
surfaces (From Zängl 2002. © American Meteorological Society. Reprinted with permission)
along slopes (Li and Atkinson 1999), but there may still be grid points near the
bottom of narrow valleys where one-sided z-diffusion is not possible either without
intersecting the ground. Moreover, Zängl (2002) found that this is recommendable
for moisture only, as one-sided z-diffusion of temperature tends to suppress the
slope wind circulation. Instead, Zängl (2002) proposed to subtract the local vertical
temperature gradient before computing diffusion along the model levels, combined
with a slope-dependent (and curvature-dependent) diffusion coefficient. The latter is
evaluated separately for the x and y directions to allow diffusion along a valley axis
but not across it. The necessity for using a slope-dependent diffusion coefficient is
illustrated in Fig. 9.13b. When the height difference of adjacent grid points greatly
exceeds the vertical layer spacing, the local vertical temperature gradient may not be
representative for the height range of the grid points and thus may induce spurious
overcorrection effects. It has also been found that the vertical temperature gradients
entering into the diffusion computation need to be limited to some threshold to
prevent runaway cooling in isolated topography depressions.
Figure 9.14 illustrates the impact numerical diffusion may have in high-
resolution simulations in mountainous topography. The results are taken from
semi-idealized simulations of the valley-wind circulation in the Alpine Inn Valley,
using realistic topography but idealized large-scale conditions without synoptic-
scale pressure gradients. The horizontal resolution in the finest model domain
(five nested grids in total) is 800 m, and the radiative forcing is prescribed for
October 15. Further details are given in Zängl (2002). Results are shown for
three simulations at a grid point in the lower Inn Valley. The simulations using
diffusion along model levels are named T-diffusion and T0 -diffusion, which means
2000
a b T-diffusion
1800 T’-diffusion
upvalley downvalley
z-diffusion
1600
Height (m ASL)
1400
1200
1000
800
600
-10 -8 -6 -4 -2 0 2 4 6 8 10 274 276 278 280 282 284 286

Wind speed (m s-1) Temperature (K)
Fig. 9.14 Wind speed (a) and temperature (b) profiles obtained from idealized simulations of the
valley-wind circulation in the Alpine Inn Valley. Results are shown for a grid point in the lower Inn
Valley lying at about 500 m ASL at 16 UTC (solid lines) and 06 UTC (dashed lines; local time is
UTC C 1 h) (See text for definition of the simulation acronyms given in the color key. From Zängl
that temperature diffusion was computed from the full temperature and perturbation
temperature (in other words, full minus reference temperature) profiles, respectively.
The temperature profiles in Fig. 9.14b for late afternoon (16 UTC) and morning (06
UTC) reveal that the valley temperature is systematically colder with T-diffusion
than with T0 -diffusion, which is consistent with the statement made above that
T-diffusion attempts to establish an isothermal temperature profile whereas T0 -
diffusion relaxes the vertical temperature gradient to a value near -6 K km1 . In both
cases, the diurnal temperature range near the ground is unrealistically small. Only z-
diffusion allows for a reasonable diurnal temperature range. As a consequence, only
the simulation using z-diffusion is able to reproduce the periodic change between
upvalley flow in the afternoon and downvalley flow at night (Fig. 9.14b). With
T-diffusion (T0 -diffusion), the valley flow is downvalley (upvalley) all day long
because the valley atmosphere always stays colder (warmer) than the atmosphere
at corresponding heights in the Alpine foreland. This demonstrates that explicit
numerical diffusion needs to be applied very carefully over steep topography
because the related errors may otherwise lead to a fundamentally wrong flow
behavior.
From this perspective, implicit numerical diffusion such as that provided by
upwind-biased advection discretizations appears to be advantageous compared to
dynamical core formulations with an explicit diffusion term. Implicit diffusion is

proportional to the local Courant number and thus is intrinsically inactive in a resting
atmosphere, whereas explicit diffusion may induce the largest errors because no ad-
vective forcing is present. In fact, experiments with the WRF model revealed that a
realistic simulation of Alpine valley wind circulations is much easier to achieve with
a dynamical core that does not rely on explicit numerical diffusion to be numerically
stable. It has also been found that a coordinate-transformed second-order diffusion
without a background diffusion coefficient, such as that implemented in WRF to
represent physical diffusion processes, behaves almost neutrally in the sense that
the simulated valley-wind circulation is essentially the same as that without any
diffusion term. However, similar to what has been found for numerical diffusion, a
limiting of the vertical temperature gradient entering into the metric terms may be
needed to reliably prevent runaway cooling at local topographic depressions.
9.9 Predictability of Atmospheric Flows over Complex

Terrain
Lorenz (1969) argued that due to the rapid upscale propagation of uncertainties in
the specification of initial conditions, the predictability of mesoscale motions with
spatial scales on the order of 10 km would be limited to time scales on the order of
1 h. However, experience with numerical models has led some to suggest that terrain
may act to constrain error growth and extend the predictability of certain mesoscale
motions (e.g. Anthes et al. 1985, Paegle et al. 1990, Mass et al. 2002). The science of
predictability seeks to understand limitations of accurate forecasts given errors in the
numerical model and uncertainties in the initial conditions. While the development
of NWP models has advanced steadily over the last two decades (see Chap. 11)
along with advancements in the understanding of mountain meteorology dynamics
(see Chaps. 2 and 3), the systematic evaluation of predictability of the mesoscale and
the influence of terrain has been limited. This is due in part to the limited tool set
available; however, the recent development of mesoscale ensemble techniques, data
assimilation, and mesoscale adjoint models has allowed researchers to investigate
the mesoscale predictability of terrain forced flows. This section will give several
examples of the tools used for mesoscale predictability and results from a handful
of recent studies.
9.9.1 Mountain Wave and Downslope Wind Predictability
One way to explore predictability of mesoscale phenomena due to initial condition

uncertainties is through the use of ensemble data assimilation and ensemble pre-
diction. An ensemble Kalman filter (EnKF) is an ensemble-based data assimilation
tool which optimally combines observations and background forecasts (Evensen
1994). Under the assumption of Gaussian error statistics, the analysis variability
generated by EnKF data assimilation represents the expected errors in NWP initial
conditions (Hamill 2006). The source of this initial error is associated with errors in
the background forecast (which may arise from both model errors and the growth
of initial condition errors from a previous forecast cycle) and uncertainty in the
observations. By integrating the initial ensemble forward in time, the predictability
of mesoscale phenomena is explored by considering the growth of the ensemble
variability over the forecast period. For detailed reviews of EnKF data assimilation,
the reader is referred to Evensen (2003) and Hamill (2006).
Reinecke and Durran (2009b) used the Coupled Ocean/Atmosphere Mesoscale
Prediction System (COAMPS) to integrate a set of 70 different initial conditions,
generated from an EnKF data assimilation system. They considered a pair of
downslope wind events which occurred during the Terrain-Induced Rotors Exper-
iment (Grubišić et al. 2008; T-REX): 25–26 March, 2006 corresponding to IOP
6 and 16–17 April, 2006 corresponding to IOP 13. Both cases exhibited large
forecasted ensemble variability of the downslope windstorm with differences in
the ensemble predicted downslope wind speed growing larger than 25 m s1 over
forecast periods shorter than 12 h. They suggested that the presence of orography
does not necessarily enhance mesoscale predictability as suggested by Anthes et al.
(1985), but instead, the predictive time scales of mesoscale perturbations may be no
longer than originally suggested by Lorenz (1969).
An example of the large error growth associated with mountain wave breaking
and downslope winds during T-REX IOP-6 is given in the remainder of this section.
For this case strong westerly flow extended through the depth of the troposphere
as a short-wave trough impinged upon the Sierra Nevada Mountains upstream
of the Owens Valley. Reinecke and Durran (2009b) simulated this event with
70 ensemble members on 27-, 9-, and 3-km one-way nested domains, centered
over the Central and Southern Sierra Nevada Mountains in California. For these
simulations the metric terms were computed in a consistent manner (Sect. 9.6) and
the horizontal advection was computed using fourth-order accurate finite difference
approximations on a staggered C grid (9.9). Figure 9.15 shows the topography of the
Owens Valley and the adjacent Sierra Nevada Mountains as represented by the 3-km
COAMPS domain. The north-south extent of the Sierra-Nevada Mountains and the
large vertical relief between the mountain crest and the valley floor make westerly
flow situations favorable for the development of large amplitude mountain waves
and strong downslope winds (Holmboe and Klieforth 1957; Grubišić and Billings
2008).
The evolution of the simulated downslope windstorm from a 20 member
ensemble subset for a forecast initialized at 18 UTC, 25 March, 2006 is shown
in Fig. 9.16. A downslope wind metric was computed by averaging the zonal
component flow below 350-m AGL throughout the boxed region shown in Fig. 9.15.
The subset was selected as the strongest- (black) and weakest-10 (grey) set of
ensemble members evaluated at 00 UTC 26 March (the 6-h forecast). The difference
between the strong- and weak-member subsets represents an upper bound of the
Fig. 9.15 The topography on the 3-km domain. The locations of the Owens-Valley metric box,
vertical cross-section, and DRI 01 mesonet station are also shown (From Reinecke and Durran
2009b. © American Meteorological Society. Reprinted with permission)
initial condition error growth of the downslope wind metric over the first 6-h of
the forecast. Apparent in Fig. 9.16 is the rapid growth between the strong and weak
subsets leading up to hour 6, followed by large variability of downslope wind speeds
in both the strong- and weak-member subsets. The difference between the strong-
and weak-member subset means (heavy lines) increases from less than 5 m s1 at
the initial time to greater than 28 m s1 for the 6-h forecast.
To better understand the source of error growth, consider vertical cross sections
located above the line segment labeled AA’ in Fig. 9.15. Figure 9.17 shows the
6-h forecast of the strong- and weak-subset means along that cross section. For
the strong subset (Fig. 9.17b, d), a tongue of high winds extends from the mid-
troposphere down the lee-slope of the Sierra Nevada Mountains and into the Owens
Valley. Associated with this strong downslope flow is a large amplitude mountain
wave as indicated by the large vertical perturbation of isentropes and the strong
vertical velocities of nearly 14 m s1 in the core of the updraft. Furthermore, an
extensive region of wave breaking is indicated by the strongly decelerated flow and
large area of turbulent mixing between 8 and 12 km. For this subset, the wave
breaking is associated with the generation of the strong downslope flow on the
Fig. 9.16 The evolution of

the zonal wind during the
simulation averaged for the
10-strongest (black) and
10-weakest (grey) ensemble
members over the
Owens-Valley metric box.
The thick line shows the mean
of each 10-member subset
(From Reinecke and Durran
2009b. © American
lee-slope (Clark and Peltier 1977; Peltier and Clark 1979). In contrast, the weak-
member subset (Fig. 9.17a, c) is characterized by high zonal-momentum air that
does not extend below crest level. However, a small region of decelerated flow, with
zonal winds less than 10 m s1 , is evident near 12-km ASL along with relatively
weak turbulent mixing with the TKE barely exceeding 10 m2 s2 between 8 and
10 km’s. While upper-level wave breaking is apparent in this subset, it is weaker
and less extensive than in the strong-member subset, and does not correspond to a
severe downslope windstorm.
It is important to note that the difference between the strong and weak downslope
wind response is due to major structural differences in the mountain wave as
opposed to small shifts in the leeward extent of some region of high surface winds.
These large differences occur even though the upstream conditions are very similar.
For example, the zonal momentum upstream of the Sierra Nevada crest is nearly
indistinguishable between the strong and weak ensemble members (compare the left
sides of Fig. 9.17a, b). Forward shear is apparent in both examples with the zonal-
wind increasing from 10 m s1 near crest level to 40 m s1 near the tropopause.
The stability of the upstream profiles is also very similar between strong and weak
members (Fig. 9.17c, d). A layer of strong crest-level stability is present and the
undisturbed tropopause height is nearly identical for both subsets. Furthermore, the
differences cannot be attributed to the presence or absence of upper-tropospheric
wave breaking since both subsets contain some degree of breaking. Despite these
similarities, the difference between the mountain waves and downslope wind
forecast for the two subsets is considerable suggesting very strong sensitivity to
the model initial conditions and predictive time scales of less than the 6-h length of
the forecast.
a b
−1
c d
−1
Fig. 9.17 The (a, b) zonal wind u (shaded) and TKE (heavy contours) as well as the (c, d)
vertical velocity w (shaded) and potential temperature ™ (heavy contours) along a vertical cross-
section across the Sierra Nevada mountains for the (a, c) weakest and (b, d) strongest ten ensemble
members for the 6-h forecast. The contour interval is 10 m s1 for u, 10 m2 s2 for TKE, 10 K for ™,
and 4 m s1 for vertical velocity. The zero contour of vertical velocity is omitted (From Reinecke
and Durran 2009b. © American Meteorological Society. Reprinted with permission)
9.9.2 Model Errors
In Section (9.4) it was demonstrated that discretization errors associated with

poorly resolved topographic features can lead to a significant overamplification of
nonhydrostatic mountain waves. In this section, an example from the T-REX will
be given to demonstrate the impact of model errors associated with poor numerical
resolution on the predictability of mountain waves and downslope winds.
A similar ensemble approach to that which was discussed in the previous section
will be used to examine the role of discretization errors for flow over the Sierra
Nevada Mountains on 16–17 April, 2006, a period when strong mountain-wave
activity was observed during IOP-13 of the TREX project. We focus on a pair
of 70-member ensemble experiments initialized at 00 UTC, 17 April, 2006. In
both experiments second-order horizontal advection is used on the 27- and 9-km
domains; however, the horizontal advection on the 3-km domain is either second-
or fourth-order accurate. This setup guarantees that the initial conditions for each
experiment are identical. Furthermore, one-way nesting ensures that the flow on the
27- and 9-km domains is consistent between the two experiments.
Figure 9.18 shows the second- and fourth-order ensemble mean 6-h forecast of
w and ™ along the same vertical cross section as presented in Fig. 9.17. A large
amplitude mountain wave is evident in the second-order solution (Fig. 9.18a) with
−1
b
−1
Fig. 9.18 The 6-h ensemble mean forecast of vertical velocity and potential temperature using
(a) second- and (b) fourth-order horizontal advection. The location of the cross-section is shown
in Fig. 9.15b (From Reinecke and Durran 2009a. © American Meteorological Society. Reprinted
with permission)
the ensemble mean vertical velocities exceeding 18 m s1 through a large depth of
the troposphere. In contrast, mountain-wave activity is considerably weaker for the
fourth-order solution (Fig. 9.18b) with the ensemble mean vertical velocities less
than 8 m s1 .
The simulated downslope wind speeds at 10 m AGL at the point labeled ‘DRI
01’ in Fig. 9.15 are shown in Fig. 9.19. Also plotted in Fig. 9.19 is the observed
wind speed data, which has been filtered with a low-pass filter to remove high-
frequency oscillations. Associated with the large-amplitude mountain wave of the
second-order solution is an over-predicted severe downslope wind storm with wind
speeds as much as 30 m s1 greater than observed. In contrast, the relatively weak
mountain-wave in the fourth-order solution leads to much better agreement between
the simulated and observed winds.
In addition to the better agreement between the observations and the fourth-order
solution, the uncertainty associated with the fourth-order downslope wind forecast is
substantially reduced. Figure 9.20 shows the ensemble derived probability densities
of the simulated downslope wind speed at the ‘DRI 01’ station 6-h into the
Fig. 9.19 The observed (dotted) 10-m downslope wind speed at the DRI 01 mesonet station (see
Fig. 9.15). Also plotted is the forecasted ensemble mean 10-m wind speed for the second-order
(solid) and fourth-order (dashed) horizontal advection schemes (From Reinecke and Durran 2009a.
simulation (valid at 06 UTC, 17 April). These probability densities have been

computed by binning each of the 70 ensemble members into 5 m s1 bins and
normalizing the resulting distribution so that the area under the curve is equal
to 1. The second-order solution has a bimodal distribution with one peak centered
on a relatively weak downslope wind event and another peak centered on a
catastrophic downslope wind event. The corresponding vertical velocities range
from 6 m s1 in the weakest member to 28 m s1 in the strongest member (not
shown). Changing to fourth-order advection, and therefore reducing the model error,
decreases the ensemble spread and concentrates the distribution around a relatively
weak downslope wind event with the minimum and maximum vertical velocities
ranging from 4 to 16 m s1 between the weakest and strongest members.
While non-linear effects cannot be neglected, the linear analysis presented in
Sect. (9.4) suggests that the second-order advection is unable to accurately propa-
gate the downstream directed wave energy in the poorly resolved nonhydrostatic
modes. Instead, this energy accumulates above the mountain crest leading to a
significant overamplification of the vertically propagating wave. For this example,
reducing the model error associated with discretization errors by increasing the
advective terms’ order of accuracy improves both the mean forecast of downslope
winds by reducing the model bias, and the predictability of the event by decreasing
the ensemble spread of the downslope wind event.
Fig. 9.20 The ensemble derived probability density function for the 6-h forecast of 10-m
downslope wind speeds at the DRI 01 station. Both the second- (solid) and fourth-order (dashed)
solutions are plotted (From Reinecke and Durran 2009a. © American Meteorological Society.
9.9.3 Predictability of Orographic Precipitation
While ensembles can be used to gain insights into the initial condition sensitivities
of mountain waves and downslope winds, they are also a valuable tool to understand
the predictability of orographic precipitation. Several groups have applied ensemble
methodologies in complex terrain to quantify the forecast uncertainty associated
with mesoscale orographic precipitation forecasts (e.g. Marsigli et al. 2001; Grimit
and Mass 2002; Hohenegger et al. 2008). An example of the forecast uncertainty
of orographic precipitation predictions was given in Grimit and Mass (2002) and
is shown in Fig. 9.21. Plotted is the 3-h accumulated precipitation from a 30-
h five-member ensemble forecast. Each ensemble member was initialized with
a different global model analysis valid at 0000 UTC 13 March, 2000, and the
lateral boundaries were driven by the corresponding global model forecast. The
numerical domain is centered over the complex topography of the United States’
Pacific Northwest region. Apparent in the figure is that the mesoscale detail of the
predicted structure and magnitude of the accumulated precipitation over regions
of complex terrain varies considerably. This forecast uncertainty is a direct result
of the uncertainty in the initial conditions and the uncertainty in the specified
lateral boundary conditions. Slightly different realizations of the synoptic-scale
flow can lead to large variations in the forecasted mesoscale details. While the
individual members show considerable variability and uncertainty, the ensemble
mean (Fig. 9.21a) is able to capture the mesoscale structure of the accumulated
precipitation and provides a better forecast than any of the individual members
(Grimit and Mass 2002).
In another example, Walser et al. (2004) examined the growth of initial condition
errors with a 12-member ensemble for several orographic precipitation forecasts
Fig. 9.21 The 3-h accumulated precipitation (color) and sea-level pressure (contours) over the
Pacific Northwest from a 30-h ensemble forecast initialized by (a) the ensemble mean and (b–f)
various global-model forecasts. The scale of precipitation ranges from 0.1 to 30 mm (From Grimit
and Mass 2002. © American Meteorological Society. Reprinted with permission)
over the European Alps. Their ensemble was initialized with time-lagged forecasts
from a larger scale model, which were then scaled so that the resulting perturbations
were smaller than observational uncertainty. In order to isolate the growth of initial
mesoscale uncertainties in the numerical domain, an identical set of lateral boundary
conditions was used for each ensemble member. They suggest that the nature of
orographic precipitation predictability is strongly case dependent with summertime
convection exhibiting large ensemble spread and thus low predictability and win-
tertime precipitation in a stably-stratified environment exhibiting little ensemble
spread and therefore high predictability (Walser et al. 2004). However, convection
alone is not sufficient to determine the predictability characteristics of orographic
precipitation. For example, in one case considered where moderate convection was
embedded in strong cross-barrier flow, the ensemble spread remained relatively
small throughout the integration. Hohenegger et al. (2006) suggested that the
relatively high predictability for this case is linked to downstream propagation of
the initial errors. They argue that if the initial errors are able to propagate upstream,
or remain stationary, then the predictability of orographic rainfall will be limited;
however, if the errors propagate downstream before they have time to grow to
an appreciable amplitude, the predictability will be controlled by synoptic-scale
uncertainties. Considerably uncertainty exists as to when initial errors will be able
to propagate upstream and this remains an active area of research.
9.9.4 Adjoint Applications
Many fundamental questions in numerical weather prediction are related to aspects

of sensitivity. For example, questions related to the sensitivity of a particular forecast
to the initial state or to specific model parameters are quite common. Addressing
sensitivity issues requires a method to quantify how particular aspects of the forecast
will change based on changes to the model formulation or to the model initial state.
To address the general question of sensitivity, an ensemble of NWP forecasts with
perturbed initial conditions or perturbed model formulations can be used. However,
this approach often requires many model forecasts to fully span all of the possible
forecast outcomes and may be computationally prohibitive. The adjoint allows one
to efficiently calculate the sensitivity of a particular forecast output to changes in
the initial state.
Adjoint techniques have proven useful in a variety of applications in NWP
and atmos-pheric research, including process studies of extratropical cyclones and
cyclogenesis (e.g., Errico and Vukicevic 1992; Langland et al. 1995), analysis error
sources (e.g., Rabier et al. 1996), and targeted observations (see review by Langland
2005). One of the most extensive uses of adjoints to date in meteorology has been in
four dimensional data assimilation applications in numerical models (e.g., Kalnay
2003).
The foundation for the adjoint methodology for meteorological applications
has been established by Marchuk (1974) and LeDimet and Talagrand (1986). The
adjoint, which is the transpose of the forward tangent linear propagator of the
forecast model (see Errico 1997 for a review), provides the capability to find
the sensitivity of a particular forecast diagnostic to changes in the initial state in
a mathematically rigorous and computationally feasible manner. Adjoint models
provide the gradient of a scalar function J of the state vector of a model xt at time
t with respect to the initial state vector of the model xt0 . The model state can be
expressed as a vector. This state vector depends on the initial and lateral boundary
conditions and it follows that
J.xt / D J ŒM.xt 0 / (9.29)
where M is the nonlinear model. For sensitivity applications, J is often referred to

as the response function. In the context of data assimilation, J is referred to as the
cost function and is a measure of the lack of fit between the model state vector and
the observations. The gradient of J with respect to the initial model state can be
approximated as
@J @J
D MT (9.30)
@xt 0 @xt
where M is the tangent linear model of M and the superscript T denotes the
transpose operation. The adjoint model MT is formulated by realizing the transpose
Fig. 9.22 Schematic showing the model components used in an adjoint sensitivity calculation
of the tangent linear model. The gradient of J is computed using the adjoint model,
while the tangent linear model is the basis for constructing the adjoint model. The
adjoint model forcing @J/@xt is straightforward to compute through differentiation
of J with respect to the model state at time t.
A schematic representation of the modeling components that are used for an
adjoint sensitivity calculation are shown in Fig. 9.22. The nonlinear forward model
is run from the initial time t0 to forecast time t, and the basic state trajectory is
saved at a predetermined frequency (light blue arrows). Based on the nonlinear
model trajectory, the adjoint model is run backwards to either t0 or an intermediate
time (sometimes referred to as the lead time). Adjoint optimal perturbations can be
formulated based on the adjoint gradients and scaled to a magnitude representative
of typical analysis errors (Errico and Raeder 1999). To address the accuracy of the
tangent linear approximation, these optimal perturbations can be added to the initial
state and evolved with both the tangent linear and nonlinear models. If the tangent
linear assumption is sufficiently valid, the perturbation evolution should be similar
in both models.
Adjoints have been extensively applied for use in four-dimensional variational
data assimilation (4D-VAR) methods in global circulation models (e.g., Rabier
2005), mesoscale models (e.g., Zupanski 1993a; Zou et al. 1995; Honda et al.
2005) and convection-resolving models (e.g., Sun 2005). The appropriateness of the
assumption of linear perturbation growth in the context of the adjoint calculation has
been studied extensively, particularly in regards to the parameterization of subgrid-
scale processes, which are often inherently nonlinear and sometimes discontinuous
(e.g., Errico and Raeder 1999; Fillion and Errico 1997; Zupanski 1993b). A great
deal of research has been done to mitigate these inherent nonlinearities through,
for example, the simplification of the physical parameterizations (e.g., Janiskova
et al. 1999; Lopez and Moreau 2005). The use of adjoints for physical processes
with inherent discontinuities has been addressed as well (e.g., Xu and Gao 1999 and
references therein).
We consider an example of two-dimensional linear hydrostatic gravity waves
excited by flow over a small-scale mountain discussed in detail in Doyle et al.
(2007). The mountain height in this case is 100 m, which leads to relatively
small perturbations resulting in a linear wave response. The initial static stability
is constant in the vertical and corresponds to a Brunt-Väisälä frequency, NBV , of
0.01 s1 . The initial wind speed is specified as 10 m s1 with no vertical shear. The
vertical velocity field obtained from the nonlinear nonhydrostatic model used in
0.10
0.08
0.06
0.04
0.02
0.00
Fig. 9.23 Vertical velocity (m s1 ) at 2.5 h is shown with a contour interval of 0.01 m s1 for a
simulation of flow over a 100-m mountain using the nonlinear model. The positive vertical velocity
is shaded at an interval of 0.01 m s1 . Dashed contours correspond to downward vertical velocity.
The tick marks are shown along the abscissa every 10 km (or ten grid points) (From Doyle et al.
2007. © Meteorologische Zeitschrift. Reprinted with permission)
this example, COAMPS (see Hodur 1997; Doyle et al. 2007), is shown in Fig. 9.23
for the 2.5-h time. The vertical velocity fields exhibit maxima and minima located
directly over the barrier with lines of equal phase tilted upstream. The vertical
wavelength,
z , apparent from the simulated wave field is approximately 6.3 km,
which is in agreement with
z obtained from analytic considerations, 2 c U/NBV or
6.3 km (e.g., Smith 1979), where U is the mean zonal windspeed.
The adjoint model for COAMPS (see Doyle et al. 2007; Amerault et al. 2008) is
applied for the linear hydrostatic gravity wave case in two dimensions. The sensi-
tivity of the kinetic energy at 3.5 h, applied over the 10 1 km2 box along the lee
slope of the mountain, to the u-wind component and potential temperature at 2.5 h
are computed and shown in Fig. 9.24. The u-component perturbations (Fig. 9.24a)
are 90ı out of phase with the potential temperature perturbations (Fig. 9.24b) for
both sensitivity maxima, consistent with the polarization equations for an internal
gravity wave. The larger values in Fig. 9.24b indicate that a 1 K change in the
a 0.4
1 2
0.3
0.2
0.1
0.0
b 0.4
0.3
0.2
0.1
0.0
Fig. 9.24 Sensitivity of the lee side kinetic energy (10 1 km2 box) to the 2.5-h (a) u-wind
component and (b) potential temperature is shown for the hydrostatic wave case based on the
2.5–3.5 h time period. The contour interval is 0.02 m s1 in (a) and 0.02 K in (b) with positive
values shaded using the gray scale. Negative sensitivity values are shown by the dashed contours.
The tick marks are shown along the abscissa every 10 km (or ten grid points) (From Doyle et al.
2007. © Meteorologische Zeitschrift. Reprinted with permission)
potential temperature field in the sensitive region can have a larger impact on the
3.5-h lee-side kinetic energy than a 1 m s1 change to the zonal wind field.
An internal gravity wave has lines of constant phase that move perpendicular to
the air-parcel trajectories and the energy flux or group velocity is oriented normal to
the phase speed (e.g., Smith 1979; Durran 1990). Considering a frame of reference
moving with the flow, the perturbations shown in Fig. 9.24 are comprised of two
superpositioned vertically propagating gravity waves characterized by downward
directed group velocities. Over the 2.5 h interval, these wave packets transport
energy toward the rectangular box in which the kinetic energy response function
is applied. The first wave packet (left packet in Fig. 9.24) is nearly stationary
in the reference frame moving with the flow and is characterized by right and
upward directed lines of constant phase and downward energy propagation. The
second wave packet (right packet in Fig. 9.24) has left- and upward directed phase
lines and also has downward-directed energy propagation. When the adjoint based
perturbations are evolved in the tangent linear model, the two beams propagate
downward into the response function box to maximize the kinetic energy at the
final time. The superposition of two linear gravity waves with downward energy
propagation satisfies the general case for the kinetic energy response function
selected. The results of this gravity wave test case are similar to a St. Andrew’s
Cross gravity wave pattern (e.g., Lighthill 1978). The wave packets can be thought
of two downward propagating branches of two adjacent crosses when viewed from
the forward model perspective.
The tangent linear and nonlinear models were initialized with adjoint-based
perturbations constructed from adjoint sensitivity fields at the 2.5-h time and scaled
following Doyle et al. (2007). The u-wind component perturbation field after 1 h of
integration is shown in Fig. 9.25, corresponding to the 3.5-h time. The perturbations
evolved in the tangent linear (Fig. 9.25a) and nonlinear (Fig. 9.25b) models are
nearly identical with a correlation coefficient of 0.99, which is a testament to the
accuracy of the adjoint and tangent linear models for the dry dynamics portion of
the model, and confirms the validity of the tangent linear assumption for the linear
hydrostatic wave regime. By the 3.5-h time, the perturbation field has contracted in
the vertical with the positive perturbations confined near or below 1 km above the
surface and horizontally positioned within the rectangular box in which the kinetic
energy response function is applied.
An example of the application of an adjoint model to Alpine lee cyclogenesis is
shown in Fig. 9.26 from a study by Vukićević and Raeder (1995). They applied
an adjoint model to provide insight into the sensitivity of a numerical model
forecast of lee cyclogenesis to the initial state. In this example, the largest influence
of the initial surface pressure on the lee cyclone over the 24-h time period is
located well upstream of the Alps over northwest France and south of Great
Britain. This sensitive region is apparently associated with a synoptic-scale trough.
Interestingly, Vukićević and Raeder highlight an additional case that shows the
largest influence of the initial surface pressure on the lee cyclone located over the
Alps and to the southeast. Additional information regarding the interpretation of the
adjoint model results can be found in Langland and Errico (1996) and Vukićević
and Raeder (1996). There have been a number of other adjoint based sensitivity
a 1.0
0.8
0.6
0.4
0.2
0.0
b 1.0
0.8
0.6
0.4
0.2
0.0
Fig. 9.25 The evolved u-wind component (m s1 ) perturbation derived from the adjoint sensitiv-
ity, introduced at the 2.5-h time and evolved to the 3.5-h time using the (a) tangent linear model and
(b) nonlinear model for the hydrostatic wave case. The contour interval is 0.05 m s1 with positive
values shown by the gray scale. Dashed contours correspond to a negative u-wind component. The
box over which the response function is applied in the lee of the crest (10 1 km2 region) is shown
by the white rhombus outline. The tick marks are shown along the abscissa every 10 km (or ten
grid points) (From Doyle et al. 2007. © Meteorologische Zeitschrift. Reprinted with permission)
Fig. 9.26 The adjoint model sensitivity of the 24-h relative vorticity averaged over the box shown
to the initial normalized surface pressure. The contour interval shown is 250 107 s1 mb1
(Adapted from Vukićević and Raeder 1995. © American Meteorological Society. Reprinted with
permission)
studies that provide additional insight into mesoscale predictability (e.g., Errico
and Vukicevic 1992; Langland et al. 1995); however, these studies do not directly
address phenomena forced by complex terrain.
In this chapter, a survey has been presented of numerical modeling techniques and
methods that are relevant for simulating and forecasting phenomena and processes
forced by complex terrain. We have given a flavor of how modeling has evolved
from the early days of numerical weather prediction and provided background on
the numerical techniques used within the modern complex modeling systems being
applied today. A number of studies have highlighted both new numerical methods
and challenges for NWP applications in complex terrain. These include basic issues
related to formulation of model vertical coordinates for sloping terrain (Durran
1999; Zängl 2007) and the rather subtle and related issue of using a consistent
formulation to represent the terrain transform metrics in the advective and pressure
gradient terms (Klemp et al. 2003). Methods have been introduced to improve
numerical simulations over steep terrain due to the pressure gradient representation
(Janjić 1989; Grell et al. 1995) and numerical filtering (Ballard and Golding 1991;
Zängl 2002). Higher-order accuracy methods have been demonstrated to improve
the prediction of nonhydrostatic mountain waves at high resolution (Reinecke and

Durran 2009a).
The rapid progress in our understanding of topographically-forced flows and
processes over the past several decades can be attributed to a large degree to
the advancement in the sophistication and fidelity of numerical models. This
new generation of numerical models are capable of simulating and predicting
orographically-forced phenomena with a degree of skill that can be successfully
used to test theories, simulate conditions found in nature, or perform predictions. In
spite of the significant advances that have been made over the past several decades
in the various components of numerical modeling systems, a number of significant
challenges remain. Recommendations for future directions in mesoscale numerical
modeling over complex terrain related to numerical methods and predictability
issues include:
• The fidelity (e.g., order of accuracy) of numerical methods for the simulation of
topographic flows has been shown to be important in some cases, probably more
than previously appreciated. Additional studies are needed to further assess the
benefits and tradeoffs of more accurate numerical methods for the prediction of
flows over terrain.
• With increasing computational capabilities, opportunities are emerging for the
development of new high-resolution dynamical cores designed for the sphere.
These approaches, such as finite volume, spectral element, and discontinuous
Galerkin methods need to be carefully evaluated for flows over complex terrain
and should build on the previous finite difference research for modeling topo-
graphic flows.
• As model resolutions increase, smaller-scale and steeper terrain features become
explicitly resolved. This raises challenges both in numerical model formulation
(e.g., coordinate systems and discretization methods) and in the treatment of
surface interactions.
• The integration of the capabilities of high-resolution limited area models with
global mesoscale models will become increasingly important as the horizontal
resolution of global models increase to convection and mountain wave permitting
scales.
• The complicated interactions between numerical methods and physical param-
eterizations in complex terrain need to be carefully considered, particularly for
cloud and orographic precipitation processes, as well as for flows influenced by
mountain boundary layer turbulence.
• The predictability of topographic flows including the sensitivity to the initial
state, boundary conditions, and model components is relatively unexplored. A
new generation of predictability tools such as adjoint and ensemble methods
need to be developed and applied to topographically-forced flows to address
basic questions related to the predictability of phenomena forced by the lower
boundary. Opportunities will exist for adjoint and ensemble based sensitivity
calculations to provide new insight into terrain-forced flow dynamics and
processes.
• With the emergence of mesoscale ensembles, new probabilistic-based tools and

metrics are needed for complex terrain flows. The communication of uncertainty
in the forecast process remains a formidable challenge.
• We are facing important questions about the point of diminishing returns as
horizontal resolution increases. The balance between high-resolution determin-
istic prediction and number of ensemble members needs to be explored for
topographically-forced flows.
Acknowledgements We gratefully acknowledge the efforts of the editors for this book. We also
acknowledge the helpful comments of three anonymous reviewers. The first (JDD) and fourth
(PAR) authors acknowledge support through the Office of Naval Research’s Program Element
0601153N. The second author’s (CCE) research was supported by National Science Foundation
Grant ATM-0242228. The third author (AP) acknowledges support from the UK Metoffice in
Exeter, UK, and also would like to acknowledge Professor Norman A. Phillips, Merrimack, NH
for fruitful discussions and for translating the 1903 Swedish newspaper articles into English. The
fifth author (GZ) acknowledges support from the German Weather Service (DWD). Computational
resources were supported in part by a grant of HPC time from the Navy DoD Supercomputing
Resource Center (Navy DSRC) at Stennis, MS. COAMPS® is a registered trademark of the Naval
Research Laboratory.
References
Amerault, C., X. Zou, and J. Doyle, 2008: Mesoscale assimilation of rain affected observations.
J. Applied Remote Sensing, 3, 033531, 1–20.
Anthes, R. A., Y. Kuo, D. P. Baumhefner, R. M. Errico, and T. W. Bettge, 1985: Prediction of
mesoscale atmospheric motions. Adv. in Geophys., S. Manabe, Ed., Academic Press, Vol. 28B,
159–202.
Arakawa A., 2000: Future development of general circulation models. General Circulation Model
Development: Past, Present and Future, D. A. Randall, Ed., Academic Press, 721–780.
Ashford, O.M., 1985: Prophet – or Professor? The life and Work of Lewis Fry Richardson. Adam
Hilger, Bristol and Boston, 303 pp.
Ballard. S. P., and B. W. Golding, 1991: Basic model formulation. Short range forecasting research,
mesoscale documentation paper No. 4. Meteorological Office, Bracknell, UK, 42 pp.
Benjamin S. G., 1989: An isentropic meso˛-scale analysis system and its sensitivity to aircraft and
surface observations. Mon. Wea. Rev, 117, 1586–1603.
Bjerknes, V. 1910: Synoptic representation of Atmospheric Motions, Quart. J. Roy. Met. Soc.
Vol. 36, pp. 267–86.
Bjerknes, V., 1904: Das Problem der Wettervorhersage, betrachtet von Standpunkt der Mechanik
und der der Physik, Meteorologische Zeitschrift, Wien, 21, 1–7. Translated by Y. Mintz in 1954
and published in Shapiro and Grønås, 1999.
Bjerknes, V., 1914: Meteorology as an exact science. Translation of “Die Meteorologie als
exakte Wissenschaft, Antrittsvorlesung gehalten am 8 Januar 1913 in der Aula der Universität
Leipzig”. Month. Weath. Rev. vol. 42, pp.11–14.
Black, T. M., 1994: The new NMC mesoscale Eta model: Description and forecast examples. Wea.
Bleck R., 1978: On the use of hybrid vertical coordinates in numerical weather prediction models.
Mon. Wea. Rev, 106, 1233–1244.
Bleck R., and S. G. Benjamin, 1993: Regional weather prediction with a model combining
terrain-following and isentropic coordinates. Part I: Model description. Mon. Wea. Rev, 121,
1770–1785.
Bolin, B., 1950: On the influence of the earth’s orography on the character of the westerlies, Tellus,
2, 184–195.
Charney, J.G. and A. Eliassen, 1949: A numerical method for predicting the perturbations of the
middle-latitude westerlies, Tellus, 1:2, 38–54.
Charney, J. G. 1955: The use of the primitive equations of motion in numerical weather prediction,
Tellus, 7, pp. 22–26.
Clark, T. L. and W. R. Peltier, 1977: On the evolution and stability of finite amplitude mountain
waves. J. Atmos. Sci., 34, 1715–1730.
Colle, B. A. and C. F. Mass, 1998: Windstorms along the western side of the Washington Cascade
Mountains. Part I: a high-resolution observational and modeling study of the 12 February 1995
event. J. Atmos. Sci., 126, 28–52.
Colle, B.A., 2008: Two-Dimensional Idealized Simulations of the Impact of Multiple Windward
Ridges on Orographic Precipitation. J. Atmos. Sci., 65, 509–523.
Corby, G. A, A. Gilchrist, and R. L. Newson, 1972: A general circulation model of the atmosphere
suitable for long period integrations. Quart. J. Roy. Met. Soc., 98, 809–832.
Davies L. A., and A. R. Brown, 2001: Assessment of which scales orography can be credibly
resolved in a numerical model. Quart. J. Roy. Meteor. Soc., 127, 1225–1237.
Doms, G., and U. Schättler, 2002: A description of the nonhydrostatic regional model LM. Part I:
Dynamics and numerics. Deutscher Wetterdienst, Offenbach, Germany, 134 pp.
Doyle J. D., and Q. Jiang, 2006: Observations and numerical simulations of mountain waves in the
presence of directional wind shear. Quart. J. Roy. Meteor. Soc., 132, 1877–1905.
Doyle, J. D. and M. A. Shapiro, 2000: A multi-scale simulation of an extreme downslope
windstorm over complex topography. Meteor. Atmos. Phys., 74, 83–101.
Doyle, J.D., C. Amerault, C.A. Reynolds, 2007. Sensitivity analysis of mountain waves using an
adjoint model. Meteorologische Zeitschrift, 16, 607–620.
Durran D. R., 1986: Mountain waves. Mesoscale Meteorology and Forecasting, P. S. Ray, Ed.,
Amer. Meteor. Soc., 472–492.
Durran, D. R., 1990: Mountain waves and downslope winds. Atmospheric Processes over Complex
Terrain, W. Blumen, Ed., AMS Monograph, 23, 59–81.
Durran, D. R., 1999: Numerical Methods for Wave Equations in Geophysical Fluid Dynamics.
Springer-Verlag, 465 pp.
Eckermann, S., 2008: Hybrid –p coordinate choices for a global model. Mon. Wea. Rev, 137,
224–245.
Eliassen, A. and E. Raustein, 1968: A numerical integration experiment with a model atmosphere
based on isentropic surfaces. Meteor. Ann., 5(2), 45–63.
Eliassen, A. and J.E. Rekustad, 1971: A numerical Study of Meso-Scale Mountain Waves,
Geofysiske Publikasjoner, 28, 13 pp.
Errico, R . M., 1997: What is an adjoint model? Bull. Amer. Meteor. Soc., 78, 2577–2591.
Errico, R . M., T. Vukicevic, 1992: Sensitivity analysis using an adjoint of the PSU-NCAR
mesoscale model. Mon. Wea. Rev. 120, 1644–1660.
Errico, RM, KD Raeder, 1999: An examination of the accuracy of the linearization of a mesoscale
model with moist physics. Q.J. Roy. Meteor. Soc., 125, 169–195.
Evensen, G., 1994: Sequential data assimilation with a nonlinear quasi-geostrophic model using
Monte Carlo methods to forecast error statistics. J. Geophys. Res., 99 (C5), 10 143–10162.
Evensen, G., 2003: The ensemble Kalman Filter: theoretical formulation and practical implemen-
tation. Ocean Dynamics, 53, 343–367.
Fels S. B., J. D. Mahlman, M. D. Schwarzkopf, and R. W. Sinclair, 1980: Stratospheric sensitivity
to perturbations in ozone and carbon dioxide: Radiative and dynamical response. J. Atmos. Sci.,
37, 2265–2297.
Ferziger, J. H. and M. Perić, 2002: Computational Methods for Fluid Dynamics. 3 rd ed. Springer-
Verlag, 423 pp.
Fillion, L, R. M. Errico, 1997: Variational assimilation of precipitation data using moist convection
parameterization schemes: A 1DVAR study. Mon. Wea. Rev., 125, 2917–2942.
Foldvik, A. and M.G. Wurtele, 1967: The computation of the transient gravity wave, Geophys. J.
vol. 14, pp. 161–185
Friedman, R.M., 1989: Appropriating the Weather: Vilhelm Bjerknes and the Construction of a
Modern Meteorology, Cornell University Press, 251 pp.
Gal-Chen, T. and R. C. Somerville, 1975: On the use of a coordinate transformation for the solution
of Navier–Stokes equations. J. Comput. Phys., 17, 209–228.
Gallus, W. A., Jr., and J.B. Klemp, 2000: Behavior of flow over step orography. Mon. Wea.
Rev., 128, 1153–1164.
Gallus, W. A., Jr., and M. Rancic, 1996: A nonhydrostatic version of the NMC’s regional eta model.
Garvert M. F., B. Smull, and C. Mass, 2007: Multiscale mountain waves influencing a major
orographic precipitation event. J. Atmos. Sci., 64, 711–737.
Gary, J. M., 1973: Estimate of truncation error in transformed coordinate, primitive equation
atmospheric models. J. Atmos. Sci., 30, 223–233.
Gill, A. E., 1982: Atmosphere-Ocean Dynamics. Academic Press, 662 pp.
Grell, G. A., J. Dudhia, and D. R. Stauffer, 1995: A description of the fifth-generation Penn
State/NCAR mesoscale model (MM5). NCAR Tech. Note NCAR/TN-398 C STR, 122 pp.
Grimit, E. P. and C. F. Mass, 2002: Initial results of a mesoscale short-range ensemble forecasting
system over the Pacific Northwest. Weather and Forecasting, 17, 192–205.
S.A. Cohn, S. Vosper, M. Weissman, S. Haimov, S. De Wekker, L. Pan, F.K. Chow, 2008:
The Terrain-Induced Rotor Experiment: A field campaign overview including observational
highlights. Bull. Amer. Meteor. Soc., 89, 1513–1533.
Grubišić, V., and B.J. Billings, 2008: Climatology of the Sierra Nevada Mountain-Wave Events.
Mon. Wea. Rev., 136, 757–768.
Hamill, T. M., 2006: Ensemble-based atmospheric data assimilation: A tutorial. Predictability of
weather and climate, T. Palmer and R. Hagedorn, Eds., Cambridge University Press, 124–156.
Harper, K.C., 2008: Weather by the Numbers, the Genesis of Modern Meteorology, The MIT Press,
Cambridge (Mass.), London. 308 pp.
Hinkelmann, K. 1959: Ein numerisches Experiment mit primitiven Gleichungen (A numerical
experiment with primitive equations) in Bolin, B. The Atmosphere and the Sea in Motion,
Rossby Memorial Volume, New York, 486–500.
Hodur, R. M., 1997: The Naval Research Laboratory’s Coupled Ocean/Atmosphere Mesoscale
Prediction System (COAMPS). Mon. Wea. Rev., 125, 1414–1430.
Hohenegger, C., A. Walser, W. Langhans, C. Schar, 2008: Cloud-resolving ensemble simulations
of the August 2005 Alpine flood. Quarterly Journal of the Royal Meteorlogical Society, 134,
889–904.
Hohenegger, C., D. Luthi, and C. Schar, 2006: Predictability mysteries in cloud resolving models.
Mon. Wea. Rev., 134, 2095–2107.
Holmboe J., and H. Klieforth, 1957: Investigation of mountain lee waves and the air flow over
the Sierra Nevada. Final Rep., Department of Meteorology, UCLA, Contract AF 19(604)-728,
283 pp.
Holton, J. R., 1993: The Second Haurwitz Memorial Lecture: Stationary planetary waves. Bull.
Amer. Meteor. Soc., 74, 1735–1742.
Holton, J. R., 2004: An Introduction to Dynamic Meteorology (Fourth Edition). Elsevier Academic
Press, 535 pp.
Honda, Y., M. Nishijima, K. Koizumi, Y. Ohta, K. Taniya, T. Kawabata, T. Tsuyuki, 2005:
A pre-operational variational data assimilation system for a nonhydrostatic model at the Japan
Meteorological Agency: Formulation and preliminary results. Quart. J. Roy. Meteor. Soc., 131,
3465–3475.
Hsu Y-J. G., and A. Arakawa, 1990: Numerical modeling of the atmosphere with an isentropic
vertical coordinate. Mon. Wea. Rev, 118, 1933–1959.
Janiskova, M., J.-N. Thepaut, J-F. Geleyn, 1999: Simplified and regular physical parameterizations
for incremental four-dimensional variational assimilation. Mon. Wea. Rev., 127, 26–45
Janjić, Z. I., 1989: On the pressure gradient force error in ¢–coordinate spectral models. Mon. Wea.
Rev., 117, 2285–2292.
Janjić, Z. I., 1990: The step-mountain coordinate: Physical package. Mon. Wea. Rev.,118,
1429–1443.
Janjić, Z. I., J. P.Gerrity, and S. Nickovic, 2001: An alternate approach to nonhydrostatic modeling.
Mon. Wea. Rev., 129, 1164–1178
Johnson D. R., A. J. Lenzen, T. H. Zapotocny, and T. K. Schaack, 2000: Numerical uncertainties
in the simulation of reversible isentropic processes and entropy conservation. J. Climate, 13,
3860–3884.
Johnson D. R., A. J. Lenzen, T. H. Zapotocny, and T. K. Schaack, 2002: Numerical uncertainties in
the simulation of reversible isentropic processes and entropy conservation: Part II. J. Climate,
15, 1777–1804.
Kalnay, E., 2003: Atmospheric Modeling, Data Assimilation and Predictability, Cambridge.
Cambridge University Press, 341 pp.
Klemp, J., and R. Wilhelmson, 1978: The simulation of three-dimensional convective storm
dynamics. J. Atmos. Sci., 35, 1070–1096.
Klemp J. B., W. C. Skamarock, and O. Furhrer, 2003: Numerical consistency of metric terms in
terrain-following coordinates. Mon. Wea. Rev., 131, 1229–1239.
Kutzbach, G. 1979: The Thermal Theory of Cyclones; A History of meteorological thought in the
nineteen century, Am Met Soc, Boston Mass, 255.
Kwa, C. 2001: The Rise and Fall of Weather Modification, in Changing the Atmosphere: Expert
Knowledge and Global Environmental Governance, eds. C.A. Miller and P.N. Edwards,
Cambridge, MA: MIT Press. 135–165.
Langland, R. H, 2005: Issues in targeted observing. Quart. J. Roy. Meteor. Soc., 131, 3409–3425.
Langland, R.H., R. L. Elsberry, R. M. Errico, 1995: Evaluation of physical processes in an idealized
extratropical cyclone using adjoint sensitivity. Quart. J. Roy. Meteor. Soc., 121, 1349–1386.
Langland, R.H., R.M. Errico, 1996: Comments on “Use of an adjoint model for finding triggers
for Alpine lee cyclogenesis”. Mon. Wea. Rev., 124, 757–760.
LeDimet F. X., and O. Talagrand, 1986: Variational algorithm for analysis and assimilation of
meteorological observations: Theoretical aspects. Tellus, 38A, 97–19.
Lewis, J.M. 1998: Clarifying the Dynamics of the General Circulation: Phillips’s 1956 Experiment,
Bull. Am. Met. Soc. Vol. 79, no.1, pp. 39–60
Li, J.-G., and B. W. Atkinson, 1999: Transition regimes in valley airflows. Boundary Layer Meteor.,
91, 385–411.
Lighthill, M. J., 1978: Waves in Fluids. Cambridge University Press, Cambridge, UK, 504 pp.
Lindzen, R.S., E.N. Lorenz, and G.W. Platzman, 1990: The atmosphere, a challenge: the science
of Jule Gregory Charney. Amer. Meteor. Soc., Boston, Mass.
Lopez, P., E. Moreau, 2005: A convection scheme for data assimilation: Description and initial
tests. Quart. J. Roy. Meteor. Soc., 131, 409–436
Lorenz, E. N., 1969: The predictability of a flow which possesses many scale of motion. Tellus, 21,
289–307.
Lorenz, E.N., 1960: Maximum Simplification of the Dynamic Equations, Tellus, vol. 12,
pp. 243–54.
Lorenz, E.N.,1966: The circulation of the Atmosphere, The Am. Scientist, vol. 54, no4, 402–418.
Lynch P. 1992: Richardson’s Barotropic Forecast: A Reappraisal, Bull. Am. Met. Soc. 73 no 1
pp. 35–47
Lynch, P. 1999: Richardson’s Marvellous Forecast, in Shapiro and Grønås, 1999, pp 61–73
Mahrer, Y., 1984: An improved numerical approximation of the horizontal gradients in a terrain-
following coordinate system. Mon. Wea. Rev., 112, 918–922.
Marchuk, G.I., 1974: The numerical solution of problems of atmospheric and ocean dynamics.
Rainbow Systems. First published in Gidrometeoizdat, 387 pp., 1967.
Marsigli, C., A Montani, F Nerozzi, T Paccagnelia, S. Tibaldi, F Molteni, and R. Buzza 2001:
A strategy for high-resolution ensemble prediction. II: Limited-area experiments in four Alpine
flood events. Quarterly Journal of the Royal Meteorological Society, 127, 2095–2115.
Mass, C.F., D. Ovens, K. Westrick, and B.A. Colle, 2002: Does Increasing Horizontal Resolution
Produce More Skillful Forecasts? Bull. Amer. Meteor. Soc., 83, 407–430.
Mesinger F., and A. Arakawa, 1976: Numerical Methods Used in Atmospheric Models. GARP
Publication Series, Vol. 17, World Meteorological Organization, 64 pp.
Mesinger, F, and Z. I. Janjić, 1985: Problems and numerical methods of incorporation of mountains
in atmospheric models. Large-Scale Computations in Fluid Mechanics, Part 2, Lectures in
Applied Mathematics, Vol. 22, Amer. Math. Soc., 81–120.
Mesinger, F., 1982: On the convergence and error problems of the calculation of the pressure
gradient force in sigma coordinate models. Geophys. Astrophys. Fluid Dyn., 19, 105–117.
Mesinger, F., 1984: A blocking technique for representation of mountains in atmospheric models.
Riv. Meteor. Aeronaut., 44, 195–202.
Mesinger, F., Z. I. Janjić, S. Nickovic, D. Gavrilov, and D. G. Deaven, 1988: The step mountain
coordinate: Model description and performance for cases of alpine cyclogenesis and for a case
of an Appalachian redevelopment. Mon. Wea. Rev., 116, 1493–1518.
Mihailović, D. T., and Z. I. Janjić 1986: Comparison of methods for reducing the error of the
pressure gradient force in sigma coordinate models. Meteor. Atmos. Phys., 35, 177–184.
Ogura, Y. and N. A. Phillips, 1962: Scale analysis of deep and shallow convection in the
atmosphere. J. Atmos. Sci., 19:173–179.
Paegle, J., R.A. Pielke, G.A. Dalu, W. Miller, J.R. Garratt, T. Vukicevic, G. Berri, and M. Nicolini,
1990: Predictability of flows over complex terrain. Atmospheric Processes over Complex
Terrain, Meteor. Monogr., No. 45, Amer. Meteor. Soc., 285–299.
Peltier, W. R. and T. L. Clark, 1979: The evolution and stability of finite-amplitude mountain waves
II: Surface wave drag and severe downslope windstorms. J. Atmos. Sci., 36, 1498–1529.
Persson, A., 2005a: Early Operational Numerical Weather Prediction outside the USA: An
historical Introduction, Part I: Internationalism and engineering: NWP in Sweden 1952-69,
Meteorological Applications, vol.12, no. 4, pp.135–59.
Persson, A. 2005b: Early Operational Numerical Weather Prediction outside the USA: An histor-
ical Introduction, Part II: Twenty Countries around the World, Meteorological Applications,
vol.12, no. 4, pp. 269–89.
Persson, A. 2005c: Early Operational Numerical Weather Prediction outside the USA: An
historical Introduction, Part III: Endurance and mathematics - British NWP, 1948-1965.
Meteorological Applications, Vol. 12, no. 4, 381–413.
Persson, A., 2000: L.F.Richardson’s inspirational source, Weather, vol. 55.no 12, pp.466–67
Phillips N. A., 1957: A coordinate system having some special advantages for numerical
forecasting. J. Meteor., 14, 184–185.
Phillips, N. A., 1956: The general circulation of the atmosphere: a numerical experiment, QJRMS,
82, 123–64.
Platzman, G.W., 1967: A retrospective View of Richardson’s Book on Weather Prediction, Bulletin
of the American Meteorological Society vol. 48, pp. 514–550.
Rabier, F, E. Klinker, P. Courtier, A. Hollingsworth, 1996: Sensitivity of forecast errors to initial
conditions. Quart. J. Roy. Meteor. Soc. 122, 121–150.
Rabier, F., 2005: Overview of global data assimilation developments in numerical weather-
prediction centres. Quart. J. Roy. Meteor. Soc., 131, 3215–3233.
Randall, D. A., editor, 2000: General circulation model development: Past, present and future.
International Geophysics Series, Vol. 70, Academic Press, New York.
Reinecke, P. A. and D. R. Durran, 2009a: The over-amplification of gravity waves in numerical
solutions to flow over topography. Mon. Wea. Rev., 137, 1533–1549.
Reinecke, P. A., and D. R. Durran, 2009b: Initial condition sensitivities and the predictability of
Richardson, L.F. 1922: Weather Prediction by Numerical Process, Cambridge University Press,
reprinted by Dover Publications, New York, 1965 with an added introduction by Sydney
Chapman.
Schär, C., D. Leuenberger, O. Fuhrer, D. Lüthi, and C. Girard, 2002: A new terrain-following
vertical coordinate formulation for atmospheric prediction models. Mon. Wea. Rev., 130,
2459–2480.
Shapiro, M. and S. Grønås (Eds), 1999: The Life Cycles of Extratropical Cyclones, American
Meteorological Society, Boston, 355 pp.
Shaw, W.N., 1914: Principia Atmospherica: a study of the circulation of the atmosphere. Mon.
Wea. Rev., 42, 196–209.
Simmons A. J., and D. M. Burridge, 1981: An energy and angular momentum conserving vertical
finite-difference scheme and hybrid vertical coordinates. Mon. Wea. Rev., 109, 758–766.
Simmons A. J., and R. Strüfing, 1983: Numerical forecasts of stratospheric warming events using
a model with a hybrid vertical coordinate. Quart. J. Roy. Meteor. Soc., 109, 81–111.
Skamarock, W.C., J.B. Klemp, 1992: The stability of time-split numerical methods for the
hydrostatic and nonhydrostatic elastic equations. Mon. Wea. Rev., 120, 2109–2127.
Smagorinsky, J. 1965: Numerical results from a nine-level general circulation model of the
atmosphere, Mon. Wea. Rev., 93, 727–68.
Smagorinsky, J., 1953: The dynamical influence of large-scale heat sources and sinks on the quasi-
stationary mean motions of the atmosphere, QJRMS, 79, 342–66
Smith, R.B., 1979: The influence of mountains on the atmosphere. Advances in Geophysics 21,
87–230.
Sun, J., 2005: Convective-scale assimilation of radar data: Progress and challenges. Quart. J. Roy.
Meteor. Soc., 131, 3439–3463.
Vukićević, T., and K. Raeder, 1995: Use of an adjoint model for finding triggers for Alpine lee
cyclogenesis. Mon. Wea. Rev., 123, 800–816.
Vukićević, T., and K. Raeder, 1996: Reply to comments on “Use of an adjoint model for finding
triggers for Alpine lee cyclogenesis”. Mon. Wea. Rev., 124, 761–763.
Walser, A., D. Luthi, and C. Schar, 2004: Predictability of precipitation in a cloud-resolving model.
Mon. Wea. Rev., 132, 560–577.
Webster S., J. T. Thuburn, G. Hoskins, and M. Rodwell, 1999: Further development of a hybrid-
isentropic GCM. Quart. J. Roy. Meteor. Soc, 125, 2305–2331.
Wicker, L. J. and W. C. Skamarock, 2002: Time-splitting methods for elastic models using forward
time schemes. Mon. Wea. Rev., 130, 2088–2097.
Williamson, D. L. and R. Laprise, 1998: Numerical approximations for global atmospheric general
circulation models. In Numerical modelling of the global atmosphere for climate prediction.
P. Mote and A. O’Neill, editors, Kluwer Academic Publishers, Dordrecht.
Xu, Q., J. Gao, 1999: Generalized adjoint for physical processes with parameterized discontinu-
ities. Part VI: Minimization problems in multidimensional space. J. Atmos. Sci., 56, 994–1002.
Zängl G., 2007: An adaptive vertical coordinate formulation for a nonhydrostatic model with flux-
form equations. Mon. Wea. Rev., 135, 228–239.C
Zängl, G., 2002: An improved method for computing horizontal diffusion in a sigma-coordinate
model and its application to simulations over mountainous topography. Mon. Wea. Rev., 130,
1423–1432.
Zängl, G., 2003: A generalized sigma coordinate system for the MM5. Mon. Wea. Rev., 131,
2875–2884.
Zängl, G., L. Gantner, G. Hartjenstein, and H. Noppel, 2004: Numerical errors above steep
topography: A model intercomparison. Met. Z., 13, 69–76.
Zou, X., Y.-H. Kuo, Y.-R. Guo, 1995: Assimilation of atmospheric radio refractivity using a
nonhydrostatic adjoint model. Mon. Wea. Rev., 123, 2229–2249.
Zupanski, D. 1993b: The effects of discontinuities in the Betts-Miller cumulus convection scheme
on four-dimensional variational data assimilation. Tellus, 45A, 511–524.
Zupanski, M., 1993a: Regional four-dimensional variational data assimilation in a quasi-
operational forecasting environment. Mon. Wea. Rev., 121, 2396–2408.
Chapter 10
Meso- and Fine-Scale Modeling over Complex
Terrain: Parameterizations and Applications
Shiyuan Zhong and Fotini Katopodes Chow
Abstract This chapter discusses the current status, success, and especially chal-
lenges of applying mesoscale numerical models to simulate atmospheric processes
over areas of complex terrain. These include thermally-induced circulations, gap
flows, mountain waves, and boundary layer structure and evolution in mountainous
regions. The choice of model configuration (e.g. choice of coordinate system,
horizontal and vertical resolution, grid nesting, and lateral boundary conditions)
and physical parameterizations (e.g. boundary layer, land surface, and radiation
parameterizations) may affect the performance of mesoscale models in complex
terrain. Application of large-eddy simulation (LES) to complex-terrain processes is
also discussed. Examples of model simulations related to several recent field studies
in mountainous areas are used to illustrate the issues and challenges.
10.1 Introduction
In the past two decades since the publication of the first comprehensive review
of atmospheric processes over complex terrain (Blumen 1990), a phenomenal
growth in mesoscale numerical modeling has been observed, due mainly to the
proliferation of relatively inexpensive high-performance computers, massively-
parallel architecture, and distributed-memory codes. The unparalleled growth is also
attributed to the significant effort in community model development with large user
groups. Increasingly finer model resolution has enabled more accurate depiction
S. Zhong ()
Department of Geography, Michigan State University, East Lansing, MI 48824, USA
e-mail: zhongs@msu.edu
F.K. Chow
Department of Civil and Environmental Engineering, University of California, Berkeley,
CA 94720-1710, USA
e-mail: tinakc@berkeley.edu

592 S. Zhong and F.K. Chow
Table 10.1 List of atmospheric mesoscale-scale numerical models commonly used in

research and operational regional weather prediction
Models Names Reference
ARPS Advanced Regional Prediction System Xue et al. (2000, 2001)
COAMPS Coupled Ocean–Atmosphere Mesoscale Hodur (1997)
Predicting System
COSMO Consortium for Small Scale Modeling Steppeler et al. (2003)
(Lokal Modell)
HIRLAM High Resolution Limited Area Model Kăllberg (1990)
GEM-LAM Global Environmental Multiscale, Local Côté (1998a, b)
Area Model
MC2 Mesoscale Compressible Community Benoit et al. (1997)
model
Meso-Eta Mesoscale Eta Model Janjic (1994), Black
(1994)
Meso-NH Mesoscale Non-Hydrostatic Model Lafore et al. (1997)
MM5 Penn State-NCAR Fifth-Generation Grell et al. (1994)
Mesoscale Model
RAMS Regional Atmospheric Modeling System Pielke et al. (1992)
WRF Weather and Research Forecasting Skamarock and Klemp
Model (2008)
Unified Model UK Met Office Unified Model Cullen (1993)
of underlying topography and associated vegetation and landuse, leading to better

predictions of local and regional terrain-induced atmospheric circulations over areas
of complex terrain. A direct result of these advances has been the improvement
in real-time operational forecasting using mesoscale models in many coastal and
mountainous regions throughout the world. The superior performance of mesoscale
numerical predictions over large-scale operational models has been demonstrated
for the Intermountain West (Horel and Gibson 1994) and the Pacific Northwest
(Colle et al. 1999; Mass and Kuo 1998; Mass et al. 2003) of the United States.
Other topographically complex regions in the world, such as the Alps of Europe, the
Andes of South America, and Antarctica, have also benefited from enhanced skills
in regional forecasting by mesoscale models (Turner et al. 2000; Seluchi and Chou
2001). A comprehensive review of these mesoscale numerical weather prediction
efforts and their success in regional weather and climate studies is given in Cotton
et al. (1994) and Mass and Kuo (1998). A list of community mesoscale models that
are widely employed for operational weather prediction is provided in Table 10.1.
In addition to improving operational regional weather predictions, mesoscale
models have been a pivotal research tool for studying local and regional atmospheric
processes, especially those associated with complex topography where observations
are either lacking or at best sparse. Numerous investigators have successfully
applied mesoscale models to simulate a range of complex terrain processes such
as slope flows (Noppel and Fiedler 2002; Schumann 1990; Smith and Skyllingstad
2005; Savage et al. 2008; Zhong and Whiteman 2008), valley winds (Kuwagata and
Kimura 1997; Zängl et al. 2001, 2004; Fast and Darby 2004; Weigel et al. 2006;
10 Meso- and Fine-Scale Modeling over Complex Terrain. . . 593
Chow et al. 2006; Bergström and Juuso 2006; Bischoff-Gauß et al. 2006), plateau
and basin circulations (Whiteman et al. 2000; Zängl and Chico 2006; De Wekker
et al. 1998), gap winds and channeled flows (Whiteman and Doran 1993; Doran and
Zhong 2000), gravity waves and rotors (Poulos et al. 2000; Doyle and Durran 2007),
and cold air pools (Kondo and Okusa 1990; Vrhovec 1991; Vrhovec and Hrabar
1996; Zhong et al. 2003; Zängl et al. 2004; Zängl 2005; Vosper and Brown 2008).
As discussed in the earlier chapters of this book, mesoscale modeling, together with
field observations, has led to significant advancement in the understanding of the
characteristics and formation mechanisms of atmospheric processes over complex
terrain.
Despite the success in operational and research applications of mesoscale models
and the steady and significant improvements in both numerical and physical aspects
of these models in the past two decades, model results still contain bias and errors
that sometimes can be relatively large. As shown in Berg and Zhong (2005),
the errors and uncertainties in mesoscale modeling are usually amplified over
complex terrain compared to over flat terrain. Zhong and Fast (2003) evaluated
high-resolution simulations of three widely used mesoscale models (RAMS, MM5,
and Meso Eta) using observations from the Vertical Transport and Mixing (VTMX)
experiment conducted in the Salt Lake Valley (Doran et al. 2002). They showed
that while the simulated structure and evolution of the valley boundary layer
and circulations systems are in good agreement with the observations, the results
from all three models exhibited lower mixed-layer depths, much weaker nocturnal
inversions, and a consistent cold bias (1–3ıC) both near the surface and throughout
the valley boundary layer. A large cold bias (3–5ıC) in surface temperature was also
found in COAMPS simulations of California’s Central Valley (Chin et al. 2001).
The generally weak nocturnal inversion in model simulations compared to observed
inversions has been identified as a common problem (Hanna and Yang 2001). This
problem is particularly severe in low-lying terrain, such as over the floor of a valley
or basin where the predicted temperature gradient can be three to four times weaker
than observed (Zhong and Fast 2003). In some cases, the models simply do not
maintain or strengthen nocturnal inversions that form on calm, clear nights as a
result of radiative flux divergence. Consequently, vertical mixing occurs too rapidly
as the surface heat flux increases in response to solar heating and the models often
have difficulty in properly capturing the rate of growth of the morning boundary
layer (White et al. 2001; Banta and White 2003). Models also are generally unable
to capture the multiple inversion layers frequently observed in valleys and basins as
a result of differential advection. Although these errors in the atmospheric boundary
layer may sometimes not be considered as significant from a weather forecasting
standpoint where the major focus is on weather systems and precipitation, they can
have substantial impacts on other weather- and climate-related applications such
as air quality, wind energy, and agriculture, especially in topographically complex
regions.
While the success of mesoscale atmospheric numerical models in regional
weather forecasting and research has been discussed in many earlier chapters in
this book, the current chapter will focus on issues and challenges that mesoscale
numerical models face when they are applied to areas of complex terrain. Studies
have attributed the biases and errors in mesoscale predictions to a number of
factors. The most commonly blamed factor is insufficient model horizontal and
vertical resolution. Another error source for mesoscale models is physical param-
eterizations. Section 10.2 will focus on how the model configuration, such as the
choice of coordinate system, horizontal and vertical resolution, grid nesting, and
lateral boundary conditions, may affect the performance of mesoscale simulations.
Section 10.3 will discuss the sensitivity of mesoscale model simulations to the
choice of physical parameterizations, with a focus primarily on boundary layer and
land-surface parameterizations. Parameterization issues associated with simulating
orographic precipitation are discussed in Chaps. 5 and 6. Section 10.4 will be
devoted to large-eddy simulations, an emerging numerical tool for simulation of
flow over complex terrain. Finally, Sect. 10.5 will summarize the challenges and
future needs in modeling complex terrain processes.
10.2 Model Configuration
The performance of a numerical model is largely dependent on the model config-

uration. This is particularly true for flow prediction in areas of complex terrain,
but little guidance exists on which model parameters will dominate under various
settings. Here, an attempt is made to shed some light on the key issues for the user
to consider when simulating flow over complex terrain. The most basic issues and
the first decision points encountered by the user involve the model domain setup,
including grid spacing, domain extent, grid nesting, and lateral boundary conditions.
10.2.1 Horizontal and Vertical Resolution
Selecting the horizontal and vertical resolution is among the first decisions to be
made in model configuration. The horizontal resolution refers to spacing between
grid points for grid-point models or the number of waves for spectral models (see
Chap. 9). Since most mesoscale models are grid point models, “grid spacing” will
be used here.
Thanks to the tremendous advances in computer performance and parallel
architecture, operational mesoscale forecasts now use a typical horizontal resolution
of 10 km. In research applications, mesoscale simulations take further advantage
of grid nesting (see Sect. 10.2.3) and commonly use several grid nesting levels
to lead to a grid spacing of 1 km or less in the finest domain (De Wekker 2008;
Chow et al. 2006; Zhong and Fast 2003). Such significant enhancement in model
resolution enables the models to resolve small-scale features in terrain and land-
use heterogeneities, producing predictions of boundary-layer structure and evolution
that are more detailed and are, in many respects, more accurate.
The model resolution should be chosen based on the types of flow phenomena
that need to be resolved (e.g. thermally-driven vs. dynamically-forced) and the
local terrain variations that affect those flows. Grid spacing needs to be sufficiently
small to resolve desired atmospheric motions, but will become too computationally
expensive if it is made too small. Generally speaking, increasing model resolution
reduces model errors (see Chap. 9, Sect. 9.3 on finite differencing). In practice,
this is not always true because of limitations in physical parameterizations and
input data, as well as other numerical errors that can be introduced due to grid
interactions with steep terrain (see Sect. 10.2.1.4 in this chapter; Mass et al. 2002;
Zängl 2007). For example, grid resolution that far exceeds the resolution of input
data sets, such as land-use or terrain data, may provide little advantage in the
model results (see Sect. 10.3.2.2 below). In addition, grid spacing that is too
fine may violate assumptions of some physical parameterizations. For example,
high resolution can violate assumptions in cumulus parameterization by allowing
a wider range of potential responses in precipitation processes and therefore
increasing errors in model precipitation (Wang and Seaman 1997). Belair et al.
(1998) assessed the impact of changing horizontal resolution from 10 to 1 km
on the simulation of surface and boundary-layer fluxes in a field experiment over
cultivated land and forest regions using aircraft observations. They found that the
high resolution captured smaller scale details in surface flux patterns, but resulted in
little improvement in the accuracy of flux predictions.
10.2.1.1 Horizontal Resolution and the Terrain and Flow Representation
The representation of topography in models depends on two factors: the hori-

zontal resolution of the model and the horizontal resolution of the digital terrain
datasets available. As digital elevation data becomes available at increasingly higher
resolution, the latter is no longer a limiting factor; for example, the USGS digital
elevation data are available at 3 arc second (100 m) spacing globally, and the
USGS National Elevation Data (ned.usgs.gov) are available for the conterminous
US at 1 arc second (30 m) and 1/3 arc-second (10 m). Consequently, how
well models represent terrain is primarily a function of model horizontal resolution.
Figure 10.1 shows an example using the Vancouver, Canada region with horizontal
grid spacings of 16, 4, and 1 km. At 16 km resolution (similar to what is typically
used in operational mesoscale forecasting), the terrain is significantly smoothed
and even major valleys are unresolved. Even at 4 km grid spacing, many of
the small valleys and tributaries are still unresolved by the model terrain. The
comparison of the terrain resolved by different grids in Fig. 10.1 also reveals
several common problems in model terrain representation by coarse resolution,
including the spread of mountain range elevation over too broad of an area and
underestimation of elevation differences between peaks and valleys. Note that for
numerical stability, some level of terrain smoothing may be necessary to remove
small-scale terrain features which cannot be represented well at the given resolution,
and most mesoscale codes include several options for smoothing.
Fig. 10.1 Topography in the vicinity of Vancouver, Canada, as seen by the WRF model with
model horizontal resolution of 16, 4, and 1 km, using 30 arc second digital elevation data (contours
in meters above sea level). The tick marks are 0.2ı latitude and longitude. The dashed lines indicate
a vertical cross section as shown in the bottom panel (Courtesy of Jerome Fast, Pacific Northwest
National Laboratory, operated by Battelle for the U.S. Department of Energy)
It is clear that the resolution determines the ability of the model to adequately
depict flows in narrow valleys and over slopes. There are many model parameters
and inputs that affect model performance in complex terrain, but if the grid
resolution is so coarse that a valley is completely smoothed out from the domain,
it becomes nearly impossible for models to produce accurate results. The smallest
resolved feature is twice the grid spacing or 2x (the Nyquist wavelength) but it
typically takes at least four grid points to define a flow feature, and the representation
would still be questionable. Skamarock (2004) and Skamarock and Klemp (2008)
have suggested 6–8 x as “effective resolution” for models. Therefore simulations
with even 1 km grid spacing would not be able to adequately capture features of
several kilometers in scale and many complex-terrain processes, especially those
forced by thermal effects of the terrain such as slope flows and valley winds, can
be smaller than 1 km in spatial scale. Banta et al. (1996), showed that small-scale,
topographically forced winds less than 2 km in width can have a strong influence on
flow over complex terrain, and thus would require resolution on the order of 100 m.
Other examples of the effects of higher grid resolution include Chen et al. (2004),
who found that increasing the horizontal resolution (to 250-m spacing) improves
wind and potential temperature predictions over the mountains in Utah’s Salt Lake
Valley. Chow et al. (2006) found that increased grid resolution (to 150-m spacing)
also improved numerical simulation results in the Swiss Alps, though only when
land-surface conditions were properly initialized. Brulfert et al. (2005) modeled
flow in two narrow valleys in the French Alps and also needed horizontal grid
spacing as fine as 300 m to represent the complex topography.
Grid spacing employed by mesoscale models will also affect how predictions
can be evaluated using point measurements. For example, Fig. 10.2 depicts actual
elevation and model terrain height at different resolutions. The discrepancy in
Fig. 10.2 A north–south terrain cross section indicated by the dashed line in Fig. 10.1 in the
vicinity of Vancouver, Canada, as seen by the WRF model with horizontal grid spacings of 16, 4,
and 1 km (Courtesy of Jerome Fast, Pacific Northwest National Laboratory, operated by Battelle
for the U.S. Department of Energy)
elevations can contribute to differences in observed and simulated temperature and

humidity profiles, in addition to the spatial variations that limit the winds that are
resolved by the model. For example, predictions of surface temperatures may be
completely wrong if the model thinks the location is at the bottom of a valley, as
opposed to on the peak of a mountain. The discrepancy in terrain elevations also
becomes important at grid nesting boundaries (discussed in Sect. 10.2.3 below) and
when initializing finer domains from coarser resolution data.
10.2.1.2 Vertical Resolution
Just as adequate horizontal resolution is necessary to depict terrain-induced meteo-

rological processes, models must also have sufficient vertical resolution to predict
the vertical structure and evolution of these processes. The vertical grid should be
structured to take advantage of the fact that certain atmospheric features or processes
are usually confined to specific layers in the atmosphere. For example, low-level
jets are confined to the lowest kilometer above ground while jet streams occur
near the tropopause. Terrain-induced circulations are typically confined to about
the height of the mountain tops. Accordingly, model users should place vertical grid
levels where they are most needed. As a default choice, the vertical grid spacing is
typically stretched such that the grid is finest near the lower boundary and coarser
at the top of the domain (see Fig. 9.6 in Chap. 9). This allows more efficient use of
computational points where large gradients are expected within the boundary layer.
When using grid stretching, each vertical grid cell should not be more than 10%
larger (or smaller) than its neighbors. This limitation on the stretching ratio limits
numerical errors that arise when the grid is non-uniform (see Thompson 1984).
In mesoscale models, the vertical spacing is typically chosen to be 10–50 m near
the ground and at least 10 grid points are within the first 1 km above the ground.
In some cases in complex terrain, 10 m grid spacing near the ground may not be
Fig. 10.3 Lidar observations from the center of the Salt Lake Valley, showing up-valley (positive)
flows in the early evening (00–04 UTC) and down-valley (negative) flows during nighttime
and early morning (0700–1600 UTC) on Oct. 17, 2000 (From Banta et al. 2004. © American
enough to resolve terrain-induced circulations, especially at night. For example,

Fig. 10.3 shows observations of valley circulation in the Salt Lake Valley using
a lidar on the valley floor during the VTMX experiment. The valley winds reverse
direction from up-valley (positive) in the afternoon and early evening to down-valley
(negative) at night and early morning. Although both the up- and down-valley flows
have maximum wind speeds near the surface, wind speed decreases with height
much more gradually in the up-valley circulation than in the down-valley flow. A
fine vertical grid with the first level at about 10 m or less is necessary for models
to capture the near-surface peak and the sharp vertical gradient of the down-valley
flows in this case. Compared to down-valley flows, downslope flows are typically
shallower and have a peak oftentimes below 10 m. Thus, to resolve downslope flows,
vertical grid spacing of 1–2 m near the surface would be necessary.
High vertical resolution simulations are extremely computationally expensive to
run because it not only increases the number of total grid points, but also puts a
stringent limitation on the time step that can be used. Fortunately, while high vertical
resolution of 1 m is certainly necessary to resolve the details of small-scale near-
surface process, many mesoscale processes (e.g., the up-valley wind in Fig. 10.3)
can be adequately resolved by a typical near-surface vertical grid spacing of 10–
30 m. For daytime convective boundary layer or mountain wave simulations, even
coarser vertical resolution (50 m) may be sufficient. Zhong and Fast (2003) found
that doubling the vertical resolution in the lower boundary layer produced little
improvement in the simulated mean boundary layer structure in the Salt Lake Valley.
They found that the impact of physical parameterizations, such as boundary-layer
and land-surface parameterizations, appears to be more significant (see Sect. 10.3).
10.2.1.3 Grid Aspect Ratio
The choice of grid aspect ratio (the ratio of horizontal and vertical grid spacing) also
becomes important over steep terrain, as discussed in Chap. 9, Sect. 9.5 on sigma
coordinate errors. If the vertical resolution is made too fine, without correspondingly
refining the horizontal resolution, most mesoscale models will either fail to run or
fail to produce adequate results. The aspect ratio must be consistent with the slopes
of the features of interest. De Wekker (2002) discusses a rule-of-thumb for slope
angle and grid aspect ratio which states that the terrain height difference over a
horizontal grid spacing x should not exceed about three to five times the vertical
grid spacing at the ground. Given a terrain slope ’, the aspect ratio is thus limited
by the relation:

bz
˛ < tan1 (10.1a)
x
where b ranges from 3 to 5. Mahrer (1984) discusses the consequences of grid

aspect ratio selection on pressure gradient errors and suggests that even b D 1 is
problematic. De Wekker et al. (2005) successfully used a horizontal grid spacing
of 333 m and a minimum vertical resolution of 70 m in the Riviera Valley with
terrain-following coordinates in the RAMS model. Chow et al. (2006) used ARPS
(Xue et al. 2003) in LES mode in the Riviera Valley with 350 m spacing and 50 m
vertical resolution at the ground, also with terrain-following coordinates, indicating
that the sensitivity of the results to the aspect ratio is also model dependent.
10.2.1.4 Vertical Coordinate System
Note that in addition to the choice of grid spacing, model results may depend
on the vertical coordinate system. Many vertical coordinate systems have been
developed and, as discussed in Chap. 9, each has its advantages as well as
limitations. Models are typically built on one specific vertical coordinate system
and the users do not have the option to choose which coordinate system to use
for their application. It is, however, important to be aware of the properties of
the coordinate system and potential effects on the simulation results. For example,
Fast (2003) illustrated the impact of the coordinate system choice on thermally-
forced flows computed using both terrain-following (sigma) and step mountain
(eta) coordinates. The largest differences in the simulation results between the
eta and sigma coordinates occurred near the surface and in stable conditions at
night, with much colder nocturnal temperature and stronger downslope and down-
canyon flows in the sigma coordinate system. The sigma or terrain-following
coordinate systems are the most commonly used in mesoscale models. Because the
lowest layers follow the topography, enhanced vertical resolution near the ground
can be easily achieved, enabling better predictions of terrain-induced processes.
Sigma coordinates, however, introduce large errors in calculation of the horizontal
pressure gradient near steep slopes, leading to unrealistic predictions of terrain-
forced circulations (Zängl 2002). Sigma coordinates are also known to vertically
exaggerate lee waves and have difficulty predicting terrain blocking and cold
air damming (Gallus 2000). More detailed discussions about the advantages and
disadvantages of sigma and other vertical coordinate systems are given in Chap. 9,
Sect. 9.7. Alternative gridding options have been proposed for complex terrain (for
example, the immersed boundary method of Lundquist et al. 2010, which eliminates
the terrain-following coordinate in WRF to allow simulation of very steep terrain).
10.2.2 Initial and Boundary Conditions
The need for generalized lateral and surface boundary conditions poses perhaps the
most serious challenge to the general application of mesoscale models over complex
terrain. Care must also be taken at the top boundary to avoid reflection of waves, but
this is usually secondary in comparison to other boundary effects.
10.2.2.1 Initial Conditions
Initial conditions must be specified at t D 0 for all prognostic variables (velocity,

temperature, pressure, moisture, etc.). This can be done using specific values or an-
alytical functions for idealized simulations, and using global or other regional model
outputs or large-scale reanalysis data (e.g., North America Regional Reanalysis
(NARR), Mesinger et al. 2006) for real case simulations. There may be sensitivity to
the initial profiles and particularly the background base state (horizontally uniform
background field that is often specified in mesoscale models). Caution is advised
to make sure that these choices do not artificially affect model results, particularly
for idealized flows. For example, specifying a uniform potential temperature with
height followed by a stable atmosphere at the top of the boundary layer as initial
conditions creates a base state that follows this shape as well. Subsequent potential
temperature profiles may show kinks near the top of the boundary layer due to the
abrupt change in slope in the base state. Test simulations can be run with different
initial conditions to determine the best choice. Idealized simulations over flat terrain
also typically require some kind of initial perturbation (e.g. random noise at the first
grid level) to trigger resolved eddies and turbulent structures. If the domain contains
heterogeneous topography then this type of perturbation is unnecessary.
For realistic initial conditions obtained from large-scale models or analyses, the
process is generally straightforward when using a community mesoscale model
(such as WRF, ARPS, etc.). In general it is wise to allow for spinup time of about
6–12 h for the model to adjust to initial and boundary conditions, and to collect
model data for analysis only after that time period. It may also help to choose the
model start time as either 00 Z or 12 Z, to coincide with times when radiosondes
are launched and when model analysis fields thus contain more field data. Further
discussion on model initialization options can be found in Pielke (2002, Chap. 11).
10.2.2.2 Lateral Boundary Conditions
The performance of mesoscale models depends to a large degree on the quality of

the information supplied at the lateral boundaries. Pre-specified inflow velocities,
velocity gradients, open radiation conditions, or periodicity can be prescribed at
lateral boundaries, but these are only useful for idealized simulations and are of
particularly limited use for simulations over complex terrain. To simulate real
events in the atmosphere, mesoscale models need time-dependent lateral boundary
conditions, which are generally supplied by large-scale or other regional model
outputs. If operational model output or analysis fields are used (e.g. the North
American Mesocale Model (NAM, which is now WRF) or NARR), velocity and
scalar fields are usually available at 3- or 6-h increments. Mesoscale models then
interpolate these values in time to provide the boundary conditions on the domain.
Nutter et al. (2004) and Denis et al. (2003) examined the effect of the frequency of
lateral boundary updates from large-scale external data on mesoscale model results
and found that more frequent updates improved the statistics of the flow results.
Lateral boundary conditions determine large-scale features and evolution over
the mesoscale domain. Because of this, mesoscale model results can be very
sensitive to lateral boundary conditions. Errors in models supplying lateral boundary
conditions may move rapidly into mesoscale model domains and contaminate
mesoscale predictions, thus losing the advantage of the finer resolution on the
mesoscale grid. Inconsistencies in the dynamics and/or physics formulations be-
tween large-scale and mesoscale models can produce spurious waves near the lateral
boundaries. An example of this inconsistency is the difference in vertical resolution
between the large-scale and mesoscale models that may create inconsistencies in
the vertical distribution of momentum, heat, and moisture near the boundaries. To
minimize boundary effects, there should be sufficient distance between the lateral
boundaries and the region of interest. This will ensure that the region of interest is
outside the relaxation and terrain smoothing zones (see Sect. 10.2.3.3), that there
is sufficient distance for smaller-scale flow features to develop, and that reflections
from the boundaries are minimized (Warner et al. 1997).
Simulations over complex terrain also call for special attention to where the
lateral boundaries should be placed with respect to terrain. Chen et al. (2004)
showed that model results in Salt Lake Valley were sensitive to the location of the
domain boundaries. To avoid such problems, it is best that the lateral boundaries be
located over smooth terrain whenever possible (Warner et al. 1997). In addition, as
mentioned earlier, model terrain is often smoothed to a degree depending on model
horizontal resolution. As a result, the terrain elevation may not be fully aligned
between the two models at the boundaries, which may lead to significant errors in the
forecasting of the flow features, especially those related to topography. Deng et al.
(2008) performed simulations for the 2006 Torino Winter Olympics site in northern
Italy using the WRF model and showed that spurious precipitation was produced at
the lateral boundaries in complex terrain. This spurious precipitation was attributed
to an inconsistency in terrain between the nested grids near the boundaries when
the inner-domain terrain is not blended well with the outer-domain terrain at the
nested domain boundaries. A better merge of terrain between nested domains helps
minimize problems like this and most mesoscale models offer various options to
merge the terrain.
10.2.2.3 Top and Bottom Boundary Conditions
The top boundary of the domain is typically placed far above the bottom boundary,
particularly when simulating flow over terrain. Numerical domain boundaries can
reflect internal waves generated by flow over topography, so domain tops of 15–
25 km are commonly used together with a damping layer or radiation boundary
condition. Rayleigh damping refers to a relaxation-type damping that is applied
to the top third (or less) of the domain (Klemp and Lilly 1978; Klemp and
Durran1983). The flow solution is relaxed back towards a homogeneous base state
or analysis field based on observations. Often numerical stability limits (based on
the damping term and choice of time stepping scheme) are reached before enough
damping can be included (Klemp et al. 2008; Zängl 2007). Alternatively, radiation
boundary conditions are designed to allow energy to transfer out of the domain, but
these are often expensive and ineffective in real atmospheric applications. Klemp
et al. (2008) introduced a new, simple Rayleigh damping term for the vertical
velocity in the WRF model which performs better than previous approaches.
Ironically, the only actual physical boundary, i.e., the earth’s surface, poses a
more difficult problem than either the lateral or top boundaries of the domain.
Because the surface beneath the atmospheric boundary layer is rough and the
flow highly turbulent, it is impractical (and currently impossible) to resolve all
of the scales of the fluid motion. Traditionally, atmospheric models use similarity
or log-law type boundary conditions to parameterize the effect of the bottom
roughness on the flow. These are discussed in Sect. 10.3 together with other model
parameterizations.
10.2.3 Grid Nesting
Given the examples in Sect. 10.2.1, it is clear that there must be a balance between
fine grid resolution and computational expense. Grid nesting is a useful means of
reducing computational cost by putting grid cells where they are really needed over
an area of interest. Thus, grid nesting provides a way of scaling down from larger
scales to smaller scales, from a coarse grid to finer grids, in essence “zooming in”
to the area of interest. Results from the coarse simulations are then interpolated to a
finer grid and lateral boundary conditions are provided at the boundaries of the fine
domain throughout the duration of the simulation in a one-way or two-way nesting
procedure.
Fig. 10.4 Example of four nested grids centering on Mt. Tsukuba in Japan, at horizontal resolu-
tions of 15, 5, 1.7 km, and 570 m (From Michioka and Chow 2008. © American Meteorological
10.2.3.1 One-Way Nesting
The simplest approach takes information from the coarse grid and passes it to the
fine grid through a one-way nesting procedure. This nesting process can be repeated
over successive nesting levels. Typically a grid ratio of 3–5 is chosen, such that
each grid is three to five times finer than the previous nesting level (see Warner
et al. 1997). Figure 10.4 gives an example of multiple nesting that has horizontal
grid spacing 15, 5, 1.7 km, and 570 m, respectively. At the lateral boundaries of
the nested grid, prognostic variables from the coarse grid are interpolated to the
fine grid, usually using linear or quadratic interpolation. If the grid ratio is higher
than about five, the forcing data on the boundaries of the fine grid may be too
uniform and thus not provide realistic inputs to the fine grid. Another potential
problem with large nesting ratios is the large inconsistency in topography between
the fine and coarse grids that may lead to spurious solutions especially near the
lateral boundaries of the fine grid (see Sect. 10.2.3.1). Despite these drawbacks and
potential problems, larger coarse to fine grid ratios have been used on occasion,
especially in operational mesoscale forecasting where coarse resolution (30 to
50 km) large-scale model output or analysis products are used to directly drive a
fine resolution (4 to 12 km) grid to gain details over a certain region of interest.
Boundary conditions must be provided for all prognostic variables, typically
including velocities and scalar quantities such as potential temperature and water
vapor. The grid nesting provides time- and space-dependent forcing for these
variables from the coarser grid.
The choice of lateral boundary update intervals is also important to provide
accurate forcing on the domain boundaries (see also Sect. 10.2.3.1). Ideally new
boundary conditions are provided on each coarse grid time step and interpolated
in between, but this usually requires that the coarse and fine grids are run simulta-
neously so data can easily be passed without requiring expensive input/output and
storage to files. Some models (e.g. ARPS) only provide options to output data to files
which are then read in by the finer grid. Since larger data storage capacity is required
when the update interval is smaller, a larger update interval is usually preferred to
conserve computational resources (storage and input/output overhead costs). The
optimal update interval has not been thoroughly investigated, as the interval choice
was highly constrained by practical computing choices in the past. The standard or
usually acceptable update interval for lateral boundaries in mesoscale simulations
is about 1–3 h (see e.g. Nutter et al. 2004 and Denis et al. 2003). For fine-scale
simulations down to 100 m horizontal resolution, Michioka and Chow (2008)
showed that using much finer update intervals (reduced from 1 h to about 5 min) for
the lateral boundaries in a one-way nesting procedure does not largely affect time-
averaged quantities (wind, wind direction, etc.), but strongly affects turbulence and
scalar transport. As the update interval was set to be smaller, the model accuracy
for scalar turbulent fluctuations and ground concentration was improved. This is
consistent with the findings by Deng et al. (2008) on the impact of lateral boundary
update interval on simulated precipitation.
10.2.3.2 Two-Way Nesting
In two-way nesting, the nested finer-domain and the outer coarser-domain interact
at every time step of the outer coarse-grid integration. The main advantage of two-
way nesting is that the fine domain solution can influence the coarse grid results by
feeding the information back to the coarse grid. The fine grid solution is used to
essentially “nudge” the coarse domain prediction towards the values computed over
the fine domain, and can consequently affect results outside of the nesting areas in
the coarse domain. Complex terrain flows are thus permitted to influence and inter-
act with the larger-scale flows to some extent, and two-way feedback is possible.
Updating the values of the coarse domain normally occurs at the time step of
the coarse domain. Flows at the last fine grid time step are used to feed back to the
coarser domain, rather than averaging over several nested grid time steps. Since the
ratio of the time steps between parent and child grids is normally 3–5, this effect
is probably not large. Two-way nested grids must be computed simultaneously as
there is continuous feedback between grid levels.
One disadvantage of existing two-way nesting schemes (such as in WRF and
MM5) is that they do not currently allow the vertical resolution to change between
nested domains. This means that resolution may be limited in the vertical direction
when two-way nesting is applied, because the vertical levels must be suitable for
all the nesting levels. The one-way nesting technique described above easily allows
adjustments in vertical resolution between grids. Two-way nesting with different
vertical grid levels should be possible but will require careful accounting of near-
surface flow properties and interpolation schemes to transfer data between grids.
Indeed, similar problems may arise with two-way nesting in the WRF model, for
example, where the resolution of the topography is allowed to change on each grid.
This means that valleys can become deeper and mountains can become higher on
the finer grid, which changes the flow domain. To account for this, the interpolation
procedure, which allows communication across the coarse and fine grids, is de-
signed to make corrections to conserve mass, usually assuming a quasi-hydrostatic
atmosphere for extrapolation (Skamarock and Klemp 2008; Moeng et al. 2007).
10.2.3.3 Relaxation Zones
Sudden jumps in grid resolution at nested domain boundaries can cause numerical
instabilities; therefore, relaxation zones are applied at the boundaries of the higher
resolution domain to gradually merge the prognostic fields on the high-resolution
domain towards the results from the coarse resolution domain. With a relaxation
boundary condition, a forcing term is added to the right-hand side of the momentum
and scalar equations to smoothly allow a prognostic variable ' to match the desired
value 'e at the boundary:
@'
D Gb .' 'e / (10.1b)
@t
The relaxation coefficient Gb is designed to ramp down from a non-zero value at
the boundary of the domain toward zero at the interior boundary of the relaxation
zone. In this way there is a smoother transition in the meteorological fields between
the coarse and fine resolution domains. The relaxation zone width is typically
chosen to be on the order of 10 grid points; even wider transition zones are
often recommended to allow a smoother transition from the coarse to the fine grid
solutions (Warner et al. 1997).
For simulations in complex terrain, the terrain resolution and how it impacts local
meteorological conditions will also contribute to the potential numerical instabilities
at the boundaries. Sudden jumps in terrain height and/or resolution of terrain
features create problems at the boundary (see also Sect. 10.2.3.1). Some models
employ the same topography dataset for nested grids where the topography in the
fine grid is simply interpolated from the coarse grid so that smoothness is not an
issue. In other models the terrain resolution is allowed to change on each grid,
therefore the finer topography must be smoothed and merged with the coarser grid
topography within the relaxation zone. Heavy terrain smoothing will remove high-
resolution information from the nested domain; therefore, as mentioned previously,
the user must ensure that important topographic features that need to be resolved lie
away from the boundaries and relaxation zones.
In addition to the issues discussed in this section, model configuration also
involves other important aspects such as the choice of numerical discretization
schemes, vertical coordinate system definition, and filters. These and other related
topics are discussed in detail in Chap. 9.
10.3 Parameterizations and Complex Terrain Applications
From the discussions above, it is clear that enhanced resolution alone does not
necessarily translate into a reduction in model biases or errors. Another important
factor in determining the performance of mesoscale models is the choice of
model physical parameterizations. Many physical processes, such as boundary-
layer turbulence and cumulus convection, occur at a scale smaller than the typical
mesoscale model grid cell of 5–10 km. These physical processes, which are not
explicitly resolved by mesoscale models, affect the mean atmospheric motions and
thus must be incorporated in the numerical solution. The approach to represent
these sub-grid scale physical processes using grid-resolved variables is called
parameterization. The main physical parameterizations in mesoscale models include
solar and terrestrial radiation, surface and boundary layer turbulence, land-surface
processes, cumulus convection, cloud microphysics, and precipitation. For each of
these, a variety of parameterization schemes has been developed and the results
vary considerably depending on the conditions to which they are applied. As a
result, mesoscale models typically include multiple parameterization schemes for
each physical process so that users have an option to choose the most appropriate
scheme or a combination of schemes for their specific applications. Although
all parameterizations are essential components of numerical weather prediction
models, mesoscale predictions over complex terrain are especially sensitive to
the parameterization of subgrid-scale physical processes near the earth’s surface;
namely, boundary-layer turbulence and land-atmosphere energy, moisture, and
momentum exchange. Accurate description of these near-surface processes is vital
for models to capture various atmospheric phenomena unique to complex terrain,
such as terrain-induced circulations, temperature inversions, and gravity waves.
In this section, we will focus our discussions on the impact of parameterizations
of boundary-layer turbulence and land-atmosphere exchange of energy, momentum,
and moisture on the performance of mesoscale modeling of complex terrain pro-
cesses. The effects of microphysics and convective parameterizations are discussed
in Chaps. 6 and 7.
10.3.1 Turbulence Parameterizations
10.3.1.1 Turbulence Parameterization in Mesoscale Models
Mesoscale models traditionally use the Reynolds-averaged Navier–Stokes (RANS)

equations to obtain the mean quantities that are of interest in describing the state
of the atmosphere. Mesoscale models directly solve for these mean flow variables
while parameterizing the effects of smaller-scale fluctuations on the mean flow. The
following briefly illustrates the parameterization process using the heat equation as
an example.
First, we separate the flow variables into mean plus fluctuating parts. For
example, for the three velocity components, uj D uN j C uj 0 , j D 1,2,3 and for the
potential temperature, D N C 0 . The Reynolds average (overbar) is defined
either as a long time average or as an ensemble average, and basic properties of
the averaging operator include uN j D uN j , uj 0 D 0, uj 0 2 ¤ 0 and uj 0 0 ¤ 0 (Stull
1988). We then substitute the mean plus fluctuating fields into the thermodynamic
equation or temperature forecasting equation and apply the Reynolds average again
to the resulting equation to obtain:
@N @N @.uj 0 0 /

C uN j D C S (10.2a)
@t @xj @xj
where S represents diabatic heating or cooling source.

Mesoscale models solve for mean variables, but the thermodynamic equation
as written contains the divergence of the unknown covariance or turbulent heat
flux u0 j 0 . Because the number of unknowns in the set of equations is larger
than the number of equations, the equations are unclosed, and this is known as
the turbulence closure problem. A similar issue arises in the Reynolds averaged
momentum equations and all other transport equations. Turbulence parameterization
consists of formulating these turbulent variance and covariance in terms of mean
(known) variables to close the system of equations.
A myriad of parameterization schemes have been developed to represent turbu-
lence in mesoscale numerical models. These schemes are categorized by the order of
closure as well as local vs. non-local, as briefly described below. First-order closures
parameterize turbulence moments (e.g. uj 0 0 ) in terms of mean variables. So-called
K-theory, where turbulent heat, moisture, and momentum fluxes are represented as
proportional to the gradients of the mean variables, with “K” as the proportionality
coefficient, is an example of first-order closure. According to K-theory, the turbulent
heat flux in (10.2a) can be expressed by
@N
uj 0 0 D Kh (10.2b)
@xj
where Kh is the eddy diffusivity for heat. Similar equations can be written for
momentum and moisture fluxes. The closure problem is now reduced to determining
the value of the eddy diffusivities represented in general by K. A variety of
formulations for K have been proposed for the momentum and scalar equations
and most of them describe K as a function of velocity gradients and atmospheric
stability. Because of their simplicity and computational efficiency, K-theory-based
first-order turbulence closures have been widely used in mesoscale modeling,
especially in operational mesoscale predictions (see Table 10.2).
In second-order closures, prognostic equations for second-order moments (e.g.,
uj 0 0 ) are formulated while the third-order moments (triple correlations of fluctuat-
ing quantities, e.g., ui 0 uj 0 0 ) are modeled using known quantities. These prognostic
608
Table 10.2 Key features of turbulence parameterizations commonly used in mesoscale numerical models. In each case, Ÿ represents prognostic variables, K is
eddy diffusivity, q2 is TKE, S is vertical wind shear, u is friction velocity, zi is boundary-layer depth, l is mixing length scale, Ri is the Richardson number and
Ric is its critical value, L is the Obukhov length, N is the Brunt-Väisälä frequency, k is the von Karman constant, and B1 and c are proportionality constants.
Subscript i is the ith model layer, 1 is the first model layer, and 0 is the surface value
Scheme Physical basis Primary closure assumption
.Ri
Blackadar (Blackadar 1979; Zhang K-theory based hybrid scheme: K D K0 C S l 2 cRiRi/
c
, Ri 6 Ric ,
and Anthes 1982) and Pleim and non-local for free convection and K D K0 , Ri > Ric , K0 is a background value
Chang (1992) local for other conditions @
N is mixing coefficient
N 1 i /, m
For free convection: @ti D m.
0
Rzi
defined by m 0
N D w v 1 f.1 Em / Œv1 v .z/d zg1 , Em (0.2) is
z1
entrainment
coefficient
p
i
MRF (PBL scheme used in the K-theory based, non-local scheme: L
K D kws z 1 zzi , p is 2.0, ws D u =.1 C 5 0:1z /,
Medium Range Forecast or MRF explicitly accounts for counter ŒU.z /2
i @ @ @
model) (Hong and Pan 1996) gradient transport in the definition zi D Ric gŒvv .zi / v0
, @t
/ @z
K. @z c /, c D b w0 0 =ws
of flux represents the counter gradient term
@ @
YSU (Yonsei University) (Noh et al. Similar to MRF, with an addition of @t
/ @z@ ŒK. @z c / w0 0 .zi / .z=zi /3 , the second term is
2003; Hong et al. 2006) an explicit treatment of entrainment flux at the capping inversion layer
entrainment at the top of PBL
NOGAPS (Burk and Thompson TKE-based, local scheme K.q; l; /N is complex algebraic function following Yamada and
1989) Mellor (1975), where q is predicted by
0 0
@q 2 @q 2 gw 2q 3
@t
D @z@
q @z 2w0 u0 @u @z
2w0 v0 @v
@z
C 2 v ^l

D 0:2 l; ƒl D 16:6 l
ETA (PBL scheme used in the Eta TKE-based, local scheme K D c l q, q is predicted by an equation similar to above

model) (Janjic 1994) 2 2
1=2
@u @v
K D l2 S @z
C @z
B 1 1 Ri f S (near surface)
Gayno-Seaman (Shafran et al. 2000) TKE-based, local scheme K D c l q, q is predicted by an equation similar to above
l q; N 2 ; S 2
S. Zhong and F.K. Chow
equations for the second-order moments are called the Reynolds stress equations
and include the processes governing the behavior of the turbulent fluxes. Although
second-order closure in theory may seem more sound, it involves solving ten
or more additional partial differential equations at each time step and thus is
computationally much more expensive compared to first-order closures for a three-
dimensional model. The tremendous increase in computational cost for second or
higher-order closures may not necessarily lead to more accurate solutions due to
the lack of appropriate experimental data from which to parameterize higher-order
moments. As a result, although numerous second- or higher-order closure schemes
have been proposed over the past three decades (see e.g. the algebraic stress closure
of Findikakis and Street 1979), few have been implemented in three-dimensional
mesoscale models.
The compromise between the simplicity of the first-order closure and the com-
plexity of the full second-order closures led to the development of closure schemes
that solve partial differential equations for only a few second-order moments while
parameterizing the rest (Holt and Raman 1988). These schemes are called 1.5-order
closures, meaning some variables are parameterized at higher and some at lower
order. Mellor and Yamada (1974) were among the first to test a 1.5-order closure
in which prognostic equations are solved for only turbulent kinetic energy (TKE
or q2 ) and a turbulent length scale (l). The values are then combined to form an
eddy viscosity as the product of a turbulent velocity scale (q) and turbulent length
scale, with appropriate coefficients of proportionality: T / q l. Mellor and Yamada
(1974) found that results from their 1.5-order closure were comparable to those of
full second-order closure, but only at a small fraction of the computational cost.
Alternatively, engineering computations often use an equation for the turbulence
dissipation rate, ", to supply the length scale necessary together with the TKE
equation to compute eddy diffusivities. This approach is known as a k-" scheme
(Harlow and Nakayama 1967). Both the k-" and q2 l 1.5-order approaches are
attractive for modeling complex mesoscale flows because the length-scale equations
recognize the effects on the length scale of turbulence transport and local sources
and sinks, and thus provides a more general parameterization than in first-order
closures. Closure parameters in the equations are empirically determined to give
accurate predictions of various benchmark flows. The Mellor-Yamada closure and
its variations are the most frequently used turbulence closure schemes in mesoscale
models. Applications of the ©-equation to atmospheric models have also been
studied but are not as common (Huang and Raman 1989; Andren 1991; Apsley
and Castro 1997; Freedman and Jacobson 2002).
Because of the large number of empirical coefficients needed, particularly in the
length scale equation, it was later found that good predictions of eddy viscosity
may be obtained by using a prescribed set of turbulence length scales with the
TKE equation (Yamada 1977; Mellor and Yamada 1982). This lowers computational
costs by eliminating one of the prognostic equations to be solved. This approach is
commonly referred to as a 1.5-order TKE-based scheme, or simply TKE scheme,
and has formed the basis for a variety of turbulence closures widely used in
Fig. 10.5 Illustration of local and non-local turbulence mixing. The horizontal dashed lines are
vertical grid levels
mesoscale model predictions. It can also be used as a closure for large-eddy sim-
ulation (see Sect. 10.4.1.2). The variations of the schemes used in different models
mainly come from how the sets of turbulence length scales are determined. In most
mesoscale models, the length scales are determined diagnostically using algebraic
formulations based on similarity theory. Modelers have made various modifications
to traditional neutral algebraic forms for stable and unstable conditions, generally
achieved by extending surface-layer profiles into the bulk of the boundary layer with
limits depending on local stability scales (Lacser and Arya 1986; Delage 1997). In
addition to the uncertainties involving how to determine the proper length scales,
all TKE-based approaches suffer from several other shortcomings, including the
neglect of redistribution of TKE by pressure and the assumption that turbulent
transport is down gradient (Moeng and Wyngaard 1989).
Besides the order of the schemes, turbulence parameterizations can also be
grouped into local, in which the parameterization is dependent on local gradients or
values, and non-local, in which turbulence is a function of larger scale differences
or values. The difference between the local and non-local concepts is illustrated in
Fig. 10.5. For local schemes, turbulent mixing at one model level is only related
to properties at immediately adjacent levels, while non-local schemes consider
exchanges of properties in turbulent flows not only between adjacent levels, but
also many other levels within the entire boundary layer. While the majority of
existing turbulence parameterization schemes are local schemes, the use of non-
local closures is suggested in situations where counter-gradient fluxes are observed
in the boundary layer, as is often the case with convective boundary layers.
Parameterizations designed by Zhang and Anthes (1982), Stull (1984), Wyngaard

and Brost (1984), Troen and Mahrt (1986), Pleim and Chang (1992), Holtslag and
Boville (1993), and Hong et al. (2006) are examples of non-local closures.
Most mesoscale numerical models contain many different types of turbulence pa-
rameterizations. For example, MM5 has seven and WRF now has ten different turbu-
lence parameterization schemes, ranging from simple first-order K-theory schemes
to more sophisticated 1.5-order TKE schemes, and from local to non-local schemes.
The turbulence closure schemes in mesoscale models are also referred to as PBL
(planetary boundary layer) schemes. Higher-order schemes (second order or higher)
are rarely used because the accuracy gained does not justify their extremely high
computational cost. Table 10.2 lists some key features of widely used turbulence
parameterization schemes in mesoscale models. The users of mesoscale models
need to be aware that the performance of these schemes can vary considerably.
Some guidance on what scheme to choose for a particular application may be found
in a number of studies that compare the performance of turbulence schemes with
observations taken in different conditions (Mahfouf et al. 1987; Zhong et al. 2007).
10.3.1.2 Limitations of Turbulence Parameterizations in Complex Terrain
Because many terrain-induced atmospheric processes occur primarily in the bound-

ary layer, the results from mesoscale numerical simulations over areas of complex
terrain are generally more sensitive to turbulence parameterization schemes and the
uncertainties and errors in these parameterizations compared to their counterparts
over flat terrain. In addition, the coefficients used in closure schemes are typically
obtained by calibration with field observation data taken over flat terrain or from
laboratory experimental data under ideal equilibrium flow conditions, and thus are
not necessarily appropriate for complex terrain.
Topography imposes a distinct imprint on turbulence properties by inducing
local circulations and waves and by altering stabilities. In complex terrain, local
conditions can vary significantly from place to place. For example, during the
VTMX field campaign in the Salt Lake Valley, tethersonde observations taken over
gentle slopes revealed strong downslope winds with peak speeds of 5–7 m s1
(Whiteman and Zhong 2008), while on the same nights not far from the slope
over the floor of the valley, winds were generally weak and variable. Canyon
outflows or gap winds are characterized by not only strong vertical wind shears,
but also horizontal shears as wind speeds decreased rapidly away from the canyon
or mountain gap. Lidar observations in the Salt Lake Valley during VTMX showed
that canyon outflows reached speeds of 4–5 m s1 , while regions separated by less
than 4 km reported wind speeds of 3–4 m s1 in the opposite direction (Fast and
Darby 2004). Topography can significantly enhance turbulent mixing. For example,
observations made during the Meteor Crater Experiment (METCRAX) (Whiteman
et al. 2008) showed that nighttime averaged TKE values reached 14 m2 s2 inside
the Meteor Crater, a small enclosed basin in northern Arizona, while it was less than
5 m2 s2 over the plain outside the crater (Yao and Zhong 2009; Fu et al. 2010).
Because of the large spatial variability of winds, stability, and strong diurnal
oscillations, some areas in complex terrain are more prone to mixing events than
others and turbulence can be localized both in space and time. In places such as those
close to mountain gaps, near canyon exits, or over slopes where terrain-induced
circulations are usually strong, turbulent exchanges may remain active throughout
the night and the characteristic scale of the turbulent eddies may be relatively large.
Over valley or basin floors, however, where a cold air pool usually develops at
night with relatively weak winds and a strong inversion, mixing activities can be
substantially reduced, or constrained to bursts of increased fluctuations known as
intermittent turbulence.
Large horizontal variability of both mean and turbulence properties over hetero-
geneous terrain may constitute a problem for turbulence parameterizations in some
mesoscale models that describe turbulence based on grid-averaged mean states.
In most parameterizations used in mesoscale models, the influence of stability is
included through the Richardson number (Ri), which is based on velocity gradients
and thermal stratification. In an environment where winds or stability exhibit large
spatial variability, a single value of Ri may not adequately represent an area even
as small as a typical grid box in mesoscale models. As a result, small areas with
sub-critical Ri can have enhanced mixing and dominate the area-averaged flux over
a region of the size of a grid cell in a mesoscale model. These small regions of
turbulent mixing may produce finite area-averaged fluxes, even if the averaged
state of the grid box derives a Richardson number that is above the critical value.
Thus, the accuracy of the turbulence parameterizations in representing turbulence
mixing in the grid box may be questionable. Compared to first-order K-theory-
based schemes that do not take into account any transport equations for turbulent
quantities, TKE-based parameterizations (1.5-order) should be more successful
over complex terrain because they contain most of the relevant physical processes.
Even in the TKE-based schemes, however, the length scales are usually derived
from essentially one-dimensional, local equilibrium ideas embodied in conventional
Monin-Obukhov (M-O) similarity theory. Large horizontal variability and temporal
intermittency that characterize turbulence in the boundary layer over mountainous
terrain make the conditions incompatible with some of the basic assumptions of
the M-O similarity theory (Fu et al. 2010). Finally, the large spatial variability of
turbulence properties over heterogeneous terrain poses a challenge for obtaining
representative measurements that can be used to evaluate or develop improved
turbulence parameterizations for mesoscale models.
In addition to horizontal wind shear, slope flows and valley winds are charac-
terized by a vertical jet-like profile with a wind maximum close to the ground.
Because the gradient is zero at a maximum, turbulence production by vertical wind
shear at the wind speed maximum would erroneously be predicted to be zero, as
would be the turbulent vertical exchange. The local scaling velocity would thus
be zero, as would the local Monin-Obukhov length, by definition. According to
parameterizations based on similarity theory, turbulent exchange across the wind
maximum would cease and a total decoupling of the flow would occur. Mesoscale
models usually overcome this total decoupling problem by placing a lower limit on
length scales. These limits are generally ad hoc and are sometimes at the discretion
of the modelers.
Another issue that arises is the impact of simplifications traditionally introduced
into the TKE-based turbulence parameterization schemes so commonly used in
mesoscale models. While some models, such as ARPS and WRF, include full three-
dimensional shear production terms in their TKE schemes, others, such as MM5
and RAMS, only retain the vertical shear production while ignoring horizontal
shear production terms. The implicit assumption is that the generation of TKE by
horizontal wind shear is much smaller than production due to vertical wind shear,
which is largely valid over large spatial scales and regions of simple terrain, but is no
longer true over irregular terrain. As an example, using the Burk-Thompson (1989)
parameterization, the time rate of change of TKE is defined as

@q 2 @ @q 2 0 0 @u 0 0 @v 0 0 @v @u gw0 v0 2q 3
D
q 2w u 2w v 2u v C C2 ;
@t @z @z @z @z @x @y ^l
„ ƒ‚ …
I
(10.2c)
p
where q is the TKE, and
and ^l are length scales. The first term on the
right-hand side of (10.2c) represents the flux divergence of TKE through the
model grid cell, the second and third terms represent production of TKE by the
vertical shear of the mean wind, the fourth term represents production by horizontal
wind shear (labeled I), the fifth term is the production or destruction of TKE by
buoyancy flux, and the final term is dissipation. Figure 10.6 shows a comparison
of simulated TKE with and without the horizontal wind shear term (labeled I)
with aircraft observations of TKE during an early morning flight in the Salt Lake
Valley. The inclusion of the horizontal wind shear led to an order of magnitude
increase in the simulated TKE and thus produced a much better agreement with the
observation. The enhanced turbulent mixing during the early morning hours would
have significant implications for boundary layer growth and air quality in the valley.
It is clear from the above discussions that conditions in complex terrain often
violate some of the assumptions used in deriving turbulence parameterizations and
therefore the model results typically are less accurate than those over flat terrain.
This is demonstrated in Fig. 10.7 which shows a comparison of simulated and
observed mixed-layer heights from two locations, one over the Oklahoma and
Kansas border in the Southern Great Plains and another in Salt Lake Valley. Three
different types of boundary-layer turbulence parameterizations were used in the
simulations for both sites and the results in general show better agreement with
observations over flat terrain than over complex terrain.
Despite the many sources of error in the turbulence schemes, mesoscale models
continue to provide valuable insight into atmospheric processes over complex
terrain. At the same time, our understanding of turbulence in complex terrain is
being enhanced through increased use of large-eddy simulations over real terrain
Fig. 10.6 Simulated TKE with horizontal wind shear (squares) and without (dots) compared to
observed TKE (line) from aircraft transect approximately 500 m AGL over the east-central portion
of the Salt Lake Valley at approximately 1300 UTC on 16 October during VTMX (Courtesy of
Jerome Fast and Larry Berg)
2000
1500
MM5 Predicted zi (m)
1000
500
BLX96 BK MRF GS
VTMX BK MRF GS
0
0 500 1000 1500 2000
Observed zi (m)
Fig. 10.7 Comparison of the MM5 predicted mixed layer heights with observations from the
Boundary Layer Experiment (BLX) over an area between Oklahoma and Kansas border in the
Southern Great Plains and from the VTMX experiment in the Salt Lake Valley. BK, MRF, and GS
denote Blackadar, MRF, and Gayno-Seaman turbulence parameterization schemes (see Table 10.2
for description) used in the simulations. The three lines represent the 1:1 line and the 10% error
(From Berg and Zhong 2005. © American Meteorological Society. Reprinted with permission)
Fig. 10.8 Schematic diagram of a two-layer land-surface model with surface and deep soil layers,
each with prescribed soil and vegetation types. The LSM provides fluxes of heat, momentum, and
moisture at the land surface for each atmospheric grid cell
(see Sect. 10.4), and through a number of recent field experiments in mountainous
areas, including the MAP-Riviera field campaign designed to look at turbulence
structure in a narrow Alpine valley (Rotach et al. 2004); the METCRAX experiment
in Arizona’s Meteor Crater (Whiteman et al. 2008), allowing for studies of
turbulence inside a closed basin, the Terrain-Induced Rotor Experiment (T-REX) in
California’s Owens Valley in the immediate lee side of the southern Sierra Nevada
Range (Grubišić et al. 2008), which provided data sets for studies of turbulence in
the lee slope of a major mountain range and in a deep valley.
10.3.2 Land-Surface Models
Mesoscale models couple the atmosphere with the earth’s surface through a land-
surface model (LSM). An LSM usually consists of a one-dimensional model of
energy, water, momentum, and in some cases carbon dioxide exchange between the
earth’s surface and the atmosphere, accounting for thermal, hydrological, and eco-
logical effects of vegetation and soil properties. The coupling is typically done by
the atmospheric model providing incident solar and terrestrial radiation and mete-
orological variables (temperature, wind, humidity, precipitation and pressure) from
the lowest model level to the LSM, and the LSM feeding the atmospheric model with
heat and moisture fluxes, surface stress, and radiation at the land surface. A single
LSM column sits beneath each atmospheric model grid cell at the land surface and
these are typically not coupled to other LSM cells in any way, so lateral water flow
is not possible, though it may be very important in steep terrain (see Fig. 10.8).
In the past two decades, many LSMs have been developed with varying degrees
of complexity. Earlier LSMs included the ‘big leaf’ parameterization scheme
(Deardorff 1978), the biosphere-atmosphere transfer scheme (BATS) (Dickinson
et al. 1986), the simple biosphere model (SiB) (Sellers et al. 1986), and the
Interaction Soil-Biosphere-Atmosphere (ISBA) model (Noilhan and Planton 1989).
More sophisticated LSMs have been developed since the 1990s. Examples of these
models include the Canadian Land Surface Scheme (CLASS) (Versephy 2007),
the ECMWF land surface parameterization scheme (Viterbo and Beljaars 1995),
the Princeton hydrological LSM (Wood et al. 1992), the Noah LSM (Chen et al.
1996; Ek et al. 2003), and CLM, the Common Land Model (Dai et al. 2003).
LSMs can differ in many ways, including the representation of subgrid spatial
variability in soil and vegetation characteristics, topography, and water storage,
treatment of surface runoff, and parameterization of snow melt, etc. The Project for
Intercomparison of Land surface Parameterization Schemes (PILPS) (Henderson-
Sellers et al. 1993, 1995; Chen et al. 1997; Wood et al. 1998; Bowling et al.
2003) evaluated the performance of up to two dozen LSMs by comparing their
output with observational data. The results revealed substantial differences among
simulated surface energy, water vapor flux, and surface runoff despite the fact
that the LSMs were driven by identical atmospheric forcing. Using data from
Cabauw, Netherland, Chen et al. (1997) found variations of 2 K for annual surface
radiative temperature, 35 and 25 Wm2 for annual mean sensible and latent heat
flux, and 315 mm for both annual total evapotranspiration and runoff across
23 LSMs. In an evaluation of the ability of 15 LSMs in simulating water balance,
Boone et al. (2004) found that while the simulated overall evapotranspiration,
runoff, and monthly change in water storage are similar among the LSMs, the
differing partitioning among the fluxes result in very different river discharges and
soil moisture equilibrium states. These differences are attributed to the different
treatments of vegetation dynamics, hydrological processes, and soil-vegetation-
atmosphere transfer processes in different LSMs. In reviewing the development
of LSMs designed for climate models, Pitman (2003) pointed out that despite the
recent rapid advances in LSM development, many problems remain to be addressed,
including the difficulties in parameterizing hydrological processes, root processes,
sub-grid-scale heterogeneity and biogeochemical cycles.
Although adequate treatment of land surface processes is key to any successful
atmospheric modeling, it is even more critical to mesoscale predictions over
complex terrain because terrain-induced processes are often driven by surface
forcing that ultimately depends on adequate descriptions of not only the mean
values of surface momentum, heat, and moisture fluxes for an area, but the spatial
variations of these fluxes across the area. In regions of complex terrain where
large spatial variations in vegetation, soil moisture, and solar exposure occur and
where the observational data are sparse at best, the success of model simulations
relies much more heavily on the accuracy of the LSM and the inputs to the LSM
compared to simulations over flat terrain. A recent model intercomparison study
for flow over an idealized valley, for example, highlighted the need for improved
LSM representations. Despite differences in numerical formulations, the largest

differences in model predictions of the daytime boundary layer were due to the
forcing at the land surface (Schmidli et al. 2010).
10.3.2.1 Land-Use Datasets
All LSMs require specification of two major classes of parameters: the primary
parameters, such as the dominant type of vegetation and soil in each grid cell, and
secondary parameters, which describe a wide range of physical properties and the
state of each type of vegetation and soil. The primary parameters are needed for each
grid cell in the modeling domain and are usually specified using satellite-derived
global or continental datasets. These are available at as fine as 30 m resolution
from USGS datasets. The secondary or physical parameters are associated with
primary parameters and are typically derived from in-situ measurements, satellite
observations, or a combination of both. As mesoscale numerical models increase in
horizontal resolution, the existing global or regional datasets can no longer provide
adequate specification of the large set of parameters and their spatial distributions
needed by LSMs. This problem is amplified over complex terrain that is often
characterized by large heterogeneity in vegetation and soil types associated with
changing terrain elevation. For example, in a modeling study in California’s Owens
Valley (Daniels et al. 2006) spurious snow was found on the valley floor due to the
coarse resolution of the NARR dataset used for model initialization, which affected
the energy budget on the valley floor and thus the valley wind circulation.
Another issue associated with LSMs that is amplified in complex terrain is the
initialization of soil moisture and temperature. The LSM needs to be initialized
with soil temperature and moisture at every grid cell and mesoscale simulations
are usually quite sensitive to these initialization values and spatial distributions.
Even for a “perfect” LSM, inadequate initialization may lead to large bias and
errors in the subsequent simulation. Studies such as PILPS have shown that LSMs
driven by the same atmospheric forcing, but slightly different initialization fields,
can produce substantially different surface water and energy budgets (Henderson-
Sellers et al. 1993, 1995; Wood et al. 1998; Bowling et al. 2003). Studies have shown
that more sophisticated LSMs do not necessarily outperform simple LSMs because
of the difficulty in accurately specifying a large set of parameters related to physical
properties and the state of vegetation and soil at the local scale (Koster and Milly
1997).
Several field studies have collected datasets specifically for the purpose of
evaluating the performance of various LSMs. These studies include HAPEX-
MOBILHY (André et al. 1986), FIFE [First International Satellite Land Surface
Climatology Project Field Experiment (Sellers et al. 1992; Betts and Ball 1992;
Betts et al. 1993)], the Southern Great Plains Hydrology Experiment 1997 (SGP-
97, Twine et al. 2000), the Energy Balance EXperiment-2000 (EBEX-2000, Oncley
et al. 2002), PILPS, and more recently ALMIP (AMMA Land Surface Model
Intercomparison Project) (Boone et al. 2009). Although data collected by these
studies encompasses a wide range of vegetation and soil types, the experiment sites
were mostly over flat terrain. The lack of detailed surface energy, water budget,
and soil datasets over mountainous terrain has hindered the ability to understand
the limitations and uncertainties of existing LSMs when applied to highly complex
terrain.
10.3.2.2 Soil Moisture Issues
Soil moisture initialization is particularly important for thermally-forced flows

because it directly affects the partitioning of total available energy at the earth’s
surface into sensible and latent heat flux which determines the heating or cooling
rate of the land surface and the near surface atmosphere (see, e.g. Banta and Gannon
1995; Jacquemin and Noilhan 1990; Ookouchi et al. 1984; Patton et al. 2005). The
impact of inadequate soil moisture initialization can be felt for a long time by the
model simulations.
Soil moisture can vary significantly over different time scales (diurnal, seasonal,
and inter-annual) in response to changes in precipitation, evaporation, and vege-
tation growth cycle. Soil moisture can also change substantially across different
spatial scales, and topography plays a pivotal role in the spatial distribution of soil
moisture. Regional-scale soil moisture gradients, such as those found between the
west and east side of the Cascade Range in the Pacific Northwest and the west and
east side of the Sierra Nevada Range in California and Nevada, are usually captured
by large-scale analyses or model output used for mesoscale model initialization.
Relatively large variation in local-scale soil moisture is often found in regions of
complex terrain as a result of differences in elevation, slope angle, sun exposure,
vegetation, etc. This small-scale variability is not captured in the mesoscale model
initialization fields provided by large-scale analyses because in-situ soil moisture
observations are typically too sparse and satellite derived soil moisture data sets
do not yet have the resolution to resolve small scale variability. The inadequate
description of the soil moisture across a region in the model initial fields has been
found to be an important source of errors in the mesoscale predictions, especially
over regions of complex terrain, as discussed below.
In a study in the Riviera Valley in the Swiss Alps, Chow et al. (2006) and De
Wekker (2002) found that the transitions between up- and down-slope and up- and
down-valley flows are very sensitive to the soil moisture initialization. The onset
of the up-valley winds in the morning was delayed by 3–4 h compared to the
observations when the soil moisture field was initialized using ECMWF regional
analyses. When a hydrological model was used to obtain a more accurate estimate
of the spatial variation of the initial soil moisture field (see Fig. 10.9), the delay in
the simulated onset of up-valley winds was cut in half, as seen in Fig. 10.10.
Another example of how an inadequate description of spatial variability in the
initial soil moisture distribution affects simulations of boundary layer development
in regions of complex terrain is given by Daniels et al. (2006). Using soil moisture
and temperature data collected at 23 sites in and around the Owens Valley during
a b
25 25
0.3 0.3
20 0.25 20 0.25
0.2 0.2
15 15
(km)
(km)
0.15 0.15
10 10
0.1 0.1
5 5
0.05 0.05
0 0 0 0
0 5 10 15 20 25 0 5 10 15 20 25
(km) (km)
Fig. 10.9 Contours of soil moisture (m3 m3 , shaded) from a hydrologic model (WaSiM) used
to initialize ARPS in the Riviera Valley, Switzerland, at (a) surface and (b) deep soil levels at
350-m resolution for simulations during August 1999. Topography contours are shown at 250-m
intervals. In contrast, standard initialization from ECMWF predicts uniform soil moisture over
the whole region (From Chow et al. 2006. © American Meteorological Society. Reprinted with
permission)
6
45
5 0
315
4
φ (degrees)
270
U (m/s)
3 225
180
2
135
1 90
45
0
0 5 10 15 20 0 5 10 15 20
Time (UTC) Time (UTC)
Fig. 10.10 Surface data time series comparisons from the Riviera Valley, Switzerland, on the
valley floor on August 25, 1999 for (a) wind speed and (b) wind direction: observations (filled
circles), ARPS simulations using soil moisture from WaSiM hydrologic model (open circles),
and simulations with reference soil moisture (times signs). Synoptic flow is quiescent and valley
circulations are thermally-forced (From Chow et al. 2006. © American Meteorological Society.
T-REX, Daniels et al. (2006) showed that the initial soil moisture values described
by coarse resolution mesoscale or large-scale operational models are up to three
times higher than the observed values in the Owens Valley. Two simulations were
performed using the ARPS model with identical configuration and parameterization,
but different soil moisture initialization data. In one simulation, the model was
initialized using the standard output from the NARR data, and in another, the initial
soil moisture fields was adjusted to better match the measurements in the Owens
Valley. Although both simulations were able to capture the overall structure of the
Owens Valley boundary layer, only the simulation with more accurate soil moisture
initialization captured the valley wind transition in the morning and the multi-
layered structure revealed by rawinsonde launches from the valley (Daniels et al.
2006). Soil moisture initialization issues such as these may perhaps be addressed by
using more sophisticated land-surface models [e.g., hydrologic models which take
into account topography gradients and impacts on spatial variability of soil moisture
(see e.g. Maxwell et al. 2007)].
10.3.3 Radiation Models
Processes of radiation, both solar and terrestrial, occur at scales much smaller
than model resolution and therefore need to be parameterized in atmospheric
models. As illustrated in Fig. 10.11, as a direct solar beam travels through the
atmosphere, its intensity is reduced as a result of absorption, reflection, and
scattering by atmospheric gases, clouds, and aerosols. Part of the scattered and
reflected solar radiation reaches the earth’s surface as diffuse solar radiation. On
average, approximately half of the solar radiation at the top of the atmosphere
reaches the earth’s surface. The earth receives solar radiation and emits energy back
to space, but at longer wavelengths. The atmospheric greenhouse gases, clouds, and
some aerosols absorb earth’s longwave radiation and reemit longwave energy in all
directions.
Topography can significantly modify radiation on the earth’s surface through
slope aspects, slope angle, shadowing, and degree of sky obstruction (sky view
factor), all of which vary over a wide range of length scales in regions of complex
topography, with important consequences for the surface energy balance (Scherer
and Parlow 1994; Oliphant et al. 2003). For example, a study by Matzinger et al.
(2003) observed a 2-h lag in maximum downwelling shortwave radiation and
an increase in intensity of nearly 200 Wm2 on sloping surfaces in an Alpine
valley compared to over the flat valley floor. Dubayah and Loechel (1997) reported
increases in downwelling radiation and an increase in spatial variability of diffuse
radiation as a result of sky view obstruction by topography. The subsequent
influence on surface energy budget by terrain-induced changes in radiation has been
demonstrated by studies such as Scherer and Parlow (1994) that showed differences
in snow melting rate as a result of different exposures to solar radiation in a drainage
basin. The net longwave radiation is often thought to be underestimated in valleys
because the radiation models are one-dimensional (in the vertical); the models do
not account for the incoming longwave emissions from the valley walls at night and
therefore allow too much cooling.
Climate and weather prediction models have generally assumed the land surface
to be flat on subgrid scales for the purpose of calculating the radiative components
Fig. 10.11 Illustration of radiation processes in the earth and atmosphere system. The yellow
wavy lines represent the shortwave solar radiation, red wavy lines represent longwave terrestrial
radiation, and the blue curves denote circulations (From the COMET Program [the source of
this material is the COMET® Website at http://meted.ucar.edu/ of the University Corporation
for Atmospheric Research (UCAR), sponsored in part through cooperative agreement(s) with
the National Oceanic and Atmospheric Administration (NOAA), U.S. Department of Commerce
(DOC). © 1997–2010 University Corporation for Atmospheric Research. All Rights Reserved])
in their surface energy budgets. This assumption may be used in coarse resolution
modeling over flat or gently sloped terrain without introducing large errors in the
surface energy balance. As the spatial resolution of mesoscale models increases,
however, the orographic effects on surface radiative fluxes and their spatial variabil-
ity can no longer be neglected. To fully account for all effects of terrain on radiation,
three-dimensional radiation modeling holds the most promise, as described in
Sect. 2.2.2.1. The tremendous computational demand of three-dimensional radiation
modeling, however, renders it impractical for mesoscale numerical simulations, at
least for the foreseeable future. In recent years, with the rapid increase in model
horizontal resolution and availability of fine-resolution global digital terrain data,
there have been some efforts to modify the radiation calculations in mesoscale
numerical models to include orographic effects on surface radiation. Most models
now take into account some terrain effects on shortwave radiation at the surface so
that surface energy budget calculation can be improved. Below we briefly describe
some of these modifications.
10.3.3.1 The Effect of Slope and Aspect
Sloped terrain is irradiated from three sources: direct irradiance from the sun, diffuse
irradiance from the sky, and direct and diffuse irradiance reflected from nearby
terrain. While slope irradiance can be neglected in numerical simulations when the
terrain resolution is coarse and the slope is low or moderate, its contribution to
surface radiation becomes important when the resolution is high (1 km or less)
and the slope is steep.
Hauge and Hole (2003) modified the surface radiation flux calculation to take
into account the effects of sloping surfaces. For given slope, aspect, and hourly
diffuse and beam irradiances, the total irradiance on a surface inclined by an angle
ˇtowards an azimuth angle may be expressed as

cos ˇ ˇ
R.ˇ; / D RB C 1 cos2 ˛ .RD C RB / C RD (10.2d)
sin 2
where RB and RD are the direct beam and diffuse sky irradiance, which are
estimated by taking into account the effects of solar zenith angle, clear air absorption
and scattering and cloud back scattering and absorption (Dudhia 1989). The second
term, Œ1 cos2 .ˇ=2/ ˛ .RD C RB /, represents ground-reflected irradiance where
˛ is surface albedo. is solar elevation given by
sin D sin ı sin C cos ı cos cos (10.2e)
with ı the Earth’s declination, local latitude, and the hour angle.
ˇ is the solar beam angle of incidence on a surface inclined by an angle ˇ, which
is calculated by (Iqbal 1983):
cos ˇ D cos sin ˇ cos . / C sin cos ˇ (10.2f)
with § solar azimuth.

The sketch in Fig. 10.12 illustrates the various angles mentioned above.
Equation 10.2f clearly relates solar radiation to the aspect and slope of the
underlying terrain. For flat surfaces where ˇ D 0, (10.2f) reduces to cos D sin
and (10.2d) reduces to R D RB C RD , which is the global irradiance on a
horizontally flat grid box.
With this modification of the shortwave radiation scheme to include slope irradi-
ance, Hauge and Hole (2003) showed significant improvement of the MM5 model
simulations for wind and temperature in complex topography with a reduction of
RMS error by 35% in wind and 13% in temperature. Their results also show better
simulations of the break up of the temperature inversion. Similarly, Müller and
Scherer (2005) found that implementing a subgrid radiation parameterization in a
mesoscale model accounting for slope irradiance gave better temperature forecasts.
Not all models consider all three sources of slope terrain irradiance. For
example, the radiation scheme in the RAMS model only considers direct solar
Fig. 10.12 Sketch

illustrating various angles: ˇ
angle of solar beam incidence
on a sloping surface inclined
by an angle “, solar
elevation, solar azimuth,
surface azimuth. The red lines
are on a plane and the green
lines are on a plane
perpendicular to the slope
beam irradiance and neglects the diffuse sky irradiance and direct and diffuse
irradiance reflected from nearby terrain. In other words, the second term in (10.2d)
which represents direct and diffuse irradiance reflected from ground is set to
zero. Neglecting direct and diffuse radiation reflected from ground may lead to
considerable errors in narrow valleys. In addition to MM5 and RAMS, several other
mesoscale models, such as WRF, ARPS, and Meso-NH, also consider slope and
aspect on surface radiation in their radiation scheme.
10.3.3.2 Topographic Shading
Another important influence of topography on radiation is topographic shading.

For example, in steep valleys, solar heating during the morning transition period
is delayed on west-facing slopes, leading to the development of asymmetric cross-
valley circulations (Whiteman 1982; Matzinger et al. 2003; see Chap. 2 in this
book). As discussed previously, most current mesoscale models take into account
self shading (i.e., the local slope relative to the sun’s inclination angle) in the
estimation of solar radiation in the surface energy balance. Whiteman (2000)
provided a simple subroutine for estimating delay in local sunrise due to self
shading. Most mesoscale models, however, do not consider the effects of shadows
cast by terrain on direct solar radiation. This may not be a problem on cloudy
days when diffuse radiation dominates, as diffuse radiation is typically uniformly
distributed over terrain. On clear days, however, when direct solar radiation prevails
and the amount of radiation received between shaded and sunny side of the terrain
can vary significantly, neglecting topographic shading may lead to relatively large
errors in the local radiation balance and spatial variation of heat flux, which
consequently affect the timing and strengths of local thermally driven circulations.
Fig. 10.13 Shading in Brush Creek Valley, Colorado, 30 min after local sunrise: self-shaded areas
(black), topographically shaded (dark gray), and illuminated sites (light gray) (From Colette et al.
The problem associated with neglecting topographic shading effects on radiation

in modeled radiation budgets has gained attention in recent years as grid resolution
has increased and the topography gradients that can be represented have become
steeper. Colette et al. (2003) introduced a new routine into the ARPS model to
modify the radiation balance in the topographically shaded areas in the modeling
domain. Their approach is relatively simple, involving drawing a line between each
model surface grid point and the sun according to changes of the sun’s position
defined by its azimuth and declination angles. If the line intersects neighboring
topography between the sun and the specific surface node, the node is considered as
shaded and the direct component of the incoming solar radiation at this node is set
to zero (see Fig. 10.13). Simulations of idealized valleys with topographic shading
show that local sunrise can be delayed for as much as 2.5 h for deep and narrow
valley, and a significant offset is added to the onset of upslope flows near the valley
floor, thus delaying the breakup of the morning inversion and affecting subsequent
boundary layer growth.
A complication of the approach proposed by Colette et al. (2003) arises near
the boundary of the modeling domain. Mountains beyond the boundary of model
domain cannot cast shadows unless specifically accounted for. This is particularly
problematic when multiple nesting levels are used as typically done in areas of
complex terrain. In addition to ARPS, some other models, such as MM5 and WRF,
also now include the effect of shadows cast by terrain on direct solar radiation (Zängl
2005).
10.3.4 Surface Flux Parameterizations
The determination of bottom boundary conditions in mesoscale models is chal-

lenging because the grid resolution can never be fine enough to capture the
large gradients and small turbulent motions that occur near the ground. In an
attempt to include the effects of the solid boundary, parameterizations are used.
The most frequently-used bottom boundary conditions in numerical simulations
are based on similarity theory, which empirically relates dimensionless groups in
the boundary layer to predict surface fluxes of heat, moisture, and momentum.
Important parameters in the surface layer are the height of roughness elements
(such as vegetation and buildings), the mean wind velocity, and the height above
the surface.
The parameter known as the aerodynamic roughness length, z0 , can be deter-
mined by wind measurements at different elevations and application of a similarity
profile (for neutral cases, this is a logarithmic profile – stable and convective
conditions are deviations from this). The parameter z0 is the height at which the
wind speed goes to zero and is less than the height of the physical roughness
elements. From these variables, dimensionless relations can be formed, after which
a functional relationship can be determined. These parameterizations are based on
flat boundary layer flows, and are not generally valid over complex terrain.
To illustrate the process of determining the bottom boundary conditions in
a mesoscale model, we consider the surface momentum flux, or surface drag,
under neutral stratification. A surface layer parameterization (also called a “wall
model”) is used to represent the stress induced by the solid boundary. For steady-
state turbulent flow over a flat solid surface, a logarithmic region will develop in
the velocity profile. The log region is usually valid from about 50 z0 to 10% of
the boundary layer depth. In the atmosphere, a typical value for z0 is 0.1 m, so the
log region extends from about 5 to 150 m (for a boundary layer of 1.5 km depth).
Because the first grid point above the wall is rarely less than 5 m away, the first
velocity point will be in the log region. If the velocity profile is assumed to be
logarithmic at this first point, the value can be used to specify the appropriate shear
stress at the bottom, or equivalently, appropriate velocity boundary conditions. The
friction velocity u (also known as the shear velocity) is defined using the bottom
stress wal l with
wal l D u2 (10.2g)

We can relate u to the wind velocity through a proportionality relation:

p
u D CD U1 (10.2h)
where CD is known as the drag coefficient and U1 is the mean velocity at the first
grid point above the wall. CD is determined using the fact that U1 satisfies the log
law, which is given by:

u z C z0
U D ln (10.2i)
z0
where is the von Karman constant (typically set equal to 0.4; note that this form
gives U D 0 if z D 0, thus placing the numerical bottom boundary at z0 ).
Evaluating this expression at z D z1 (the height of the first point above the wall)
and solving for CD using (10.2i) yields

1 z1 C z0 2
CD D ln (10.2j)
z0
We now have a relationship for CD given the grid spacing and the roughness
height. Now combining (10.2g) and (10.2h) we have
wal l D u2 D CD jU1 j U1 (10.2k)
which relates the shear stress at the bottom to the velocity at the first point using a
log-law-derived drag coefficient (CD ) (the absolute value is needed to give the right
direction for the bottom stress which will always be opposing the mean flow). Since
the log law is only valid under neutral stratification, the determination of CD must
be adjusted to account for stable and convective conditions (see Pielke 2002, p. 176)
Common practice is to specify the wall stress as a boundary condition directly.
This is very convenient with a staggered grid (see Chap. 9, Sect. 9.3.6) since the
mean U-momentum equation requires the term @13 =@z, where ij is the turbulent
stress:

@U @U @U @U 1 @P @11 @12 @13
CU CV CW D C C (10.2l)
@t @x @y @z @x @x @y @z
For example, since 13 and U are staggered, we can calculate the z derivative of
13 using a central difference (2nd order) centered at the U point (see Fig. 10.14).
The vertical advection term on the left-hand side of the equation also requires a
z derivative, but this contribution is zero at the wall, so we can set @U=@z D 0
using a ghost point (fictitious point) below the wall. The exact boundary condition
implementation depends on the coordinate system, and is code dependent. The effect
of the solid boundary in generating shear in the velocity profile thus appears through
the specification of the stress on the solid boundary. In this manner the stress appears
Fig. 10.14 Sketch of grid cells showing implementation of the log law through specification of
the wall shear stress on a staggered grid. Shear stress 13 is specified on the wall, and above that, a
standard finite difference can be used. The momentum equation for U1 requires @13 =@z, which can
be calculated using 13 values on the wall and at one grid cell above (for a second order scheme)
in the right-hand side of the U momentum equation when solving for the velocity at
the first grid point above the wall.
The log-law boundary condition (in the formulation above) has the drawback
that it assumes the velocity profile to be instantaneously logarithmic at each grid
cell. While this is generally not exactly true, the log-law condition provides a
useful and practical way to represent rough surfaces (Nakayama et al. 2004b).
Whether the log-law assumption over complex terrain is physically realistic is
questionable. Similarity or log-law approximate boundary conditions are usually
inadequate even for flow over smooth surfaces (Piomelli and Balaras 2002; Cabot
and Moin 2000) let alone the high Reynolds number roughness-influenced flow in
the atmospheric boundary layer. In complex terrain, the bottom boundary problem is
further exacerbated by small terrain features which are not resolved on coarser grids.
For example, a mesoscale model at 10 km resolution will not be able to resolve
a 2 km wide valley and will assign boundary conditions on a smoothed bottom
boundary. This smoothed boundary will be unable to capture the true surface fluxes
from the valley. It may be helpful for land-surface fluxes at coarser grid levels to
include effects of valley circulations which modify vertical fluxes near ridge tops
(Weigel et al. 2007a; De Wekker et al. 2005; Noppel and Fiedler 2002). There are
also open questions regarding the issue of filtering in LES and how it affects the
bottom boundary conditions and near-surface drag (Nakayama et al. 2004a; Chow
et al. 2005). Several alternative models have been proposed for surface momentum
flux, but none have yet been proven to be robust enough or significantly better in a
wide range of flow conditions than the similarity theory commonly used today.
10.4 Large-Eddy Simulation for Flow over Complex Terrain
Large-eddy simulation (LES) is a modeling technique which separates the large,

energy-containing eddies from smaller turbulent motions through the application
of a spatial filter. The effect of the small scales on the larger scales is represented
through a turbulence closure model. LES can be applied to high Reynolds number
flows because it can accommodate a range of filter widths to separate resolved and
subfilter scales. LES was first introduced by Lilly (1962) and Deardorff (1970) using
the turbulence closure of Smagorinsky (1963). Deardorff (1972) extended LES from
applications in the atmospheric boundary layer to engineering plane channel flow.
It has since been heavily used as a turbulent flow research tool in engineering fluid
mechanics (Lesieur et al. 2005). Several recent textbooks have been published on
LES (see e.g. Sagaut 2002; Lesieur et al. 2005), and overview chapters can be found
in Pope (2000, Chap. 13) and Ferziger and Peric (2002, Chap. 9).
The LES approach is fundamentally different than the RANS approach used in
mesoscale models because it uses spatial filtering rather than ensemble averages.
As such, the goal in LES is to resolve the energy-containing eddies in the turbulent
flow, rather than to represent only the mean flow as in RANS. Interest in LES
for the atmospheric boundary layer has been growing over the last two decades
as simulations have moved to higher grid resolutions which can resolve more
turbulence structures (see e.g. Andren et al. 1994). As shown in recent studies, LES
is a very promising numerical technique for atmospheric boundary layer studies
over complex terrain, even though it has so far primarily been applied to idealized
flows over flat terrain. Here we provide an overview of LES methods, describe some
successful applications to flow over mountainous terrain, and discuss recent trends
in turbulence modeling and the need for future work. A discussion of practical
differences between LES and RANS is also included.
10.4.1 LES Basics
10.4.1.1 Filtered LES Equations
Large-eddy simulation employs a spatial filter to separate the small scales from
the large scales. The large eddies are explicitly calculated by prognostic equations,
while the effect of the smaller eddies must be modeled. The filter function G
(typically a Gaussian or tophat filter) is applied to a flow variable f in physical
space as
Z 1

f .x; ; t / D G x; x 0 ; f x 0 ; t dx 0 (10.3a)
1
where is the filter width. The application of this filter (denoted by an overbar) to
the incompressible Navier–Stokes and continuity equations results in
@ui @ui uj 1 @p @2 ui @ui

C D C ; and D 0; (10.3b)
@t @xj @xi @xj @xj @xi
where ui denotes the filtered velocity field for which we seek a solution, p is
the filtered pressure, is the molecular viscosity (which is usually neglected in
atmospheric flows), and Einstein summation notation is used. The filtered product
of the velocities ui uj appearing in the advection term creates a closure problem
for the equations because the unfiltered velocity field ui is unknown. Bringing this
term to the right-hand side and rearranging, we obtain the LES equations in their
traditional form:
@ui @ui uj 1 @p @2 ui @ij

C D C (10.3c)
@t @xj @xi @xj @xj @xj
where ij D ui uj ui uj is defined as the subfilter-scale (SFS) turbulent stress. This

SFS term must be parameterized as a function of the filtered velocity ui to close the
equations.
The SFS motions are responsible for energy dissipation, among other processes,
and it is crucial that the SFS stress representation perform well so that the resolved
quantities (represented by ui ) are predicted accurately. The difficulty in formulating
a closure model for LES is that modeling of the unresolved motions (contained in
the correlation term ui uj ) must be based only on knowledge of the resolved motions
(ui ). Such a requirement means that the closure model is inherently imperfect,
as the exact flow is influenced by higher-order velocity moments not present in
the resolved field. The LES closure problem is similar to the closure problem
encountered in the Reynolds-averaged Navier–Stokes (RANS) equations, though
in LES the range of turbulent length scales being modeled depends on the choice of
filter width (rather than a time average or ensemble average). Note that because a
local spatial filter is applied, double application of the filter gives a different result
such that ui ¤ ui .
10.4.1.2 LES Turbulence Closures
The vast majority of LES studies of the atmospheric boundary layer (ABL) have
used eddy-viscosity models. The first SFS models mimicked the RANS closure
idea of an eddy-viscosity
gradient-transport model, setting ij 2T Sij ,
or
@u
where Sij D 12 @x @ui
j
C @xji . A constant eddy viscosity, T , can be chosen but
this gives very poor results. Smagorinsky (1963) introduced a model of the form
1
T D .Cs g /2 .Sij Sij / 2 where g is the grid spacing, and the Smagorinsky
coefficient CS is a constant that must be chosen (e.g. CS 0.18 is often used in ABL
flows) to give the best results. In this model the filter width is based on the grid
spacing, which determines the eddy motions resolved in the LES. The Smagorinsky
model is still widely used (e.g. it is used for horizontal diffusion in WRF) and is the
basis for the popular dynamic Smagorinsky model described later.
Closures using a prognostic equation for the turbulent kinetic energy (TKE)
are also common in ABL simulations. These 1.5-order TKE closures (see e.g.
Deardorff 1980; Moeng 1984) are similar to RANS approaches such as k-" or q2 -
q2 l approaches (see Sect. 10.3.1.1 in this chapter; Yamada and Mellor 1975; Stull
1988). The eddy viscosity is represented as a function of a velocity scale and a
length scale. The TKE equation (q2 or k) is solved to obtain a representative velocity
scale. In RANS closures, the length scale (l) is derived using a diagnostic algebraic
formula or another prognostic equation, such as for dissipation (") or q2 l. When
a TKE approach is used in LES applications, the velocity scale used in the eddy
viscosity is again based on solution of the TKE equation; however, instead of solving
an additional prognostic equation for the length scale, the 1.5-order TKE models use
the LES filter width (f ) as the basis for determining the length scale.
Eddy viscosity models perform reasonably well in predicting mean flows away
from boundaries, but often overpredict shear stress because they require the SFS
stress to be aligned with the resolved strain rate tensor. Field experiments indicate
there is little correlation between the modeled and actual SFS stresses with eddy
viscosity closures (Horst et al. 2004; Sullivan et al. 2003). These models are also
entirely dissipative and do not allow energy transfer from small to large scales which
can be significant as well. Various other eddy-viscosity models have been proposed,
some with nonlinear dependence on the resolved strain rate, Sij (see e.g., Kosović
1997).
Due to the limited success of eddy-viscosity closures, an alternative approach
was introduced by Bardina et al. (1983) and was based on the hypothesis that the
largest SFS scale and the smallest resolved scale are similar. The Bardina model, or
scale-similarity model, makes the approximation ui ui (i.e., that the full velocity
is equal to the filtered velocity) to close the SFS stress term, resulting in ij
ui uj ui uj . This model exhibits relatively high correlations with the exact SFS
stress and allows backscatter of energy from the small to the large scales, making
it a desirable component of any SFS model. The model does not dissipate enough
energy, however, so Bardina et al. (1983) also introduced the mixed model, which
combined the Smagorinsky and the scale-similarity model, simply giving ij
ui uj ui uj 2T S ij . This gives much improved correlations while allowing for
adequate dissipation.
Dynamic eddy viscosity models allow the coefficient (e.g. CS in the Smagorinsky
model) to vary in space and time dynamically, without the use of prognostic
equations or predetermined coefficients. The first such model was developed by
Germano et al. (1991) and was applied to the Smagorinsky eddy-viscosity model,
which became known as the dynamic Smagorinsky model (DSM). It is based on
the assumption that the same Smagorinsky coefficient can be applied at the given
filter width and at a test filter level on a coarser grid (usually twice the original filter
width). The coefficient is determined by applying an extra test filter and using a
least-squares fit to choose CS (Lilly 1992). Fluctuations in the dynamic Smagorinsky
coefficient allow positive and negative energy transfer, which simulates the forward-
and back-scatter of energy between large and small scales. Large fluctuations in the
coefficient are often observed, however, which lead to instabilities. To overcome
this problem, clipping is usually applied to limit how negative the eddy viscosity
can become, and/or averaging of the dynamic constant is applied over planar cross-
sections which means that back-scatter or forward-scatter must occur uniformly
over that entire plane. Newer formulations allow various forms of local averaging
which restores some of the spatial variability to the model (Ghosal et al. 1995; Bou-
Zeid et al. 2005; Zang et al. 1993; Kirkpatrick et al. 2006). Meneveau et al. (1996)
introduced the idea of Lagrangian averaging, where the averaging is performed over
a fluid path line (Anderson and Meneveau 1999; Sarghini et al. 1999; Bou-Zeid
et al. 2005). This type of averaging is particularly appropriate for flow over complex
terrain (Bou-Zeid et al. 2005). Porte-Agel et al. (2000) introduced a scale-dependent
dynamic Smagorinsky model, where the Smagorinsky constant is not required to be
invariant with grid resolution, thus introducing greater flexibility in the model and
better agreement in the near-wall region. Further variations of dynamic models have
also been developed, such as those of Wong and Lilly (1994), Vreman et al. (1996),
and Cederwall and Street (1999).
Zang et al. (1993) extended the idea of dynamic coefficients to the mixed
model (Bardina et al. 1983) and created the dynamic mixed model (DMM). This
incorporated the scale-similarity term which is capable of accounting for backscatter
on its own, thus greatly decreasing the required contribution of the eddy-viscosity
component. The dynamic coefficient in the DMM does not fluctuate as much as in
the DSM; hence local averaging is sometimes adequate to smooth the fluctuations.
Gullbrand and Chow (2003) and Chow et al. (2005) developed the dynamic
reconstruction model (DRM) which is a mixed model that combines a higher-
order scale-similarity model with a dynamic eddy viscosity model. This approach
has been shown to reduce numerical errors in finite difference or finite volume
applications of LES and to improve predictions of SFS stresses.
While dynamic models are considered the state-of-the-art, their disadvantage is
that they require increased computational time as two levels of filtering must be
employed, as well as averaging of the dynamic coefficient. The performance of
dynamic models over a rough wall is a topic of continued study (see e.g., Chow
et al. 2005) (Table 10.3).
10.4.2 Application to Atmospheric Flows
LES was originally developed for ABL studies and has been used extensively for
research in semi-idealized ABL configurations. Numerous studies have applied LES
to convective, neutral, and stably-stratified flows over flat terrain (see Fig. 10.15).
Here we focus on neutral cases as illustrative of the challenges that arise with LES.
632

1 @ui @uj
Table 10.3 Key features of LES turbulence closures that are applied to atmospheric models. In each case T is eddy viscosity, q2 is TKE, Sij D 2 @xj
C @xi
.
is the grid spacing or filter width
Scheme Type Primary closure assumption
1
Smagorinsky (1963) Eddy viscosity ij 2T Sij T D .Cs /2 .Sij Sij / 2
CS , the Smagorinsky coefficient must be specified
TKE-1.5 order (Deardorff 1980; Moeng Eddy viscosity ij 2T Sij where T q and q is obtained from a prognostic
1984) equation for TKE
Bardina et al. (1983) Scale-similarity ij ui uj ui uj , and explicit filter must be defined
1
Mixed model (Bardina et al. 1983) Scale-similarity plus eddy ij ui uj ui uj 2T Sij , T D .Cs /2 .Sij Sij / 2
viscosity
1
Dynamic Smagorinsky model (Germano Dynamic eddy viscosity ij 2T Sij T D .Cs /2 .Sij Sij / 2
et al. 1991) the Smagorinsky coefficient CS is determined dynamically through a
test filter procedure
1
Dynamic mixed model (Zang et al. 1993) Dynamic eddy viscosity ij ui uj ui uj 2T Sij , with T D .Cs /2 .Sij Sij / 2
plus scale similarity the Smagorinsky coefficient CS is determined dynamically through a
test filter procedure
1
Lagrangian-averaged scale-dependent Dynamic eddy viscosity ij 2T Sij , T D .Cs /2 .Sij Sij / 2
dynamic Smagorinsky model CS is determined dynamically through a test filter procedure but is
(Porte-Agel et al. 2000) allowed to vary with resolution, requiring two test filter operations
Dynamic reconstruction model Dynamic eddy viscosity ij ui uj ui uj 2T Sij , where T is calculated dynamically and
(Gullbrand and Chow 2003) plus scale similarity the scale-similarity piece is represented with high-order series
expansion reconstruction
S. Zhong and F.K. Chow
Fig. 10.15 Horizontal cross section of the vertical velocity field in a convective PBL at half the
boundary layer depth from a large-eddy simulation with 5123 computational cells. The color bar is
in units of m s1 (From Sullivan and Patton 2008. © American Meteorological Society. Reprinted
with permission)
10.4.2.1 Flat Terrain
Even with simple flat terrain, there are challenges in correctly representing ABL
development. A well-known problem in atmospheric boundary layer simulations in
idealized domains is their lack of agreement with similarity theory in the near-wall
(near-surface) region, where the contribution of the turbulence model dominates
that of the resolved terms (Andren et al. 1994; Sullivan et al. 1994; Kosović 1997;
Juneja and Brasseur 1999). Standard eddy-viscosity models (e.g. Smagorinsky
(1963) or TKE-1.5 (Deardorff 1980; Moeng 1984)) do not have correct near-wall
behavior. The study of four different LES codes by Andren et al. (1994) confirmed
that the challenge for eddy-viscosity models has long been to obtain the correct
logarithmic velocity profile near the wall. Traditional simulations result in excessive
shear at the wall and hence incorrectly predict the velocity profile. Using a stochastic
backscatter model [i.e., the model of Mason and Thomson (1992) tested by Andren
et al. (1994)] seemed to improve results near the wall. As scale-similar models allow
for backscatter, these also greatly improve the flow profile in the ABL, without
introducing the random (and non-physically based) fluctuations of the stochastic

model (Esau 2004; Chow et al. 2005).
To attempt to correct the poor near-wall behavior of traditional eddy viscos-
ity models, dynamic models have also been applied to the ABL. Esau (2004)
simulated the large-scale neutral atmospheric boundary layer using the dynamic
Smagorinsky model and the dynamic mixed model (DMM) of Zang, Street, and
Koseff (1993) with some improvement over the standard Smagorinsky model.
Porte-Agel, Meneveau, and Parlange (2000) developed a scale-dependent dynamic
Smagorinsky model to account for the near-wall region where motions are under-
resolved. When applied to a neutral atmospheric boundary layer, they achieved
improved logarithmic velocity profiles, using an extra filtering step in the dynamic
procedure and an empirical function to determine the scale dependence.
Explicit filtering and reconstruction can also improve LES results, as demon-
strated by Chow et al. (2005) for neutral boundary layer flow. Simulations with
the dynamic reconstruction model showed good agreement with similarity theory
and the ability to better represent SFS momentum fluxes. The spatial patterns
in the resolved turbulent structures also depend on the choice of the turbulence
model, which could be of importance in studies of turbulent transport and mixing
(Ludwig et al. 2009). The explicit filtering approach needs further development
for complex terrain and for non-neutral conditions as difficulties with turbulence
modeling increase significantly over complex terrain, as discussed further below.
LES of the stable boundary layer has always been a challenge and is a topic of
current research interest (Mason and Derbyshire 1990; Kosović and Curry 2000;
Saiki et al. 2000; Cederwall 2001; Basu and Porte-Agel 2006; Beare et al. 2006,
Zhou and Chow 2011). SBL flows contain smaller turbulent length scales and the
turbulence is often intermittent, making it particularly challenging for SFS models
to properly model the evolution of the ABL. The difficulties in simulating idealized
stable boundary layers are likely to be amplified in some situations of complex
terrain where stability effects will be combined with terrain and synoptic forcing.
10.4.2.2 Idealized Topography LES Simulations
LES has been used to investigate detailed turbulent flow over idealized topography.
These studies have often looked at small-scale flows to examine topography-
influenced flow dynamics. For example, Ding and Street (2003) and Ding, Calhoun,
and Street (2003) studied flow over a 3D hill under neutral and stratified conditions.
Smith and Skyllingstad (2005) performed simulations over uniform and compound
angle slopes in an idealized study to examine the TKE budget and slope flow devel-
opment using a mesoscale and an LES configuration. Smith and Skyllingstad (2009)
performed LES of rotors to study heating effects using an idealized topography.
Numerous other modeling studies have examined idealized topography such as
valleys or plateaus, but most are either two-dimensional or based on RANS closures.
Studies with simplified terrain can also focus on the turbulence dynamics and
the detailed performance of the LES closure without worrying about the impact
of complex atmospheric forcing or terrain conditions. Brown, Hobson, and Wood

(2001) performed LES of flow over rough sinusoidal ridges, showing that LES
agreed better with wind tunnel experiments than a simple RANS mixing length
closure. Iizuka and Kondo (2004) investigated various LES closures and their ability
to capture flow separation in the lee of idealized hills.
10.4.3 Challenges for LES in Complex Terrain
Traditionally LES has been applied to idealized flows over simple terrain, so no one
has fully addressed the issues related to turbulence modeling, land surface fluxes,
grid nesting, and other parameterizations over real terrain for LES. Accommodating
real, complex terrain means expanding the range of scales that must be represented
in the simulation, often requiring coarser resolution and larger domain sizes than
typically used by idealized LES studies. LES was not used at large scales in
the past because current subfilter-scale turbulence models were not designed to
accommodate this range of scales. The fundamental equations used in large-
eddy simulation, however, are applicable at all scales. There are no mathematical
limitations which dictate that LES be applied only at “fine” resolutions where the
grid cutoff falls within the inertial range of the energy spectrum, as is commonly
believed. Rather, application of LES to meso-scales simply requires increasing the
filter width which separates large, resolved motions from subfilter-scale motions.
Recent developments in SFS modeling (as discussed above) enable LES to be
confidently applied to more complex flows, including atmospheric flow over
complex terrain. Thus, many of the challenges for applying LES in complex terrain
are the same challenges that any real modeling application (e.g. RANS mesoscale
models) to the atmospheric boundary layer faces. The consequences of moving
LES to larger scales, and the practical differences between LES and RANS are
discussed in the sections below, including examples of LES applied to flow over
highly complex terrain.
10.4.3.1 Moving to Finer Resolution and Larger Scales
For a given turbulent flow, one can consider a range of modeling options, from direct
numerical simulation (DNS) where all length scales would be resolved, to LES,
which resolves much of the turbulence and only parameterizes dissipation-scale
eddy motions, to RANS which parameterizes the turbulence at all scales. With
DNS, all scales are resolved and none are modeled, thus limiting applications to
microscale flows because of computational cost limitations on domain size. With
RANS, traditionally all turbulent scales are parameterized, as all of the turbulence
is averaged out so that hu0 i D 0 thus allowing application at mesoscales. LES
falls in between, because it resolves some range of turbulence and models the rest.
When LES is pushed to higher Reynolds numbers at a fixed resolution, only the
Fig. 10.16 A schematic of the turbulence spectrum (kinetic energy E versus wavenumber k)
showing traditional extent of resolved (solid) and unresolved (dotted) for each modeling strategy.
RANS models typically operate at the mesoscale, where most if not all of the turbulence must
be modeled with a closure scheme. Direct numerical simulations (DNS) resolves all scales
down to the viscous scales, so no turbulence model is required. LES traditionally resolve most
energy containing scales, with the filtering cutoff placed somewhere in the inertial subrange.
VLES/URANS fall somewhere in between traditional RANS and LES models. With a fixed domain
size, moving from RANS to LES to DNS would require enormous increases in grid resolution and
thus computational cost. With a grid nesting strategy, typically the domain size decreases and thus
the larger flow scales (larger than the given domain) would not be able to be represented by the
LES or DNS
largest eddies are resolved and more of the energy falls in the subfilter scales,
prompting some groups to label this very large-eddy simulation (VLES). Others
have argued for an unsteady RANS (URANS) approach, which moves RANS
applications to finer resolution and time variation. A modeling hierarchy could be
formed from DNS to LES to VLES/URANS to RANS, where at each level more
of the turbulence would need to be parameterized. The distinctions between LES
and RANS methodologies lie primarily in the subfilter-scale model, as discussed in
Sect. 10.4.3.2 below. The extent of the resolved scales is shown schematically in
Fig. 10.16.
The power of the LES formulation is that it allows the use of spatially varying
filters and therefore the opportunity for seamless transitions between grid nesting
levels. Subfilter-scale turbulence models can accommodate changing grid resolution
by adjusting the chosen filter width. This is perhaps more convenient than the RANS
approach typically used for meso-scale modeling, where the turbulence model has
no direct link to grid resolution or a filter width and there is therefore little guidance
on how to adjust for different resolutions. Thus, large-eddy simulation can be used
to provide the framework for simulations at all scales, from meso-“ (20–200 km)
to meso-” (2–20 km) or urban scales where buildings are resolved with 1 m grid
spacing (see Lundquist et al. 2010).
Taking the WRF model, for example, nested grid configurations could be used
to investigate flow over the Rocky Mountains starting at traditional mesoscale
resolutions of 12 km spacing, and ending up at 100 m spacing for LES. The big
question is what to do about transitioning between RANS and LES in terms of
turbulence modeling and other physical parameterizations. This transition region
has been named the terra incognita by Wyngaard (2004), and aptly describes
the challenges faced in bridging the scales at which mesoscale and LES models
have been traditionally used. Advances in turbulence modeling (discussed in
Sect. 10.4.1.2 above) have made LES more capable of moving into this transition
region, which represents an active area of research.
In addition to issues with turbulence closure modeling, other boundary conditions
and parameterizations need improvement as the simulation is moved into the terra
incognita, particularly over complex terrain. There is evidence that traditional
surface similarity theory needs significant modification for scaling of turbulent
kinetic energy profiles in steep terrain (Weigel and Rotach 2004; Weigel et al.
2007b). Also, current implementations of lateral boundary forcing place artificial
constraints on the development of ABL turbulence (see the two-way nesting study
by Moeng et al. 2007, and the effect of lateral boundary update interval studied
by Michioka and Chow 2008). LES of the ABL, like RANS, suffers from poorly
resolved topography, and typical grid construction tends to distort flow and generate
numerical errors that must also be addressed (Lundquist et al. 2010). Finally, the
representation of land-surface processes at the bottom boundary becomes more
problematic with LES due to the lack of high-resolution surface data and the
use of empirical models which provide surface fluxes to the atmosphere. Other
challenges specific to steep terrain have been addressed in previous sections.
Thus, in addition to needing improved turbulence closure schemes, LES at coarse
resolution suffers from errors in land-surface models and in lateral boundary forcing
just like any numerical model used at large scales. While LES on the meso-scale
is mathematically allowable, the quality of the simulations decreases at coarse
resolutions unless appropriate adjustments are made. Clearly, large-eddy simulation
must address several challenges before it can be universally applied to ABL flows,
but successful simulations are possible and are described in Sect. 10.4.3.3 below.
10.4.3.2 Practical Differences Between LES and RANS
The continued improvement in grid resolution due to the increase in computer

power has begun to diminish the distinction between RANS and LES approaches in
practice, and the same code can be used for LES and RANS simply by changing the
turbulence model option. LES applied at grid spacings larger than about 1 km will
behave very similarly to the traditional Reynolds-Averaged Navier–Stokes (RANS)
approach. In fact, the only real difference between the turbulence closures applied
for each is in the representation of the turbulent length scale used in the definition
of the eddy viscosity. While RANS is strictly based on time averaging, using a
smaller time period for the average (such as in unsteady RANS, or URANS) is akin
to using a smaller filter width for the spatial average in LES. The LES approach
is, however, ultimately tied to the definition of the filter width. In numerical codes
which use eddy-viscosity closures, the distinction can be made by the choice of
the representative length scale. Thus, while a RANS TKE-1.5 closure will specify
the length scale based on the ABL depth (Sun and Chang 1986), an LES TKE-1.5
closure will use a length scale proportional to the grid spacing (Deardorff 1980).
Thus, mesoscale models, such as WRF, ARPS, and RAMS, can be run in both LES
and RANS mode simply by changing the turbulence closure model (in particular the
choice of length scale). The advantage of LES is a smoother transition between grid
resolutions because calculations are based on spatial filtering which can be used to
adjust the closure model. The challenge is in deciding which closure method is most
appropriate at particular length scales, as discussed previously.
Additionally, extra care must be used in constructing an LES grid that limits
the grid aspect ratio so that resolved eddies are not badly distorted (Chow et al.
2005; Wyngaard et al. 1998). Because the filter width which separates resolved
and subfilter scales is typically connected to the grid, a large grid aspect ratio
means that eddies would be filtered at very different scales in the horizontal vs.
the vertical and, therefore, should be avoided to allow for 3D turbulent structures
to develop accurately. The choice of grid aspect ratio also becomes important over
steep terrain, as discussed in Chap. 9 on sigma coordinate errors. Generally LES
aspect ratio requirements are not any more restrictive than the guidelines given in
Sect. 10.2.1.3 in this chapter; e.g. an aspect ratio on the order of 10 (i.e. Dx/Dz 10)
is satisfactory. Ratios on the order of 100 are used in practice but not ideal. Note that
with horizontal resolution of 10 km, a ratio of 100 effectively limits the grid spacing
near the ground to 100 m, which is much coarser than typical RANS applications at
the mesoscale. This issue of the choice of near-surface vertical grid spacing for LES
at mesoscales requires further study.
10.4.3.3 LES for Real Terrain
Despite remaining open questions regarding LES over complex terrain, several
research teams have successfully applied LES to real atmospheric flows with
topography at high resolutions. These studies have been performed primarily using
the ARPS model with time-varying lateral boundary conditions in a grid nesting
configuration typically used in mesoscale studies. As the grid is refined, various
aspects of the simulation may become important and dominate the performance of
the model. Several of these studies therefore focus on improving other parameteri-
zations (such as land-surface models) so that the LES results will be improved.
Some LES studies described here were able to explain contradictions in the field
data or show sensitivity to model parameters which had been previously overlooked.
Chen, Ludwig, and Street (2004) found that increasing the horizontal resolution
(to 250 m spacing) improved wind and potential temperature in simulations over
the mountains around the Salt Lake City area (Utah, USA) as part of the VTMX
field experiment. They were able to observe atmospheric recirculations or rotors
Fig. 10.17 Simulated cross-valley wind vectors on 22 August, 1999 in two slices across the
Riviera Valley (as indicated on the small topography panels): (left) in the northern half of the
valley and (right) close to the southern valley mouth. The cross-valley flow is shown at 1230
UTC. The shading indicates vertical wind velocity. The black line is the contour of zero vertical
velocity. Asymmetric circulations of different orientations were observed at different cross sections
of the Riviera Valley, due to thermal forcing and secondary circulations (From Weigel et al. 2006.
in the lee of the Jordan Narrows, which is at the southern end of the Salt Lake
Valley. Chow et al. (2006) found that increased horizontal grid resolution (to 150 m
spacing) also improved numerical simulation results in the Riviera Valley as part
of the Mesoscale Alpine Programme (MAP-Riviera) experiment in the Swiss Alps,
though only when land surface conditions were properly initialized. The authors
demonstrated the sensitivity of the valley boundary layer development to land-
surface properties, soil moisture distribution, topographic shading, and the choice
of turbulence model. Weigel et al. (2006) performed a heat budget analysis using
LES in the Riviera Valley, demonstrating the role of upvalley winds in re-stabilizing
the valley atmosphere despite strong thermal forcing at the ground. A secondary
circulation in the cross-valley direction was generated due to curvature in the valley
axis, explaining the contradictory downslope winds observed at surface stations
despite strong surface radiation (see Figs. 10.17 and 10.18).
Further studies include Brulfert et al. (2005), who modeled flow in two narrow
valleys in the French Alps and needed spacing as fine as 300 m to represent the
complex topography; they used a photochemistry model coupled with a mesoscale
model to predict air quality in the valleys. High-resolution simulations with time-
varying lateral boundary forcing are particularly important for accurately simulating
the detailed concentration field resulting from atmospheric tracer or contaminant
releases over complex terrain. Michioka and Chow (2008) demonstrated the ability
to apply LES in a mesoscale setting near Mt. Tsukuba in Japan using up to eight grid
nesting levels from horizontal resolutions of 45 km down to as fine as 25 m. They
found sensitivity to model configuration, such as computational mixing parameters,
lateral boundary update intervals, and choice of grid resolution, but were able to
obtain good agreement with maximum concentrations along each sampling arc
Fig. 10.18 Snapshots of online movies of slope flows in the Riviera Valley available at http://www.
ce.berkeley.edu/ chow/movies. Left: Surface winds (vectors) and solar radiation (shading) during
the sunrise transition showing shift to up-slope and up-valley flows. Right: Cross-valley vertical
cross section of the Riviera Valley, showing along valley winds (color contours, m s1 ) and cross-
valley flow direction (vectors) at various times of the day. A clockwise secondary circulation is
visible within the upvalley jet (red) in the afternoon
when compared to field data. Schmidli et al. (2009) applied ARPS in LES mode to
flow in the Owens Valley, California to investigate nocturnal thermally driven flows.
With horizontal resolution of 350 m the results showed good agreement with field
observations and explained the sensitivity of the valley flow to mid-level external
pressure gradients which on one occasion led to a unique three-layer wind structure
in the valley.
Note that the coarser grids in all of these nested domain setups are more typical
of mesoscale simulations, but can use the same LES equations. As discussed
previously, the differences between LES and RANS become small when similar
space and time resolutions are used; often the only difference in implementation
is the formulation of the turbulence model. The LES formulation is preferred
for studies of turbulent flows because it is clear which physical features (length
scales) are resolvable and which must be modeled. There are several RANS studies
(e.g. De Wekker et al. 2005) which performed simulations at similar resolutions
over complex terrain where similar issues arise, but the studies described above
specifically used LES closures.
Given that the main difference between LES and RANS applications over
complex terrain is the choice of the turbulence model, a few words are needed
to discuss differences in the resulting representation of turbulence. The potential
for improvement of predictions of turbulence structure is large given the generally
mediocre agreement of TKE values from simulations and observations in complex
terrain. Weigel and Rotach (2004) and Weigel et al. (2007a) investigated the scaling
of TKE in the Riviera Valley and showed that traditional similarity theory was
inapplicable. Rotach et al. (2004) have discussed this as well. There is a clear
need for further analysis of turbulence structure over complex terrain, but because
computational domains are limited in size, a big challenge in turbulence model
Fig. 10.19 Comparisons of velocity speedup (fraction of wind increase based on upstream
velocity) (left) and normalized turbulence intensities (right) for flow over Askervein Hill using
a standard TKE-1.5 closure, a mixed model TKE-1.5 plus reconstruction (TKE-ADM0), and a
dynamic reconstruction model (DRM-ADM0). Hill elevation is outlined at the bottom of left figure.
Flow is from left to right and HT hill top. See also http://www.ce.berkeley.edu/chow/movies for
other images (From Chow and Street 2009. © American Meteorological Society. Reprinted with
permission)
comparisons is the influence of external factors. In the real terrain configurations

described above, for example, the influence of the lateral boundary conditions
dominates the flow forcing even under high-pressure synoptic conditions where
thermal surface heating drives valley circulations. Especially in a one-way nested
configuration where the smaller domains cannot encompass details from the entire
thermally-driven valley flow network, errors in the lateral boundary conditions
propagate from one nesting level to the next. The effect of the turbulence model
in these cases is usually overshadowed by these external factors, as discussed in
Chow et al. (2006).
To investigate the role of the turbulence model more closely, semi-idealized
lateral forcing is useful. For example, Chow and Street (2009) used simplified
inflow turbulent boundary conditions to study flow over Askervein Hill, Scotland.
Using the explicit filtering mixed model approach of Chow et al. (2005), Chow and
Street (2009) showed the dynamic reconstruction model was able to match observed
TKE values much more closely compared to the standard TKE-1.5 closure (see
Fig. 10.19). The dynamic model thus shows promise for complex terrain studies.
The comparison was performed over short time scales (15–30 min) so the influence
of the turbulence model was easily apparent. A similar effect was seen in the Riviera
Valley – the turbulence model mattered at short time scales, but observation data for
such validation and comparison exercises is difficult to come by (Chow et al. 2006).
Lopes, Palma, and Castro (2007) also investigated flow over Askervein Hill using
a hybrid RANS/LES model, though their focus was on grid aspect ratio and effects
on recirculation in the lee of the hill [see also Golaz et al. (2009)].
Guidelines for lateral grid nesting procedures are heuristic [see e.g., Warner,
Peterson, and Treadon (1997)] and not well established, particularly for LES
(Moeng et al. 2007). In two way nesting over complex terrain, the nesting algorithm
will need to allow changes in the vertical grid resolution (to keep the grid aspect ratio
close to unity). Furthermore, the size of turbulent eddies that can be represented will
vary by a factor of 3 across a typical 3-to-1 grid nesting boundary. It is unknown how
simulated turbulence will behave under these circumstances, but these issues are
likely to be critically important given that the grid is refined precisely in regions
where more detailed predictions of turbulent stress are required. For example,
Michioka and Chow (2008) found a dependence of turbulence spectra on lateral
boundary update interval and distance from the lateral boundary showing that while
high frequency motions can be generated locally, they are also generally affected by
the boundary update interval.
10.5 Future Research Needs
Since their incipience in the 1970s, mesoscale numerical models have gone through
considerable improvements and have become an essential tool for a wide range
of applications ranging from operational forecasting of regional weather and air
pollution to studies of regional climate and climate change. Over the last two
decades [in particular since the publication of the last mountain meteorology
review Blumen (1990)], mesoscale models have made a tremendous contribution
to the understanding of atmospheric dynamics. In the current state of mountain
meteorology, numerical models are considered an indispensable component in
any prediction system, whether for hindcasting, nowcasting, or forecasting. For
example, field studies can now be designed to take advantage of high resolution
models, even in complex terrain, to guide field operations and analysis (e.g. T-REX
in 2006). The Winter Olympics now rely on detailed, highly local forecasts in
complex terrain for storm conditions. Even operational weather forecasting models
for mountain regions can now resolve some of the more narrow valleys (2 km in
scale) which were before invisible to the models.
Despite the steady progress towards more accurate forecasts or simulations, ap-
plications of mesoscale models are especially challenging over regions of complex
terrain. This can be attributed to both the various numerical issues that arise with
complex terrain and to the lack of observational data for model initialization and
validation. The modeling challenges include sharp orographic gradients and the
associated heterogeneity in vegetation, soil properties, and landuse that are often
under-resolved by current model resolution. In addition, there are numerical errors
associated with calculations of horizontal gradients and violations of some under-
lying assumptions used in many physical parameterization schemes. Compared to
flat terrain, observational data in complex terrain are often sparse and the existing
observations hardly represent the mean properties over an area of model grid cell,
thus making direct comparisons between model and observations more problematic
than over flat terrain.
Some of these challenges will likely be addressed as model resolution increases

considerably in the next decade with the continued advancement of high perfor-
mance computing. As pointed out earlier in this chapter, the benefits of increased
model horizontal resolution can only be fully realized if the increase in model
resolution is also matched by the increase in resolution of the geographic data such
as topography, vegetation, and soil characteristics, especially soil moisture, that are
used to define lower boundary conditions. It is also important that model perfor-
mance with increasing horizontal and vertical resolution be carefully evaluated to
avoid creating a false sense of improvement, because increased resolution may also
raise new issues with existing coordinate systems and model parameterizations.
Clearly, more studies are needed to understand and document model performance
and behaviors at high spatial resolution, especially at the intermediate scale or
the scale of terra incognita (Wyngaard 2004). Studies are also needed to address
possible issues that may arise in applying current physical parameterizations at
scales smaller than those intended. Finally, the use of ensemble forecast techniques
(see discussions in Chap. 9) is a promising means of dealing with uncertainty due to
model configuration and model errors and it will significantly advance both forecasts
and simulations of processes over complex terrain.
Acknowledgments The authors are grateful for the support of National Science Foundation
Grants ATM-0646299 and 0837860 (for SZ) and 0645784 (for FKC) (Physical and Dynamic
Meteorology Program). A special thanks to Dr. Jerome Fast at the Pacific Northwest National
Laboratory for preparing several of the figures used in the chapter and for useful discussions. The
authors also thank the anonymous reviewers for their thorough review and constructive comments,
as well as Mike Kiefer, Megan Daniels, Nikola Marjanovic, Jason Simon, and Bowen Zhou for
their careful proofreading and suggestions.
References
Anderson, R, and C. Meneveau. 1999. Effects of the similarity model in finite-difference LES of
isotropic turbulence using a Lagrangian dynamic mixed model. Flow, Turb. and Combustion.
62,: 201–225.
André, J. C., J. P. Goutorbe and A. Perrier, 1986: HAPEX-MOBLIHY: a hydrological atmospheric
experiment for the study of water budget and evaporation flux at the climatic scale. Bull. Amer.
Meteor. Soc., 67, 138–144.
Andre, A., 1991: A TKE-dissipation model for the atmospheric boundary layer. Bound.-Layer
Meteorol., 56, 207–221
Andren, A., A. R. Brown, J. Graf, P. J. Mason, C.-H. Moeng, F. T. M. Nieuwstadt, and
U. Schumann. 1994: Large-eddy simulation of a neutrally stratified boundary layer: A com-
parison of four computer codes. Q. J. R. Meteorol. Soc. 120, 1457–1484.
Apsley, D. D., and I. P. Castro, 1997: A limited length scale k-" model for the neutral and stably
stratified atmospheric boundary layer. Bound.-Layer Meteorol. 83, 75–98.
Banta, R. M., L. S. Darby, J. D. Fast, J. O. Pinto, C. D. Whiteman, W. J. Shaw, B. W. Orr, 2004:
Nocturnal Low-Level Jet in a Mountain Basin Complex. Part I: Evolution and Effects on Local
Flows. J. Appl. Meteor., 43, 1348–1365
Banta, R. M. and R. T. Gannon, 1995: Influence of soil moisture on simulations of katabatic flow.
Theor. Appl. Climatol. 52, 85–94.
Banta, R. M., and A. B. White, 2003: Mixed-height differences between land use types: Depedence
on wind speed. J. Geophys. Res. 108, No. D10,4321
Banta, R., L. Olivier, P. Gudiksen, and R. Lange, 1996: Implications of Small-Scale Flow Features
to Modeling Dispersion over Complex Terrain. J. Appl. Meteor., 35, 330–342.
Bardina, J., J. H. Ferziger, and W. C. Reynolds, 1983: Improved turbulence models based on
large eddy simulation of homogeneous, incompressible, turbulent flows. Technical Report.
Department of Mechanical Engineering, Stanford University, Stanford, California.
Basu, S, and F Porte-Agel, 2006: Large-eddy simulation of stably stratified atmospheric bound-
ary layer turbulence: A scale-dependent dynamic modeling approach. J. Atmos. Sci., 63,
2074–2091.
Beare, RJ, MK MacVean, AAM Holtslag, J Cuxart, I Esau, JC Golaz, MA Jimenez, et al. 2006: An
intercomparison of large-eddy simulations of the stable boundary layer. Bound.-Layer Meteor.,
118, 247–272.
Bélair, S. P. Lacarrère, J. Noilhan, V. Masson, and J. Stein, 1998: High-Resolution Simulation of
Surface and Turbulent Fluxes during HAPEX-MOBILHY. Mon. Wea. Rev., 126: 2234–2253.
Benoit, R., M. Desgagné, P. Pellerin, S. Pellerin, Y. Chartier, and S. Desjardins, 1997: The
Canadian MC2: A semi-Lagrangian, semi-implicit wideband atmospheric model suited for
finescale process studies and simulation. Mon. Wea. Rev., 125, 2382–2415.
Berg, L.K., and S. Zhong, 2005: Sensitivity of MM5-Simulated Boundary Layer Characteristics to
Turbulence Parameterizations. J. Appl. Meteor., 44, 1467–1483.
Bergström, H., and N. Juuso, 2006: A study of valley winds using the MIUU meso-scale model.
Wind Energy, 9, 109–129.
Betts, A. K., and J. H. Ball, 1992: FIFE-1987 mean time series, Data diskette. (available from
Atmospheric Research, R.D. #3, Box 3125, Pittsford, VT 05763.
Betts, A. K., J. I. Ball, and Beljaars, 1993: Comparison between the land surface response of the
ECMWF model and the FIFE-1987 data. Quart. J. Roy. Meteor. Soc., 119, 975–1001.
Bischoff-Gauß, I. N. Kalthoff, S. Khodayar, M. Fiebig-Wittmaack, S. Montecinos, 2006: Model
simulations of the boundary-layer evolution over an arid Andes valley, Bound.-Layer Meteor.,
128, 357–379.
Black, T. L., 1994: The new NMC mesoscale Eta Model: Description and forecast examples. Wea.
Blackadar A. K., 1979: High resolution models of the planetary boundary layer. Advances in
Environmental Science and Engineering, J. Pfafflin and E. Ziegler, Eds., Vol. 1, Gordon and
Breach, 50–85.
Blumen, W. (ed.), 1990: Atmospheric Processes Over Complex Terrain. Meteor. Monogr., Vol. 23,
No. 45. Amer. Meteor. Soc., Boston, MA, 323pp
Boone, A. B. Decharme, F. Guichard, P. de Rosnay, G. Balsamo, A. Beljaars, F. Chopin, T. Orgeval,
J. Polcher, C. Delire, A. Ducharne, S. Gascoin, M. Grippa, L. Jarlan, L. Kergoat, E. Mougin,
Y. Gusev, O. Nasonova, P. Harris, C. Taylor, A. Norgaard, I. Sandholt, C. Ottlé, I. Poccard-
Leclercq, S. Saux-Picart, and Y. Xue, 2009: The AMMA land surface model intercomparison
project (ALMIP), Bull. Amer. Meteor. Soc., 90, 1865–1880.
———, F. Habets, J. Noilhan, D. Clark, P. Dirmeyer, S. Fox, Y. Gusev, I. Haddeland, R. Koster,
D. Lohmann, S. Mahanama, K. Mitchell, O. Nosonova, G.-Y. Neu, A. Pitman, J. Polcher, A. B.
Shmakin, K. Tanaka, B. van den Hurk, S. Verant, D. Verseghy, P. Viterbo, and Z.-L. Yang, 2004:
The Rhône-aggregation land surface scheme intercomparison project: An overview. J. Climate,
17, 187–208.
Bou-Zeid, E, C Meneveau, and M Parlange. 2005: A scale-dependent Lagrangian dynamic model
for large eddy simulation of complex turbulent flows. Phys. Fluids, 17.
Bowling, L. C., and Coauthors 2003: Simulation of high latitude hydrological processes in the
Torne-Kalix basin: PILPS Phase 2e. 1: Experimental description and summary intercompar-
isons. Global Planet.Change, 38, 1–30.
Brown, A. R., J. M. Hobson, and N. Wood, 2001: Large-eddy simulation of neutral turbulent flow
over rough sinusoidal ridges. Bound.-Layer Meteor., 98, 411–441.
Brulfert, G, C Chemel, E Chaxel, and JP Chollet, 2005: Modelling photochemistry in alpine

valleys. Atmos. Chem. Phys., 5, 2341–2355.
Burk, S.D., and W.T. Thompson, 1989: A vertically nested regional numerical weather prediction
model with second-order closure physics. Mon. Wea. Rev., 117, 2305–2324.
Cabot W. and P. Moin, 2000: Approximate wall boundary conditions in the large-eddy simulation
of high Reynolds number flow. Flow, Turb. and Combustion. 63, 269–291.
Cederwall, R. T., 2001: Large-eddy simulation of the evolving stable boundary layer over flat
terrain. Stanford University.
———, and R. L. Street, 1999: Turbulence modification in the evolving stable boundary layer: a
large-eddy simulation. 13th Symposium on Boundary Layers and Turbulence, Amer. Meteorol.
Soc.: 223–226.
Chen, F., K. Mitchell, J. Schaake, Y. Xue, H.-L. Pan, V. Koren, Q.Y. Duan, M. Ek, and A. Betts,
1996: Modeling of land-surface evaporation by four schemes and comparison with FIFE
observations. J. Geophys. Res., 101, 7251–7268.
Chen, T. H. and Coauthors, 1997: Cabauw experimental results from the project for intercompari-
son of land-surface parameterization schemes. J. Climate, 10, 1194–1215.
Chen, Y., F. L. Ludwig, and R. L. Street, 2004: Stably stratified flows near a notched transverse
ridge across the Salt Lake Valley. J. Appl. Meteor. 43, 1308–1328.
Chin, S. H-N, M.J. Leach, G. A. Sugiyama, and J. A. Fernando, 2001: A preliminary study of
surface cold bias in COAMPS. Preprints, Ninth Conference on Mesoscale Processes, Amer.
Meteorol. Soc.: 30 July – 2 August, Fort Lauderdale, FL.
Chow, F.K., R. L. Street, M. Xue, and J. H. Ferziger, 2005: Explicit filtering and reconstruction
turbulence modeling for large-eddy simulation of neutral boundary layer flow. J. Atmos. Sci.,
62, 2058–2077.
———, A. P. Weigel, R. L. Street, M. W. Rotach, and M. Xue. 2006: High-resolution large-eddy
simulations of flow in a steep Alpine valley. Part I: Methodology, verification, and sensitivity
studies. J. Appl. Meteor. Climatol., 45, 63–86.
———, and R. L. Street, 2009: Evaluation of turbulence closure models for large-eddy simulation
over complex terrain: flow over Askervein Hill. J. Appl. Meteor. Climatol. 48, 1050–1065.
Colette, A., F.K. Chow, and R.L. Street, 2003: A Numerical Study of Inversion-Layer Breakup and
the Effects of Topographic Shading in Idealized Valleys. J. Appl. Meteor., 42, 1255–1272.
Colle, B.A., K.J. Westrick, and C.F. Mass, 1999: Evaluation of MM5 and Eta-10 Precipitation
Forecasts over the Pacific Northwest during the Cool Season. Wea. Forecasting, 14, 137–154.
Côté, J., J.G. Desmarais, S. Gravel, A. Méthot, A. Patoine, M. Roch, and A. Staniforth, 1998:
The Operational CMC–MRB Global Environmental Multiscale (GEM) Model. Part II: Results.
Mon. Wea. Rev., 126, 1397–1418
———, S. Gravel, A. Méthot, A. Patoine, M. Roch, and A. Staniforth, 1998: The Operational
CMC–MRB Global Environmental Multiscale (GEM) Model. Part I: Design Considerations
and Formulation. Mon. Wea. Rev., 126, 1373–1395.
Cotton, W.R., G. Thompson, and P.W. Mieike, 1994: Real-Time Mesoscale Prediction on
workstations. Bull. Amer. Meteor. Soc., 75, 349–362.
Cullen, M. J. P., 1993: The unified forecast/climate model. Meteor. Mag., 122, 81–94.
Dai, Y. X. Zeng, R. E. Dickinson, I. Baker, G. B. Bonan, M. G. Bosilovich, A. S. Denning, P. A.
Dirmeyer, P. R. Houser, G. Niu, K. W. Oleson, C. A. Schlosser, Z-L Yang, 2003: The Common
Land Model, Bull. Amer. Meteor. Soc., 84, 1013–1023
Daniels, M. H., F. K. Chow, and G. S. Poulos, 2006: Effects of soil moisture initialization
on simulations of atmospheric boundary layer evolution in owens valley. Paper 7.2. 12th
Conference on Mountain Meteorology, Amer. Meteor. Soc.
Deardorff, J. W., 1980: Stratocumulus-capped mixed layers derived from a 3-dimensional model.
Delege, Y. 1997: Parameterising sub-grid scale vertical transport in atmospheric models under
statistically stable conditions. Bound.-Layer Meteor. 82, 23–48.
Deng, A., D. R. Stauffer, J. R. Zielonka and G. K. Hunter, 2008: A Comparison of high-resolution
mesoscale forecasts using MM5 and WRF-ARW. The 9th WRF Users’ Workshop, June 23–27,
Boulder, CO
Denis, B., R. Laprise, and D. Caya, 2003: Sensitivity of a regional climate model to the resolution
of the lateral boundary conditions. Climate Dynamics, 20, 107–126.
De Wekker, S. F. J., 2002: Structure and morphology of the convective boundary layer in
mountainous terrain. Dissertation, University of British Columbia, Department of Earth and
Ocean Sciences, Vancouver, Canada, 191 pp.
———, 2008: Observational and Numerical Evidence of Depressed Convective Boundary Layer
Heights near a Mountain Base. J. Appl. Meteor. Clim. 47, 1017–1026.
———, S. Zhong, J. D. Fast, and C. D. Whiteman, 1998: A numerical study of the thermally
driven plain-to-basin wind over idealized basin topographies. J. Appl. Meteor., 37, 606–622.
———, D.G. Steyn, J.D. Fast, M.W. Rotach, and S. Zhong, 2005: The performance of RAMS in
representing the convective boundary layer structure in a very steep valley. Env. Fluid Mech. 5,
35–62.
Deardorff, J. W. 1970: A numerical study of three-dimensional turbulent channel flow at large
Reynolds numbers. J. Fluid Mech. 41, 453–480.
———, 1972. Numerical investigation of neutral and unstable planetary boundary layers. J. Atmos.
Sci., 29, 91–115.
———. 1978: Efficient prediction of ground surface temperature and moisture with inclusion of
layer vegetation. J. Geophy. Res., 83, 1889–1903.
———. 1980: Stratocumulus-capped mixed layers derived from a 3-dimensional model. Bound.-
Layer Meteor. 18, 495–527.
Dickinson, R. E., Henderson-Sellers, A., Kennedy, P. J. and Wilson, M. F. 1986: Biosphere-
atmosphere transfer scheme (BATS) for the NCAR Community Climate Model. National
Center for Atmospheric Research, Boulder, CO. Tech Note/TN-275CSTR.
Ding, L., R. Calhoun, and R. Street, 2003: Numerical simulation of strongly stratified flow over a
three-dimensional hill. Bound.-Layer Meteor., 107, 81–114.
Ding, L., and R. L. Street, 2003: Numerical study of the wake structure behind a three-dimensional
hill. J. Atmos. Sci., 60, 1678–1690.
Esau, I., 2004: Simulation of Ekman boundary layers by large eddy model with dynamic mixed
subfilter closure. Environmental Fluid Mechanics, 4, 273–303.
Doran, J.C., and S. Zhong, 2000: Thermally Driven Gap Winds into the Mexico City Basin. J. Appl.
Meteor., 39, 1330–1340.
———, J.D. Fast, and J. Horel, 2002: The VTMX 2000 Campaign. Bull. Amer. Meteor. Soc., 83,
537–551.
Doyle, J.D., and D.R. Durran, 2007: Rotor and Subrotor Dynamics in the Lee of Three-
Dimensional Terrain. J. Atmos. Sci., 64, 4202–4221
Dubayah, R., and S. Loechel, 1997: Modeling topographic solar radiation using GOES data.
J. Appl. Meteor., 36, 141–154.
Dudhia, J. 1989: Numerical study of convection observed during the winter monsoon experiment
using a mesoscale two-dimensional model. J. Atmos. Sci. 46, 3077–3107.
Ek, M. B., K. E. Mitchell, Y. Lin, E. Rogers, P. Grunmann, V. Koren, G. Gayno, and J. D.
Tarpley, 2003: Implementation of Noah land surface model advances in the National Centers
for Environmental Prediction operational mesoscale Eta model, J. Geophys. Res., 108(D22),
8851.
Fast, J.D., 2003: Forecasts of Valley Circulations Using the Terrain-Following and Step-Mountain
Vertical Coordinates in the Meso-Eta Model. Wea. Forecasting, 18, 1192–1206.
———, and L.S. Darby, 2004: An Evaluation of Mesoscale Model Predictions of Down-Valley
and Canyon Flows and Their Consequences Using Doppler Lidar Measurements during VTMX
2000. J. Appl. Meteor., 43, 420–436.
Ferziger, J.H. and M. Peric. 2002: Computational Methods for Fluid Dynamics. Springer-Verlag,
3rd edition: Berlin.
Findikakis, A.N. and R.L. Street. 1979: Algebraic model for subgrid-scale turbulence in stratified
flows. J. Atmos. Sci., 36, 1934–1949.
Freedman, F. R., and M. Z. Jacobson, 2002: Modification of the standard "-equation for the
stable ABL through enforced consistency with Monin-Obukhov similarity theory. Bound.-
Layer Meteor., 106, 383–410.
Fu, P., S. Zhong, C. D. Whiteman, T. M. Horst, and X. Bian, 2010: An observational study of
turbulence inside a closed basin. J. Geophys. Res., 115, D23106, 15 pp.
Gallus, W. A., 2000: The impact of step orography on flow in the Eta model: two contrasting
examples. Weather and Forecasting, 15, 630–637.
Germano, M., U. Piomelli, P. Moin, and W. H. Cabot, 1991: A dynamic subgrid-scale eddy
viscosity model. Phys. Fluids, 3, 1760–1765.
Ghosal, S., T. S. Lund, P. Moin, and K. Akselvoll, 1995: A dynamic localization model for large-
eddy simulation of turbulent flows. J. Fluid Mech., 286, 229–255.
Golaz, J., J. D. Doyle, and S. Wang, 2009: One-Way Nested Large-Eddy Simulation over the
Askervein Hill. J. Adv. Model Earth Syst., 1, 1–6.
Grell, G. A., J. Dudhia, and D. R. Stauffer, 1994: A description of the fifthgeneration Penn
State/NCAR mesoscale model (MM5), NCAR Tech. Not., NCAR/TN-397CIA, 114 pp.
Grubišic, V., J.D. Doyle, J. Kuettner, S. Mobbs, R.B. Smith, C.D. Whiteman, R. Dirks, S. Czyzyk,
S.A. Cohn, S. Vosper, M. Weissmann, S. Haimov, S.F.J. De Wekker, L.L. Pan, and F.K. Chow,
2008: The Terrain-Induced Rotor Experiment. Bull. Amer. Meteor. Soc., 89, 1513–1533.
Gullbrand, J., and F. K. Chow, 2003: The effect of numerical errors and turbulence models in
large-eddy simulations of channel flow, with and without explicit filtering. J. Fluid Mech., 495,
323–341.
Hanna, S. R. and R. Yang, 2001: Evaluations of mesoscale models’ simulations of near-surface
winds, temperature gradients, and mixing depths. J. Appl. Meteor., 40, 1095–1104.
Harlow, F. H. and Nakayama, P. I. 1967. Turbulence transport equations, Phys. Fluids 10, 2323.
Hauge, G., and L. R. Hole, 2003: Implementation of slope irradiance in Mesoscale Model version
5 and its effect on temperature and wind fields during the breakup of a temperature inversion.
J. Geophys. Res., 108, D2, 4058.
Henderson-Sellers, A., Z. L. Yang, and R. Dickinson, 1993: The Project for Intercomparison of
Land-surface Parameterization Schemes. Bull. Amer. Meteor. Soc., 74, 1335–1349.
———, A. J. Pitman, P. K. Love, P. Irannejad, and T. Chen, 1995: The Project for Intercomparison
of Land Surface Parameterization Schemes (PILPS): Phases 2 and 3. Bull. Amer. Meteor. Soc.,
76, 489–503.
Hodur, R.M., 1997: The Naval Research Laboratory’s Coupled Ocean/Atmosphere Mesoscale
Prediction System (COAMPS). Mon. Wea. Rev., 125, 1414–1430.
Holt, T., and S. Raman, 1988: A review and comparative evaluation of multilevel boundary layer
parameterization for first-order and turbulent kinetic energy closure schemes. Rev. Geophys.,
26, 761–780.
Holtslag, A. A. M., and B. A. Boville, 1993: Local versus nonlocal boundary-layer diffusion in a
global climate model. J. Climate, 6, 1825–1842.
Hong, S.Y., and H.L. Pan, 1996: Nonlocal Boundary Layer Vertical Diffusion in a Medium-Range
Forecast Model. Mon. Wea. Rev., 124, 2322–2339.
———, Y. Noh, and J. Dudhia, 2006: A New Vertical Diffusion Package with an Explicit
Treatment of Entrainment Processes. Mon. Wea. Rev., 134, 2318–2341.
Horel, J. D., and C. V. Gibson, 1994: Analysis and simulation of winter storm over Utah, Weather
and Forecasting, 9, 479–494.
Horst, T., J. Kleissl, D. Lenschow, C. Meneveau, C. Moeng, M. Parlange, P. Sullivan, and J. Weil,
2004: HATS: Field observations to obtain spatially filtered turbulence fields from crosswind
Arrays of sonic anemometers in the atmospheric surface layer. J. Atmos. Sci., 61, 1566–1581.
Huang, C.-Y., and S. Raman, 1989: Application of the E-" closure model to simulations of
mesoscale topographic effects. Bound.-Layer Meteor., 49, 169–195.
Iizuka, S., and H. Kondo, 2004: Performance of various sub-grid scale models in large-eddy
simulations of turbulent flow over complex terrain. Atmos. Envir., 38, 7083–7091.
Iqbal, M, 1983: An Introduction to Solar Radiation, Academic, San Diego, California.
Jacquemin B. and J. Noilhan, 1990: Sensitivity study and validation of a land surface parameteri-
zation using the HAPEX-MOBILHY data set. Bound. Layer. Meteor., 52, 93–134.
Janjić, Z. I., 1994: The Step-Mountain Eta Coordinate Model: Further Developments of the
Convection, Viscous Sublayer, and Turbulence Closure Schemes. Mon. Wea. Rev. 122,
927–945.
Juneja, A., and J. G. Brasseur, 1999: Characteristics of subgrid-resolved-scale dynamics in
anistropic turbulence, with application to rough-wall boundary layers. Phys. Fluids, 11,
3054–3068.
Kăllberg, P. Ed., 1990: The HIRLAM level 1 system. Documentation manual,160 pp. [Available
from SMHI, S 60176 Norrköping, Sweden]
Kirkpatrick, M. P., A. S. Ackerman, D. E. Stevens, and N. N. Mansour, 2006: On the application
of the dynamic Smagorinsky model to large-eddy simulations of the cloud-topped atmospheric
boundary layer. J. Atmos. Sci., 63, 526–546.
Klemp, J. B. and D. K. Lilly, 1978: Numerical simulation of hydrostatic mountain waves. J. Atmos.
Sci., 35, 78–107.
Klemp, J. B. and D. R. Durran, 1983: An upper boundary condition permitting internal gravity
wave radiation in numerical mesoscale models. Mon. Wea. Rev., 111, 403–444.
Klemp, J., Dudhia, J. & Hassiotis, A., 2008. An Upper Gravity-Wave Absorbing Layer for NWP
Applications. Mon. Wea. Rev., 136(10), 3987–4004.
Kondo, J., and N. Okusa, 1990: A Simple Numerical Prediction Model of Nocturnal Cooling in a
Basin with Various Topographic Parameters. J. Appl. Meteor., 29, 604–619.
Kosović, B., 1997: Subgrid-scale modelling for the large-eddy simulation of high-Reynolds-
number boundary layers. J. Fluid Mech., 336, 151–182.
Kosović, B., and J. A. Curry, 2000: A large eddy simulation study of a quasi-steady, stably stratified
atmospheric boundary layer. J. Atmos. Sci., 57, 1052–1068.
Koster, R.D., and P.C.D. Milly, 1997: The Interplay between Transpiration and Runoff Formula-
tions in Land Surface Schemes Used with Atmospheric Models. J. Climate, 10, 1578–1591.
Kuwagata, T., and F. Kimura, 1997: Daytime Boundary Layer Evolution in a Deep Valley. Part II:
Numerical Simulation of the Cross-Valley Circulation. J. Appl. Meteor., 36, 883–895.
Lacser A. and S. P. S. Arya, 1986: A comparative assessment of mixing-length parameterizations
in the stably stratified nocturnal boundary layer (NBL). Bound.-Layer Meteor., 36, 53–70.
Lafore, J. P., J. Stein, N. Asencio, P. Bougeault, V. Ducrocq, J. Duron, C. Fischer, P. Héreil, P.
Mascart, V. Masson, J. P. Pinty, J. L. Redelsperger, E. Richard and J. Vilà-Guerau de Arellano,
1997: The Meso-NH atmospheric simulation system. Part I: adiabatic formulation and control
simulations. Annals Geophysicae. 16, 90–109.
Lesieur, M., Metais, O., and P. Comte, 2005: Large-eddy simulations of turbulence. Cambridge
University Press, Cambridge.
Lilly, D. K., 1962: On the numerical simulation of buoyant convection. Tellus, 14, 148–172.
Lilly, D. K., 1992: A proposed modification of the Germano subgrid-scale closure method. Phys.
Fluids, 4, 633–635.
Lopes, A., J. Palma, and F. Castro, 2007: Simulation of the askervein flow. Part 2: Large-eddy
simulations. Bound.-Layer Meteor., 125, 85–108.
Ludwig, F., F. Chow, and R. Street, 2009: Effect of Turbulence Models and Spatial Resolution
on Resolved Velocity Structure and Momentum Fluxes in Large-Eddy Simulations of Neutral
Boundary Layer Flow. J. Appl. Meteor. Clim., 48, 1161–1180.
Lundquist, K.A., Chow, F.K., and J.K. Lundquist. 2010. An immersed boundary method for the
Weather Research and Forecasting model. Mon. Wea. Rev., 138(3), 796–817.
Mahrer, Y., 1984: An Improved Numerical Approximation of the Horizontal Gradients in a Terrain-
Following Coordinate System, Mon. Wea. Rev., 112, 918–922.
Mahfouf, J. F., E. Richard, P. Mascart, E. C. Nickerson, and R. Rosset, 1987: A comparative study
of various parameterization of the planetary boundary layer in a numerical mesoscale model,
J. Appl. Meteor.., 26, 1671–1695.
Mason, P. J., and D. J. Thomson, 1992: Stochastic backscatter in large-eddy simulations of
boundary layers. J. Fluid Mech., 242, 51–78.
Mason, P., and S. Derbyshire, 1990: Large-Eddy Simulation of the Stably-Stratified Atmospheric
Boundary-Layer. Bound.-Layer Meteor., 53, 117–162.
Mass, C.F., and Y.H. Kuo, 1998: Regional Real-Time Numerical Weather Prediction: Current
Status and Future Potential. Bull. Amer. Meteor. Soc., 79, 253–263.
———,D. Ovens, K. Westrick, and B.A. Colle, 2002: Does Increasing Horizontal Resolution
Produce More Skillful Forecasts? Bull. Amer. Meteor. Soc., 83, 407–430.
———, M. Albright, D. Ovens, R. Steed, M. MacIver, E. Grimit, T. Eckel, B. Lamb, J. Vaughan, K.
Westrick, P. Storck, B. Colman, C. Hill, N. Maykut, M. Gilroy, S.A. Ferguson, J. Yetter, J.M.
Sierchio, C. Bowman, R. Stender, R. Wilson, and W. Brown, 2003: Regional Environmental
Prediction Over the Pacific Northwest. Bull. Amer. Meteor. Soc., 84, 1353–1366.
Matzinger, N., M. Andretta, E. Van Gorsel, R. Vogt, A. Ohmura, and M. W. Rotach, 2003: Surface
radiation budget in an alpine valley. Quart. J.Roy. Meteor. Soc., 129, 877–895.
Maxwell, R.M., Chow, F.K., and S.J. Kollet. 2007. The groundwater-land-surface-atmosphere con-
nection: soil moisture effects on the atmospheric boundary layer in fully-coupled simulations.
Adv. in Water Res. 30, 2447–2466.
Mellor, G. L., and T. Yamada, 1982: Development of a turbulence closure model for geophysical
fluid problems, Rev. Geophys., 20, 851– 875.
———, and ———, 1974: A Hierarchy of Turbulence Closure Models for Planetary Boundary
Layers. J. Atmos. Sci., 31, 1791–1806.
Meneveau, C., T. Lund, and W. Cabot, 1996: A Lagrangian dynamic subgrid-scale model of
turbulence. J. Fluid Mech., 319, 353–385.
Mesinger, F., G. DiMego, E. Kalnay, K. Mitchell, P.C. Shafran, W. Ebisuzaki, D. Jović, J. Woollen,
E. Rogers, E.H. Berbery, M.B. Ek, Y. Fan, R. Grumbine, W. Higgins, H. Li, Y. Lin, G. Manikin,
D. Parrish, and W. Shi, 2006: North American Regional Reanalysis, Bull. Amer. Meteorol. Soc.,
87, 343–360.
Michioka, T., and F. Chow, 2008: High-Resolution Large-Eddy Simulations of Scalar Transport in
Atmospheric Boundary Layer Flow over Complex Terrain. J. Appl. Meteor. Clim., 47, 3150–
3169.
Moeng, C. -H., 1984: A large-eddy-simulation model for the study of planetary boundary-layer
turbulence. J. Atmos. Sci. 41, 2052–2062.
Moeng, C., J. Dudhia, J. Klemp, and P. Sullivan, 2007: Examining two-way grid nesting for large
eddy simulation of the PBL using the WRF model. Mon. Wea. Rev., 135, 2295–2311.
———, and J.C. Wyngaard, 1984: Statistics of Conservative Scalars in the Convective Boundary
Layer. J. Atmos. Sci., 41, 3161–3169.
———, and ———, 1989: Evaluation of Turbulent Transport and Dissipation Closures in Second-
Order Modeling. J. Atmos. Sci., 46, 2311–2330
———, J Dudhia, J Klemp, and P Sullivan, 2007: Examining two-way grid nesting for large eddy
simulation of the PBL using the WRF model. Mon. Wea. Rev., 135, 2295–2311.
Müller, M.D., and D. Scherer, 2005: A Grid- and Subgrid-Scale Radiation Parameterization of
Topographic Effects for Mesoscale Weather Forecast Models. Mon. Wea. Rev., 133, 1431–1442.
Nakayama, A., K. Hori, and R. L. Street, 2004a: Filtering and large eddy simulation of flow
over irregular rough surface. Proc. 2004 Summer Program, Center for Turbulence Research,
Stanford, CA, NASA Ames–Stanford University, 145–156.
Nakayama, A., H. Noda, and K. Maeda, 2004b: Similarity of instantaneous and filtered velocity
fields in the near wall region of zero-pressure gradient boundary layer. Fluid Dynamics
Research 35, 299–321.
Noh, Y., W. G. Cheon, S.-Y. Hong, and S. Raasch, 2003: Improvement of the K-profile model for
the planetary boundary layer based on large eddy simulation data. Bound.-Layer Meteor., 107,
401–427.
Noilhan, J. and S. Planton, 1989: A simple parameterzation of land surface process for mesoscale
models. Mon. Wea. Rev., 117, 536–549.
Noppel, H., and F. Fiedler, 2002: Mesoscale Heat Transport Over Complex Terrain By Slope
Winds – A Conceptual Model And Numerical Simulations. Bound.-Layer Meteor., 104,
73–97.
Nutter, P., M. Xue, and D. Stensrud, 2004: Application of Lateral Boundary Condition Perturba-
tions to Help Restore Dispersion in Limited-Area Ensemble Forecasts. Mon. Wea. Rev., 132,
2378–2390.
Oliphant, A. J., R. A. Spronken-Smith, A. P. Sturman, and I. F. Owens, 2003: Spatial variability of
surface radiation fluxes in mountainous terrain. J. Appl. Meteor., 42, 113–128.
Oncley, S. P., T. Foken, R. Vogt, C. Bernhofer, W. Kohsiek, H. Liu, A. Pitacco, D. Grantz, and L.
Riberio, 2002: The energy balance experiment EBEX-2000. 25th Conference on Agriculture
and Forest Meteorology. Amer. Meteor. Soc., May 19–24, Norfolk, VA.
Ookouchi, Y., M. Segal, R.C. Kessler, and R.A. Pielke, 1984: Evaluation of Soil Moisture
Effects on the Generation and Modification of Mesoscale Circulations. Mon. Wea. Rev., 112,
2281–2292.
Patton, E.G., P.P. Sullivan, and C.H. Moeng, 2005: The Influence of Idealized Heterogeneity on
Wet and Dry Planetary Boundary Layers Coupled to the Land Surface. J. Atmos. Sci., 62,
2078–2097.
Pielke, R. A., 2002: Mesoscale meteorological modeling. Academic Press, 704 pp.
Pielke, R. A., W. R. Cotton, R. L. Walko, C. J. Tremback, W. A. Lyons, L. D. Grasso, M. E.
Nicholls, M. D. Moran, D. A. Wesley, T. J .Lee, and J. H. Copeland, 1992: A comprehensive
meteorological modeling system – RAMS. Meteor. Atmos. Phys., 49, 69–91.
Piomelli, U., and E. Balaras, 2002: Wall-layer models for large-eddy simulations. Annual Review
of Fluid Mechanics, 34, 349–374.
Pitman, A. J., 2003: The evolution of, and revolution in, land surface schemes designed for climate
models. Int. J. Climatol., 23, 479–510.
Pleim, J. E. and J. S. Chang, 1992: A non-local closure model for vertical mixing in the convective
boundary layer. Atmos. Envir., 26, 965–981.
Pope, S. B. 2000: Turbulent flows. Cambridge University Press, Cambridge.
Porte-Agel, F., C. Meneveau, and M. B. Parlange, 2000: A scale-dependent dynamic model for
large-eddy simulation: application to a neutral atmospheric boundary layer. J. Fluid Mech.,
415, 261–284.
Poulos, G.S., J. E. Bossert, T. B. McKee, R. A. Pielke, 2000: The Interaction of Katabatic
Flow and Mountain Waves. Part I: Observations and Idealized Simulations, J. Atmos. Sci, 57,
1919–1936,
Rotach, M. W. and Coauthors, 2004: Turbulence structure and exchange processes in an Alpine
Valley: The Riviera Project. Bull. Amer. Met. Soc., 85, 1367–1385.
Sagaut, P., 2002: Large eddy simulation for incompressible flows. Springer-Verlag, 2nd edition:
Berlin.
Saiki, E., C. Moeng, and P. Sullivan, 2000: Large-eddy simulation of the stably stratified planetary
boundary layer. Bound.-Layer Meteor., 95, 1–30.
Sarghini, F., U. Piomelli, and E. Balaras, 1999: Scale-similar models for large-eddy simulations.
Phys. Fluids, 11, 1596–1607.
Savage, L., S. Zhong, W. Yao, W. Brown, T. Horst, and C. Whiteman, 2008: An observational
and numerical study of a regional-scale downslope flow in northern Arizona. J. Geophys. Res.
Atmos., 113, D14114, 17pp.
Scherer, D., and E. Parlow, 1994: Terrain as an important controlling factor for climatological,
meterological and hydrological process in NW-Spitsbergen. Ann. of Geomorphology, 97,
175–193.
Schmidli, J., G. S. Poulos, M. H. Daniels, and F. K. Chow, 2009: External influences on nocturnal
thermally driven flows in a deep valley. J. Appl. Meteor. Climatol., 48, 3–23.
Schmidli, J., Billings, B.J., Chow, F.K., De Wekker, S.F.J., Doyle, J.D., Grubisic, V., Holt, T.R.,
Jiang, Q., Lundquist, K.A., Sheridan, P., Vosper, S., Whiteman, C.D., Wyszogrodzki, A.A.,
Zaengl, G. 2010. Intercomparison of mesoscale model simulations of the daytime valley wind
system. Mon. Wea. Rev., 139, 1389–1409.
Schumann, U., 1990: Large-Eddy Simulation of the Up-Slope Boundary-Layer. Quart. J. Roy.
Meteor. Soc., 116, 637–670.
Sellers, P. J., F. G. Hall, G. Asrar, D. E. Strebel, and R. E. Murphy, 1992: An Overview of the First
International Satellite Land Surface Climatology Project (ISLSCP) Field Experiment (FIFE),
J. Geophys. Res., 97, 345–18,371.
———, Y. Mintz, Y. Sud, and A. Dalcher, 1986: A Simple Biosphere Model (SIB) for Use within
General Circulation Models. J. Atmos. Sci., 43, 505–531.
Seluchi, M. E. and S. C. Chou, 2001: Evaluation of two Eta model versions for weather forecasting
over South America. Geofisica International, 40, 219–237.
Shafran, P. C., N. L. Seaman, and G. A. Gayno, 2000: Evaluation of numerical predictions of
boundary layer structure during the Lake Michigan Ozone Study (LMOS). J. Appl. Meteorol.
39, 412–426.
Skamarock, W. C., 2004: Evaluating mesoscale NWP models using kinetic energy spectra. Mon.
Wea. Rev., 132, 3019–3032.
Skamarock, W., and J. Klemp, 2008: A time-split nonhydrostatic atmospheric model for weather
research and forecasting applications. J. Comput. Phys., 227, 3465–3485.
Smagorinsky, J., 1963: General circulation experiments with the primitive equations. Mon. Wea.
Rev., 91, 99–152.
Smith, C., and E. Skyllingstad, 2005: Numerical simulation of katabatic flow with changing slope
angle. Mon. Wea. Rev., 133, 3065–3080.
Smith, C., and E. Skyllingstad, 2009: Investigation of Upstream Boundary Layer Influence on
Mountain Wave Breaking and Lee Wave Rotors Using a Large-Eddy Simulation. J. Atmos.
Sci., 66, 3147–3164.
Steppeler, J., G. Doms, U. Schattler, H.W. Bitzer, A. Gassmann, U. Damrath, and G. Gregoric,
2003: Meso-gamma scale forecasts using the nonhydrostatic model LM, Meteorol. Atmos.
Phys., 82, 75–96.
Stull, R. B., 1984: Transilient Turbulence Theory. Part I: The Concept of Eddy-Mixing across
Finite Distances. J. Atmos. Sci., 41, 3351–3367.
Stull, R. B., 1988: An introduction to boundary layer meteorology. Kluwer Academic Publishers,
Boston,.
Sullivan, P. P., J. C. McWilliams, and C.-H. Moeng, 1994: A subgrid-scale model for large-eddy
simulation of planetary boundary-layer flows. Bound.-Layer Meteor., 71, 247–276.
Sullivan, P. P., T. W. Horst, D. H. Lenschow, C. H. Moeng, and J. C. Weil, 2003: Structure of
subfilter-scale fluxes in the atmospheric surface layer with application to large-eddy simulation
modelling. J. Fluid Mech., 482, 101–139.
Sullivan, P.P. and E.G. Patton, 2008: A highly parallel algorithm for turbulence simulations in
planetary boundary layers: Results with meshes up to 10243 . 18th Conference on Boundary
Layer and Turbulence, Amer. Meteor. Soc., Stockholm, Sweden.
Sun, W. Y., and C. Z. Chang, 1986: Diffusion model for a convective layer. 1. Numerical simulation
of convective boundary layer. J. Clim. Appl. Meteorol., 25, 1445–1453.
Twine, T. E., W. P. Kustas, J. M. Norman, D. R. Cook, P. R. Houser, T. P. Meyers, J. H. Prueger,
and P. J. Starke, 2000: Correcting eddy covariance flux underestimates over a grassland. Agric.
For. Meteor., 103, 279–300.
Thompson, J. F., 1984: Grid generation techniques in computational fluid dynamics. AIAA Journal,
22, 1505–1523
Troen, I. B., and L. Mahrt, 1986: A simple model of the atmospheric boundary layer; sensitivity to
surface evaporation. Bound.-Layer Meteor., 37, 129–148.
Turner, J., S. Pendlebury, L. Cowled, K. Jacka, M. Jones, and P. Targett, 2000: Report on the
First International Symposium on Operational Weather Forecasting in Antarctica. Bull. Amer.
Meteor. Soc., 81, 75–94.
Versephy D. L., 2007: Class – A Canadian land surface scheme for GCMs. I. Soil Model, Int. J.
Climatol., 11, 111–133.
Viterbo, P., and A.C. Beljaars, 1995: An Improved Land Surface Parameterization Scheme in the
ECMWF Model and Its Validation. J. Climate, 8, 2716–2748.
Vosper, S. B. and A. R. Brown, 2008: Numerical simulations of sheltering in valley: the formation
of nighttime cold air pools. Bound.-Layer Meteor., 127, 429–448.
Vreman, B., B. Geurts, and H. Kuerten, 1996: Large-eddy simulation of the temporal mixing layer
using the Clark model. Theoretical and Computational Fluid Dynamics, 8, 309–324.
Vrhovec, T., 1991: A cold air lake formation in a basin—A simulation with a mesoscale numerical
model. Meteor. Atmos. Phys., 46, 91–99.
——, and A. Hrabar, 1996: Numerical simulations of dissipation of dry temperature inversions in
basins. Geofizika, 13, 81–96.
Wang, W., and N.L. Seaman, 1997: A Comparison Study of Convective Parameterization Schemes
in a Mesoscale Model. Mon. Wea. Rev., 125, 252–278.
Warner, T. T., R. A. Peterson, and R. E. Treadon, 1997: A tutorial on lateral boundary conditions as
a basic and potentially serious limitation to regional numerical weather prediction. Bull. Amer.
Met. Soc., 78, 2599–2617.
Weigel, A. P., and M. W. Rotach, 2004: Flow structure and turbulence characteristics of the daytime
atmosphere in a steep and narrow Alpine valley. Quart. J. Roy. Meteor. Soc., 130, 2605–2628.
Weigel, A. P., F. K. Chow, M. W. Rotach, R. L. Street, and M. Xue, 2006: High-resolution large-
eddy simulations of flow in a steep Alpine valley. Part II: Flow Structure and Heat Budgets.
J. Appl. Meteor. Clim., 45, 87–107.
Weigel, A. P., F. K. Chow, and M. W. Rotach, 2007a: The effect of mountainous topog-
raphy on moisture exchange between the “surface” and the free atmosphere. BLM, 125,
227–244.
Weigel, A. P., F. K. Chow, and M. W. Rotach, 2007b: On the nature of turbulent kinetic energy in
a steep and narrow Alpine valley. BLM, 123, 177–199.
White, A. B.,R. J. Zamora, K. J. Olszyna, C. A. Russell, B. D. Templeman, and J.-W. Bao, 2001:
Observations and numerical study of the morning transition: A case study from SOS99. 14th
Symposium on Boundary Layer and Turbulence.
Whiteman, 2000: Mountain Meteorology: Fundamentals and Applications. Oxford University
Press, New York, 355pp.
———, 1982: Breakup of Temperature Inversions in Deep Mountain Valleys: Part I. Observations.
J. Appl. Meteor., 21, 270–289.
———, and J.C. Doran, 1993: The Relationship between Overlying Synoptic-Scale Flows and
Winds within a Valley. J. Appl. Meteor., 32, 1669–1682.
———, and S. Zhong, 2008: Downslope Flows on a Low-Angle Slope and Their Interactions with
Valley Inversions. Part I: Observations. J. Appl. Meteor. Climatol., 47, 2023–2038.
———, C. D., S. Zhong, X. Bian, J. D. Fast, and J. C. Doran, 2000: Boundary-layer evolution and
regional scale diurnal circulation over the Mexico Basin and Mexico Plateau. J. Geophy. Res.,
105, 10081–10102.
———, A. Muschinski, S. Zhong, D. Fritts, S. W. Hoch, M. Hahnenberger, W. Yao, V. Hohreiter,
M. Behn, Y. Cheon, C. B. Clements, T. W. Horst, W. O. J. Brown, and S. P. Oncley, 2008:
METCRAX 2006 – Meteorological experiments in Arizona’s Meteor Crater. Bull. Amer.
Meteor. Soc., 89, 1665–1680.
Wong, V. C., and D. K. Lilly, 1994: A comparison of two dynamic subgrid closure methods for
turbulent thermal-convection. Phys. Fluids, 6, 1016–1023.
Wood, E. F., and Coauthors, 1998: The project for Intercomparison of Landsurface Parameter-
ization Schemes (PILPS). Phase 2(c) Red–Arkansas River Basin experiment: 1. Experiment
description and summary intercomparisons. Global Planet. Change, 19, 115–136
———, D. P. Lettenmaier, and V. G. Zartarian, 1992: A landsurface parameterization with subgrid
for general circulation models. J. Geophys. Res., 97, 2717–2728
Wyngaard, J. C. and R.A. Brost, 1984: Top-Down and Bottom-Up Diffusion of a Scalar in the
Convective Boundary Layer. J. Atmos. Sci., 41, 102–112.
Wyngaard, J. C., L. J. Peltier, and S. Khanna, 1998: LES in the surface layer: Surface fluxes,
scaling, and SGS modeling. J. Atmos. Sci., 55, 1733–1754.
Wyngaard, J. C., 2004: Toward numerical modeling in the “Terra Incognita.” JAS, 61, 1816–1826.
Xue, M., D. Wong, J. Gao, K. Brewsker, K. Droegemier, 2003: The advanced regional prediction
system (ARPS), storm-scale numerical weather prediction and data assimilation. Meteorol.
Atmos. Phys., 82, 139–170.
———, K. K. Droegemeier and V. Wong, 2000: The Advanced Regional Prediction System
(ARPS) – A multi-scale nonhydrostatic atmospheric simulation and prediction model. Part I:
Model dynamics and verification. Meteor. Atmos. Phys., 75, 161–193.
Xue, M., K. K. Droegemeier, V. Wong, A. Shapiro, K. Brewster, F. Carr, D. Weber, Y. Liu, and
D. Wang, 2001: The Advanced Regional Prediction System (ARPS)—A multiscale nonhy-
drostatic atmospheric simulation and prediction tool. Part II: Model physics and applications.
Meteor. Atmos. Phys., 76, 143–165.
Yamada, T., and G. Mellor, 1975: A simulation of the Wangara atmospheric boundary layer data.
J. Atmos. Sci., 32, 2309–2329.
———, 1977: A numerical experiment on pollutant dispersion in a horizontally homogeneous
atmospheric boundary layer. Atmos. Environ. 11, 1015–1024.
Yao, W. and S. Zhong, 2009: Nocturnal temperature inversions in a small, enclosed basin and their
relationship to ambient atmospheric conditions. J. Meteor. Atmos. Phys., 103, 195–210.
Zang, Y., R. L. Street, and J. R. Koseff, 1993: A Dynamic Mixed Subgrid-Scale Model and Its
Application to Turbulent Recirculating Flows. Phys. Fluids, 5, 3186–3196.
Zängl, G., 2002: An Improved Method for Computing Horizontal Diffusion in a Sigma-Coordinate
Model and Its Application to Simulations over Mountainous Topography. Mon. Wea. Rev., 139,
1423–1432.
———, 2005: Formation of Extreme Cold-Air Pools in Elevated Sinkholes: An Idealized
Numerical Process Study, Mon. Wea. Rev. 133, 925–941.
———, 2007: To what extent does increased model resolution improve simulated precipitation
fields? A case study of two north-Alpine heavy-rainfall events. Meteorologische Zeitschrift, 16,
571–580.
———, and S.G. Chico, 2006: The Thermal Circulation of a Grand Plateau: Sensitivity to the
Height, Width, and Shape of the Plateau. Mon. Wea. Rev., 134, 2581–2600.
———, B. Chimani, and C. Häberli, 2004: Numerical Simulations of the Foehn in the Rhine Valley
on 24 October 1999 (MAP IOP 10). Mon. Wea. Rev., 132, 368–389.
———, J. Egger, and V. Wirth, 2001: Diurnal Winds in the Himalayan Kali Gandaki Valley. Part
II: Modeling. Mon. Wea. Rev., 129, 1062–1080.
Zhang, D.-L., and R. A. Anthes, 1982: A high-resolution model of the planetary boundary layer—
Sensitivity tests and comparisons with SESAME-79 data. J. Appl. Meteor., 21, 1594–1609.
Zhong, S., and J. D. Fast, 2003: An evaluation of MM5, RAMS, and Meso Eta at sub-kilometer
resolution using the VTMX field campaign data in the Salt Lake Valley. Mon. Wea. Rev., 131,
1301–1322.
———, X. Bian, and C. D. Whiteman, 2003: Time scale for cold-air pool breakup by turbulent
erosion. Meteorologische Zeitschrift. 12, 229–233.
———, H-J. In, and C. B. Clements, 2007: The impact of physical parameterizations on simulated
boundary layer structure in a coastal environment. J. Geophy. Res., 112, D13110.
Zhong, S., and C. Whiteman, 2008: Downslope flows on a low-angle slope and their interactions
with valley inversions. Part II: Numerical modeling. J. Appl. Meteor. Clim., 47, 2039–2057.
Zhou, B., and F. K. Chow, 2011: Large-eddy simulation of the stable boundary layer with explicit
filtering and reconstruction turbulence modeling. J. Atmos. Sci., 68(9), 2142–2155.
Chapter 11
Numerical Weather Prediction and Weather
Forecasting in Complex Terrain
Brad Colman, Kirby Cook, and Bradley J. Snyder
Abstract This chapter describes the advancement of numerical weather prediction

(NWP) models and the role of the mountain weather forecaster in the forecast
process. The chapter begins with a historical perspective of the roles NWP and the
forecaster have played. This ranges from the time when models could only resolve
the largest terrain features, to the present where models have shed light on processes
and helped forecasters better understand features over complex terrain and improve
their conceptual models. The issue of atmospheric predictability is addressed along
with the inherent challenges facing the mountain weather forecaster. In some cases
complex terrain actually extends predictability while for others predictability limits
are relatively short and skillful forecasts are limited to hours rather than days. For
each case the forecasters need to be able to recognize the salient processes and
understand the differences in predictability. They must also understand the strengths
and weaknesses of NWP and communicate appropriately the uncertainty of the
forecast. As forecast interests expand to specific locations over a larger domain –
especially in complex terrain – post processing of mesoscale NWP has become
even more important. This is discussed together with a number of methodologies for
downscaling NWP. Commensurate with this trend in NWP is the need for detailed
observations that match current forecast resolutions. This is followed by a discussion
of forecast tools and the role of the forecaster with respect to NWP. It is argued
that the optimal forecast approach today is a blend of subjective techniques and
statistical post-processing of NWP solutions, combined with local forecasting tools
and cognitive experience. Finally, the role of the mountain weather forecaster in
field programs (e.g., the Winter Olympics) gives a glimpse into the future for what
may become routine. These programs shed insight into predictability limits and
B. Colman () • K. Cook

NOAA/National Weather Service, 7600 Sand Point Way NE, Seattle-Tacoma, WA 98115, USA
e-mail: Brad.Colman@noaa.gov
B.J. Snyder
EC/Meteorological Service of Canada, Environment Canada, Vancouver, BC, Canada

656 B. Colman et al.
provide an opportunity to evaluate new observing systems, tools and approaches.

The chapter ends with a vision for the future for the research and operational
communities. Further exploration into mesoscale ensemble prediction and objective
analysis in complex terrain, among other areas, is needed so researchers can improve
mountain weather forecasting. Meanwhile, forecasters need to be open to these new
advances and recognize that while their role will continue to evolve, it will also
continue to be of value to society.
11.1 Introduction
Numerical weather prediction (NWP) is the foundation for modern day weather
forecasting. The focus of this chapter is on what NWP brings to the table when
forecasting in areas of complex terrain – where it succeeds, where it fails, and
how the forecaster can best use the guidance it provides. The role of the forecaster
is highlighted throughout, together with reflections on alternative methods and
forecasting approaches that can be used to optimize forecast quality.
It is an exciting time to be forecasting in complex terrain. Remarkable advances
in understanding, observations, and mesoscale numerical weather prediction pro-
vide an incredibly fertile, yet challenging, arena – arguably unsurpassed across
today’s weather forecasting challenges. The governing physics of mountain weather
processes are complicated and varied across time and space scales. Some processes
are relatively predictable, others are not. The impacts of an incorrect prediction
can be dramatic and costly. As a result, the excitement of facing this challenge
can, at times, turn to frustration as a forecaster tries to unravel the complicated
interaction of multiple scales and numerous salient processes, decaying forecast
skill with forecast lead time, and an increasingly demanding consumer.
Section 11.2 gives a general overview and history of mountain weather fore-
casting and describes how it has evolved since the time that numerical weather
prediction began to play an important role. Section 11.3 describes the current state of
operational NWP and weather forecasting in complex terrain. Section 11.4 examines
atmospheric predictability and explores what light it can shed on forecasting
approaches and forecaster decisions; followed in Sect. 11.5 by a consideration of
the various ways in which mesoscale NWP in complex terrain can be downscaled or
post-processed to optimize its benefit to the forecast process. Section 11.6 describes
the representativeness of observations and forecasts in mountainous areas, which is
a growing challenge as demands and expectations for forecast specificity increase.
Section 11.7 presents thoughts on the overall forecast process given the current
state of mesoscale NWP in complex terrain. It illustrates some of the situations in
which alternative forecast methodologies are warranted, and/or when a blend of
methodologies is necessary, to optimize forecast skill.
Field programs and special venues provide a unique opportunity to explore new
weather forecasting ideas and tools. This is particularly true for mountain meteo-
rology, for example through the special support efforts for the Winter Olympics in
Salt Lake City, Utah, USA, in 2000, and Vancouver, British Colombia, Canada in
11 Numerical Weather Prediction and Weather Forecasting in Complex Terrain 657
2010. Both of these efforts demanded forecasters challenge the limits of mountain
weather forecasting. Section 11.8 examines these efforts to find new approaches and
techniques, and outlines some of the lessons learned.
There is little reason to think that advances in the area of mountain weather
forecasting will slow or stop. Section 11.9 thus presents a priority of needs,
suggested studies, and benefits to be gained by a continued investment into mountain
meteorology and forecasting. Finally, Sect. 11.10 presents a summary of the chapter
and includes a few final thoughts and ideas.
11.2 Historical Perspective
Numerical weather prediction (NWP) is the cornerstone of modern day weather

forecasting. This prominent role arises from a long series of advances in numerical
methods, data assimilation, parameterizations, and computer processing over the
nearly 60 years since the first daily NWP forecasts were generated at the joint US
Weather Bureau, Navy, and Air Force numerical forecasting unit in Washington,
D.C., in 1954. The impact of NWP on the role of the forecaster has been dramatic
over these same years. Yet recurring forecasts of the demise of the human forecaster
have been wrong. It is true that there has been a beneficial and growing reliance
by forecasters on NWP solutions over other tools and techniques with dramatic
improvements in skill. However, rather than eliminating the forecaster, the NWP
improvements are providing a valuable foundation that enables the forecasters to
play an equally important role of providing professional interpretation and stability
within the forecast process. Mesoscale modeling, in both operational and research
applications, has also contributed to significant advances in scientific understanding.
As a result, forecasters now have a more complete hydrodynamic understanding of
mesoscale processes that can be intelligently applied regardless of whether the given
NWP solution is fully accurate in location or timing. These changes in role and focus
will continue.
Improvements in computing power and numerical methods (among others) have
led to overall increased horizontal and vertical resolutions, which have had a
significant impact on model performance in complex terrain. These changes have
brought an important added dimension to the forecast problem. Picture, for example,
the dramatic evolution of an extratropical cyclone over the ocean as it approaches
and moves into an area of complex terrain (Fig. 11.1); its initial classic structure,
with coherent precipitation and wind fields, rapidly evolves into a complicated
highly fragmented structure in which terrain influences dominate. At a minimum,
the picture has increased in complexity, even if the predictability has remained the
same, or improved as some have suggested (Mass et al. 2002).
Terrain places a significant control on local weather, often reducing degrees of
freedom, yet forcing a unique spectrum of terrain-matching scales. For example,
winds rarely blow across confined valleys; rather they blow up or down valley,
or remain stagnant as free atmospheric winds ride over the top. Strong terrain
Fig. 11.1 University of Washington’s MM5-GFS simulation showing a land-falling mid-latitude

cyclone. Surface wind (speed is color fill) at (a) 12 h and (b) 24 h; and 3-h integrated precipitation
at (c) 12 h and (d) 24 h. The well defined frontal band just offshore at 12 h is nearly lost 12 h later
as it transits the complex terrain of the Pacific Northwest. Courtesy of Cliff Mass
forcing, although unique to each location, remains fixed, creating many coherent and
recurring features that are important to local weather, and which may lend enhanced
predictability. Historically, these terrain-locked features have been targeted in the
forecast process by a priori identifying the associated larger-scale forcing pattern
and then monitoring for it in the larger-scale NWP model forecasts (e.g., Colman
and Dierking 1992), the identification of the relevant pattern typically being done
through an analysis of a large sample of past events.
The most robust of these unique terrain-induced features have been recognizable
in the real-time operational observing data stream even if details and subtleties
were missing or unknown. As such, many of the features can be found within the
local forecaster’s catalog, since, in many cases, these “classic” patterns have been
known by forecasters for decades. This interpretation and subsequent accounting
for unresolved terrain-induced circulation serves as a baseline for subjective forecast

skill. Recognizing this forecast methodology is important when one begins to assess
the added value that NWP brings to the table regarding any particularly local event.
For the features already known by the forecasters, the model simulations primarily
add detail to earlier diagnoses, identify salient physical processes, and ultimately
help refine or replace existing forecast approaches. Of course there are also a number
of new features discovered through the use of mesoscale NWP for which the NWP
solutions offer hope of new predictive skill.
Early in the evolution of NWP, models were only capable of resolving the largest-
scale terrain influences, such as lee-side troughs and large-scale mountain-range
influences. As such, local terrain influences were left to a forecaster’s own means:
experience, rules of thumb, and decision trees (as discussed below). Relatively
quickly the power of statistical post-processing was also recognized and many of the
more robust terrain signatures began to emerge through the use of tools like MOS
(Model Output Statistics) and a few simple models. Figure 11.2 shows how the MOS
application can reliably correct for a terrain influence lacking in the NWP solution.
In the case shown, the MOS application correctly adjusts the surface winds to be
more channeled and ageostrophic. At 12 h, the gradient wind over SeaTac airport is
strong southeasterly, yet MOS recognizes that the wind is much more likely to be
light northerly. Similarly, as the surface low jumps inland at 30 h, MOS has correctly
identified southerly down-gradient flow.
Through the 1970s and 1980s there was a reasonable balance between the
improving forecast skill of NWP models, which were resolving processes with
wavelengths on the order of hundreds of kilometers and forecast projections to a
few days, and the ability of the forecaster to incorporate separately the local terrain
influences. This was partly because there was reasonable scale-separation. For
example, the NWP models gave the forecasters a good sense of larger-scale forcing
not influenced or contaminated by mesoscale terrain features. The forecaster was
then able to “add in” the impact of anticipated terrain-induced structures (channeled
flow and eddies, convergence bands, mountain waves, orographically enhanced
precipitation, etc.). Probably in many ways this reflected a mature synoptic-scale
forecast system where the large-scale patterns were provided by numerical models
(with a gradual decay in forecast skill with increasing forecast projection) and
forecasters incorporated detailed terrain features in the shorter ranges and broad-
brushed the same features at greater projections in recognition of the overall
decrease in forecast skill.
Of course this seemingly balanced and mature forecast system reflects a rela-
tionship grounded over a relatively short period of time, and does not reflect the true
nature of constantly evolving and advancing NWP. The march of NWP downscaling
continued in full force through the 1990s and 2000s as modelers continued to
explore mesoscale modeling in complex terrain with ongoing dramatic changes to
the suite of guidance products available to the forecasters (Mass and Kuo 1998).
These years were a heyday for researchers and forecasters alike. Initially, non-real-
time simulations revealed terrain features that were previously only conceptualized
in their most simple form. Mesoscale NWP runs between 10 and 20 km grid
Fig. 11.2 Panels (a) 12 hour forecast and (b) 30 hour forecast show the Nested Grid Model (NGM)
forecast of mean seal-level pressure (contours) and boundary layer wind for a land-falling cyclone.
The NGM had particularly smooth terrain for its 40 km grid spacing. The surface wind time series
shown in panel (c) contrast the model solution, or perfect prog, to the statistical post-processing
with MOS
spacing provided new insights into local weather patterns and circulations. Local
terrain influences that had been recognized for decades by their impacts, and loosely
incorporated into local forecasts using pattern recognition and decision trees, were
now revealed in remarkable detail. The simulations were also remarkable in their
physical realism, yet their predictive value was still unknown.
Fast computing platforms became commonplace and inexpensive, with many
local National Weather Service forecast offices running their own, in-house,
mesoscale models. There were also several regional collaborations that developed
during this time period. The most notable was the Northwest Modeling Consortium
centered at the University of Washington that explored many facets of mesoscale
NWP, including environmental applications to air quality, and hydrologic prediction
(Mass et al. 2003). These early efforts provided tremendous insights into the benefits
and challenges of mesoscale NWP, especially in complex terrain.
Most of these local and regional efforts derived their model initial conditions
by downscaling a coarser operational model to a finer, limited-domain grid without
any additional data assimilation. In almost every case there were brilliant successes
or “hits” with the finer scale NWP; yet, the early sheen quickly tarnished as
forecasters who used the highly deterministic details directly were often frustrated
by large errors. The critical operational difference between physical realism and
true predictability was paramount to the slow acceptance of the fine-scale models
in real-time weather forecasting efforts; the simple and direct use of a single model
run as the forecast was not the answer.
As such, the initial and highest value of these simulations was not in their
predictive skill, but rather in their ability to aid in unraveling complex processes and
to better explain known features in local weather patterns. This added knowledge
and understanding proved to be highly beneficial to the forecaster, especially in
adjusting short-term forecasts to account for such features. Without this additional
physical insight, the forecaster, when faced with some unrecognized feature, was
left with few means other than to use persistence or simply ignore it; while with the
newly gained understanding, cause and effect, life-cycle evolution, and impacts, all
became clearer. This gain in understanding was by far the most significant and clear
initial benefit of mesoscale NWP in mountain weather forecasting.
The primary challenge that forecasters initially faced in using the NWP sim-
ulations in their forecasts was how to determine which features or impacts were
predictable and which ones were not. Post-event critiques were often frustrating
because they would, at times, reveal a forecast feature in one of the simulations that
verified well; as such, the direct use of the model in that situation would have led
to a significantly improved forecast. But what about the other features that were
just as “significant” in appearance in the simulations yet turned out to be ephemeral
and nowhere to be found as the event unfolded? How was a forecaster to be able
to differentiate between the two types of features? In addition, model simulations
were often clearly marred by model biases, most often related to the handling of the
boundary layer, as well as the radiation and precipitation parameterizations. With
few objective bias-correcting tools available, how was the forecaster expected to
adjust for these biases?
While the first decade of mesoscale NWP was characterized by clear gains in
understanding, there remained considerable frustration with regard to how to take
advantage of these new, physically realistic mesoscale model simulations in the
operational forecast process. Run-to-run consistency was highly variable and true
forecast skill was difficult to determine. In addition model biases were significant
and clearly required additional adjustments. Much remained to be learned.
11.3 Current State of Operational Mesoscale Numerical

Weather Prediction in Complex Terrain
To a meteorologist, there are few things more exciting than seeing the detail and
physical realism presented by a mesoscale NWP simulation of an evolving weather
event in an area of complex terrain. Features that were previously known primarily
through only simple conceptual models and gross simplifications can be revealed
in a mesoscale model with surprisingly realistic detail and evolution. Add to that
a spectrum of other swirls and eddies that were only in the minds of the most
imaginative and it is tempting to accept that these features are indeed predictable. No
more will there be surprise windstorms or convergence band precipitation events; no
more will forecasters be left scratching their heads wondering what process caused
the latest surprise event. While it is understandable, in many ways it appears almost
magical how one can simply use a larger-scale simulation for initial and boundary
conditions in a nested high-resolution run and, over the matter of a few hours
of simulated time, have recognizable mesoscale features emerge from the smooth
larger-scale patterns.
Even today there is disagreement on whether forecasters should be taking a
hands-off approach to the use of mesoscale NWP in the forecast process. This
occurs despite ongoing experiences of operational forecasters who feel the direct
use of these simulations often results in missed forecasts (and a high level of
frustration). The various features revealed in the simulations bring with them
stronger gradients and an exactness in space and time that apparently goes beyond
realistic predictability limits – at least for some of the features, at least some of
the time. The use of the model to predict the timing and location of a feature often
results in larger errors than a forecast that blurs the feature or ignores it altogether. In
fact, this is likely the reason why conventional verification scores frequently fail to
show improved skill for models run on the finest grids over those run on somewhat
coarser grids.
In post-analysis one can identify features that were predicted well and make
significant claims of success. But this frequently involves ignoring other features,
often even in the same simulation that were not so well handled. Forecasters quickly
reach a belief that to use mesoscale NWP directly, as a single deterministic solution,
to construct a forecast in complex terrain is just as likely to result in a large error
as an impressive hit. While the physical realism in mesoscale NWP is amazing, it is
clear that more needs to be understood about the predictability at these scales before
these simulations will fully earn their position as the primary tool in the forecasters’
tool box.
One immediate, and ongoing, benefit of today’s mesoscale NWP simulations is
their tremendous ability to reveal the details of local circulations. While forecasters
frequently are aware of many local terrain-induced phenomena through their
recurring impacts, often their conceptual models of these phenomena are vague,
incomplete, or simply wrong. At best they are simplifications of what are typically
very complex circulations with many different layers, processes, and subtleties.
The fine-scale NWP solutions are the proverbial “light bulbs” as they reveal
coherent time-dependent evolutions and structures. The physical realism is amazing
to both researchers and forecasters, shedding tremendous insights into the struc-
ture of many terrain-controlled circulations, precipitation bands, and temperature
patterns. Considerable advances in the scientific understanding of multiple unique
phenomena are being achieved as a result.
The early successes in mesoscale NWP encouraged operational centers to invest
their growing computing resources into resolution (over ensemble prediction) and
move toward finer grid spacing. As a result, the National Weather Service in the
US now runs the North American Model (NAM), four times daily, at 12-km grid
spacing. They also have moveable fine mesh models down to 4-km grid spacing.
Essentially all large numerical modeling centers (ECMWF, UKMET, CMC, etc.)
are running similar resolutions of between 10 and 15 km grid spacing, with
windowed finer grids. These finer grid models, along with a few mesoscale ensemble
forecasting systems, are the primary guidance for short-range forecasting around the
world.
Interestingly, simple verification results (e.g., RMSE of temperature or wind
speed) generally show significant benefit derived by moving toward finer grids down
to about 10 km, and then a leveling off of forecast skill for some elements for finer
grids. For some metrics there is even an apparent degradation of skill with grid
spacing of just a few km over slightly coarser model grids.
There are several factors that could account for these somewhat discouraging
results. First, the multitude of increasingly detailed features resolved in the finer
grid models may not necessarily be predictable given the associated large-scale
uncertainty – they may well be forecast noise but are of large enough magnitude
that they negatively impact verification scores. Second, the standard verification
approaches unfairly penalize the finer-grid solutions due to placement and timing
errors – even when the model has the correct features. Third, is the parameterization
of subgrid-scale processes, including convection. There has been a long-standing
discussion about whether it is better to resolve convection or to parameterize for
its impacts. Below 10 km the models begin to reveal explicit convection but it is
typically not well resolved. There is a “no-modeler’s” domain between 1–2 km, and
10–15 km, grid spacing where partially resolving convection potentially penalizes
the solutions. Similarly, the finer-mesh models are pushing the limitations of
model microphysics, boundary layer schemes, and other modeling approaches with
associated model biases.
The most-recent significant developments in mesoscale NWP in complex terrain
are several real-time operational mesoscale ensemble forecast systems. Prior to the
last few years, limitations in computational resources prevented ensemble systems
from running at scales necessary to resolve terrain-induced mesoscale structures.
Now, most operational centers are running ensemble prediction systems with at least
some members on the order of 20 km; experimental systems are now running at grid
spacings of around 10 km.
The Met Office has recently introduced a short-range ensemble prediction system
known as MOGREPS (Bowler et al. 2008). This system consists of global and
regional ensembles, with the global ensemble providing the boundary conditions
and initial-condition perturbations for the regional ensemble; a model construc-
tion very similar to the National Center for Environmental Prediction’s (NCEP)
Downscaled Global Forecast System with WRF EXtension (DGEX). Perturbations
to the initial conditions are calculated using the ensemble transform Kalman filter,
which is a computationally-efficient version of the ensemble Kalman filter. Model
uncertainties are represented in the system through a series of schemes designed to
tackle the structural and subgrid-scale sources of model error.
Clearly there have been many successes with the use of mesoscale NWP
in complex terrain. Processes are better understood and verification scores are
generally better than those obtained with coarser grid NWP models. The primary
challenge for operational weather forecasters from the start has been predictability.
“Is this particular scenario likely to occur?” “Should I go with this area of enhanced
precipitation or indicated wind switch?” “What if the low tracks farther north, how
will that impact the mesoscale structure?” To get a better handle on this critical
aspect, the next section looks at the topic of predictability in complex terrain.
11.4 Atmospheric Predictability
Lorenz (1969) presented the viewpoint that for each scale of motion there is a limit
on its predictability. The study looked at error growth and how quickly the errors
saturated, or reached the same magnitude as the signal, across a spectrum of scales
using the k5/3 energy spectrum from classical turbulence theory. The outcome
of the study was that predictability at a given scale decreases as the scale of the
eddy turnover time decreases. For processes on the order of 20 km scale, the errors
grow sufficiently to mask the signal within approximately 1 h. Large planetary-scale
waves on the order of 10,000 km or so have much longer predictability limits on the
order of 5–10 days. Of course it is the smaller, terrain-induced circulations that are
the most important to the mountain weather forecaster and these early projections
regarding predictability were rather discouraging.
Anthes (1985) was one of the first to suggest that these classical estimates
were too pessimistic and that predictability limits would be longer for at least
some processes, including coherent structures like supercell thunderstorms and
hurricanes. Anthes also concluded “physical forcing at the earth’s surface, such as
mountains, may contribute to extended predictability.” Another important aspect of
mesoscale numerical weather prediction is that fine-scale events can be forced by
the terrain from large-scale initial conditions.
There have been, in fact, a few studies of both global models and Limited Area
Models (LAM) that show an apparent positive impact on predictability by complex
terrain. Vukicevic and Errico (1990) ran case study simulations using a LAM over
Europe and found that without terrain (a smooth Europe) errors grew quickly, yet
when the Alps were included, error growth was essentially flat (Fig. 11.3) for many
of the case studies.
Fig. 11.3 Error growth reflected in multiple runs of a complex Limited-Area Model over
Europe. The 23 March case is for a barrier jet and Yugoslavian Bora; each of the simulations
represents a perturbation in the initial conditions. The terrain dramatically improves forecast skill
through 100 h. (Vukicevic and Errico 1990. © American Meteorological Society. Reprinted with
permission)
The Canadian Meteorological Center (CMC) global model was also found to
display an extension of predictability for certain scales that was attributed by Boer
(2003) to terrain influences. Figure 11.4 shows a band of the mean component at
intermediate/high wave numbers that exhibits an apparent region of enhanced skill –
apparently connected to the topography. Additional studies have shown that, at least
for some scales and events, the error growth is actually diminished and predictability
is extended. Berri and Paegle (1990) demonstrated that error growth could be
controlled by increasing the size of a terrain obstacle. For real terrain (Fig. 11.5a)
the simulations showed significant error growth through the 24-h forecast period as
wind speed and direction were perturbed. Yet, when the terrain elevation is increased
tenfold the results show little or no error growth for perturbations in wind speed
(Fig. 11.5b).
Recent verification studies of the National Weather Service’s National Digital
Forecast Database (NDFD) (a blend of NWP and forecaster input) are somewhat
supportive of the idea that terrain does extend predictability for at least some
features and scales. A comparison between forecasts issued in the central United
States to those issued in the western United States (Figs. 11.6a, b, and 11.7a, b)
shows the central US forecasts are more skillful for the short-ranges but are less
skillful at longer ranges (to day 7). The near lack of error growth in forecasts
for the western US, an area dominated by complex terrain and boundary-forced
circulations, is at least supportive of the idea that there is some benefit from terrain
to forecast skill.
Not all predictability studies of terrain-induced phenomena have been as en-
couraging. For example, Reinecke and Durran (2009) showed extreme sensitivity
to initial conditions using an ensemble to forecast downslope windstorms in the lee
of the Sierra Nevada Mountains in their analysis of data from the Terrain-induced
Fig. 11.4 Predictability timescale by wave number for the Canadian NWP Model (Boer 2003)
Rotor Experiment (T-REX). Looking at the extreme members of the ensemble

(weakest and strongest subsets ranked by windstorm amplitude; Fig. 11.8) revealed
very small differences in initial-condition soundings (that were not substantive and
were close in magnitude to sampling and instrument error (Fig. 11.9)). This resulted
in forecast wind speed differences of 29 m s1 in just 6 h. This study offers very
little support for the skillful predictability of this particular downslope windstorm
event. The development of this class of storms is highly non-linear and, perhaps, it
shouldn’t be surprising that this sensitivity to initial conditions exists, and may, in
fact, represent one of the more extreme examples of poor predictability.
There are several key operational weather forecasting considerations that must
be considered when accounting for predictability in complex terrain. First, the
forecaster needs to appreciate that some features have good predictability (e.g.,
channeled flow, lee-side rain shadows), whereas others have very poor predictability
(e.g., high amplitude, lee-side waves). Each class should be recognized and handled
differently. Another consideration is how to blend the robust forced features, and
their associated details, into a forecast that necessarily needs to reflect a synoptic-
scale signal with decaying forecast confidence at longer lead times. The final
consideration is how best to communicate confidence (or lack thereof) within a
forecast framework that is highly detailed by its very nature since it reflects fine
terrain and coastal features.
Fig. 11.5 (a) RMS vector differences (m s1 ) with respect to the initial state at 119 meters above
the surface vs. time of integration (h). Left panel, solid curve is signal (RMSC) for the NE 6 m
s1 control run; dashed and dotted curves are noise (RMSP) for the NE 5 m s1 and NE 1 m
s1 perturbed runs, respectively. Right panel same as left except dashed and dotted curves are
noise (RMSP) for the 6 m s1 perturbed runs with wind direction rotated 15 degrees clockwise
and counterclockwise, respectively. (b) Same as panel (a) except that the topography has been
increased by a factor of 10. Solid curve is signal (RMSC) for the NE 6 m s1 control run; dashed
curve is noise (RMSP) for the NE 6 m s1 perturbed run with the wind direction rotated 15 degrees
clockwise; and dotted curve is noise for the perturbed wind speed run (4 m s1 ). (Berri and Paegle
(1990). © American Meteorological Society. Reprinted with permission)
The importance of forecaster sensitivity to resolvable scales and predictability is-

sues (spatial and temporal) cannot be overstated. In the operational use of mesoscale
NWP, the forecaster must be able to recognize which features in the simulation
should be accepted with reasonably high confidence and which features should be
rejected. This important distinction was not initially recognized and contributed
significantly to the rather slow acceptance of the guidance by forecasters.
Fig. 11.6 (a) Western Region, (b) Central Region. Mean absolute error (deg. F) for maximum
temperature forecasts from 24-h through 168-h as verified against in situ observations. MOS, is the
Model Output Statistics (MOS) forecast for those locations; GMOS is the gridded MOS forecast;
NDFD is the National Digital Forecast Database (human/model blended forecast) and HPC is a
centralized national forecast
Fig. 11.7 (a) Western Region, (b) Central Region. Mean absolute error (deg. F) for wind speed
forecasts from 24-h through 168-h as verified against the Real Time Mesoscale Analysis (RTMA)
on a 5-km grid. GMOS is the gridded MOS forecast; NDFD is the National Digital Forecast
Database (human/model blended forecast) and DNG is a centrally produced model downscaled
national forecast
For those events that are highly predictable, the forecaster should accommodate
them within the forecast. These tend to be the most robust and resilient features, e.g.,
gap flows, melting levels, and coastal thermal gradients. These are the features most
likely responsible for the extended predictability associated with complex terrain.
For example, predicting wind direction in a long channel is easier than if that terrain
Fig. 11.8 The evolution of the zonal wind averaged over the Owens Valley metric box during
the IOP 6 simulation for the (a) 10 strongest and (b) 10 weakest ensemble members. The thick
line shows the mean of each 10-member subset (From Reinecke and Durran 2009. © American
feature were not there. This is because, in the simplest order, the wind is almost
always parallel to the channel, responding to the along-channel component of the
total pressure gradient vector.
Today’s primary approach to capturing these elements in the forecast is to run
LAMs forced by larger-scale global models through the entire forecast period to
get a dynamic down-scaling of the global model solution. In the U.S., the National
Weather Service runs the Downscaled Global Forecast System with WRF EXtension
(DGEX) through Day 8; many local forecast offices run their own simulations.
There are two additional important considerations when using mesoscale NWP
in operational mountain weather forecasting. First is determining if the finer details
associated with a robust and predictable feature are also predictable. Gap flow is
along-channel and highly predictable, but what about the lateral boundary eddies,
cross-channel variations, or frequently observed hydraulic jumps? These additional
details are important but there is much less information on whether the NWP
simulations are skillful in their prediction. For example, it is well demonstrated that
mesoscale models can capture realistic convergence bands downstream from terrain
features. Forecaster experience also supports the idea that some characteristics, like
positioning and overall strength, can be skillfully captured. It is less understood, and
Fig. 11.9 Composite model soundings for the strong subset (solid) and weak subset (dashed) for
IOP 6. The soundings are valid at forecast hour 5 (one hour before the time of maximum wind).
Plotted is the (a) cross-barrier component of the wind U, (b) potential temperature u, (c) Brunt-
Vaisala frequency N, and (d) RH (From Reinecke and Durran 2009. © American Meteorological
less likely, that additional details like secondary band formation from associated
convective outflow or subtle deviations resulting from minor eddies are skillfully
predicted.
The second consideration has to do with the handling of the complex, likely
highly non-linear features, where there appears to be very little associated pre-
dictability. These might include terrain-induced shower and thunderstorm regimes
or mountain wave windstorms. These are the same events that, when accepted early
on in the forecasters’ use of mesoscale models, lead to frustration and reluctance to
use the model guidance.
After more than a decade of experience with mesocale NWP models it is clear
that to use them operationally, the forecaster must recognize the mix of features
described above and treat each one separately. The next two sections look at
alternative, and in some ways more optimal, ways to deal with the wealth of
information provided by the models.
11.5 Post-processing Mesoscale NWP
As discussed earlier one method of improving point-specific forecasts is by post-

processing the synoptic-scale NWP output using statistical techniques such as
Model Output Statistics (MOS). This technique successfully corrects for model
biases. It can also downscale the information to the exact location of the observing
site by adding in the microscale characteristics of the observations that are not
captured by the smoother numerical model solution (see Fig. 11.2). Forecasters
can then adjust further to accommodate other model simulations (different forecast
scenarios) or can apply subjective interpretation about changes in anticipated
conditions or known model issues. They can also broaden these point forecasts to
a somewhat less detailed area forecast covering expected conditions over a zone
or region. MOS has certainly given the forecaster a significant advantage over the
raw model output and has been a benchmark against which forecaster skill has been
measured. However, this forecast paradigm no longer meets the needs of the weather
information user community. This is especially true in areas of complex terrain
where the weather conditions may be dramatically different between two adjacent
valleys or across a ridgeline.
In many cases today, forecast interests are for very small areas or for specific
locations, especially in areas of complex terrain. The demand for “personalized”
forecasts is in large part a result of the overall significant improvement in overall
forecast skill. As forecasts have improved, users are recognizing the value of
detailed and specific forecasts and want their own customized forecast where they
need it. Also playing significant roles in increasing expectations are advances
in communication technology and the, at times unreasonable, promises of some
weather forecast providers.
Certainly the recognition of this growing demand was part of the impetus for the
US National Weather Service (NWS) to deliver their forecast on a fine-scale grid
across all areas of responsibility. The NWS is currently delivering these forecasts
with grid spacing ranging from 2.5 to 5 km. These very fine resolutions are
necessary to resolve realistically the terrain and land/water interface. This gridded
format essentially eliminates the past focus on a few locations and broadens the
forecast to cover the entire area, all the while trying to retain the detail of very
specific point forecasts.
With this newly evolving forecast paradigm, there is significant demand to post-
process mesoscale NWP across the full domain rather than for a relatively small
set of exact observation locations as was done with MOS. It must now be done
across broad areas where there are no in situ observations to help drive the bias
removal or downscaling. Given that most operational models are not run at the
resolution of these new gridded forecasts, not only do model biases need to be
addressed but the model solutions also need to be downscaled to the grid-spacing
of the forecast. Similar to point-MOS, the post-processing technique needs to
address issues like underlying surface characteristics (water/land boundaries, land
use, forest, etc.), terrain elevation and aspect, and model biases; however it does not
Fig. 11.10 Forecast surface temperature (ı F) from NWS GFS model. (a) GFS 40 km depiction
linearly interpolated to a 2.5 km grid. (b) The SmartInit GFS obtained by considering 2.5 km grid
box elevation and model lapse rate
need to capture the microscale characteristics at an exact location, since, at least in

an academic sense, the grid pixel represents processes larger than at a single point
(more analogous to a grid box average).
The spatial downscaling of forecast models to resolutions necessary to meet the
fine-grid forecast paradigm is currently being done in essentially three different
ways in the NWS. The first is a simple interpolation and elevation adjustment
(using high resolution terrain and model lapse rate), which is known as the SmartInit
(Smart Initialization) process. The results of this process can be seen in Fig. 11.10,
which contrasts the GFS (native resolution of approximately 40 km grid-spacing-
equivalent since it is a spectral model) to the SmartInit GFS. The impact is
dramatic as finer terrain features are reflected in the temperature grid through this
downscaling process.
A second methodology being used is to down-scale models dynamically by
nesting a finer scale model inside the coarser model (which provides the initial
and boundary conditions). For the NWS, this is the DGEX (Downscaled GFS with
WRF EXtension), which is run at 12 km for forecast ranges from 3.5 to 8 days. (The
NWS’ NAM is only run through 3.5 days, but the NWS routine forecast package
extends through day 7.) The NWS forecast offices further downscale the DGEX
output using the SmartInit process to match their forecast resolution (either 2.5 km
or 1.5 km). Figure 11.11 shows the DGEX and SmartInit DGEX for the same time as
Fig. 11.10.
A third methodology for downscaling (and bias-correcting) NWP is gridded
MOS. Dallavalle and Glahn (1995) discussed the development of a new gridded
Fig. 11.11 Forecast surface temperature (ı F) for the same forecast projection and time as
Fig. 11.10, except from: (a) the native 12 km DGEX, (b) the SmartInit DGEX on a 2.5 km grid
MOS system within the National Weather Service. Gilbert et al. (2009) reported
on recent improvements to the same initial system. Their new method is based on
an objective technique using successive corrections to extend the site-specific MOS
forecasts to the desired grid (Glahn et al. 2009). The scheme accounts for differences
between land and water forecast points and for changes in an element’s value
due to elevation. Figure 11.12 shows a gridded MOS forecast for the contiguous
U.S. and Fig. 11.13 shows a regional depiction of gridded MOS on a 2.5 km grid
spacing.
There are certain limitations with each of these downscaling procedures that
forecasters must be aware of and accommodate in the forecast process. For example,
dynamic atmospheric processes or significant terrain influences not resolved by the
initial model cannot appear through simple interpolation processes. However, they
can be generated through dynamic downscaling approaches (e.g., DGEX). Simple
downscaling can account for altitude change – assuming the model vertical lapse
rate of temperature is reasonable and representative. (See Fig. 11.10.) Wind is more
challenging, yet, if a model vertical wind shear is used, remapping to a finer terrain
can be done with some success by simple downscaling. However these downscaling
tools cannot accurately account for drainage flows or other locally controlled
circulations. Simple downscaling can sharpen a coastal gradient by using a finer
resolution coastline but it cannot create a strong sea breeze or correctly redistribute
precipitation across complex terrain. A comparison of Figs. 11.10 and 11.11 shows
that the DGEX was able to account for the inland penetration of cooler marine
air through the Strait of Juan de Fuca (resolved at 12 km) but not by the coarser
native GFS. As such, the GFS SmartInit is not able to recover this feature, whereas
Fig. 11.12 Example of Gridded MOS surface temperature forecast (ı F) on a 5 km grid spacing
for the contiguous United States
the DGEX and DGEX SmartInit both reflect its influence. These are differences
and subtleties that forecasters must be aware of and take into consideration when
preparing the forecasts.
Mass et al. (2008) took an alternative approach to removing model bias for
mesoscale simulations over the Pacific Northwest. The approach identifies model
bias at each observing location and then clusters model grid points with similar ele-
vation and land-use characteristics. The elements include gridded 2-m temperature,
2-m dewpoint temperature, and 12-h precipitation forecasts. The correction is then
applied to the model grid points as the final guidance product.
All of these early approaches show some skill in addressing model bias and
downscaling issues. Their use in operational forecasting is still early so exactly what
kinds of features are retained through the post-processing remains to be explored
further. It is likely, however, that the geoclimatic signal, which is most dramatically
revealed by changes in elevation and land/water boundaries, is one element that
is captured, especially in the temperature fields by the statistical and dynamical
approaches. On the other hand, the features resulting from mesoscale dynamics are
most likely not captured by these post processing techniques. It is quite possible that
these statistically post-processed fields will be excellent starting points for the fore-
caster to which they can add in the impacts of additional anticipated local features.
Finally, systems like the North American Ensemble Forecast System (NAEFS)
(Candille 2009) could be post-processed statistically to account for the effects of
Fig. 11.13 Example of Gridded MOS surface temperature forecast (ı F) on a 2.5 km grid centered
over Utah
small-scale dynamical features captured in the verification dataset, assuming that the
strength and position of the features in the forecast period will be the same as they
were during the verification period. This approach has great promise since it both
downscales and corrects for model biases – assuming that the observed conditions
used are correct and representative.
11.6 Representativeness of Forecasts and Observations
Among the greatest challenges of forecasting in complex terrain are maintaining

high situational awareness (overall appreciation of the current conditions and
anticipated short-term changes) and achieving a realistic balance between detail
and uncertainty in the forecast. As demand for detailed forecasts for essentially
all locations has grown, there is the commensurate increase in the requirement for
the forecasters to know what is currently occurring at all locations. It is no longer
sufficient to know in detail the weather at a few locations (where there are in situ
observations) and cover the surrounding areas with less detailed forecasts. Rather
there is the growing expectation from the users, and the growing need from the new
forecast paradigm (gridded forecasts), to know the weather in detail across the area.
This is particularly challenging in complex terrain where patterns are complicated
and under-observed, or not observed at all.
Observed weather plays several critical roles in the forecast process. First, it is
the starting point for every forecast. There are few more embarrassing moments
for a forecaster than to put out a forecast that is not consistent with the current
conditions; it seriously undermines credibility. Second, the observed weather is the
verification for past forecasts; it is what the forecaster uses to learn how to improve
or optimize the forecast process. If you cannot determine the accuracy of your
forecast, how can you improve? Third, the observed weather is used objectively
to improve NWP tools through model validation of parameterization approaches or
initialization methodologies. Observed weather is also important to the process of
deriving gridded MOS or model bias adjustments as discussed in the last section.
Today numerous forecasts are being produced (with NWP and/or subjectively by
forecasters) on fine grids with spacings of only a few kilometers, yet we do not have
high-quality matching weather analyses of the current conditions (either in real time
or post analysis). Observations alone are woefully inadequate. There are not enough
of them and the ones we do have (in situ observations) capture all scales down to the
very location of the weather sensors. As such, the point observation ideally needs
to be upscaled to remove the microscale impacts of local surface characteristics,
buildings, terrain, and other local conditions that are not necessarily representative
of the area over even a few square kilometers. Furthermore, the observations contain
sensor errors.
On the other side of the problem, relatively large areas are un-sampled. In
some areas, there may be no surface observations and/or radar beams may be
blocked by terrain. Moreover, satellite observations cannot be used effectively
to capture fine-scale detail given their relatively coarse sampling resolution. The
simple interpolation of conditions between surface observations is risky, especially
in complex terrain where changes in elevation, exposure, ridge versus valley, etc.,
all play to decrease observation representativeness.
The determination of current weather conditions (even in a non-real-time
approach) on a scale that matches current forecast resolutions is one of the greatest
challenges facing today’s forecast and research communities. Ultimately, it will
require a sophisticated data assimilation method that blends the full set of in situ
Fig. 11.14 The Real Time Mesoscale Analysis of surface air temperature (ı F, 2.5 km pixels)
and remote observations with a background full-physics model. Fortunately, there

are a number of ongoing efforts exploring techniques of this type with the goal being
quality mesoscale analyses that include terrain variation and influences.
One such technique is the NWS’ Real-Time Mesoscale Analysis (RTMA) that is
currently produced hourly on a 2.5 km grid for temperature, dew-point temperature,
and wind. The RTMA uses Grid-point Statistical Interpolation (GSI), which is a
variational data assimilation methodology developed at the National Center for
Environmental Prediction (NCEP), to blend all available mesonet data with the
WRF background field. Figure 11.14 shows a temperature analysis from the RTMA.
Figure 11.15a shows an analysis produced with a different methodology called
Match Observations All (MatchObsAll). This methodology uses a background
Fig. 11.15 (a) MatchObsAll analysis of surface air temperature (ı F, 2.5 km pixels) for same time
as Fig. 11.14. (b) Difference field between RTMA (Fig. 11.14) and MatchObsAll (Fig. 11.15a)
Fig. 11.16 Domain-average difference (ı F) between RTMA and MOA over the northern Cascade
Mountains of Washington
downscaled model first guess and then a serpentine curve to match observations
exactly. While they both clearly show a signature of the terrain and have many
qualitatively similar characteristics, when they are compared over areas of complex
terrain over extended periods of time it becomes very clear that the accuracy of these
experimental analyses is low. Figure 11.15b is a difference field between the two
analyses, while Fig. 11.16 shows a multi-month time series of average differences
over complex terrain (north Cascade Mountains of Washington state). Note the
magnitude of these differences substantially exceeds what would be considered an
acceptable forecast error.
Without a quality, mesoscale analysis of weather conditions across forecast
areas (including complex terrain), situational awareness, forecast methodologies,
statistical post processing, and improved scientific understanding are compromised
and restricted. This deficiency must be recognized and appropriate scientific energy
and resources invested into advancing the science of mesoscale analysis.
11.7 The Forecast Process and Complementary

Forecasting Tools
Today’s mesoscale NWP guidance is contributing significantly to skillful mountain

weather forecasts and is responsible for much of our recent gains. While NWP
solutions (global to mesoscale) dominate today’s weather forecast process, the
output is not yet mature enough to simply pass on to the user community unfiltered.
For example, the NWP output still contains model biases. Furthermore, it reflects
atmospheric events that are not predictable, as well as those that are not predicted
well on a given day. Calibrated probability density functions (PDFs) for estimates
of uncertainty are lacking (even in today’s ensemble prediction systems). Progress
toward a more mature system is being made every day (including Bayesian Model
Averaging (BMA) approaches and other ensemble post-processing techniques) and
will be discussed in Sect. 11.9.
Model biases are especially significant in the model boundary layer, which
is critical to so many aspects of weather forecasting from surface wind and
temperature, to air quality forecasts. The NWP solutions also contain some rapidly
growing features that are unpredictable given the state of today’s models and
data assimilation; yet, these features are reflected in, e.g., precipitation, wind, and
temperature patterns, and can dramatically impact the forecast output.
In complex terrain there is essentially a fractal of forced scales and a model
is never of fine enough resolution to capture all of the possible phenomena. For
example, consider the richness of scales identifiable in the visible satellite image
shown in Fig. 11.17. Just as the larger terrain features bordering the Strait of Juan
de Fuca produce their own bow wake clouds superposed on the down channel flow,
there will be other smaller features with their own local impacts. For NWP, there will
always be sub-grid-scale terrain features with impacts on the wind field, or poorly
represented steep terrain that forces shear-induced eddies that may have downstream
coherence, or small ridges that serve to focus cumulus updrafts, or local drainage
winds that impact the location of fog patches. Critical questions about these sub-grid
features remain unanswered. If model resolution is increased sufficiently to capture
them, are they predictable? Of those that are predictable, is it necessary to apply
the full force of NWP to achieve optimal forecast skill, or is there a more simple
statistical approach or downscaling methodology that can be as successful with far
fewer resources? Recall that well before NWP became the valuable tool it is today,
forecasters were able to identify and forecast some of the most robust of the terrain
induced features using rules of thumb and decision trees; albeit with generally less
skill than they can today with the added benefit of the finer scale models.
So, in the operational arena today, the mountain weather forecaster is faced
with a very complicated set of highly deterministic NWP solutions that (a) contain
both important features and noise, (b) are contaminated by model biases, (c) are
too coarse to resolve all features, and (d) are influenced by uncertainty inherent
in the initial conditions. Ensemble systems are improving, as are post-processing
applications like BMA, but they are still unable to generate calibrated probabilities
and maintain sufficient resolution through the required forecast period for most
operational requirements. Until better objective tools can be developed forecasters
will need to face the challenge of subjectively filtering this information to generate
the best possible weather forecast.
The forecaster’s primary objective is to separate the predictable processes from
the noise (those features with no hope of predictive skill), discount the noise, and
Fig. 11.17 A high resolution visible satellite image showing an onshore flow of marine air with
areas of stratus and stratocumulus clouds pushing eastward from the Pacific Ocean through the
Strait of Juan de Fuca
then adjust for any known biases or errors associated with the salient processes.
The predictability is not only scale and phenomena based but also dependent upon
the forecast range. What might be predictable in a Day 1 forecast may not be
at Day 3, but some robust terrain features seem to have very high predictability
nearly independent of forecast range (e.g., channeled flow as long as the pressure
gradient is well forecast). Today, identifying predictable features of model forecasts
is primarily a subjective process, although model post processing, gridded MOS,
and ensemble applications are already helping address this issue and will contribute
even more in the future.
To accomplish this diagnosis of NWP guidance, the forecaster must have a clear
understanding of what features are possible (along with their impact on the weather),
the underlying physics, and know how the various phenomena are manifest in
the observational database and in the NWP simulations. This is primarily done
through the use of a set of conceptual models. Without this understanding and
associated conceptual models the NWP simulations quickly become black boxes
and the forecaster is very limited as to what kind of adjustments can be made to the
forecast. In this case, the forecast is essentially left to the objective forecast system.
Conceptual models are also critical to the short-term forecast process. For
example, an NWP solution may show a well understood event that is on-average
skillfully forecast by the models and as a result generally reflected in the local
forecasts. Yet, due to poor initial conditions, the model develops the event with
an incorrect forecast timing or location. In this case, the forecaster can use the
conceptual model of the event (including evolution and associated weather impacts)
to adjust the local forecast.
A key benefit of routine real-time mesoscale NWP in an area of complex
terrain is the ability to build climatologies of recurring events and structures –
essentially adding to the tool box of conceptual models. The value of this process
cannot be overstated. The forecaster’s situational awareness is improved as in
situ observations of under-sampled structures can now make sense in light of this
heightened understanding. Adjustments to the model solutions for poor timing or
placement can now be done with much improved physical insight. Over the course
of this process, forecasters also develop a sense of biases in the way the models
predict specific features or handle the boundary layer and sensible weather elements.
For example, models may consistently over-predict the intensity of some features
and under-predict others (especially those that are not fully resolved), or they may
incorrectly decouple the nocturnal boundary layer, resulting in a warm bias in low
temperatures and a high bias in wind speeds. These observations and understandings
go side-by-side with the model solutions in the forecaster’s tool box.
On a day-by-day basis, forecasters survey the NWP output for confirmation of
what their regional forecast experience suggests to them will happen. They are also
looking for “events” in the model solution that they may not be anticipating and
which may need additional thought or scrutiny. Overall, forecasters generally view
the solution broadly in an effort to filter out the finest details and retain the more
robust signals. They are likely to look at a sequence of runs, or other parallel runs, in
order to obtain a sense of uncertainty or confidence in a given solution. They must
recognize that each of these runs is but one realization of a spectrum of different
possible evolutions. The forecaster must routinely ask: “What feature is most likely
to occur?” “What can I say about its timing, location, and intensity?” “How do
I successfully incorporate this information into my forecast and communicate it
effectively?” For some processes, the forecasters can be more effective by using
simple rules of thumb, or forecast guidelines developed through past studies or
climatologies of events, than if they used a single deterministic run; these rules tend
to smooth features and, as such, better represent the underlying uncertainty.
In summary, the optimal forecast approach today is a blend of subjective tech-
niques and statistical post-processing of NWP solutions, combined with local fore-
casting tools and cognitive experience. Until more mature tools are developed that
better address forecast uncertainty, model bias, and situational awareness, the fore-
casters’ key role will be spread between the generation of the forecast through this
process, and professional, value-added interpretation of the forecast for the end user.
11.8 Forecast Applications and Special Field Programs

in Complex Terrain
As pointed out in Sect. 11.3, most operational NWP centers are running models with
grid spacings that lie between 10 and 15 km. Generally speaking, this resolution
supports reasonably well today’s operational forecasters. In comparison, field
programs and special venues bring much greater spatial and temporal challenges
to the operational meteorologist and provide an excellent opportunity to explore
future observing systems, forecast techniques, and communication. In Sect. 11.7,
an example was used that showed the utility of special modeling and observing
efforts in building conceptual models and increasing understanding.
The Winter Olympic Games, by their nature, are sensitive to the complex weather
found in mountain environments. Race officials, athletes, and media demand
specialized weather services. These require forecasts with spatial scales on the order
of meters and temporal scales on the order of minutes. Weather elements critical to
the successful completion of events include precipitation type/amount, visibility,
wind, temperature, and humidity.
As already discussed, high-resolution NWP simulations can help reveal local
circulations and orographic forcing, especially in areas of complex terrain, thus
adding to the knowledge gathering for meteorologists. Indeed, researchers have tried
to address the spatial challenges facing meteorologists, by implementing ultra-high
resolution models during recent Winter Olympic Games (Horel et al. 2002; Mailhot
et al. 2010). These ultra-high resolution runs when combined with a corresponding
enhanced observational network have provided an excellent opportunity to evaluate
the utility of these models.
For the 2002 Winter Olympics period in Salt Lake City, Utah, Hart (2005)
found that wind and precipitation simulations from a 4-km MM5 model were
measurably better than those produced by the 12-km domain. The improvements
were particularly evident in areas with higher precipitation amounts, where oro-
graphic forcing was aided by the improved model topography. Using data from the
2006 Winter Olympic period in Torino, Stauffer et al. (2007) studied the impact
of four dimensional data assimilation and horizontal resolution on MM5 model
performance. Here, too, wind errors were reduced when going from a coarser
(12-km) domain to one at 4-km; interestingly, the differences in predictive skill
were relatively small between 4-km and the finest scale 1.4-km.
Preliminary results from the 2010 Winter Olympics have provided further
evidence of the quality of high-resolution numerical model output (Mailhot 2010).
Preliminary findings have indicated significant improvements with a 1-km grid for
near surface winds when compared to coarser grids. Notable improvements were
also found for temperature from 1- and 2.5-km grids.
Leading up to events such as the Winter Olympics, there is a heightened state of
interaction between researchers and forecasters (see Chap. 12). Both are motivated
by a common goal to succeed and, as a result, we see an enhanced suite of guidance
products. For venues at Winter Olympics, site specific (spot) forecasts are demanded
by users. At these fine spatial scales even the best available topographic data
can be insufficient to address forecast requirements and so post-processed model
output is absolutely necessary. During the 2002 Winter Olympics, MM5 model
output statistics (MOS) provided hourly values of temperature, dewpoint, wind
speed, and wind direction (Horel et al. 2002). Leading up to the 2010 Games,
Environment Canada (EC) developed three experimental NWP systems in support
of the forecasters (Mailhot 2010): (1) a regional ensemble prediction system, (2)
a high-resolution deterministic prediction system (including grid spacings of 2.5-
and 1-km) with improved model dynamics, geophysical fields, cloud microphysics
and radiation schemes and (3) an external land surface modeling system integrated
on a 100-m grid. Consultations with venue forecasters led to a specialized list of
products. Examples included 2D maps, time series or meteograms at a number of
surface stations (Figs. 11.18 and 11.19), cross-sections along specific lines, and
vertical soundings at standard and additional Olympic locations.
While information on model performance and an enhanced suite of products sup-
port the meteorologist in providing better forecasts for special venues, ultimately,
considerable in-situ experience is required to optimally apply this information.
Often, a climatology or even local knowledge is missing at special venues and
field projects. An iterative process to understanding meteorological processes is
inevitably required in these situations. Forecasters begin with a solid footing in
mountain meteorology theory and by gaining local experience; they establish a
sense of the climatology of the area. With this information in combination with
the NWP models, rules of thumb and conceptual models (Sect. 11.6) are developed
and refined as experience grows.
A good example of the development of new forecast tools was seen at the
Callaghan Valley venue during the 2010 Vancouver Winter Olympics. This venue
hosted the ski jump competition, which is extremely sensitive to variations in wind
speed and direction. In the years leading up to the 2010 Games, little was known
of the meteorology in the Callaghan Valley. For three years prior to the Games,
meteorologists spent the winter season on-site, gaining valuable experience and
learning as much as possible about the local weather. In addition, experience with
the NWP models and learning to identify the key synoptic forcings led to some
simple conceptual models of wind patterns in the valley. Further on-site experience,
developing a climatology, and most important of all, exposure to high temporal
resolution data helped establish finer-scale conceptual models and tools to enable
the forecasting of winds at the ski jump itself.
Interestingly, the NWP models, with grid spacings as low as 1-km, were not
deemed critical (Teakles et al. 2010) in the forecast process for the ski jump. This
is not to say that the NWP models were not used. Rather, the output was used to
guide the meteorologist toward the appropriate conceptual model based on the local
scale forcing. Due to the extremely fine spatial scales for the ski jump (100 m),
the meteorologist relied upon refining conceptual models in light of real time high
resolution wind (Swiss timing) data.
A simple example of how observations and models were used to develop valuable
conceptual models can be illustrated using data from the Olympics (Mailhot et al.
Fig. 11.18 Example of 2-dimensional output for Whistler, BC area from the GEM-LAM (1 km
resolution) during the 2010 Winter Olympics. Contours represent terrain. Color shading represents
precipitation type (Reproduced with the permission of Environment Canada)
2010). In Fig. 11.20, a multi-season collection of surface wind observations at a

location in a Vancouver Olympics’ venue shows highly preferred regimes, yet is
lacking the physical insight. However, when a 1-km simulation is used (Fig. 11.21)
there is a greater appreciation for the mountain-valley circulation. Given this strong
preference for repeating regimes, one can legitimately ask “Does one need a full
physics model to forecast its occurrence or can it be handled more simply with
statistical methods or even rules of thumb applied to a series of observations?” One
would also need to evaluate whether a mesoscale NWP solution would actually have
any skill in forecasting day-to-day variations or if optimal skill would be most easily
obtained using the diurnal climatology.
Special venues such as those at Winter Olympics have provided excellent
learning experiences for meteorologists faced with forecasting in complex terrain.
They have also provided an opportunity to evaluate new observing systems, tools
Fig. 11.19 Example of time series (meteogram) from the GEM-LAM (1 km resolution) during
the 2010 Winter Olympics (Reproduced with the permission of Environment Canada)
and approaches – essentially a glimpse into the future of what may become the
routine. The Olympics are about legacies and this has extended to the meteorological
community with the development of highly refined numerical models and products
as well as new forecast techniques.
Fig. 11.20 A frequency histogram (of wind speed versus direction) for the Vancouver Olympics
Callaghan Valley venue. The clustering of direction/speed pairs shows the dramatic local control
placed by the terrain (Reproduced with the permission of Carl Dierking)
Fig. 11.21 GEM-LAM 1 km resolution simulation (1500 UTC left panel and 1800 UTC right
panel) of the cross-barrier wind speed and wind direction on 25 January 2009. Near the surface,
drainage flow along the slope contrasts with the westerly flow below the descending mountain
wave. As the boundary layer stability weakened, a shift to more upslope conditions was seen along
the slope (Reproduced with the permission of Environment Canada)
11.9 Vision for the Future
Critical decisions are being made each day based on today’s operational mountain
weather forecasts (Chap. 1). This growing dependence on, and increasing demand
for, skillful forecasts is strong testimony to their recognized value in making
weather-sensitive decisions. The research, academic, and operational communities
have delivered NWP successes that were unimaginable just decades ago and are
the foundation for these improvements. In step, the forecasters have adjusted their
forecast process to best utilize the increasingly skillful guidance products to develop
today’s operational forecasts. In addition, with the increasing sophistication and
complexity of today’s forecasts, forecasters are playing an increasingly important
role of explaining and interpreting the results for those in the user community.
Clearly there is much to be proud of in what has already been accomplished in this
area. Looking forward, there should be much optimism as the future of mountain
weather forecasting is very bright.
There are a few areas where improvements can clearly be made in order to
develop a more mature mesoscale mountain weather forecasting process. Pre-
dictability and uncertainty remain critical to the future of mountain weather
forecasting and are among the top challenges remaining to be met. Central to this
area is the productive use of ensemble prediction systems to better estimate the
uncertainty of the NWP forecast. For example, a calibrated PDF of an element
would be extremely valuable to the forecaster in determining the correct forecast
message, and to the user in estimating risk and taking appropriate actions. The
well constructed PDF should correctly filter out the elements for which there is
not forecast skill and indicate a likelihood of those for which there is forecast skill
(a task currently in the hands of the forecaster). The system needs to be run at
scales sufficient to resolve those features where there is a reasonable expectation
for forecast skill. This is based both on the large-scale uncertainties from the
initial conditions and the unique predictability characteristics of the phenomena
themselves. As mentioned earlier, many emerging ensemble forecasting systems are
starting to reach into the resolution and specialization required to meet these needs
and it is likely much progress will be made over the next few years.
In addition to the unfiltered or bulk performance statistics and PDFs, forecasters
need additional focused studies on performance metrics regarding specific phenom-
ena or elements. These would be similar to Reinecke and Durran’s (2009) study of
lee-side mountain waves and would be completed to reveal objective skill metrics
by key features. These additional predictability studies are important tools for the
forecasters to use in deciding how to interpret the NWP solutions. They are also
critical to the research and operational communities in establishing a long-term
strategic plan regarding investment of future computing and scientific resources.
Additionally, statistical post-processing can be brought to bear on the problem with
likely significant benefits. For example, sophisticated uncertainty tools would be
helpful in the effort to address today’s overly deterministic forecasts.
Another key area where improvements will bring great benefit is in defining
current conditions with high-resolution, real-time analyses (on the order of 1 km
grid spacing). These real-time analyses are an absolute necessity for improving
mountain weather forecasting. Skillful analyses (on a very fine grid) would allow
better and more robust verification, better statistical post-processing and bias
removal, and serve as a critical tool to maintain forecaster situational awareness.
Related observation representativeness issues also need to be addressed.
An improved awareness of what is currently going on across an area is incredibly
valuable to everyone, from the forecaster to the end user. Observations alone cannot
meet this need for obvious reasons. The system of the future must have a state-
of-the-science mesoscale data assimilation system that is continually ingesting all
in situ and remotely sensed data and blending these data with a background field
to provide a gridded (very fine scale) estimate of current conditions across all
areas that is both physically consistent across variables, and includes spatially and
temporally varying estimates of uncertainty. This must be provided in near real
time for operational use and decision-making. An optimized analysis (non-real-
time) would also be needed to provide the truth analysis necessary to identify model
biases, forecast skill, and post-processing metrics.
A final key area in which gains can be made is in the communication and
interpretation of the forecasts. When the improved analysis and modeling systems
are in place, the forecaster’s role will be one of monitoring both systems to maintain
a high degree of situational awareness with respect to the current conditions and an
understanding of the potential scenarios that may play out. Each scenario would
be represented statistically in the ensemble guidance products but not necessarily
in a clear depiction of the actual phenomena or physical process. The forecaster
will need to identify these processes or scenarios by surveying the individual
ensemble members and relying on their experience and scientific understanding and
appreciation of the spectrum of possible processes. The forecaster’s experience and
knowledge will benefit those needing to interpret the various forecast products. As
forecast ability and skill improve, so will the complexity of the data. Forecasters will
need to provide professional interpretation in support of optimal decision making by
those who are weather sensitive but not necessarily weather savvy.
A key role for the forecaster will be one of highlighting likely events through
various storm watches and warnings that will also include objective uncertainty
information. There will also be frequent opportunities for the forecaster to offer
additional insights, update short-term forecasts based on the various observing
systems and real-time analyses, and provide enhanced decision support services to
those confronted with making critical decisions.
11.10 Summary
The recent and ongoing advances in mesoscale numerical weather prediction and
fundamental scientific understanding of mountain weather have been nothing short
of remarkable. If forecasters are going to continue to add value to the forecast
process, the advances clearly demand forecasters become even better scientists and
communicators. This is happening. With over a decade of near-real-time mesoscale
NWP simulations, and many lessons learned, forecasters are now embracing the use
of these data in the day-to-day forecast process – and many use them with a great
deal of success.
Past are the days when significant terrain-induced features are absent from
NWP guidance and the forecaster could simply “add in” qualitatively the feature’s
anticipated impacts. Many of these features are now being captured by the oper-
ational NWP models, at least in a simple sense – some skillfully and others not.
Today’s best forecasters are those who have an excellent understanding of these
processes, grasp their reflection in the guidance, and can make sound modifications
and adjustments to the forecast. While there remain considerable opportunities for
forecasters to add value to the forecast process, the temptation for forecasters to
try to improve on NWP guidance with simple hunches needs to be resisted. Each
change is ideally based on a solid footing of a known model deficiency or bias,
predictability concerns, or user requirements.
In many ways the steep learning curve for the operational use of mesoscale
NWP in weather forecasting for complex terrain had to do with predictability
issues. The richness of the solutions and gains in physical insights were relatively
easy. It was the inconsistent day-to-day performance (many remarkable successes
interspersed with equally impressive failures) that presented the greatest challenge.
How was a forecaster to recognize a priori the likely success of the NWP forecast
on any given day? It was a frustrating story of sorting out the reliable signal
from the chaff. It was also the most significant obstacle to collaboration between
the academic and operational forecasting communities. Some in the academic
community were too quick to claim success and failed to fully recognize the critical
difference between physical realism and skillful prediction.
There are still many opportunities for improvements in mountain weather
forecasting spread across the research and operational communities (Chap. 12).
For the research community, in addition to ongoing enhancements to full physics
NWP models (see Chaps. 9 and 10), they can clearly bring significant benefit
to the table by exploring further (1) mesoscale ensemble prediction in complex
terrain, (2) mesoscale objective analysis techniques, (3) statistical post-processing
and estimates of uncertainty, and (4) new and more appropriate verification tools.
Operational forecasters need to be open to these advances, recognizing that
the advances are not diminishing their forecast efforts but rather, are opening
new doors and opportunities for them to do an even better job. If embraced,
the advances will help forecasters deliver more skillful forecasts and significantly
improved services. No doubt, as the research community pushes ahead, the role
of the forecaster in making mountain weather forecasts will continue to evolve.
That role will not be less critical to the forecast process, rather, as forecast quality
improves, there will be more opportunities for forecasters to provide valuable
information and support. There will be significant opportunities to deliver detailed
(with uncertainty), customized, and truly beneficial weather forecasts to a broad
spectrum of those with weather sensitive interests.
The academic and operational communities must work closely to address the
remaining challenges confronting mountain weather forecasting. Each must also
recognize the critical and positive role played by the other, and be sensitive to
each community’s priorities and motivations. If we can move ahead through a truly
collaborative process, the ultimate winners will be those impacted by weather in the
mountains – for example, economic impacts on industry and transportation will be
minimized, outdoor enthusiasts will be safer, and lives will be saved.
Acknowledgements The authors would like to thank the Editors (Fotini Katopodes Chow,
Stephan F.J. de Wekker, and Bradley J. Snyder) for their invitation to write this chapter, their
tireless efforts to get it to press, and their leadership in helping to organize the Workshop and
meetings. We would also like to thank reviewers Mike Angove, Clifford Mass and Ron Goodson
for their many helpful comments.
References
Anthes, R. A., Y.-H. Kuo, D. Baumhefner, R. M. Errico and T. W. Bettge, 1985: Predictability of
mesoscale atmospheric motions. Advances in Geophysics, 28B, 159–202.
Berri, G. J., and J. Paegle, 1990: Sensitivity of Local Predictions to Initial Conditions. J. Applied
Meteor., 29, 256-267.
Boer, G.J., 2003: Predictability as a Function of Scale. Atmosphere-Ocean, 41 (3) 2003, 203–215.
Bowler, N. E., A. Arribas, S. E. Beare, K. R. Mylne, and G. J. Shutts, 2009. The local ETKF
and SKEB: Upgrades to the MOGREPS short-range ensemble prediction system. QJRMS, 35:
767-776.
Candille, G., 2009: The Multiensemble Approach: The NAEFS Example. Mon. Wea. Rev., 137, 5,
1655-1665
Colman, B. R., and C.F. Dierking, 1992: The Taku Wind of Southeast Alaska: Its Identification
and Prediction. Wea. Forecasting, 7, 49-64.
Dallavalle, J. P., and B. Glahn, 2005: Toward a Gridded MOS System. Preprints, 21st Conference
on Weather Analysis and Forecasting, Washington, D. C., Amer. Meteor. Soc.
Gilbert, K. K., B. Glahn, R. L. Cosgrove, K. L. Sheets, and G. A. Wagner, 2009: Gridded
model output statistics: Improving and expanding. Preprints, 23rd Conf. Weather Analysis and
Forecasting and 19th Conf. Numerical Prediction, Omaha, NE, Amer. Meteor. Soc.
Glahn, B., K. Gilbert, R. Cosgrove, D. Ruth, and K. Sheets (2009): The Gridding of MOS. Wea.
Forecasting, 2, 520-519.
Hart, K. A., W. J. Steenburgh, and D. J. Onton, 2005: Model Forecast Improvements with
Decreased Horizontal Grid Spacing over Finescale Intermountain Orography during the 2002
Olympic Winter Games, Wea. Forecasting, 20, 558-576.
Horel, J., T. Potter, L. Dunn, W. J. Steenburgh, M. Eubank, M. Splitt, and D. J. Onton, 2002:
Weather Support for the 2002 Winter Olympic and Paralympic Games. Bull. Amer. Meteor.
Soc., Feb., 227-240.
Lorenz, E. N., 1969: The predictability of a flow which possesses many scales of motion.
Tellus, 21, 289-307.
Mailhot, J., S. Bélair, M. Charron, P. Joe, M. Abrahamowicz, N. B. Bernier, G. A. Isaac, R.
McTaggart-Cowan, J. Milbrandt, C. Doyle, B. Denis, A. Erfani, R. Frenette, A. Giguère,
N. McLennan, and L. Tong, 2010: Environment Canada’s Experimental Numerical Weather
Prediction Systems for the Vancouver 2010 Winter Olympic and Paralympic Games. Bull.
Amer. Meteor. Soc., 91, Issue 1, 31-36
Mass, C. F., D. Ovens, K. Westrick, B. Colle, 2002: Does Increasing Horizontal Resolution Produce
More Skillful Forecasts? Bul. Amer. Meteor. Soc., Volume 83, Issue 3, 407-430.
Mass, C. F., M. Albright, D. Ovens, R. Steed, M. MacIver, E. Grimit, T. Eckel, B. Lamb, J.
Vaughan, K. Westrick, P. Storck, B. Colman, C. Hill, N. Maykut, M. Gilroy, S. A. Ferguson,
J. Yetter, J. M. Sierchio, C. Bowman, R. Stender, R. Wilson, and W. Brown, 2003: Regional

Environmental Prediction Over the Pacific Northwest. Bull. Amer. Meteor. Soc., 84, 1353-1366.
Mass, C. F., J. Baars, G. Wedam, E. Grimit, and R. Steed, 2008: Removal of Systematic Model
Bias on a Model Grid. Wea. Forecasting, 23, 438-459.
Mass, C.F. and Y.-H. Kuo, 1998: Regional real-time numerical weather prediction: Current status
and future potential. Bull. Amer. Meteor. Soc., 79, 253–263.
Raftery, A. E., T. Gneiting, F. Balabdaoui, M.Polakowski, 2005: Using Bayesian Model Averaging
to Calibrate Forecast Ensembles. Mon. Wea. Rev., 133, 5, 1155-1174
Reinecke, P.A., and D. R. Durran, 2009: Initial condition sensitivities and the predictability of
downslope winds. J. Atmos. Sci., 66, 3401-3418.
Stauffer, D. R., G. K. Hunter, A. Deng, J. R. Zielonka, K. Tinklepaugh, P. Hayes, and C. Kiley,
2007: On the role of atmospheric data assimilation and model resolution on model forecast
accuracy for the Torino Winter Games. Preprints, 22nd Conference on Weather Analysis and
Forecasting and 18th Conference on Numerical Weather Prediction, 25-29 June 2007, Park City,
UT, Amer. Meteor. Soc.
Teakles, A, C. Dierking, and N. McLennan, 2009. Strong outflow events affecting the 2010
Olympic venues in the Callaghan Valley. Canadian Meteorological and Oceanographic
Society, Halifax, Nova Scotia, 31 May – 4 June 2009.
Vukicevic, T., and R.M. Errico. 1990. The Influence of Artificial and Physical Factors upon
Predictability Estimates Using a Complex Limited-Area Model. Mon. Wea. Rev., 118, 1460-
1482.
Chapter 12
Bridging the Gap Between Operations
and Research to Improve Weather Prediction
in Mountainous Regions
W. James Steenburgh, David M. Schultz, Bradley J. Snyder,

and Michael P. Meyers
Abstract A gap between operational and research meteorologists has existed

since the infancy of weather forecasting and represents an obstacle to progress
in meteorology. This gap is related to the profoundly different perspectives and
professional expectations of operational and research meteorologists. For the
knowledge, observations, tools, and models described in this book to reach their
full potential, the mountain meteorology community must work more effectively to
bridge this gap, as described in this chapter. Essential to this effort are advocates
who are capable of interacting, communicating, and commanding respect with both
the operational and research communities. As a result, the mountain meteorology
community should provide the attention and resources needed to ensure that
future advocates are created from the pool of young scientists and forecasters.
The community should also ensure that knowledge and technological advances
from field programs and other research efforts are effectively transferred into
operations and, at least in North America, explore the development of an integrated
research and forecast center to tackle challenges in mountain hydrometeorology
W.J. Steenburgh ()

Department of Atmospheric Sciences, University of Utah, Salt Lake City, UT, USA
e-mail: jim.steenburgh@utah.edu
D.M. Schultz
Division of Atmospheric Sciences, Department of Physics, University of Helsinki,
Helsinki, Finland
Finnish Meteorological Institute, Helsinki, Finland
Centre for Atmospheric Science, School of Earth, Atmospheric and Environmental Sciences,
University of Manchester, Manchester, UK
B.J. Snyder
Meteorological Service of Canada, Vancouver, Canada
M.P. Meyers
NOAA/National Weather Service, Grand Junction, CO, USA

694 W.J. Steenburgh et al.
and fire–atmosphere prediction. Although the existence of a modest gap reflects a

healthy scientific and forecasting enterprise, these and other gap-bridging activities
and incentives described in this chapter should benefit the entire mountain weather
community, its operational and research sectors, and, via improved forecasts, society
at large.
12.1 Introduction
“One of the greatest obstacles to the progress in meteorology is undoubtedly to be

found in the wide gulf between the mathematical theory on one hand and the applied
science weather-map analysis and forecasting, on the other.” As true now as when
Rossby (1934) said it, the gap between operational and research meteorologists
prevents forecasters from extracting maximum benefit from today’s sophisticated
observations, forecast tools, and numerical models, and inhibits researchers from
fully evaluating weaknesses in current scientific understanding and capabilities.
History suggests that forecast improvements and scientific advances accelerate
when operational and research meteorologists respect their unique perspectives
and interact productively (e.g., Board on Atmospheric Sciences and Climate 2000;
Waldstreicher 2005; Mass 2006; Volkert and Gutermann 2007). If the observations,
models, tools, and knowledge described in this book are to reach their full potential,
the operational and research communities must develop a closer, more integrated
collaboration to address critical challenges for weather prediction in mountainous
regions.
This chapter provides a roadmap for bridging the gap and accelerating progress in
mountain meteorology by examining the differing perspectives of operational and
research meteorologists, identifying ingredients for successful gap bridging, and
providing specific examples upon which to pattern future collaborative efforts. We
refer to a single gap, although in reality there may be multiple gaps in the chain from
basic research to operations. Although the operational and research communities
each stand to benefit from increased collaboration, the primary reason to bridge
the gap is to produce better weather forecasts for the benefit of society. As such, a
gap must also be bridged to decision makers and forecast consumers (Morss et al.
2005), but, for the purposes of this chapter, we focus on improving interactions
between operational and research meteorologists in the mountain meteorology
community.
12.2 Causes of the Gap
Winston Churchill said, “true genius resides in the capacity for evaluation of
uncertain, hazardous, and conflicting information.” If so, Churchill would con-
sider forecasters the epitome of true genius. Every working day forecasters face
uncertainty. They never have all the data they want, are confronted with several
12 Bridging the Gap Between Operations and Research. . . 695
computer models (and ensembles) that produce different forecasts, and have
imperfect knowledge of a chaotic atmosphere. Forecast deadlines and the urgent
nature of severe weather demand that forecasters be comfortable making decisions
in the face of uncertainty, even when the data and forecast guidance cannot be fully
analyzed and interpreted. The best forecasters develop schema to quickly organize
the wide array of observational data and model guidance, identify critical issues,
and make good decisions (e.g., Doswell 2004). This ability to make decisions from
limited evidence is known as forecaster intuition, a combination of experience,
conceptual model application, and educated guesses. Such skills are most apparent
in the engaged forecaster (e.g., Roebber et al. 2002; Pliske et al. 2004; Stuart
et al. 2007). Weather forecasting is a scientific endeavor involving hypothesis
formulation, hypothesis testing, and prediction (e.g., Roebber et al. 2004), but it
is inherently less rigorous and more speculative than scientific research.
Although the best forecasters are comfortable dealing with incomplete infor-
mation and evidence, conscientious scientists are not. Scientists must extensively
test hypotheses and determine the generality of their conclusions. They demand
that conclusions be justified and consider statements based on limited evidence
(i.e., the forecaster’s intuition) to be nothing more than speculation. Just as an
Olympic alpine skier may not make the best Nordic skier, despite both disciplines
requiring strength, cardiovascular fitness, and balance on skis, operational and
research meteorologists, despite their common skills and knowledge in meteorology,
are not necessarily interchangeable.
The gap between operations and research originates from these profoundly
different worldviews, which allow forecasters and researchers to succeed within
their individual communities by meeting the expectations of their colleagues and
organizations, but make crossing boundaries difficult. Forecasting is challenging for
most researchers, who are unable to fully ponder and investigate under the deadline
of getting the forecast out. In contrast, research and publication are challenging for
forecasters—the intuition that works well for them in day-to-day operations cannot
be relied upon in formal publications to convince a critical audience. Those who
accommodate and accept these two worldviews recognize that forecaster experience
and practice cannot always be justified with a citation or easily quantified through
calculations, but that evidence and logic are essential for both weather forecasting
and scientific research.
12.3 Perceptions of the Gap
The gap between operational and research meteorologists has been recognized since
at least the early twentieth century (e.g., Rossby 1934; Bergeron 1959) and remains
considerable even today (Doswell et al. 1981; Ramage 1993; Doswell 2007). Snyder
and Loney (2007) reported that operational meteorologists in the Meteorological
Service of Canada (MSC) and the National Weather Service (NWS) of the United
States believe that there are three major contributors to the current gap.
• Limited exposure of researchers to the operational forecast environment. Many

of today’s young research scientists have limited (if any) experience with
synoptic weather analysis and have never worked a forecast shift, so they have
little exposure to the unique demands of operational weather forecasting, as
argued by Doswell (1986). This lack of exposure to operations is ironic because
many of their scientific forebears began their careers as weather forecasters.
For example, analysis and forecasting was essential to the scientific advances of
the Bergen School (Friedman 1989, 1999), and many distinguished atmospheric
scientists of the mid to late twentieth century served as weather observers or
forecasters in World War II [e.g., Heinz Lettau (Besser 2008), Edward Lorenz
(Palmer 2008), Sverre Petterssen (Bundgaard 1979; Petterssen 2001), Richard
Reed (Reed 2003), Frederick Sanders (Sanders 2008), Joanne Simpson (Lewis
1995)]. Nearly all universities emphasize atmospheric dynamics and physics at
the expense of applications and forecasting. When taught, weather analysis and
forecasting is often decoupled from dynamics and physics, when, in practice,
forecasting (and research) requires a broad view that integrates across these
atmospheric subdisciplines (e.g., Doswell et al. 1981; Doswell 1986).
• A lack of training and opportunities for forecasters to participate in research.
Forecasters need training and encouragement to read the meteorological
literature, identify scientific articles relevant to their job, and participate in
research (e.g., Doswell et al. 1981; Doswell 1986). Many forecast organizations
have positions designed to address this need (e.g., the NWS Science and
Operations Officer), but administrative obstacles and other work demands
frequently limit the time available for scientific research and forecaster training.
As a result, many forecasters are unable to keep up with recent progress in the
atmospheric sciences or maintain the mathematical and theoretical knowledge
needed to communicate effectively with the research community. Exacerbating
this problem are bureaucratic rules that prevent or limit forecaster participation
in research and training projects, a lack of release time from forecast shifts
for research and training, and limited rewards and funding for forecaster
participation in research.
• The need for improved methods to transfer results from research to operations
(e.g., Smith et al. 2001; Stuart et al. 2006; Snyder et al. 2006; Rauhala and Schultz
2009). Because of contrasting cultures and professional expectations, forecasters
and researchers prefer different modes of communication. Researchers typically
disseminate research results in journal articles and conference presentations.
Such approaches appeal to scientists and university faculty, but are not favored
by students and forecasters who prefer concrete examples and active participation
(Roebber 2005; Stuart et al. 2007). In fact, only 40% of the US students who
responded to an American Meteorological Society (AMS) member survey said
that they “read printed research literature on a daily or weekly basis” (Stanitski
and Charlevoix 2008). Further, forecasters consider reading journal articles
and attending scientific conferences amongst the least effective training modes
and consider collaborative forecasting with experts (i.e., “double banking”),
weather-event simulators, and residency training courses with close and regular
Fig. 12.1 Annual threat scores for 24-h, 1-in. (2.54-cm), day 1 quantitative precipitation forecasts
produced by the NCEP North American Mesoscale (NAM) model, Global Forecast System (GFS),
and forecasters at the HPC (Courtesy HPC)
interaction with scientific experts to be the most effective training modes (Snyder
and Loney 2007). Such responses are not unexpected as forecasters would much
rather learn in ways that directly relate to their job rather than through ways
that researchers learn best, and many training programs may not contain enough
operationally relevant material for forecasters (e.g., Smith et al. 2001). Therefore,
such differences in learning contribute to creating the gap.
Addressing these three issues head-on is essential for the mountain meteorology
community to extract maximum societal benefit from today’s sophisticated obser-
vations and forecast tools. Although concerns that numerical weather prediction
and automation will eliminate humans from the forecast process have been raised
since at least the 1950s (e.g., Harper 2008, pp. 229–231; 2009), well-trained and
fully engaged forecasters continue to play “a clear role in the forecast process by
contributing a wealth of knowledge, tools, and techniques that cannot be duplicated
by computers or [numerical weather prediction]” (McCarthy et al. 2007). This
is particularly true in short-range forecasting, as illustrated by comparing day-1,
24-h quantitative precipitation forecasts produced by operational numerical weather
prediction models run by the National Centers for Environmental Prediction
(NCEP) with those produced by forecasters at the NWS Hydrometeorological
Prediction Center (HPC, Fig. 12.1). Over the past 15 years, even as model forecasts
have improved, HPC forecasters have maintained a consistent annual threat score
advantage of 0.05 over the NCEP models. Clearly, advances in numerical model
skill do not necessarily result in a reduction in the value added by humans. Instead,
well-trained forecasters take advantage of these improvements and extend forecast
skill (e.g., Bosart 2003; Harper et al. 2007; Erkkilä 2007; Sills 2009).
On the other hand, forecasters that lack sufficient training, fail to keep up with
scientific progress, and rely excessively on numerical model guidance provide
little benefit to the forecast process, particularly in areas of complex terrain where
numerical weather prediction models inadequately resolve orographic processes and
fine-scale weather variability. Snellman (1977) coined the term “meteorological
cancer” to describe the insidious overreliance on information generated by com-
puters at the expense of human interpretation and cognition. As stated forcefully
by Bosart (2003), forecasters who grow accustomed to letting MOS [model output
statistics] and the models do their thinking for them on a regular basis during the
course of their daily activities are at high risk of “going down in flames” when
the atmosphere is in an outlier mode. Clearly, the mountain weather forecasting
community can only reach its full potential if forecasters are both highly educated
and fully engaged in the forecast process. This goal can only be accomplished if the
operational and research communities work together.
To summarize, these perspectives highlight three critical challenges that individ-
uals seeking to bridge the gap must address:
• Forecasters must be motivated, fully engaged, and encouraged to participate in
research and training on a regular basis. To a large extent, this motivation must be
internal, but also bureaucratic barriers preventing regular participation in research
and training can be removed.
• Researchers must recognize and be willing to communicate the forecast rele-
vance of their research. This requires knowledge and respect of the operational
forecast environment, understanding of the role of humans in the forecast pro-
cess, and the use of effective methods for transferring knowledge and techniques
from research to operations.
• Forecasters and their managers must recognize the value of a mathematical
and theoretical education and develop the necessary educational foundation to
pursue one, while researchers must embrace the potential for weather analysis
and forecasting to enable them to formulate and test scientific hypotheses using
real weather data.
12.4 Why Bridge the Gap?
Addressing these three challenges requires effort—often considerable effort. Why

should anyone take this hard road to bridge the gap, when remaining isolated
within one’s community can produce relatively easy personal success? There are
two reasons.
The first is the shared desire among forecasters and research scientists to have
a positive impact on our profession and society. The second is that the mountain
weather community simply cannot reach its full potential without productive
collaboration between operational and research meteorologists. Forecasters must
be well educated and trained to extract maximum benefit from today’s observations
and forecast tools and to prevent avoidable forecast errors related to the misuse
or ignorance of scientific understanding (Doswell 1986, 2007; Bosart 2003).
Researchers can benefit from the regular analysis of evolving weather systems,
which, even given the inherent limitations of the operational data stream, enables
improved formulation and testing of scientific hypotheses, stimulates the broadening
of research results, and prevents the overgeneralization of conclusions based on
comprehensively sampled but limited sample size field-program studies (Doswell
1986, 2007). Finally, increasingly tight budgets demand the rigorous and efficient
testing of research advances, including quantification of their impact on the forecast
cycle and decision making (e.g., Doswell and Brooks 1998). Such a cycle requires
increased collaboration between researchers, forecasters, and decision makers
through an end-to-end research approach (Morss et al. 2005).
12.5 Ingredients of Successful Gap Bridging
At the heart of successful gap-bridging efforts are forecasters and scientists

determined to be advocates for improved collaboration. At least two advocates
(preferably more) are generally needed (one from each side) and, if the collaboration
is between organizations, the advocates may occupy supervisory, managerial, or
administrative roles within their organizations. Without advocates on each side, no
way exists to make or enforce collaboration, especially when resources (personnel,
computer support, release time from forecasting for collaborative projects, etc.) are
needed. Advocates must be able to communicate with and command respect from
both the forecast and research communities. These individuals may be doctoral-
level scientists if they are forecasters, perhaps with considerable research experience
before becoming a forecaster, or researchers who began their career as a forecaster
or are passionate about weather and forecasting. Advocates may also be hands-on
managers who have the ability to change reward structures for participants or reduce
bureaucratic barriers to collaboration.
In addition to an advocate or advocates, additional ingredients for successful
bridging of the gap are typically required. Although not all need be present, these
ingredients may include:
• Buy-in and commitment. Both the operational and research sides must embrace
each others’ different perspectives, learn to speak each others’ language, and
possess the staying power to see the work to completion. Not everyone in the
project or organization needs to be “on board,” but a critical mass is needed for
success, particularly in larger projects. Researchers must recognize that spending
time educating forecasters about new research approaches will be required
(e.g., Morss and Ralph 2007). Likewise, forecasters must be open to learning and
Fig. 12.2 Attendees of the 2008 American Meteorological Society/Cooperative Program

for Operational Meteorology, Education and Training/Meteorological Service of Canada
(AMS/COMET/MSC) Mountain Weather Workshop: Bridging the Gap Between Research and
Forecasting, Whistler, BC, Canada, included research, operational, and student meteorologists.
Presentations and interactions at the workshop stimulated this book and chapter
applying new approaches. Setbacks during collaboration are common, so buy-in

is essential for long-term success.
• Stimulation at the grassroots level. The cliché “if the people lead, eventually
the leaders will follow” applies here. The most successful collaborations begin
or are stimulated at the bottom and are not simply imposed by management.
Grassroots collaborations can succeed with either the support or neglect of
management. (Active interference by management is likely to lead to failure.)
Forced collaborations might be successful in the short term, but total top-down
efforts are rarely successful (Doswell 2007, p. 16).
• Collocation. Collocation enables more regular and routine interaction of research
and forecast meteorologists. It facilitates communication and collaboration. The
collocation may be temporary (e.g., joint forecast production, residency courses,
Fig. 12.2), or permanent where forecast and research meteorologists are housed
in a common facility (and perhaps share some job responsibilities). Collocation
alone, however, does not guarantee success. As noted by Doswell (2007),
“organizational structure and proximity do not necessarily result in productive
collaboration. People chose to collaborate, or not.” Given collocation, advocates

must still provide leadership, encouragement, and a framework for collaboration.
• Securing the time, resources, and personnel needed for meaningful collabora-
tion. Discussion and interaction must be actively promoted, but talking alone
is not sufficient. Forecasters must gain release time from operational respon-
sibilities to participate in research, education, and training on a regular basis.
Similarly, researchers must be patient because the progress of working with shift
forecasters may take longer than a typical research project. Because scientists are
more likely to have experience with applying for grants and funding, most of the
onus is upon them to supply students and personnel to help with the research,
although some operational avenues may be available.
• Clearly defined priorities and goals. Key scientific issues are not necessarily key
operational issues, so compromises and adjustments must be made to ensure that
they are “win–win” for both operations and research. Identifying goals for the
operational community and mapping them onto the capabilities of the researchers
is one way to propose achievable goals and articulate collaboration priorities.
• Establishing incentives. Because the institutional reward systems for researchers
and forecasters are different, each community must recognize that bridging
the gap may require nontraditional incentives for the individuals involved.
Oftentimes the advocate is a senior person for whom the traditional reward
system is no longer a motivator—the reward is in seeing the success of the
interaction between researchers and forecasters and the fruits of their labor.
12.6 Examples of Successful Gap Bridging
How do these ingredients work in practice to produce a fruitful collaboration?

Here we provide examples of successful gap bridging both inside and outside the
mountain weather community. Examples from the latter are used in areas where the
mountain meteorology community can learn from the efforts in other meteorological
subdisciplines (e.g., severe convective weather, tropical meteorology).
12.6.1 Advocates for Collaboration
One powerful example of what advocates can accomplish is provided by Dr. Cliff
Mass, professor of atmospheric sciences at the University of Washington, and Dr.
Brad Colman, Science and Operations Officer (since promoted to Meteorologist-
In-Charge) at the NWS Forecast Office in Seattle, who have forged a two-decade
relationship to advance local weather knowledge and prediction. In addition to
passion, these two individuals possess the key characteristics identified above as
necessary for bridging the gap. Dr. Mass is a research scientist with a passion
for weather forecasting and an ability to communicate to operational forecasters,

whereas Dr. Colman is a doctoral-level forecaster who commands the respect of the
research community.
Efforts led by these two advocates have: (1) established a mesoscale surface
observing network, (2) created local high-resolution and regional-scale ensemble
modeling systems, and (3) contributed to the successful execution of the COAST
and IMPROVE field programs examining front–mountain interactions and oro-
graphic precipitation processes over the Olympic, Coast, and Cascade Mountains
(e.g., Bond et al. 1997; Mass et al. 2003; Stoelinga et al. 2003). Mass et al.
(2003) describe how this interaction has stimulated research at the University of
Washington, including efforts to improve model parameterization and ensemble
prediction techniques. In turn, NWS forecasters (and ultimately the general public)
have benefitted from the transfer of knowledge and forecast tools into operations.
For example, the powerful “Hanukkah Eve Wind Storm” (14–15 Dec 2006), which
killed 15 people in western Washington and left an estimated 4.08 million people
without power, was forecast with remarkable specificity, urgency, and lead time by
the NWS Forecast Office in Seattle (Washington State Military Department 2007,
pp. 10–11). Such forecasts, which minimized loss of life and enabled a timely emer-
gency response and recovery, were enabled by knowledge of windstorms produced
by landfalling cyclones spawned by Mass’ research group (e.g., Steenburgh and
Mass 1996), advances in local numerical forecast modeling (e.g., Mass et al. 2003),
and well-educated and fully engaged forecasters.
There are two major lessons to be learned from this example. The first is
the power of bringing two talented and motivated advocates together from the
operational and research communities. In this case, the late Dr. Tom Potter, former
director of NWS Western Region, helped stimulate the partnership by encouraging
Dr. Colman to move to Seattle, become the science and operations officer, and
improve training and science within the ranks of the NWS. The second is the
benefit provided to each respective community when the gap is bridged. The
research program built at the University of Washington has benefited from drawing
motivation from applied forecasting problems (Mass et al. 2003), whereas NWS
forecasters benefit from new knowledge and state-of-the art weather analysis and
forecast tools.
12.6.2 Buy in and Commitment Stimulated by a Common

Research and Forecast Problem
12.6.2.1 Olympic Winter Games
Over a 17-day period every 4 years, the Olympic Winter Games require extremely
precise mountain weather forecasts at high temporal and spatial resolution (Horel
et al. 2002a). Significant and sometimes hazardous weather has impacted nearly
every Olympic Winter Games, and nearly all outdoor competitions are sensitive
Fig. 12.3 The Olympic Winter Games provide a uniting goal and an exceptional opportunity for
bridging as exhibited by the closer interactions between research and operational meteorologists
(pictured) prior to and during Vancouver 2010
to subtle variations in otherwise “garden variety weather.” Knowledge and skill

predicting all of the phenomena described in this book (and more), each of which
occurs in a region with unique climatological and topographic characteristics, are
needed for a successful Games.
The Olympic Winter Games provide a uniting goal for operational and research
meteorologists, with the “Olympic Spirit” stimulating buy-in and commitment
(Fig. 12.3). During the 2002 Olympic Games in Salt Lake, for example, three
disparate groups were brought together—the University of Utah to develop and
provide a regional mesonet and modeling system, the NWS to provide public
forecasts and warnings to protect lives and property, and a team of 13 private sector
meteorologists to provide detailed forecasts for the outdoor sports venues (Horel
et al. 2002a).
These three groups collaborated to ensure forecast-relevant research and devel-
opment and to provide the necessary education and training to take advantage of
new knowledge and forecast tools. For example, the mesoscale modeling system
developed for the games was used by forecasters for 3 years prior to the Games. This
long-term experience with the modeling system enabled the forecasters to determine
that the direct model output was insufficient to provide the detailed forecasts
required at outdoor venues and other locations. As a result, the University of Utah
developed model output statistics (MOS) for 18 sites in the Olympic region (see
Chap. 11 for details on the use of MOS). This mesoscale-model-based MOS was
relatively straightforward to develop, provided forecasts at multiple sites at several
outdoor venues, including three along the men’s downhill, and proved extremely
beneficial for the forecast effort during the Olympics (Hart et al. 2004).
Similar gap-bridging activities have occurred during other Olympic Games.
Sydney 2000 and Beijing 2008 included World Weather Research Program Forecast
Demonstration Projects in which researchers shared a work area with forecasters to
test leading-edge nowcasting tools (e.g., Keenan et al. 2003; May et al. 2004; Ebert
et al. 2004; Wilson et al. 2004; Joe et al. 2010; Mailhot et al. 2010). Testing and
evaluation of a multimodel superensemble and high-resolution limited-area models
occurred during Torino 2006 (Cane and Milelli 2006; Stauffer et al. 2007), with the
superensemble used subsequently for operational weather prediction over the Pied-
mont region of Italy (Cane and Milelli 2008). During the 2010 Vancouver Games, a
team of researchers participated in the Science and Nowcasting of Olympic Weather
for Vancouver 2010 (SNOW-V10) in order to improve the understanding of and
ability to forecast and nowcast low cloud, visibility, precipitation (amount and
type), and winds in complex terrain (Joe et al. 2010). The project involved the
deployment of an enhanced network of instruments at specific venues, as well as
the production of prototype forecast products. Researchers collaborated extensively
with forecasters on the instrument siting and assisted in pre-Games training.
For the Vancouver 2010 Olympic Winter Games, a comprehensive training
strategy was also devised based in part on the survey findings described in
Sect. 12.3 (Snyder et al. 2006). One aspect of this training involved creating a
mountain weather residency course in cooperation with the University Corporation
for Atmospheric Research Cooperative Program for Operational Meteorology,
Education and Training (COMET) program. Subject matter experts, many of
whom have contributed to chapters in this compilation (e.g., Chaps. 2, 6, and 7),
delivered lectures on the latest knowledge in mountain meteorology and applied this
knowledge through case studies and forecast labs. As a testament to the quality of
instruction and the focus on operationally relevant research findings, course ratings
were some of the highest ever given to a COMET course. In another effort to
facilitate interaction between the operational and research teams, a meteorologist
was assigned the dual-role of venue forecaster and applied research meteorologist
for Olympic forecast product development. Buy-in and commitment were not an
issue during the Games but adding a person who could speak the others’ language
solidified the gap-bridging effort
12.6.2.2 MAP D-PHASE
The Mesoscale Alpine Programme Demonstration of Probabilistic Hydrological

and Atmospheric Simulation of Flood Events (MAP D-PHASE) is a project to
demonstrate and evaluate potential improvements in the operational forecasting of
flood events in the European Alps (Zappa et al. 2008; Rotach et al. 2009a, b).
MAP D-PHASE seeks to demonstrate forecast advances derived from the Mesoscale
Alpine Programme (MAP), which involves substantial collaboration between the
research and operational sectors (Volkert and Gutermann 2007). Participants include
17 countries, 18 operational centers, and 7 research institutions (Arpagaus et al.
2009). MAP D-PHASE has many goals including: (1) assessing high-resolution
deterministic and probabilistic hydrological and atmospheric modeling systems, (2)
delivering advanced flood warnings and background information for end users, (3)
developing nowcasting tools, (4) improving radar observations of precipitation over
complex terrain, and (5) improving decision making by civil protection authorities.
As such, MAP D-PHASE involves not only bridging the gap between research and
operations, but also between meteorologists and policy makers, end-users, and other
scientific disciplines.
Integral to MAP D-PHASE is the testing and evaluation of an end-to-end forecast
system based on new methods of assessing forecast uncertainty, specifically,
ensembles of hydrologic forecasts created using ensembles of weather forecasts.
The hydrological forecasts are produced for 43 catchments, and warnings are issued
if a deterministic forecast or a third of the ensemble members exceed one of three
criteria based on flood return periods. MAP D-PHASE also incorporates a Web-
based visualization platform containing all MAP D-PHASE graphical information
(e.g., nowcasting products, warning maps, validation products). Although this
common framework requires compromise by the participants, all warnings are based
on the same thresholds and procedures, allowing different regions and models to be
fairly and uniformly intercompared.
Although the results of MAP D-PHASE are only beginning to be revealed, the
effort required to integrate data, model ensembles, and warnings into a common
framework suggests an immense level of cooperation and planning between re-
search, operations, and end users. Rotach et al. (2009a) suggest that the use of
common formats, warning levels, and routines amongst different forecast models
is essential for program success. MAP D-PHASE was designed from the beginning
to address a research and operational challenge, ensuring that both research and
operational centers would have vested interests in seeing success. It sought the input
and participation of end users. In short, MAP D-PHASE is perhaps the best single
example from the mountain meteorology community of a large-scale collaboration
between research and operations.
12.6.3 Stimulating and Funding Grass-Roots Efforts:

CSTAR and COMET
Research requires resources, even efforts that begin at the grassroots. The NWS
has developed two successful programs to support these grassroots efforts: (1) the
Collaborative Science, Technology, and Applied Research (CSTAR) program and
(2) the COMET program.
The CSTAR program provides funding to university scientists to support highly

collaborative applied research activities with the NWS. CSTAR partnerships are col-
laborative efforts requiring a buy-in from researchers and forecasters, consequently
providing a foundation for the ongoing infusion of science and technology into the
forecast office. The importance of CSTAR funding is clearly evident in the examples
noted earlier as it has helped support collaborative activities between the University
of Washington and the NWS Forecast Office in Seattle (Sect. 12.6.1) and between
the University of Utah, NWS, and private forecasters for the 2002 Olympic Winter
Games (Sect. 12.6.2.1). The COMET program addresses meteorological education
and training, including mountain-related topics, through distance learning, residence
classes, and an outreach program that facilitates the transfer of research results to
operations and provides funding for the academic and operational communities to
participate in collaborative research. A similar European effort for education and
training project is called Eumetcal (http://www.eumetcal.org).
CSTAR and COMET provide the funding needed to bring together researchers
and forecasters who otherwise would not typically interact. These two programs
force the two groups into a middle ground where both an operational and research
focus is achieved, accomplishing the mandates of both groups. Sometimes, it
requires the researchers, forecasters, or both to go beyond their comfort level. One
example is the COMET partnership between the NWS Forecast Office in Grand
Junction and researchers from the Desert Research Institute Storm Peak Laboratory
and Colorado State University. These three groups worked from 2002 to 2006 to
investigate orographic precipitation over the Park Range of north-central Colorado
(Wetzel et al. 2004). To minimize any potential reluctance by the forecasters,
the researchers visited the Grand Junction NWS forecast office to describe their
research. The session included a short seminar followed by one-on-one interaction
between the forecasters and the researchers, which enhanced the personal con-
nection between the two groups. The operational staff also demonstrated for the
researchers the forecast system and methodology. Some of the forecasters then
visited the Storm Peak Laboratory, enabling direct communication between the
groups in a laboratory setting. These interactions increased buy-in and commitment
of both groups.
Because of the communication and team-building efforts of the researchers, the
forecasters were committed to success during and following the field study. One of
the project objectives was to increase meteorological understanding of the physical
controls on precipitation over this relatively data-sparse mountainous region in a
way that would directly benefit operational forecasting. Because of the availability
of supplemental data during the field study, the forecasters gained invaluable
insights into the local orographic forcing over the Park Range, knowledge that
they have expanded upon after the project conclusion. For example, the forecasters
developed a better understanding of the predominance of heavy snowfall during
moist, post-frontal flow and the sensitivity of quantitative snowfall forecasts to the
snow-to-liquid ratio. The forecasters learned that snow-to-liquid ratios of 30:1 are
not uncommon in this region and that an underestimation of this ratio, as happened
in the past, has a significant negative impact on the prediction of snowfall amount
and winter storm potential. The impact of improved understanding was evident
during the 2009 winter season when 24 winter storms in the Park Range region
were forecast with a 100% probability of detection at a lead time of 32 h.
By 2005, over 250 COMET-funded research projects had been completed, a
result of collaborations between over 70 universities and over 90 NWS forecast
offices (Waldstreicher 2005). Importantly, the research funded by COMET has led
to demonstrable improvements in forecasts, as Waldstreicher (2005) has shown.
For example, NWS Eastern Region offices with COMET projects aimed at better
understanding severe thunderstorms or tornadoes improved twice as much in
probability of detection for severe thunderstorm and tornado warnings as the region
as a whole. The rate of improvement was also strong for lead times for severe
thunderstorms and tornadoes (eight times), lead times for winter storm warnings
(two times), and lead times for flash flood warnings (four times). Furthermore,
an office with a long-term history of collaborative research, Raleigh, NC, also
demonstrated remarkable improvement in their metrics compared to the other offices
in the region. Consequently, Waldstreicher’s (2005) results provide quantitative
evidence showing the positive impact of research–operations collaborations on
forecast quality.
12.6.4 Collocation: The Research Support Desk
An example of the value of collocation for gap bridging is the so-called Research
Support Desk (RSD), which was implemented in some MSC forecast offices to
increase real-time interaction between forecasters and researchers (Sills and Taylor
2008). The RSD involves the collocation of a researcher with operational forecasters
in the weather center. Although the RSD is staffed only during busy seasons,
it has proven to be an effective knowledge transfer mechanism. Education and
training about new techniques and technologies is passed from the researcher to
the forecasters through direct real-time interaction and formal briefings. In turn, the
researcher is able to also evaluate experimental products and identify science needs
in an operational environment.
The first real challenge for this initiative was getting permission to have the RSD
located in the operational environment. Forecasters were concerned about increased
demands on their time, particularly during severe weather (Sills 2005). Once
established, however, the overwhelming majority of forecasters were comfortable
with the researcher in operations and 72% of forecasters responding to a survey
believed the RSD enhanced the learning environment. This experience exemplifies
some of the other key ingredients to bridging the gap (advocates for collaboration,
stimulation at the grassroots level). With motivated individuals and the political will,
the mountain meteorology community could benefit by applying this model in other
weather centers.
As part of the SNOW-V10 project, a RSD was put in place in Vancouver for
the 2010 Olympic Games. This desk fostered communication between researchers
and forecasters and was especially helpful in exposing forecasters to a vast array
of new data. To facilitate collaboration between a larger pool of researchers and
forecasters, daily web-based briefings were also held over a 2-month period before
and during the Games.
12.6.5 Collocation: Forecasting and Research Teams

in Mountain Meteorology
Examples of collocated operational and research groups that concentrate on moun-

tain weather applications exist in Europe (i.e., MeteoSwiss where the radar applica-
tion and research group is collocated with the forecasters in Locarno) and Canada
(the Pacific and Yukon Region National Lab for Coastal and Mountain Meteorol-
ogy), but no major center for mountain weather research and operations exists in
the United States. This is unfortunate, because collocation of motivated individuals
and groups has enabled dramatic advances in knowledge and forecasting in other
meteorological subdisciplines. Perhaps the earliest and best-known example is the
Bergen School of Meteorology (e.g., Friedman 1989, 1999). To address the needs
of farmers and fishermen in Norway, Vilhelm Bjerknes developed a system and
received funding to collect and analyze mesoscale observations of weather systems.
Study of these weather systems led to the formulation of the Norwegian cyclone
model. Bergen School members were trained physicists and mathematicians; their
emphasis was on understanding the relevant weather processes to produce the best
forecast. As such, theirs is arguably the first scientific effort to bridge the gap
between research and operations in atmospheric science.
In the 1950s, a dominant center for bridging the gap between research and
operations emerged in Chicago (e.g., Allen 2001). Gordon Dunn, an employee
of the Weather Bureau (predecessor of the NWS), convinced the Weather Bureau
to locate the Chicago Forecast Office at the Department of Meteorology at the
University of Chicago (Burpee 1989, pp. 576, 581–582). Dunn, concerned about
the gap between researchers and forecasters, arranged daily map discussions with
both groups participating. In such an environment, understanding of the jet stream
improved, field research programs on tropical and midlatitude convection began,
and convective-storm and tornado expert Ted Fujita prospered.
In 1959, Dunn also contributed to the collocation of the National Hurricane
Center (now the Tropical Prediction Center) and the National Hurricane Research
Project (later the National Hurricane Research Laboratory and then the Hurricane
Research Division of the NOAA/Atlantic Oceanographic and Meteorology Labora-
tory) (Burpee 1989, p. 580). The Joint Hurricane Testbed (Knabb et al. 2005) was
the eventual result of this interaction. One of the benefits of the cooperation and
collocation was the development of statistical hurricane track prediction models,
assimilation of data from reconnaissance aircraft, and operational targeting of
additional observations.
Collocation does not, however, guarantee collaboration. Doswell (2007) de-

scribes a situation at the National Severe Storms Forecasting Center (predecessor
to the Storm Prediction Center) where the initial collaboration between researchers
and forecasters developed into a schism where the researchers and forecasters
were placed on separate floors in the same building because of some of the
researchers’ disdain of having to do shift work. The schism in the severe weather
community would later be repaired by the collocation of the National Severe
Storms Laboratory and Storm Prediction Center in Norman, Oklahoma, in the late
1990s. This collocation has subsequently led to a flourish of collaborative research
and forecasting endeavors (e.g., Kain et al. 2003a, b, 2006, 2008) including the
development and application of convection-allowing models in operations (e.g.,
Kain et al. 2006, 2008), efforts to improve the interpretation of forecast-model
soundings (e.g., Baldwin et al. 2002), research into the use of ensembles for
convective-storm forecasting (e.g., Bright et al. 2004; Weiss et al. 2006, 2007), and
other research projects on convective storms (e.g., Craven et al. 2002; Banacos and
Schultz 2005; Coniglio et al. 2007). Because of these largely grassroots successes,
other individuals, groups, and companies—both domestic and international—have
collaborated with these two groups in Norman, now renamed the Hazardous
Weather Testbed, showing the power that an initial interaction can have for growing
the research–operations connection. Indeed, a small initial success can grow into a
sustained and much larger success, given the right ingredients.
12.7 Fostering Improved Future Coordination
The successful gap-bridging examples described above raises hope for the future,
but individuals, agencies, and organizations must increase their efforts if the moun-
tain weather community is to meet its full potential. Mass (2006) raises concerns
about a lack of coordination between the research and operational communities
in the United States, and he provides a substantial list of recommendations for
improvement, which we expand on here for the mountain weather community. In
particular, there are four critical areas in which we should focus our attention and
resources:
1. Creating tomorrow’s advocates. As noted earlier, the most important factor to
ensure the overall success of the mountain-meteorology enterprise are advocates
who work to bridge the gap between research and operations (and ultimately
decision makers; Morss et al. 2005). As such, greater emphasis must be placed
on educating, mentoring, and grooming young scientists and forecasters to bridge
the gap between mathematical theory and scientific forecasting as they move
through their careers (e.g., Doswell 1986). On the research side, universities
should provide curricula and opportunities for tomorrow’s scientific leaders
(i.e., current graduate students) to participate in scientific forecasting and gain
exposure to how theory is used in practice (as argued by Ramage 1978).
It should not be acceptable for a student to earn an advanced degree without

exposure to weather analysis and forecasting. Such exposure can be achieved
through courses in synoptic–dynamic meteorology and forecasting, internships
at forecasting agencies, and collaborative research projects leading to advanced
degrees. Approaches that challenge the student to test, apply, and refine their
research in an operational environment should be strongly encouraged. On
the operational side, forecasting agencies must provide additional time and
incentives for forecasters to participate in research by creating more opportunities
for young forecasters to return to graduate school to pursue graduate degrees or
regular release time to participate in research projects and personal training. In
either case, it is essential that the forecaster work with research meteorologists
and conduct substantive research to gain exposure to the scientific method.
A forecaster with a doctoral education or substantive prior research experience
could serve as the scientific mentor in these efforts, but the participation of
research scientists from universities and government labs should be encouraged.
2. Prioritizing and providing support for the transfer of field-program advances
into operations. Field research in mountain meteorology continues to provide
valuable data for scientific analyses and publications. MAP, for example, has
spawned more than 220 scientific publications (Volkert and Gutermann 2007).
For the operational meteorological community, however, the challenge has
been transferring this wealth of knowledge to the forecast environment so that
theoretical concepts can be applied to forecasting in complex terrain (Snyder
et al. 2006). Historically, many mountain meteorology field programs have
contained no component or clear mechanism for transferring field-program
advances into operations (although many field-program proposals state that
forecast improvements are a likely broader impact of the research). Frequently,
operational meteorologists provide support for field-program operations, but are
less involved in the early field-program planning or subsequent analysis.
The MAP D-PHASE program, however, represents a recent major under-
taking to transfer knowledge and achievements from MAP into the forecast
and decision-making process. Similarly, the Hydrometeorological Testbed in
the American Fork River Basin of northern California, a collaboration between
the Earth Systems Research Laboratory and California NWS offices (Ralph
et al. 2005; http://hmt.noaa.gov) explicitly integrates operational, research, and
decision-making activities. The strengths and weaknesses of these programs
should be identified and used to improve the impact of future field programs
on operations. Within the United States, greater support from NOAA to mine the
successes of National Science Foundation–sponsored field programs would also
help.
Outside of these major undertakings, Smith et al. (2001) created a list of 14
more ordinary, but quite specific, items to increase the transfer of knowledge
between forecasters and researchers. This list includes such items as “(ii) ask
forecasters to flag interesting/important observed cases, noting the nature of the
event to alert researchers to cases for possible study; (iii) ask researchers to
flag interesting new results in terms accessible to forecasters;” “(v) commission
researchers to write articles on phenomena and issues in language intelligible

to forecasters; (vi) encourage the World Meteorological Organization (WMO)
and National Weather Services to fund visits of researchers to forecast offices
for immersion in the culture/science of forecasting, not procedures; (vii) iden-
tify forecast offices as possible sites for study leave.” These and the other
recommendations do not require significant amounts of funding, if any, yet,
with enthusiastic advocates, could lead to greater and more positive interactions
between researchers and forecasters.
3. Leveraging the power of proximity through collocation. Despite the tremendous
impact that the collocation of motivated individuals and groups has had on
knowledge and forecasting advances in severe-convective weather and tropical
storms (e.g., Hazardous Weather Testbed, Joint Hurricane Testbed), no major
collocated center for mountain weather research and forecasting exists in the
United States. This situation contrasts with many countries in Europe, which per-
haps due to smaller size and less diverse national forecast challenges, often have
weather services with stronger cooperation between the operational and research
communities. In North America, however, there are some strong collaborations
fostered by geographic proximity or shear will, such as collaborations between
the University of Washington and NWS Forecast Office in Seattle (Sect. 12.6a);
University of Utah, NWS Forecast Office in Salt Lake City, and NWS Western
Region Headquarters (Sect. 12.6b; Horel et al. 2002a, b); and the Earth Systems
Research Laboratory and California NWS forecast offices via the Western US
Hydrometeorological Testbed (Sect. 12.7; Ralph et al. 2005). Given the growing
financial losses imposed by flooding and wildfire events in the western United
States (Ross and Lott 2003), the time is ripe to evaluate the need for an integrated
research and forecast center concentrating on hydrometeorological and fire–
atmosphere prediction in areas of complex terrain.
4. Establishing incentives for collaboration. Although forced collaborations are
rarely effective, there are a number of ways to create opportunities for increased
collaboration. These include providing support for operational forecasters
to obtain continuing education or an advanced degree, increasing funding
(both number and size of awards) for successful collaborative programs like
CSTAR and COMET, providing salary for University faculty to take sabbaticals
in forecast offices, and providing rewards (e.g., promotion, pay, awards) to
individuals who engage or encourage successful collaboration. Such incentives
will help create the culture of collaboration needed for the mountain weather
enterprise to reach its full potential.
Acknowledgments We thank the participants in the 2008 AMS/COMET/MSC Mountain Weather

Workshop: Bridging the Gap between Research and Forecasting for 4 days of lectures and
discussion that stimulated this chapter, as well as our editors and chapter coauthors for their
contributions to this book. We also thank Katja Friedrich, John Lewis, Andrea Rossa, Mathias
Rotach, Andrew Russell, David Stensrud, Hans Volkert, and three anonymous reviewers for their
contributions to the manuscript. Participants in the panel discussion “Enhancing the Connectivity
between Research and Applications for the Benefit of Society” at the 2008 AMS Annual Meeting
also provided thoughts and ideas that influenced parts of this chapter. Contributing author
Steenburgh acknowledges the support of the National Science Foundation and National Weather
Service. Contributing author Schultz acknowledges the support of Vaisala Oyj. Any opinions,
findings, and conclusions or recommendations expressed in this material are those of the authors
and do not necessarily reflect the views of the National Science Foundation, National Weather
Service, or Vaisala Oyj.
References
Allen, D. R., 2001: The genesis of meteorology at the University of Chicago. Bull. Amer. Meteor.
Soc., 82, 1905–1909.
Arpagaus, M., and Coauthors, 2009: MAP D-PHASE: Demonstrating forecast capabilities for
flood events in the Alpine region. Veröffentlichung der MeteoSchweiz, 78, 75 pp. [Available
online at http://www.meteoschweiz.admin.ch/web/de/forschung/publikationen/meteoschweiz
publikationen/veroeffentlichungen.html.]
Baldwin, M. E., J. S. Kain, and M. P. Kay, 2002: Properties of the convection scheme in NCEP’s
Eta Model that affect forecast sounding interpretation. Wea. Forecasting, 17, 1063–1079.
Banacos, P. C., and D. M. Schultz, 2005: The use of moisture flux convergence in forecasting
convective initiation: Historical and operational perspectives. Wea. Forecasting, 20, 351–366.
Bergeron, T., 1959: Weather forecasting: Methods in scientific weather analysis: An outline in the
history of ideas and hints at a program. The Atmosphere and the Sea in Motion, B. Bolin, Ed.,
The Rockefeller Institute Press, 440–474.
Besser, B., 2008: Development of meteorology and geophysics at the University of Graz.
Proceedings of the First European History of Physics Conference, Graz, Austria, Sep 18–21
2006, P. M. Schuster and D. Weaire, Eds., Living Edition Publishers, 159–170. [Available
online at http://www.livingedition.at/en/titles/science/proceedings.]
Board on Atmospheric Sciences and Climate, 2000: From Research to Operations in Weather
Satellites and Numerical Weather Prediction: Crossing the Valley of Death. National Academy
Press, 96 pp. [Available online at http://www.nap.edu/catalog.php?record id=9948.]
Bond, N. A. and Coauthors, 1997: The Coastal Observation and Simulation with Topography
(COAST) Experiment. Bull. Amer. Meteor. Soc., 78, 1941–1955.
Bosart, L. F., 2003: Whither the weather analysis and forecasting process? Wea. Forecasting, 18,
520–529.
Bright, D. R., S. J. Weiss, J. J. Levit, M. S. Wandishin, J. S. Kain, and D. J. Stensrud, 2004:
Evaluation of short-range ensemble forecasts during the 2003 SPC/NSSL Spring Program.
Preprints, 22nd Conference on Severe Local Storms, Hyannis, MA, Amer. Meteor. Soc., CD-
ROM, P15.5.
Bundgaard, R. C., 1979: Sverre Petterssen, weather forecaster. Bull. Amer. Meteor. Soc., 60,
182–195.
Burpee, R. W., 1989: Gordon E. Dunn: Preeminent forecaster of midlatitude storms and tropical
cyclones. Wea. Forecasting, 4, 573–584.
Cane, D., and M. Milelli, 2006: Weather forecasts obtained with a multimodel superensemble
technique in a complex orography region. Met. Zeitschrift, 15, 207–214.
Cane, D., and M. Milelli, 2008: Comparison of COSMO models and multimodel superensemble
outputs in Piemonte. COSMO Newsletter No. 9, 69–79. [Available online at http://www.
cosmo-model.org/content/model/documentation/newsLetters/newsLetter09/cnl9-13.pdf.]
Coniglio, M. C., H. E. Brooks, S. J. Weiss, and S. F. Corfidi, 2007: Forecasting the maintenance of
quasi-linear mesoscale convective systems. Wea. Forecasting, 22, 556–570.
Craven, J. P., R. E. Jewell, and H. E. Brooks, 2002: Comparison between observed convective
cloud-base heights and lifting condensation level for two different lifted parcels. Wea.
Doswell, C. A. III, 1986: The human element in weather forecasting. Nat. Wea. Dig., 11 (2), 6–17.
Doswell, C. A. III, 2004: Weather forecasting by humans—Heuristics and decision making. Wea.
Doswell, C. A. III, 2007: Historical overview of severe convective storms research. Electr. J. Severe
Storms Meteor., 2(1), 1–25.
Doswell, C. A. III, and H. E. Brooks, 1998: Budget cutting and the value of weather services. Wea.
Doswell, C. A. III, L. R. Lemon, and R. A. Maddox, 1981: Forecaster training—A review and
analysis. Bull. Amer. Meteor. Soc., 62, 983–988.
Ebert, E., L. J. Wilson, B. G. Brown, P. Nurmi, H. E. Brooks, J. Bally, and M. Jaeneke, 2004:
Verification of nowcasts from the WWRP Sydney 2000 Forecast Demonstration Project. Wea.
Erkkilä, T., 2007: About the nature of the forecaster profession and the human contribution to very
short range forecasts. The European Forecaster, 14, 6–11. [Available online at http://www.
euroforecaster.org/latenews/newsletter.html.]
Friedman, R. M., 1989: Appropriating the Weather: Vilhelm Bjerknes and the Construction of a
Modern Meteorology. Cornell Univ. Press, 251 pp.
Friedman, R. M., 1999: Constituting the polar front, 1919–1920. The Life Cycles of Extratropical
Cyclones. M. A. Shapiro and S. Grønås, Eds., Amer. Meteor. Soc., 29–40.
Harper, K. C., 2008: Weather by the Numbers: The Genesis of Modern Meteorology. MIT Press,
308 pp.
Harper, K. C., 2009: Will meteorologists lose their jobs? NWP and automation fears in the Fifties.
Presidential History Symposium, 89th American Meteorological Society Annual Meeting, P2.3.
[Available online at http://ams.confex.com/ams/89annual/techprogram/session 22297.htm.]
Harper, K., L. W. Uccellini, E. Kalnay, K. Carey, and L. Morone, 2007: 50th anniversary of
operational numerical weather prediction. Bull. Amer. Meteor. Soc., 88, 639–650.
Hart, K. A., W. J. Steenburgh, D. J. Onton, and A. J. Siffert, 2004: An evaluation of mesoscale-
model-based Model Output Statistics (MOS) during the 2002 Olympic and Paralympic Winter
Games. Wea. Forecasting, 19, 200–218.
Horel, J., T. Potter, L. Dunn, W. J. Steenburgh, M. Eubank, M. Splitt, and D. J. Onton, 2002a:
Weather support for the 2002 Winter Olympic and Paralympic Games. Bull. Amer. Meteor.
Soc., 83, 227–240.
Horel, J., M. Splitt, L. Dunn, J. Pechmann, B. White, C. Ciliberti, S. Lazarus, J. Slemmer, D. Zaff,
and J. Burks, 2002b: Mesowest: Cooperative mesonets in the western United States. Bull. Amer.
Meteor. Soc., 83, 211–225.
Joe, P., and Coauthors, 2010: Weather services, science advances, and the Vancouver 2010 Olympic
and Paralympic Winter Games. Bull. Amer. Meteor. Soc., 91, 31–36.
Kain, J. S., M. E. Baldwin, P. R. Janish, S. J. Weiss, M. P. Kay, and G. W. Carbin, 2003a: Subjective
verification of numerical models as a component of a broader interaction between research and
operations. Wea. Forecasting, 18, 847–860.
Kain, J. S., P. R. Janish, S. J. Weiss, M. E. Baldwin, R. S. Schneider, and H. E. Brooks, 2003b:
Collaboration between forecasters and research scientists at the NSSL and SPC: The Spring
Program. Bull. Amer. Meteor. Soc., 84, 1797–1806.
Kain, J. S., S. J. Weiss, J. J. Levit, M. E. Baldwin, and D. R. Bright, 2006: Examination of
convection-allowing configurations of the WRF model for the prediction of severe convective
weather: The SPC/NSSL Spring Program 2004. Wea. Forecasting, 21, 167–181.
Kain, J. S., S. J. Weiss, D. R. Bright, M. E. Baldwin, J. J. Levit, G. W. Carbin, C. S. Schwartz,
M. L. Weisman, K. K. Droegemeier, D. B. Weber, and K. W. Thomas, 2008: Some practical
considerations regarding horizontal resolution in the first generation of operational convection-
allowing NWP. Wea. Forecasting, 23, 931–952.
Keenan, T., and Coauthors, 2003: The Sydney 2000 World Weather Research Programme Forecast
Demonstration Project: Overview and current status. Bull. Amer. Meteor. Soc., 84, 1041–1054.
Knabb, R. D., J. G. Jiing, C. W. Landsea, and W. R. Seguin, 2005: The Joint Hurricane Testbed
(JHT): Progress and future plans. Preprints, Ninth Symposium on Integrated Observing and
Assimilation Systems for the Atmosphere, Oceans, and Land Surface, San Diego, CA, Amer.
Meteor. Soc., 2.2. [Available online at http://ams.confex.com/ams/Annual2005/techprogram/
paper 84938.htm.]
Lewis, J. M., 1995: WAVES forecasters in World War II (with a brief survey of other women
meteorologists in World War II). Bull. Amer. Meteor. Soc., 76, 2187–2202.
Mailhot, J., and Coauthors, 2010: Environment Canada’s experimental numerical weather predic-
tion systems for the Vancouver 2010 Olympic and Paralympic Games. Bull. Amer. Meteor. Soc.,
in press.
Mass, C., 2006: The uncoordinated giant: Why U.S. weather research and prediction are not
achieving their potential. Bull. Amer. Meteor. Soc., 87, 573–584.
Mass, C. F., and Coauthors, 2003: Regional environmental prediction over the Pacific Northwest.
May, P. T., and Coauthors, 2004: The Sydney 2000 Olympic Games Forecast Demonstration
Project: Forecasting, observing network infrastructure, and data processing issues. Wea.
McCarthy, P. J., D. Ball, and W. Purcell, 2007: Project Phoenix—Optimizing the machine-person
mix in high-impact weather forecasting. Preprints, 22nd Conference on Weather Analysis and
Forecasting/18th Conference on Numerical Weather Prediction, Park City, UT, Amer. Meteor.
Soc., P6A.5. [Available online at http://ams.confex.com/ams/22WAF18NWP/techprogram/
paper 122657.htm.]
Morss, R. E., and F. M. Ralph, 2007: Use of information by National Weather Service forecasters
and emergency managers during CALJET and PACJET-2001. Wea. Forecasting, 22, 539–555.
Morss, R. E., O. V. Wilhelmi, M. W. Downton, and E. Gruntfest, 2005: Flood risk, uncertainty,
and scientific information for decision making: Lessons from an interdisciplinary project. Bull.
Amer. Meteor. Soc., 86, 1593–1601.
Palmer, T., 2008: Edward Norton Lorenz. Physics Today, 61, 81–82.
Petterssen, S., 2001: Weathering the Storm: Sverre Petterssen, the D-Day Forecast, and the Rise of
Modern Meteorology, J. R. Fleming, Ed., Amer. Meteor. Soc., 329 pp.
Pliske, R. M., B. Crandall, and G. Klein, 2004: Competence in weather forecasting. Psychological
Investigations of Competence in Decision Making, K. Smith, J. Shanteau, and P. Johnson, Eds.,
Cambridge University Press, 40–68.
Ralph, F. M., and Coauthors, 2005: Improving short-term (0–48 h) cool-season quantitative
precipitation forecasting: Recommendations from a USWRP workshop. Bull. Amer. Meteor.
Soc., 86, 1619–1632.
Ramage, C. S., 1978: Further outlook—Hazy. Bull. Amer. Meteor. Soc., 59, 18–21.
Ramage, C. S., 1993: Forecasting in meteorology. Bull. Amer. Meteor. Soc., 74, 1863–1871.
Rauhala, J., and D. M. Schultz, 2009: Severe thunderstorm and tornado warnings in Europe. Atmos.
Res., 93, 369–380,
Reed, R. J., 2003: A short account of my education, career choice, and research motivation. A Half
Century of Progress in Meteorology: A Tribute to Richard Reed. R. H. Johnson and R. A. Houze
Jr., Eds., Amer. Meteor. Soc., 1–12.
Roebber, P. J., 2005: Bridging the gap between theory and applications: An inquiry into
atmospheric science teaching. Bull. Amer. Meteor. Soc., 86, 507–517.
Roebber, P. J., D. M. Schultz, and R. Romero, 2002: Synoptic regulation of the 3 May 1999 tornado
outbreak. Wea. Forecasting, 17, 399–429.
Roebber, P. J., D. M. Schultz, B. A. Colle, and D. J. Stensrud, 2004: Toward improved prediction:
High-resolution and ensemble modeling systems in operations. Wea. Forecasting, 19, 936–949.
Ross, T., and N. Lott, 2003: A climatology of 1980–2003 extreme weather and climate events.
National Climatic Data Center Technical Report No. 2003-01. [Available online at http://ols.
nndc.noaa.gov/plolstore/plsql/olstore.prodspecific?prodnum=C00580-PUB-A0001.]
Rossby, C.-G., 1934: Comments on meteorological research. J. Aeronaut. Sci., 1, 32–34.
Rotach, M. W., and Coauthors, 2009a: MAP D-PHASE: Real-time demonstration of weather
forecast quality in the Alpine region. Bull. Amer. Meteor. Soc., 90, 1321–1336.
Rotach, M. W., and Coauthors, 2009b: Supplement to MAP D-PHASE: Real-time demonstration of
weather forecast quality in the Alpine region: Additional applications of the D-PHASE datasets.
Bull. Amer. Meteor. Soc., 90, S28–S32.
Sanders, F., 2008: A career with fronts: Real ones and bogus ones. Synoptic–Dynamic Meteo-
rology and Weather Analysis and Forecasting. A Tribute to Fred Sanders, L. F. Bosart and
H. B. Bluestein, Eds., Amer. Meteor. Soc., 421–422.
Sills, D. M. L., 2005: The Research Support Desk Initiative at the Ontario Storm Prediction Centre.
Meteorological Research Branch Technical Note #-2005-001, Environment Canada. 30 pp.
Sills, D. M. L., 2009: On the MSC forecasters forums and the future role of the human forecaster.
Sills, D. M. L., and N. M. Taylor, 2008: The Research Support Desk (RSD) initiative at
Environment Canada: Linking severe weather researchers and forecasters in a real-time
operational setting. Preprints, 24th AMS Conference on Severe Local Storms, Savannah, GA,
Amer. Meteor. Soc., Paper 9A.1. [Available online at http://ams.confex.com/ams/pdfpapers/
142033.pdf.]
Smith, R. K., G. Garden, J. Molinari, and R. K. Morton, 2001: Proceedings of an International
Workshop on the Dynamics and Forecasting of Tropical Weather Systems. Bull. Amer. Meteor.
Soc., 82, 2825–2829.
Snellman, L. W., 1977: Operational forecasting using automated guidance. Bull. Amer. Meteor.
Soc., 58, 1036–1044.
Snyder, B. J., C. Doyle, D. A. Wesley, J. D. Cummine, and M. Meyers, 2006: The first
MSC/COMET mountain weather course. Preprints, 12th Conf. on Mountain Meteorology,
Santa Fe, NM, Amer. Meteor. Soc., P16.2.
Snyder, B. J., and M. Loney, 2007. Survey of Forecaster Training, 2006 Results. Meteorological
Service of Canada. Unpublished.
Stanitski, D. M., and D. J. Charlevoix, 2008: AMS membership survey results: Who are the student
members of the AMS? Bull. Amer. Meteor. Soc., 89, 892–895.
Stauffer, D. R., G. K. Hunter, A. Deng, J. R. Zielonka, K. Tinklepaugh, P. Hayes, and C. Kiley,
2007: On the role of atmospheric data assimilation and model resolution on model forecast
accuracy for the Torino Winter Olympics. Preprints, 22nd Conference on Weather Analysis and
Forecasting/18th Conference on Numerical Weather Prediction, Park City, UT, Amer. Meteor.
Soc., P11A.6. [Available online at http://ams.confex.com/ams/22WAF18NWP/techprogram/
paper 124791.htm.]
Steenburgh, W. J., and C. F. Mass, 1996: Interaction of an intense extratropical cyclone with coastal
orography. Mon. Wea. Rev., 124, 1329–1352.
Stoelinga, M. T., and Coauthors, 2003: Improvement of Microphysical Parameterization through
Observational Verification Experiment. Bull. Amer. Meteor. Soc., 84, 1807–1826.
Stuart, N. A., P. S. Market, B. Telfeyan, G. M. Lackmann, K. Carey, H. E. Brooks, D. Nietfeld,
B. C. Motta, and K. Reeves, 2006: The future of humans in an increasingly automated forecast
process. Bull. Amer. Meteor. Soc., 87, 1497–1502.
Stuart, N. A., D. M. Schultz, and G. Klein, 2007: Maintaining the role of humans in the forecast
process: Analyzing the psyche of expert forecasters. Bull. Amer. Meteor. Soc., 88, 1893–1898.
Volkert, H., and T. Gutermann, 2007: Inter-domain cooperation for mesoscale atmospheric
laboratories: The Mesoscale Alpine Program as a rich study case. Quart. J. Roy. Meteor. Soc.,
133, 949–967.
Waldstreicher, J. S., 2005: Assessing the impact of collaborative research projects on NWS warning
performance. Bull. Amer. Meteor. Soc., 86, 193–203.
Washington State Military Department, 2007: Windstorm Response after Action Report: A
Statewide Report to the Governor. 77 pp. [Available online at http://www.emd.wa.gov/
publications/documents/FINAL AAR 040407.pdf.]
Weiss, S. J., D. R. Bright, J. S. Kain, J. J. Levit, M. E. Pyle, Z. I. Janjic, B. S. Ferrier, and J. Du,
2006: Complementary use of short-range ensemble and 4.5 km WRF-NMM model guidance
for severe weather forecasting at the Storm Prediction Center. Preprints, 23rd Conference on
Severe Local Storms, St. Louis, MO, Amer. Meteor. Soc., CD-ROM 8.5.
Weiss, S. J., J. S. Kain, D. R. Bright, J. J. Levit, G. W. Carbin, M. E. Pyle, Z. I. Janjic, B. S.
Ferrier, J. Du, M. L. Weisman, and M. Xue, 2007: The NOAA Hazardous Weather Testbed:
Collaborative testing of ensemble and convection-allowing WRF models and subsequent
transfer to operations at the Storm Prediction Center. Preprints, 22nd Conference on Weather
Analysis and Forecasting/18th Conference on Numerical Weather Prediction, Park City, UT,
Amer. Meteor. Soc., CD-ROM, 6B.4.
Wetzel, M., and Coauthors, 2004: Mesoscale snowfall prediction and verification in mountainous
terrain. Wea. Forecasting, 19, 806–828.
Wilson, J. W., E. Ebert, T. Saxen, C. Pierce, M. Sleigh, A. Seed, R. Roberts and C. Mueller, 2004:
Sydney 2000 Forecast Demonstration Project: Convective storm nowcasting. Wea. Forecasting,
19, 131–150.
Zappa, M., and Coauthors, 2008: MAP D-PHASE: Real-time demonstration of hydrological
ensemble prediction systems. Atmos. Sci. Lett., 9, 80–87.
Author Index
A Beard, K.V., 350

Acevedo, O.C., 76 Bélair, S.P., 595
Afanasyev, Y.D., 167, 191 Benjamin, T.B., 186
Albright, M.D., 162, 166 Benoit, R., 323, 592
Alcott, T.I., 8, 384, 386 Beran, D.W., 236
Alpers, W., 175, 183 Berg, L.K., 593, 614
Andren, A., 609, 628, 633 Bergeron, T., 315, 371, 375, 695
Andretta, M., 49, 70 Bergström, H., 73, 76, 83, 84, 593
Anquetin, S., 76, 83 Berri, G., 665, 666
Anthes, R.A., 171, 567, 568, 608, 611, 664 Bilbro, J., 493
Arakawa, S., 163, 183 Billwiller, R., 223, 233
Armi, L., 135, 137, 175, 178, 179, 181–184, Binder, H.-J., 281
186–189, 191, 194 Bischoff-Gauß, I., 76, 83, 84, 271, 593
Atkinson, B.W., 38, 55, 76, 83, 564, 565 Bitencourt, D.P., 76
Atlas, D., 377, 379, 414, 450, 452 Bjerknes, V., 532–534, 536, 708
Auer, A.H., 354 Black, T.L., 592
Austin, G.L., 371 Blackadar, A.K., 608, 614
Boer, G.J., 665
Bolin, B., 536
B Bond, N.A., 162, 182, 183, 196, 197, 202–204,
Bacmeister, J.T., 166 292, 702
Bader, D.C., 82, 83 Boone, A.B., 616, 617
Bader, M.J., 364, 371 Borys, R.D., 358, 361
Bailey, M.P., 353 Bosart, L.F., 2, 5, 11, 13, 19, 196, 199, 295,
Baines, P.G., 46, 126, 134, 163, 170, 187, 191, 310, 319, 375, 698, 699
196, 200 Bossert, J.E., 85, 87, 94–96, 505
Baker, M., 365, 366 Bougeault, P., 126, 127, 157, 167, 206, 295,
Banta, R.M., 52, 69, 73, 75, 94, 144, 264, 271, 348, 504
368, 409, 415–417, 419, 436, 441, 442, Bousquet, O., 314, 367
448, 460, 466, 476, 481–485, 487, 488, Boville, B.A., 611
490, 491, 494, 499–501, 505, 511, 514, Braun, S.A., 196, 204, 206
515, 593, 596, 598, 618 Brehm, M., 44, 82
Bao, J.-W., 269–271, 466, 505 Breidenbach, J.P., 16, 17
Bardina, J., 630–632 Brewer, W.A., 359, 409, 481
Barry, R.G., 38, 44, 59, 225 Brinkmann, W.A.R., 158, 170, 220, 234, 255
Barstad, I., 204, 320, 322, 332, 365 Brost, R.A., 47, 611
Battan, L.J., 426, 427, 443, 519 Brown, A.R., 90, 133, 549, 635

Springer Atmospheric Sciences, DOI 10.1007/978-94-007-4098-3,
718 Author Index
Brown, W., 409 D

Browning, K.A., 315, 371, 373, 450, 452 Dallavalle, J.P, 672
Bruintjes, R.T., 303, 323, 348, 372, 502 Daniels, M.H., 617–620
Brulfert, G., 596, 639 Darby, L.S., 69, 79–81, 143, 409, 426, 469,
Buchot, C., 230 470, 480, 481, 484, 490–493, 497–499,
Buettner, K.J.K., 64 505–509, 592, 611
Burger, A., 86 Davies, L.A., 549
Burk, S.D., 608, 613 Davis, T., 8
Burri, K., 234, 244–246 de Franceschi, M., 49, 54, 70, 71, 73, 77, 79,
Burridge, D.M., 555 80, 83
Byun, K.-Y., 384 De Wekker, S.F.J., 55, 62, 71, 76, 80, 84, 90,
91, 94, 99, 261, 265–267, 270, 272,
278, 280, 281, 283, 479, 593, 594, 599,
C 618, 627, 640
Calhoun, R., 491, 634 Deardorff, J.K., 628
Carruthers, D.J., 371 Defant, F., 38, 44, 45
Casson, G.D., 356, 383–385 Deng, A., 601, 604
Casswell, S.A., 153 Denis, B., 601, 604
Castro, F., 641 Ding, L., 634
Cederwall, R.T., 631, 634 Doesken, N., 3, 303, 382, 384
Chang, J.S., 608, 611 Doran, J.C., 40, 42, 44, 49, 57–62, 69–72, 94,
Charney, J.G., 535, 536 95, 184, 466, 593
Chen, T.H., 616 Doswell, C.A. III, 11, 13, 18, 22, 192, 695,
Chen, Y., 48, 596, 601, 638 696, 699, 700, 709
Chen, Y.-L., 196, 204, 295, 313 Douglas, C.K.M., 305
Choularton, T.W., 371, 372 Douglas, W., 291
Chow, F.K., 54, 76, 591, 593, 594, 596, 599, Doyle, J.D., 71, 142, 144, 149–151, 155, 158,
603, 604, 618, 619, 627, 631, 632, 634, 161, 165, 168, 169, 171, 172, 196, 197,
637–639, 641, 642 201, 204, 367, 531, 549, 577–581, 593
Clark, T.L., 157, 158, 163, 165–167, 170, 474, Drechsel, S., 192, 243
483, 490, 505, 549, 570 Dreiseitl, E., 38, 56, 61, 63, 222
Cobb, D.K., 386 Drobinski, P., 62, 227, 229, 238, 481, 505
Colette, A., 43, 52, 76, 83, 624 Dubayah, R., 620
Colle, B.A., vi, 20, 166, 171, 177, 182, 191, Dubé, I., 384, 386
196, 199, 201, 204–206, 291, 292, 295, Dudhia, J., 327, 622
297, 300, 303–305, 307, 309, 311–317, Duguay, C.R., 51
323, 324, 330, 331, 351, 361, 363, Dürr, B., 242
366–369, 371, 372, 388, 390–392, 397, Durran, D.R., 18, 124, 133, 134, 144, 148–151,
452, 549, 592 155, 157, 158, 160, 163–166, 172, 189,
Colman, B.R., 10, 162, 172, 182, 655, 657, 191, 301, 310, 380, 382, 481, 538, 539,
701 542, 549–553, 568–574, 580, 582, 583,
Colson, D., 170 593, 602, 666, 669, 688
Conway, H., 383
Cook, K., 655
Cornman, L., 515 E
Côté, J., 592 Eckman, R.M., 81, 83, 268
Cotton, W.R., 85, 94–96, 323, 351, 357, 397, Efstathiou, G.A., 75, 83
460–462, 592 Egger, J., 36, 38, 63, 68, 75, 94–95
Coulter, R.L., 57, 69, 73, 74, 81, 429, 437–442, Ekhart, E., 36–38, 68, 85, 86
496, 514, 516 Ekholm, N., 532, 534
Courvoisier, H.W., 241 Eliassen, A., 536, 552–553
Cullen, M.J.P., 592 Emeis, S., 78, 414, 473, 513, 514
Cupp, R.E., 468, 469, 472–474, 481 Enger, L., 76, 127, 184, 268
Cuxart, J., 38, 44, 53, 88, 92 Epifanio, C.C., 151–152, 531
Author Index 719
Errico, R.E., 576, 577, 580–582, 664 Gr¥nås, S., 165, 204, 532, 533
Esau, I., 633–634 Grimit, E.P., 574, 575
Evensen, G., 567–568 Grisogono, B., 39, 42, 44, 63, 70, 127, 158,
183, 184
Gross, G., 63, 70, 76, 283
F Grubisic, V., 323, 324
Farley, R.D., 158, 165, 167 Grund, C.J., 468–469, 472, 474, 476, 480–481
Farmer, D.M., 189, 191 Gubser, S., 253
Fast, J.D., 43, 69, 76, 79–81, 88, 94, 98, Gullbrand, J., 631, 632
272, 506, 508–510, 592–594, 598, Gutermann, Th., 223, 239–242
599, 611
Fernando, H.J.S., 45, 46, 55, 71
Ferretti, R., 301, 313, 317, 319 H
Ferrier, B.S., 392 Hächler, P., 219
Ferziger, J.H., 539, 628 Hahn, R.S., 298, 392
Ficker, H., 222–223, 231, 233, 251 Haiden, T., 42, 44, 57–59, 61, 67, 81, 88, 92
Fiedler, F., 73, 75, 81, 265–267, 278, 282, Hallett, J., 353, 359
592, 627 Hamberg, H.E., 533
Field, P.R., 357, 395 Hamill, T.M., 568
Foldvik, A., 536 Hann, J., 222, 224
Frame, J., 310 Hardesty, R.M., 409, 426, 469, 470, 480
Frehlich, R., 515 Harimaya, T., 90, 361
Frey, K., 224, 232, 233, 250 Hashino, T., 397
Freytag, C., 67–68, 73, 82 Hauge, G., 43, 622
Froude, F., 271, 418, 493 Hawkes, H.B., 38
Fuhrer, O., 307, 369 Helmis, C.G., 44, 58, 60–62
Fujita, T., 11, 448, 708 Hertenstein, R.F., 138, 139, 152, 153
Hill, F.F., 315, 371, 452
Hiller, R., 49, 70
G Hinkelmann, K., 536
Gabersek, S., 189, 481, 504 Hobbs, P.V., 292, 315, 323, 347, 352, 357–360,
Gal-Chen, T., 515–516, 553–555 365, 369, 373, 374
Galewsky, J., 301, 314, 317 Hobson, J.M., 635
Garreaud, R., 298 Hoch, S.W., 48, 49, 91
Garvert, M.F., 295, 303, 305, 308, 314, 323, Hodur, R.M., 578, 592
324, 331, 336–337, 362, 369, 370, 392, Hohenegger, C., 295, 324, 325, 574, 575
397, 549 Hoinka, K.P., 169, 223, 234, 249, 250
Gayno, G.A., 608 Hole, L.R., 43, 622
Geiger, R., 12, 15, 38 Holton, J.R., 36, 535, 537
Georgelin, M., 167 Holtslag, A.A.M., 611
Germano, M., 630, 632 Hong, S.Y., 608, 611
Gilbert, K.K., 673 Horel, J.D., 15, 25, 592, 683, 702, 703, 711
Gill, A.E., 537 Horst, T.W., 40, 42, 44, 49, 57–59, 62, 630
Givati, A., 351 Houghton, D.D., 163
Gladwell, M., 13, 23 Houze, R.A. Jr., 11, 23, 295, 296, 302, 303,
Glahn, B., 672–673 307, 309, 314, 315, 336, 351, 357, 360,
Glasspoole, J., 305 366–368, 370, 379, 456
Gocho, Y., 371 Hunt, J.C.R., 46, 55, 127
Gohm, A., 71, 160, 162, 183, 188–191, Huppert, H.E, 165
222–223, 238, 493, 497, 499
Gossard, E.E., 429, 437, 448
Grabowski, W.W., 368, 371, 397 I
Grell, G.A., 71, 295, 324, 560, 561, 564, Iizuka, S., 635
582 Ikeda, K., 350
720 Author Index
Isaacson, E., 163 Li, J.-G., 76, 196, 564, 565

Iziomon, M.G., 52 Lilly, D.K., 157–160, 163, 169, 171, 184, 233,
259, 602, 628, 630, 631
Liu, H., 175
J Loechel, S., 620
Jackson, P.L., 121, 182, 188, 189, 191, 202 Loescher, K.A., 196–198, 201, 202, 205
Janjic, Z.I., 556, 592 Löffler-Mang, M., 68, 265, 281
Jaubert, G., 228, 238 Lohmann, U., 351
Jiang, Q., 71, 133, 148, 150, 152, 158, 161, Loney, M., 695, 697
168, 172, 183, 301, 305, 310, 311, 319, Long, R.R., 157, 163, 184
364, 549 Lopes, A., 641
Jirak, I.L., 351 Lorenz, E.M., 171
Judson, A., 3, 382, 384 Lorenz, E.N., 536, 567, 568, 664
Juuso, N., 73, 76, 83, 84, 593 Lothon, M., 229, 515
Ludwig, F.L., 15, 98, 634, 638
Lundquist, J.D., 376
K Lynn, B., 351
Kaimal, J.C., 49, 69, 70, 418 Lyra, G., 149, 233
Kallistratova, M., 437–442, 516
Kang, D., 283
Kasahara, A., 163 M
Kimura, F., 75, 79, 88, 271, 592 Magono, C., 90, 353
King, C.W., 81, 90, 440, 441, 466 Mahrt, L., 38, 41, 42, 47, 53, 62, 512, 513, 611
Kirshbaum, D.J., 310, 352, 366, 369, 370 Maki, M., 90
Klein, G., 13 Manins, P.C., 40–42, 44, 47, 49, 57
Klemp, J.B., 148, 157–160, 163, 166, 169, 171, Marchuk, G.I., 576
184, 189, 233, 301, 541, 556, 557, 559, Markowski, P., 310
582, 592, 596, 602, 605 Martilli, A., 261
Klimowski, B.A., 502 Martı́nez, D., 37, 44, 70, 88
Knupp, K.R., 460, 462 Martner, B.E., 453, 454, 457, 458, 464, 515
Kondo, H., 72, 90, 635 Marty, C., 52
Kondo, J., 55, 72, 88, 90, 593 Marwitz, J.D., 196, 199, 302, 314, 360, 367,
Koracin, D., 76 373, 379, 382
Kossmann, M., 71, 261, 264, 267, 268, 271, Mass, C.F., 18, 74, 162, 166, 171, 177,
272, 278, 281–283 182–183, 191, 196, 199, 202, 204, 206,
Kropfli, R.A., 448, 515 292, 295, 313, 317, 323, 324, 390–392,
Kuettner, J.P., 138, 139, 152, 153 398, 567, 574, 575, 592, 595, 649, 657,
Kuhn, M., 222 660, 675, 694, 701, 702, 709
Kuo, Y.H., 201, 295, 592, 660 Matsuo, T., 376, 378
Kuwagata, T., 55, 75, 79, 88, 271, 592 Matzinger, N., 47, 51, 52, 83, 620, 623
Mayr, G.J., 121, 160, 179, 181–184, 188, 192,
194, 223, 238, 243, 252, 481, 493, 503
L McCarthy, J., 448
Lafore, J.P., 592 McCormick, H.S., 354, 355, 358, 361, 397
Langland, R.H., 576, 580, 582 McFarquhar, G.M., 378
Laprise, R., 537 McGowan, H.A., 63, 72, 73, 84, 220, 271
Law, D.C., 409 McKee, T.B., 66, 75, 82, 83, 98, 265
LeDimet, F.X., 576 Medina, S., 23, 302, 303, 309, 314, 317, 318,
Lee, C.W., 353 336, 357, 367–370, 373, 379, 456
Lee, R.L., 76 Mellor, G.L., 608, 609, 630
Lenschow, D.H., 267, 414, 512, 515 Meneveau, C., 631, 634
Leone, J.M. Jr., 76 Mercer, A.E., 173
Letkewicz, C.E., 310 Mesinger, F., 349, 556, 560, 600
Author Index 721
Meyers, M.P., 1, 10, 11, 24, 295, 352, 693 Peltier, W.R., 157, 158, 163, 165–167, 170,
Michelson, S.A., 466, 505 191, 549, 569–570
Michioka, T., 603, 604, 637, 639, 642 Peng, M.S., 133, 167
Miglietta, M.M., 305, 307, 310 Peric, M., 628
Milbrandt, J.A., 390, 397 Perry, S.J., 372
Miles, J.W., 165 Persson, A., 531, 534
Minder, J.R., 324, 366, 372, 380, 382 Philipona, R., 52
Mobbs, S.D., 143, 154, 183, 192 Phillips, N.A., 124, 532, 535–536, 541–542,
Mohorovicic, A., 142 552–553
Monti, P., 59, 61 Pichugina, Y.L., 412, 442, 515, 516
Morrison, H., 397 Pielke, R.A., 50, 592, 600, 626
Muhlbauer, A., 351 Pierrehumbert, R.T., 166, 203
Müller, M.D., 52, 622 Pitman, A.J., 616
Mursch-Radlgruber, E., 71, 72 Pleim, J.E., 608, 611
Plüss, C., 51–52
Pope, S.B., 89, 628
N Porch, W.M., 62, 73, 74, 81
Nance, L.B., 10, 172 Porte-Agel, F., 631, 632, 634
Nappo, C.J., 126 Post, M.J., 409, 468–469, 472–475, 478,
Neff, W.D., 49, 69, 70, 90, 414, 429, 431, 437, 480–481, 484
438, 440, 441, 466, 484 Poulos, G.S., vi–vii, 10, 38, 62, 69, 143, 264,
Neiman, P.J., 158, 182–183, 292, 298, 299, 271, 295, 493, 505–508, 513, 593
302, 311, 317, 327, 366, 452 Power, B.A., 354, 384
Nutter, P., 601, 604 Prandtl, L., 44, 45, 57, 63, 184
Princevac, M., 42–43, 45, 46, 55, 58, 62
O
Ochs, H.T., 350 Q
Oerlemans, J., 44, 62, 63 Queney, M.P., 184, 233
Ogura, Y., 124, 542
Ohmura, A., 51–52
Okusa, N., 90, 593 R
Ólafsson, H., 127, 167, 206 Raeder, K., 580
Oliphant, A., 52, 620 Rakovec, J., 91–92, 173
Olson, J.B., 196, 197, 199–201, 204–206 Ralph, F.M., 20, 24, 292, 295, 305, 311, 317,
Olyphant, G.A., 51 347–348, 375, 452, 699–700, 710, 711
Oncley, S.P., 47, 50, 617 Ramanathan, N., 76, 83–84
O’Neal, R.D., 66, 75, 98, 265 Rampanelli, G., 64, 65, 67, 76, 83
Orgill, M.M., 60, 72 Randall, D.A., 537
Orlic, M., 142, 184 Rangno, A.L., 349, 352, 358–359
O’Steen, L.B., 73, 76 Raphael, M.N., 162, 220
Overland, J.E., 175, 176, 182–183, 196, 197, Rasmussen, R.M., 347–348, 350, 356
203, 204 Rauber, R., 302, 348, 356–357, 359–360, 373
Reed, R.J., 162, 166, 696
Reeves, H.D., vii, 93, 301, 311
P Reinecke, P.A., vii, 18–19, 172, 191, 531,
Paegle, J., 567, 665, 666 549–552, 568–574, 582–583, 666–669,
Palma, J., 641 688
Pan, F., 175, 182–184, 188 Reinking, R.F., 358, 372, 409, 446, 458, 459,
Papadopoulos, K.H., 44–45, 57–62 502, 503
Parker, M.D., 310 Reisner, J., 323, 388, 392
Parlange, M.B., 634 Reitebuch, O., 493
Parlow, E., 620 Rekustad, E., 536
722 Author Index
Reuten, C., 45, 55, 274, 279 Simpson, J.R., 38, 696
Reynolds, C.A., 172 Sinclair, M.R., 302, 365–367
Richard, E., 133, 167, 323, 324, 371 Skamarock, W.C., 295, 541, 592, 596, 605
Richardson, L.F., 534–536 Skyllingstad, E., 40, 44, 54, 57, 60, 61, 152,
Richner, H., 219, 227, 239–242 592, 634
Riemenschneider, U., 137, 187, 188 Smagorinsky, J., 536, 628, 629, 632, 633
Rinehart, R.E., 426–429, 443, 446, 448–449, Smith, C., 634
519 Smith, R.B., 84–85, 184, 196, 203–204, 364,
Roach, W.T., 363–364, 371 368–369, 536, 578, 580
Robichaud, A.J., 371 Smith, R.K.G., 710
Roe, G.H., 296, 365, 366, 372 Smull, B.F., 196, 314, 367
Roebber, P.J., 6, 12, 18, 19, 382–384, 386, 695, Smull, M., 314, 367
696 Snellman, L.W., 2, 13, 698
Rosenfeld, D., 351 Snyder, B.J., 665, 693, 695–697, 704, 710
Rossby, C.-G., 122, 195, 202, 203, 206, 311, Sobel, A., 301, 314, 317
535, 541, 694, 695 Somerville, R.C., 553, 555
Rossmann, F., 232, 233 Sprenger, M., 127, 189, 238
Rotach, M.W., 50, 51, 69, 70, 75, 83, 84, 615, Srinivasan, K., 76, 83, 84
637, 640, 704–705 Steenburgh, W.J., 2, 6, 13, 19, 22, 183, 197,
Rothermel, J., 493 317, 363, 373–374, 384, 386, 702
Rotunno, R., 23, 67, 76, 133–134, 295, 296, Stein, J., 228, 238
301, 305–307, 309–311, 313, 317, 319, Steinacker, R., 60, 63, 66, 67, 85, 86, 88, 230
367, 456 Steiner, M., 314, 367
Rucker, M., 67, 68, 481 Steppeler, J., 592
Stewart, R.E., 69, 345, 376, 378, 379
Steyn, D.G., 182–183, 188, 189, 191, 201–202,
S 261, 276, 277, 473
Saleeby, S., 358, 397 Stiperski, I., 148
Sassen, K., 348, 371–373, 476 Stoelinga, M.T., 292, 295, 345, 348, 702
Sasyo, Y., 378 Storebo, P.B., 315, 371
Sawford, B.L., 40–42, 44, 49, 57 Street, R.L., 609, 631, 634, 638, 641
Sawyer, J.S., 148, 292, 305, 346 Streiff-Becker, R., 222–223, 230, 231, 233,
Schamp, H., 253 234, 251, 253
Schär, C., 20, 127, 133–134, 171, 188, 189, Stull, R.B., 38, 82, 98, 267, 607, 611, 630
238, 307, 319, 324, 369, 553–555, 562 Sun, J., 50, 165, 166, 265, 577, 638
Scherer, D., 52, 620, 622
Schmidli, J., 67, 72, 76, 617, 640
Schüepp, W., 222–223, 233 T
Schultz, D.M., 183, 184, 295, 348, 693, 696, Talagrand, O., 576
709 Tatarskii, V.I., 429, 437
Schumann, U., 43, 55, 56, 592–593 Thériault, J.M., 345
Schweitzer, H., 232–234 Thompson, G., 592
Schwiesow, R.L., 414, 421, 469, 480 Thompson, W.T., 133, 167, 608, 613
Scinocca, J.F., 167 Thyer, N., 64
Scorer, R.S., 183, 222 Tollerud, E.I., 493
Seaman, N.L., 325, 595 Toth, J., 196, 199, 382
Seibert, P., 78, 98, 184, 222–223, 238, 251, Tripoli, G., 397
253, 265, 513–514 Troen, I.B., 611
Senff, C.J., 409, 494, 515
Serafin, S., 54, 55, 67, 70, 76, 83, 100
Sheridan, P.F., 143, 144, 236 V
Shun, C.M., 409, 466, 485, 487–489, 491 Vergeiner, I., 38, 61, 75
Shutts, G.J., 154 Vickers, D., 512, 513
Simmons, A.J., 555 Vivekanandan, D.S., 447, 456
Author Index 723
Vogt, S., 63–64, 67, 73 Williams, R., 175, 178, 179, 186, 189
von Neumann, J., 535 Williamson, D.L., 537
Vosper, S.B., 90, 133, 140, 148–150, 152, 154, Winstead, N.S., 175, 196, 197
190, 593 Wong, V.C., 631
Vreman, B., 631 Wood, I.R., 178, 180, 181, 185, 186, 189
Vrhovec, T., 90, 593 Wood, N., 635
Vukicevic, T., 576, 582, 664 Woods, C.P., 314, 360, 372, 374, 378
Wurtele, M.G., 536
Wyngaard, J.C., 514, 516, 611, 637, 638, 643
W
Wagner, A., 38, 63, 66, 76, 77, 86
Wagner, W., 25 X
Waldstreicher, J.S., 386, 694, 707 Xue, M., 592, 599
Wallace, J.M., 347
Walser, A., 574, 575
Ware, E.C., 383, 386 Y
Wasula, A.C., 310 Yamada, T., 608, 609, 630
Weigel, A.P., 54, 67, 69, 70, 76, 83, 84, 592, Yau, M.K., 11, 347, 390, 397
627, 637, 639, 640 Yoshino, M.M., 38, 158, 160
Weitkamp, C., 414, 469 Yuter, S.E., 315, 316, 351, 360, 366, 371
Werner, C., 480
Wesley, D.A., 291, 303, 315
Westrick, K., 295 Z
Wetzel, M., 16, 384, 443, 706 Zang, Y., 631, 632, 634
Wexler, R., 378, 450 Zängl, G., 37, 63–64, 67–69, 72, 73, 84, 88, 92,
White, A.B., 295, 311, 317, 350, 351, 360, 94, 95, 171, 184, 189–191, 193, 238,
443, 452, 454, 455, 473, 514, 593 352, 366, 397, 531, 555, 556, 562, 564,
White, J.M., 354 565, 582, 599
Whiteman, C.D., 3, 12, 35, 37, 38, 40, 44, 47, Zardi, D., 35, 49, 51, 54, 55, 67, 69–71, 73, 75,
52, 57–62, 64, 66, 68, 70, 71, 74, 75, 76, 83
78–83, 87–90, 92–95, 98, 264, 265, Zeng, Y., 323, 324, 361, 366, 388
272, 512, 592, 593, 611, 615 Zhang, D.-L., 608, 611
Wicker, L.J., 541 Zhong, S.Y., 38, 43, 58–60, 68, 76, 78, 81, 88,
Wilbour, C., 383 92, 94, 95, 184, 264, 268, 272, 466,
Wilczak, J.M., 53, 271, 465 505, 506, 508–510, 591–594, 598, 611,
Wild, H., 230 614
Wilhelmson, R., 541 Zoumakis, N.M., 50, 75, 83
Subject Index
A role in precipitation initiation and riming

Accidents, 3, 10, 137, 236, 247–248 growth, 366, 374
cable car, 248 rotors, 137–138
traffic, 247–248 scattering, 468, 469
train, 248 size distributions, 469, 478, 479
Adiabatic liquid water production, 349, 350 wavelength, 425, 430, 468
Adjoint model, 567, 576–583 Ageostrophic wind, 189, 192, 193, 195, 197,
optimal perturbations, 577 233, 658
Advanced Regional Prediction System Agriculture, 8, 12, 593
(ARPS), 592, 599, 600, 604, 613, Airborne remote-sensing, 494
619, 623–625, 638, 640 Aircraft
Advection airborne sensors, 273
air mass, 192, 193, 267 aviation forecasters, 27, 28
buoyancy, 557–559 drop-size distributions, 428
cold air, 78, 79, 83, 93, 272 HKIA, 485, 487
ice storms, 2–3, 5–6 icing, 10, 350, 456, 462
inversions, 93 in-situ sensors, 410
phase-speed error, 546 lee waves, 140–142
pollution, 273, 277 measurements, 273, 278, 293, 294, 398,
steady state, 42–43, 55, 321 506
surface energy budget, 47 MWISP, 456
temperature, 44, 57, 67, 74–75, 227 NOAA-P3, 175–177
three-dimensional, 387 observation, 140, 158, 166, 197, 199, 392,
Advective venting, 267 595, 613
Advocates, 699, 701–702, 707, 709–711 PSD, 395
Aerosol layers (AL), 266, 267, 279, 469, 473, radiometers, 420
479, 493 rotors, 137, 143, 144
Aerosols SARJET, 197, 199
airborne, 493–495 TKE, 229, 613, 614
backscatter windstorm, 166
intensity, 268–270 wind velocity, 69
measurement, 473–479, 514 Air mass
concentration, 246, 358, 473, 474, 477, 480 advection, 267
ice particles, 352 aerosols, 246, 493
lidar, 469–473 Bragg scattering, 428–429
radiation, 620 CCN, 349, 350
representation in models, 283–284 Chinook, 157

Springer Atmospheric Sciences, DOI 10.1007/978-94-007-4098-3,
726 Subject Index
Air mass (cont.) Aspect ratio, 446, 468, 598–599, 638, 641, 642
foehn breakdown, 226 Atmospheric
gaps, 173–176, 182–187, 190–195, 207 boundary layer (see Boundary layer)
in-situ measurements, 416, 417 river, 294, 305, 317, 325, 327, 334, 335
interaction, 246 transport, 493, 494
LES simulations, 56 Atmospheric Emitted Radiance Interferometer
microphysical evolution, 373 (AERI), 420
radiative process, 79 Attenuation
scale interaction, 13–14 depolarization, 476
temperature, 89, 224, 237 description, 429–430
Air pollutants. See Pollutants DIAL, 470, 471
Airport lee waves, 148
aviation forecasters, 27 measurements, 478
CAT, 506 remote sensing, 425
HKIA, 466, 485–490, 517 scattering, 438, 473
laser emission, 468 Avalanche forecasts, 26–27, 346
MOS, 658 Aviation, 10, 16, 27, 28, 126, 137, 152, 207,
QPF, 382 220, 466, 468, 485
RWPs, 466 Aviation forecasts, 27–28
snowstorms, 3
Air quality
boundary layer processes, 263–266 B
cold-air pools and, 12 Background flows, 62
forecasts, 274, 275, 283, 284, 466, 679 Backscatter
management strategies, 274–275 aerosol, 268–270, 469, 470, 514
societal impact, foehn, 246 Bardina model, 630
Albedo, 26, 48, 50, 52, 62, 97, 622 CW systems, 480
Alpine Experiment (ALPEX), 157, 160, 222, direct detection, 472, 473, 480
224, 238 DMM, 631
Alpine massif, 220, 221, 238 hydrometeors, 175
Ambient stability, 55, 57, 58, 60, 61, 80, 90, measurement, 141, 142, 473–479, 513
99, 265, 334 NBB, 454
Anelastic equations, 125, 127, 542 polarization, 446
Anomalous winds, 76, 78 PRF, 424
Antenna pulse transmission, 422
beam width, 425 SNR, 432
characteristics, 430–432 sodar, 435, 437–442, 513–514
continuous-wave (CW) signal, 518 turbulence, 513
DBS, 464 wavelength, 425–430
lidar, 468 Backscatter cross section, 426–428
phased array, 432, 464 Balloons, 140, 236, 280, 296, 410, 418, 532,
radomes, 450 534
transmitter, 438 Baroclinic storms/systems/waves, 295, 317,
Anticyclonic, 37, 47, 162, 189, 205, 249, 319, 335, 348, 353, 371
275–276, 562 Barrier dimension impacts on precipitation,
Antitriptic balance, 195, 199, 202 310
Anti-winds, 65 Barrier jets, 122, 123, 195–208, 302, 311, 314,
Appalachian Mountains, 199, 275, 310, 317 317, 664
Arctic, 160, 162, 476, 497 classic, 201
Arizona, 372, 500, 502, 503, 611 Barrier winds, 268
ARPS. See Advanced Regional Prediction Basins, 262, 271, 272, 461, 512
System (ARPS) Aizu, 88
Artificial cloud seeding, 352 Colorado Plateaus basins, 88, 93
ASCOT project, 72 Columbia, 88
Subject Index 727
Danube Valley, 88 plateaus, 94

Dolines, 88 QPF, 324
Duero Basin, 88 radiative flux divergence, 47
Grünloch, 88 reflectivity, 466
Ina Basins, 88 rotors, 149–152
Los Angeles basin, 88, 262, 493 SBL, 512–513
Majorca, 88 separation, 137, 148–150, 152, 206
McKenzie Lake basin, 88 temperature, 36, 600
Meteor Crater, 88 TKE, 610, 613
Peter Sinks basin, 88 turbulence, 53, 69–71, 190, 336, 606,
Sinbad Basin, 88 610–613
Basin-to-plain wind, 88, 94, 271, 272 urban areas, 71
Bernoulli, 136, 168, 186, 194 vertical resolution, 597, 598
Bifurcating flows, 73 Boussinesq approximation, 42, 124, 128, 132,
Biometeorological effects, 248–249 146
Bistatic, 438 Bragg scattering, 428–429, 437, 443, 448, 463,
Blocked flow 465, 513
barrier jets, 195, 208, 302 Brenner Pass, 179, 181, 183, 188–190, 221,
MAP, 302 222, 224, 238, 252
orographic precipitation, 303, 311, 364, Bridging the gap, vi, 325, 693–712. See also
381 Gap bridging
structure, 315 Bright band, 303, 315, 317, 350, 357, 375, 452,
Blocking, 92, 167, 196, 197, 203, 224, 227, 453, 456, 503, 504
256, 367, 382, 599 Brooks Range, 196
BMPs. See Bulk microphysical Brunt Väisälä
parameterizations (BMPs) buoyancy frequency, 122, 128, 130, 203,
Bora, 10, 142, 157, 158, 160, 162, 167, 184, 368, 549, 557, 558, 562, 577, 608,
188, 189, 664 669
Bottom boundary, 602, 625–627, 637 period, 237–238
Boulder, Colorado, 69, 138, 139, 144, 158, Bulk microphysical parameterizations (BMPs)
160, 171–172, 473–474, 483, 506 aerosols and CCN, 393
Boulder windstorm, 157, 158, 167, 173, 254 autoconversion, 393
Boundary layer (BL) double-moment, 390
aerosols, 469, 470, 473, 493 fall speed-diameter relationships, 395–396
and air quality, 261–284 growth due to riming, 396
Alps, 86 ice initiation in numerical models, 394
Boussinesq approximation, 128 mass-diameter relationships, 395–396
convective eddies, 83–84 prediction of degree of riming, 397
cross-slope flow, 61 single-moment, 389–390
damping processes, 133, 134 size sorting, 390
depth, 69, 162, 262, 263, 608, 625, 633 Buoyancy forces, 39, 42, 47, 63, 89, 125, 127,
development, 618 170
dynamic adaptation, 173 Burger number, 203
effects, processes, 69–71 Butterfly effect, 533
fire weather forecasters, 25
forecasting, 97
gap flow, 189 C
interaction, 337 California Land-falling Jets (CALJET), 292,
internal, 73, 224, 235, 237 294, 295, 298, 302, 348, 452
lee waves, 137, 148 Canyon outflows, 442, 481, 491, 508, 611
LES, 43–44, 56, 628, 634, 635 Canyons, 69, 74–75, 162, 482, 508
mesoscale models, 593 Columbia Gorge, 74
NWP, 547, 661, 681–682 Grand Canyon, 74–75, 78, 484, 491
parameterizations, 75, 99, 625 Carbon monoxide (CO), 416, 500
728 Subject Index
Cascade Range, 175, 318, 392, 618 reflectivity, 429

CBL. See Convective boundary layers (CBL) RWP, 463
Central California Ozone Study, 466 radiometers, 420, 503
Channeled flow, 61, 71, 79, 80, 268, 269, 271, rotor, 138, 139, 142
310, 466, 476, 481, 494, 658 satellite, 16, 18, 173
Channeling, 5, 6, 39, 71, 72, 133, 225, 268, venting, 267
283, 332, 460 Coastally trapped
Chemical species, 283, 415, 416, 420, 470 disturbances, 201
Chinook wind, 10, 157, 220, 249, 255–258 wind reversals, 201, 202
Circulation Coastal plains, 262
air pollutants, 271 Cold-air
asymmetric, 623 avalanches, 62
background flows, 60 damming, 3, 5, 199–200, 204, 206, 208,
boundary layer, 36, 593 599–600
cross-valley, 72, 83, 639 damming precipitation impacts, 303
depression, 55 pools, 3, 12, 67, 74, 92, 93, 100, 264, 266,
idealized simulations, 562, 563, 565, 566 268, 560, 593, 612 (see also Cold
mass flux, 73 pool)
mountain-plain, 84–87, 100 Cold bias, 43, 509, 510, 593
mountain waves, 126 Cold front
NWP, 657, 660, 662, 683 advection, 193
organize, 36, 535 air mass, 192
plateaus, 94, 95, 592–593 Doppler lidar, 497, 499
precipitation enhancement, 316 foehn breakdown, 226, 245, 251
remote effects, mountains, 2 orographic storm life cycle, 373–375
rotor, 137–138, 466 Cold pool
terrain-induced, 597–598, 606, 612, 657, Chinook, 157
664 Colorado, 12, 93
valley winds, 63–66, 565, 566 development, 12
Clear-air turbulence (CAT), 126, 506 depth, 90
Climate impact, 246 downslope flow, 90
Cloud condensation nuclei (CCN), 349–352, forecasts, 93
374, 386, 393, 475–476 gap flow, 205
Clouds idealized simulations, 310
Alps, 86–87 interaction, 235
backscatter measurement, 475–476 MAP, 252
bulk microphysical parameterizations multi-day, 266
(BMPs), 323 removal, 235
cold-air pools, 93 shear-induced turbulence, 194
dimmerfoehn, 234, 249–250 temperature inversions, 90–92
foehn, 173, 221, 233–235, 249–252, 362 waves, 235–238
lee waves, 137, 140, 144 Collaboration, 99, 337, 361, 660, 690, 694,
lenticularis, 235–236, 249 696, 699–702, 704–711
linear precipitation model, 320–322 Collaborative Science, Technology, and
Maloja wind, 76 Applied Research (CSTAR),
microphysical 705–707, 711
impacts, 313–316 Collision-coalescence (CC), 349–352, 358,
processes, 292, 313, 345–399 360, 361, 365, 441, 442, 454
NWP, 535, 537, 680 Colorado Front Range, 3, 51, 196, 199, 294,
persistent cold-air pools, 93 295, 303, 315, 441, 505
radar Colorado Springs, 386, 423, 460, 474, 481,
ice particles, 443 483, 491, 493, 496, 506, 507
measurement, 445–448 Columbia, v, 88, 174, 196, 197, 202, 277, 414
Subject Index 729
Columbia Gorge, 74, 191 interaction, 168

COMET. See Cooperative Program for lee-wave amplitude, 150
Operational Meteorology, Education levels, 26, 130, 165
and Training (COMET) melting process, 377
Compensation currents, 63–64, 86 non-dimensional mountain height, 165
Compensation flows, 36, 253 pressure gradient, 559
Computational cost, 173, 389, 397, 602, 609, streamline, 165, 166
611, 635, 636 turbulence, 53, 57, 512
Condensation rate, 321 value, 165, 205, 610, 612
Conditionally unstable flows, 310 wavelength, 425
Conical scans, 423, 450, 451, 486, 490, 492, Cross-barrier pressure differences, 243
499, 501, 515 Cross-slope flow, 61
Continuous stratification, 137, 163, 178, 186, Cross-valley circulation, 72, 623
189 Cross-valley flows, 72, 83, 639, 640
Continuous-wave (CW), 421, 448, 480, 481, deep convection, 83
518 CSTAR. See Collaborative Science,
Convective boundary layers (CBL) Technology, and Applied Research
breakup, 75, 82 (CSTAR)
CBL depth, 263, 283, 514 Cyclones, 193, 250, 251, 292, 303, 317–319,
eddies, 83 411, 441, 442, 534, 576, 580,
LES, 43, 631, 633 657–659, 702, 708
LFV, 277 Cyclonic shear, 234
transport, pollutants, 263
turbulence, 53
types, 279, 281 D
Convective velocity scale, 70 Dashikaze, 183
Convergence zone, 311, 448, 508 Data assimilation, 283, 324, 336, 531, 538,
Cooling rates, 48, 57, 74, 80, 90, 618 567, 576, 656, 660, 676, 679
Cooperative Program for Operational Dead zone, 170, 422, 423
Meteorology, Education and Debris flows, 3, 6–7, 325
Training (COMET), v, vi, 14, 170, Decorrelation time, 519
621, 700, 704–707, 711 Deep valleys, 81, 83, 265, 330, 615
COSMO, 592 Density currents, 310, 440, 448, 497–499
COSMO-2, 244–245 Density of accumulated snow, 356, 383–385
Counter flow, 237 Denver cyclone, 271, 441, 442
Coupled Ocean/Atmosphere Mesoscale Depletion of precipitation
Prediction System (COAMPS), evaporation and sublimation, 361, 363,
155, 156, 168, 169, 568, 578, 592, 364, 378, 380, 387
593 Depolarization ratio, 446, 447, 476, 478
Crest Descending air, 173, 230
Bernoulli equation, 136 DGEX. See Downscaled Global Forecast
coastally trapped wind reversals (CTWRs), System with WRF EXtension
202 (DGEX)
Kelvin-Helmholtz instability, 167 Diabatic impacts on orographic precipitation,
lee waves, 137, 138, 149, 150, 235, 236 313–314
rotors, 138, 149, 206 Differential-absorption lidar (DIAL), 470–473,
Crest height, 67, 223, 224, 235, 236, 366 494, 497
Critical Differential Optical Absorption Spectroscopy
amplitude, 150 (DOAS), 420
avalanche forecasts, 27 Diffluence, 225
challenges, 694, 698 Diffuse radiation, 50, 620, 623
conceptual models, 22, 681 Dimensionless governing parameters, 274
conditions, 27, 187, 394 Dimmerfoehn, 234, 249–250
fire weather forecasts, 25 Direct detection, 471–473, 480
730 Subject Index
Discretization errors, 556, 559, 561, 564, 571, VERTIKATOR field study, 87
573. See also Numerical errors wind-profile measurements, 464
Dispersion Downscaled Global Forecast System with
air pollutants, 12, 264, 271, 283 WRF EXtension (DGEX), 663, 669,
convective boundary layer (CBL), 265 671–675
gravity wave motion, 128 Downslope
numerical, 545–546 flows
processes, 261–262, 264, 268 Mt. Hymettos, 58
turbulence, 512 Oquirrh Mountains, 61
Dissipation Rattlesnake Mountain, 58
clouds, 252 snow surface, 62
fogs, 38 wind, 10, 59, 63, 133, 151, 157, 159, 161,
mountain waves, 126 164–166, 168–173, 220, 233, 268,
numerical, 545–546 493, 572–574
temperature inversions, 36 wind predictability, 172–173, 567–571
turbulent kinetic energy (TKE), 515, 609, windstorms, 10, 27, 122, 123, 126, 130,
613, 630 134, 136, 155–173, 180, 181, 183,
turbulence, 57, 133, 440, 609 191, 206–207, 220, 246, 251, 505,
Distributed targets, 427 568, 570, 666–668
Diurnal east winds, 157
cycle, 71, 87, 89–92, 97, 281, 505, 512 Down-valley jet (DVJ), 81
distribution, 240 Down-valley winds
flows, 36, 96, 98 Elqui Valley, 68
mountain winds, 35–100 Drag coefficient, 626
wind systems, 37, 38, 47, 88–98, 512 Drainage flow oscillations and overshooting,
diurnal reversal, 37 62
Dividing streamline, 166, 268 Droughts, 3, 8, 25
Doppler Dry-adiabatically, 220–221, 252
shift, 434, 435, 464, 472, 519 Dry deposition, 500
sodar, 412, 503 Drying ratio (DR) examples, 300, 310, 314
velocity, 425, 434–435, 445, 446, 449, 450, Dynamical cores, 537, 538, 567, 583
456, 463, 472, 474, 480–482, 502, Dynamically driven flows, 71, 123–125
507, 519 Dynamical rotor, 268
Doppler lidar, 413, 415, 435, 458, 469, 476, Dynamic mixed model (DMM), 631, 632,
482–493, 496, 499, 500, 505–507, 634. See also Scale-similarity model
515–517 Dynamic reconstruction model (DRM), 631,
atmospheric particulates, 410 632, 634, 641
Boulder windstorm, 158 Dynamics
conical scans, radial velocity, 501 diurnal valley winds, 69, 76
continuous wave (CW), 480–481 foehn, 228, 230–239
direct-detection systems, 480 gravity wave, 167
dual-Doppler lidar datasets, 484 mountain waves, 126–128, 166
fixed-beam profile display, 436 rotors, 143, 148, 149, 151, 152, 155
high-resolution, wind speed, 412 turbulence, 99, 512, 634–635
lee-wave rotor event, life cycle, 506 Dynamic Smagorinsky model
pulsed, 480, 481, 516 scale-dependent, 630–632, 634
pulse-to-pulse frequency, 468
surface-based, 490, 491
thermally-forced LLJ, 481 E
velocity field, 143, 145 East winds, 157
vertical mixing effects, small-scale cold Eddies
front, 497 convective boundary layer (CBL), 83
vertical-slice, Colorado Springs airport, formation, 269–271
423, 474
Subject Index 731
large-eddy simulation (LES), 43, 56, 628, IPEX, 294, 295, 302, 348
635–636, 638 MAP, 69, 189, 220, 227, 294, 295, 348
shear-induced, 194, 680 MAP D PHASE, 704–705
turbulent, 43, 465, 513, 612 METCRAX, 611, 615
Eddy covariance, 70 orographic precipitation, 296, 302, 305,
Eddy diffusivities, 44, 510, 607–609 334, 337
Eddy viscosity, 46, 609, 629–634, 637, 638 PYREX, 126, 140, 157, 167
Electromagnetic signal, 426, 428, 429 SARJET, 196, 197, 199
Emergency response, 28, 410, 511, 702 SRP, 143
Energy budget T-REX, 139–141, 143–145, 155, 157, 172,
conceptual and analytical models, 75 466, 568, 615, 667
surface (see Surface energy budget) VERTIKATOR, 86, 87
turbulent kinetic energy (TKE), 56 VTMX, 466, 481, 490–492, 506, 593, 611,
Energy conservation, 46, 50, 136 638
Ensemble averages, 607, 628, 629 Filtering, 531, 537, 545–546, 562–567, 572,
Ensemble Kalman filter (EnKF), 190, 567–568, 582, 627–632, 634, 636–639, 641,
663 682, 688
Ensembles Filter width, 628–630, 632, 635–638
average, 607, 628, 629 Finite differencing, 154, 191, 504, 509, 510,
data assimilation, 567–568 536–540, 543–550, 557–559, 568,
forecast systems, 18–19 583, 595, 627, 631
MAP D-PHASE, 705 Fires, 247–248
mean vertical velocities, 571–572 Fire weather forecasts, 25
NWP, 576, 662, 663, 688 Flat terrain
orographic precipitation, 325, 336, aerosols, 266
574–575 conceptual model, 263
probability densities, 572–574 dispersion processes, 262
simulations, 172, 568, 570 idealized simulations, 600
Entrainment, 56–58, 191, 235, 267, 608 large-eddy simulation (LES), 628, 631
Equilibrium flow, 57, 611 land-surface model (LSM), 616
Error growth, 324, 567–569, 664–666 measurements, 273
Euler equations, 124 radiative flux divergence, 47, 48
Evening transition period, 71, 80 stable boundary layer (SBL), 512–513
Exner pressure, 124 turbulence parameterization, 611, 613
Explicit microphysical schemes, 348 Floods, 2–3, 6–7
Exposure, 37, 52, 82, 347, 443, 616, 619, 620, flash flooding, 19–20, 295, 500, 501, 504
676, 685–686, 696, 709, 710 Flow aloft, 80, 87, 175, 180, 182, 189, 190,
Extended sea breeze (ESB), 72, 272 207, 224, 300, 315, 457, 458, 493
External governing variables, 274 Flow blocking and precipitation, 302–303,
Eye safety, 468, 469 320, 335
Flow convergence, 264, 270, 310
Flow dynamics, 126, 166, 173, 194, 230, 320,
F 583, 634
Fall speed of ice particles, 300, 354, 357, 358, Flow separation, 137, 138, 143, 148, 150–152,
366, 390, 393, 395, 396, 445, 446, 193, 268, 556, 635
452 Flow splitting, 162, 238, 264, 282
Fast modes, 541–542 Foehn
Field programs, 656, 682–687 Austrian, 224, 234
Field studies breakdown, 225, 226, 242, 245, 251
ALPEX, 157, 160, 238 Century, 222, 248
ASCOT, 72, 73 cloud, 173, 221, 233–235, 249–252, 362
CALJET, 292, 294, 295, 302, 305, 348 definition, 220
flow blocking, 302, 303 dimmerfoehn, 234, 249–250
IMPROVE-2, 294, 305, 324, 362 dynamics, 223–224, 228, 230–239
732 Subject Index
Foehn (cont.) HPC, 697

extreme, 222 intuition, 695
flow, 252 lightning, 18
frequency, 225, 240, 249 MAP D-PHASE, 704–705
harbor, 248 models and, 152–155, 168–173, 188–191,
knee, 245, 249–252 204–206
nose, 245, 250, 251 MOS, 19
North, 223, 230 NWP and weather, 655–690
occurrence, 173, 220, 227, 239–242, 249, NWP, 18–19, 387, 655–690
251 NWS, 702
onset, 225, 244, 245, 268 pattern recognition, 14, 20–22, 98, 386,
open problems, 245–246 398, 660
regions, 158, 220, 223, 230, 239 radars, 16, 443
research, 220–224 RSD, 707–708
simultaneous south and north, 222, 233 rule of thumb, 19–20
societal impact, 220, 246–249, 346 snowstorms, 3
South, 223, 230, 235, 238, 249, 250, 252, surface-based observations, 15
255, 257 training and opportunities, 696
storm, 158, 220, 246–248, 251 Forecast process, complex terrain
surface pressure pattern, 245 description, 12
Swiss, 224 forecast funnel, 13–14
test, 241 mountain weather forecasting, 23–24
textbook theory, 221–224, 227, 230, 238 objective tools (see Objective forecast
topography effect, 225 tools)
type I, 224, 227 specialized forecast (see Specialized
type II, 224, 246 forecast, mountains)
West, 223 subjective tools (see Subjective forecast
Foehn-wall cloud, 362 tools)
Fog, 237, 252, 279 Forest fires, 230, 448, 476
Forced channeling, 71–72. See also Channeling Fourth-order diffusion, 563–564
Forced convection, 220 Fraser River Valley, 414, 415, 484, 499
Forecast, 233 Freeze warnings, 511
funnel, 13–14, 18, 19 Freezing level, 6, 11, 301, 325, 327, 332, 447,
process, 2, 12–23, 296, 398, 584, 656–657, 452–454, 456, 503, 504
662, 673, 675, 676, 679–682, Fresno Eddy, 269
686–690, 697, 698 Friction
skill, 19, 656, 659–664, 671, 680, 688 boundary layer, 134, 267
tools for orographic precipitation, 296 effect, 167–168
Forecasting surface stress, 57
air quality, 265, 274, 275, 284, 466 turbulent, 188–190, 195
Alpine foehn, 219–258 Frontal systems, 196, 200, 275–276, 292, 295,
avalanche, 26–27 303, 305, 315, 374
aviation, 27
collocation, 708–709 G
complex terrain, 23–28 Gap-bridging (between research and
conceptual models, 22, 23 forecasting)
CSTAR, 706 causes, 694–695
diurnal mountain winds, 97–98 ingredients, bridge, 699–701
fire weather, 25 perceptions, 695–698
foehn knee, 245 successful bridging
gap-bridging, 699–701 advocates, collaboration, 701–702
gap flows, 192–195 CSTAR and COMET, 705–707
Geographic Information Systems (GIS), forecasting and research teams,
19, 98 mountain meteorology, 708–709
Subject Index 733
MAP D-PHASE, 704–705 motion, 128, 129

Olympic winter games, 702–704 nonhydrostatic, 550
research support desk, 707–708 restoring force, 170
research to operations, 335, 694, 696, 698 sidewall heating, 83
Gap flows. See also Passes stratosphere, 126
channel acceleration, 122, 206 structures, 337
characterization, 622 vertical motion, 305
dashikaze, 183 warm air, 95
downslope windstorms, 10, 130, 136 windward ridges, 305
dynamics, 165 Graz City Eddy, 271
foehn, 254 Great Plains, 36, 87, 100, 411, 412, 459, 460,
forecasting, 192–195 464, 466, 613, 614, 617
high-amplitude mountain waves, 10 Great Salt Lake, 423, 440, 481, 482, 490, 491,
NOAA-P3 aircraft, 176 506
nocturnal downvalley, 193 Grid
observations, 176–186 nesting, 155, 547, 594, 597, 602–605, 635,
outflow jets, 485 636, 638, 639, 641–642
prototypes, 181 resolution, 54, 273, 546–548, 595, 596,
SAR, 175 602, 605, 624, 625, 628, 631,
shear-induced turbulence, 194 636–640, 642, 671
theory, 184–188 stretching, 506, 547, 597
valleys, 122 Ground clutter, 16, 443, 449, 462–463
wind observations, 174–184 Ground heat flux, 47, 50, 89
wind profile, 177 Group velocity vector, 551
Gayno-Seaman, 510, 608, 614 Gust, 158, 167, 243, 248, 460, 485, 490,
GEM-LAM. See Global environmental 497
multiscale, local area model fronts, 248, 460, 485, 490
(GEM-LAM) winds, 222, 243, 248
General circulation, 2, 75, 240, 535–536 Gütsch, 222, 227, 229, 242, 245
Geostrophic adjustment, 195–197, 201, 205,
208
GFS. See Global forecast system (GFS) H
Glaciation, 352, 394 Half width, 195, 203, 204, 305, 311, 320, 347,
Glacier wind, 7, 64–65 364, 550, 558
Glarus Alps, 246 Hazardous Weather Testbed, 709, 711
Glide path scan (GPScan), 485, 489, 490 Haze, 12, 234, 235, 237, 246
Gliders, 27, 152, 236 Heat flux, 49–52
Global environmental multiscale, local area Heating
model (GEM-LAM), 592, 684, 685, air parcels, 125
687 atmospheric layers, 36
Global forecast system (GFS), 325, 327, 663, by convection, 235
670–672, 674–675, 697 day-night, 37
Global-positioning system (GPS), 324, 420 mountain atmosphere, 85
Gotthard, 221, 222, 224, 225, 227 non-uniform, 536
Grand Canyon, 74, 78, 484, 491 surface, 440, 441
Gravity waves topography, 634
cold pool shear, 236 and upslope advection, 55
and convective motions, 541 upstream cloud-cover, 193
distribution, 126 valley atmosphere, 67, 193
downslope windstorms, 123 Heat of condensation, 222, 301
dynamics, 167 High Spectral Resolution Lidar (HSRL), 470,
internal, 123–125 476, 503
and Kelvin–Helmholtz waves, 81 Himalayas, 36, 307, 536
microphysical processes, 367–368 HIRLAM, 592
734 Subject Index
Hong Kong International Airport (HKIA), 466, I

485–490, 517 Ice nucleation (IN)
Hong Kong Observatory (HKO), 485 heterogeneous freezing, 352
Horizontal homogeneous freezing, 352, 394
advection, 82, 83, 171, 262, 267, 548, 555, ice nuclei, 352, 354, 358, 359, 369, 373
557–559, 568, 571 Ice particle types
heat transfer, 82, 83 crystal habits, 356, 358, 395, 397
pressure gradient, 37, 54, 63, 68, 78, 171, dendrites, 353, 358, 374, 384, 394
188, 553, 556, 557, 599 graupel, 357, 358, 384, 397, 456
resolution, 97, 153–155, 171, 190, 191, graupel-like snow, 358, 384
204, 283, 323, 549, 565, 583, Ice-storms, 2–3, 5–6
584, 594–599, 601, 604, 617, 621, Idealized
638–640, 643, 683 precipitation simulations, 305, 310, 337,
Horizontally homogeneous, 411 368, 370
HPC. See Hydrometeorological Prediction simulations, 38, 44, 48, 56, 65, 75, 76,
Center (HPC) 83, 94, 166, 167, 170–171, 188,
HSRL. See High Spectral Resolution Lidar 189, 204, 205, 305, 310, 337,
(HSRL) 359, 368, 370, 380, 389, 561, 563,
Hybrid 565, 566, 600, 601, 616–617, 624,
barrier jets, 197, 199, 201, 204, 205, 633–635
208 topography, 38, 48, 52, 56, 61, 62, 75, 76,
vertical coordinate, 555 189, 204, 205, 310, 370, 565, 600,
Hydraulic flows, 61, 124, 134, 157, 160, 161, 624, 634–635
163–166, 170, 187, 207 IMPROVE-2 field experiment, 307, 362
Hydraulic jumps Inertial subrange, 43, 429, 513, 515, 636
downslope winds, 207 Initial conditions
downstream air mass, 207 complex terrain, areas, 336
downstream conditions, 134 ensemble forecast systems, 18
flow types, 138 ensemble Kalman filter (EnKF), 567–568
gap flow jets, 189 meso-and fine-scale modeling, complex
isentropes, 158 terrain, 600
large amplitude lee waves, 234 NWP forecasts, perturbed, 576
strong downslope wind, 172 precipitation error growth, 324–325
supercritical flow, 159 realistic gap flows, 190
supercritical shooting, 160 Innsbruck, Austria, 45, 222, 497, 499
theory, 234 Inn Valley, 63, 66, 68, 69, 78, 179, 235, 565,
turbulent, 125 566
Hydraulic models, 26, 44, 57, 75, 188–190 In-situ sensors, 292, 410, 412, 414–419, 422,
Hydraulic theory, 123, 127, 134–137, 163, 165, 442, 499
166, 184, 493 Interaction. See also Scale interaction
Hydrological runoff forecasts, 26 air-sea, 184
Hydrologic models, 620 boundary layers and air quality,
Hydrometeorological Prediction Center (HPC), mountainous terrain, 271–272
697 drainage flow oscillations, 62
Hydrometeorological Testbed, 710, 711 hybrid barrier jets, 197
Hydrostatic microphysics, kinematic and dynamic
balance, 42 aspects
balanced reference profiles, 124 blocking, 367
mountain wave response, 131 gravity waves, 367–368
and non-hydrostatic models, 542 mountain geometry and flow
and non-hydrostatic wave fields, 131 characteristics, 365–366
quasi-hydrostatic balance, 42, 605 simple terrain-following flow and
and semigeostrophic balance, 317 fundamental time scales, 363–365
two-dimensional linear, 577 and small-scale cellularity, 368–369
Subject Index 735
midlatitude frontal and coastal orography, entrainment, 56

200 equations, 54, 628–629, 640
model microphysics with turbulence, 369 flat terrain, 628, 631, 632
model microphysics with positive-definite idealized simulations, 44, 628
advection, 397–398 idealized topography, 75
model processes, 397–398 stable boundary layers, 43, 634
radiation field and topography, 51, 52 terra incognita, 637, 643
slope and valley winds turbulence closure, 54, 628–632, 637
daytime phase, 78–79 Laser, 410, 468, 470–472, 481
evening transition phase, 79–80 Latent heat
morning transition phase, 81–84 cloud, 314
nighttime phase, 80–81 foehn, 220
turbulent boundary layer and lee-wave mountain wave generation, 168
rotors, 151–152 nonlinear resonance, 169
Intermittent orographic precipitation, 307 parameterized processes, NWP, 537
Intermountain barrier precipitation, 325 release, mountain barrier, 157
Internal boundary (Foehn), 224, 235, 237 vegetated areas and mountainous regions,
Internal boundary layers, 73. See also water bodies, 47
Boundary layers Lateral boundary conditions, 324, 325,
Internal gravity wave, 123–126, 128, 578, 580 574–576, 594, 601–602, 638, 641
Inversion. See Temperature inversion update interval, 603, 604, 637, 639, 642
Ion concentrations, 248 Lee of barrier precipitation, 317
Isothermalcy, 74 Lee waves
Isothermal layer, 170, 377, 379 boundary layer, 148
ISS2, 466 Boussinesq approximation, 146
cloud patterns, 140
forecasting, 155
J horizontally propagating gravity waves,
Joint Hurricane Testbed, 708, 711 123
Juan de Fuca, 162, 175, 177, 186, 190–191, interface, 146
197, 674, 680 linear theory, 148
Jura Mountains, 11, 235 measurements, activity, 140
resonant waves, 147
rotor, 122, 123, 126, 137–155, 206, 236,
K 249, 268, 506
Kanto Plain Eddy, 271 sigma coordinates, 599
Katabatic winds, 39, 160, 263, 271, 274 strong downslope wind, 233
Kelvin–Helmholtz waves, 81 theory, 126, 145–150, 154, 234
trapped
behaviour, 148
L decay, 133
Laboratory water tanks, 45 description, 133, 137
Lake breezes, 60, 72, 73, 77 flow separation, 150
Land breezes, 162, 271, 272, 278 formation, 145
Land/sea breeze circulations, 276. See also Sea horizontal wavelengths, 137, 148, 153
breeze internal dynamics, 151
Land-surface model, 615–617, 620, 637, 638 rotor clouds, 139
Lantau Island, 485, 488 structure, 138
Large-eddy simulation (LES) wave energy propagation, 149
complex terrain, 43, 44, 336, 628–642 turbulence, 236
convective boundary layers, 43 vertical displacements, 141, 142
detrainment, 56 vertical velocity, 141
difference from RANS, 628–630, 634–638, Lenticularis clouds, 236, 249, 250, 252
640, 641 LES. See Large-eddy simulation (LES)
736 Subject Index
Level gap, 174, 175, 178, 179, 185, 187, 190. DIAL, 471, 494
See also Gap flows heat flux divergence, 50
Lidar in-situ
aerosol, 426, 473, 479 advantages and limitations, 418
aerosol backscatter intensity, 269, 270 atmospheric process, 414
coherent, 471, 472, 480 extreme turbulence, 144
CW, 421 lower-O3 air mass, 416
description, 468 Pitt Lake valley, 414
Doopler (see Doppler lidar) precipitation and snowpack,
eye safety, 468–469 hydrological applications, 418
incoherent, 471–472, 480 tower-mounted, 512
low-level flow, Owens Valley, 145 isotope, 293, 294
“pencil beam”, 468 lee-wave
scattering, 468, 469 activity, Scotland, 140
solar light scatter, 432 Alps, 233
temporal variability, 144 ground based, 143
types, 469–473 Sierra Nevada mountains, 141
Lidar Windshear Alerting System (LIWAS), lidar (see Lidar)
485, 488, 490 limitations, 411
Linear precipitation model radar, 463
advantages, 323 radar wind profiler measurements, 468
“precipitation distribution”, 322 rain gauge, 365
source terms, 321 research aircraft, 140, 506
supersaturation, 321 snow water equivalent (SWE), 26
time scales, 320 sodar, 439, 442
Linth Valley, 246 surface wind, 416
LIWAS. See Lidar Windshear Alerting System synthetic aperture radar (SAR), 175
(LIWAS) terrain, 273, 410
Local climate, 246 trace atmospheric chemical species,
Local hydraulics, 165, 166 415
Local wind systems, 240 turbulence
Logarithmic profile, 625 intrinsic, 513
Log-law, 602, 626, 627 mountainous terrain, 512
Log-normal distribution, 478 slope flows, 49
Longwave radiation, 50–52, 74, 79, 81–82, 93, valley flows, 69
510, 620 upper-air, 532
Colorado Front Range, 51 wind, 464
Lower Fraser Valley, 262, 265, 275–278 Mediterranean Sea, 235, 257
Melbourne Eddy, 271
Melt cooling, 367, 377–380
M Melting level, 347, 351, 360, 361, 363, 364,
Maloja wind, 76 366, 369, 370, 375, 378–380, 669
MAP. See Mesoscale Alpine Programme Merge, 72, 73, 175, 187, 246, 602, 605
(MAP) Merging flows, 73
MAPR, 466 Meso-Eta, 43, 506, 510, 592, 593
Mass flux, 60, 68, 73, 87, 190, 484 Meso-NH, 229, 592, 623
Match Observations All (MatchObsAll), 676, Mesoscale Alpine Programme (MAP)
678 Alpine foehn downslope, 158
MC2, 592 European Alps, 481
Mean-state critical layer, 165, 166, 170, 207 foehn studies, Rhine valley, 227, 229
Measurements MAP D-PHASE, 704–705, 710
accuracy and precision, 435 mountain waves simulation, 172
backscatter, 473–479 precipitation enhancement vs. flow
complex terrain, 504–511 blocking, 302–303
Subject Index 737
precipitation process, cold-season rain, Microphysical growth processes

454, 456 collision and coalescence, 349–351
Riviera field campaign, 615 Microphysical impacts on precipitation, 315
Mesoscale Alpine Programme Demonstration Microphysics, 346
of Probabilistic Hydrological and coalescence, 349–351
Atmospheric Simulation of Flood collision, 349–351
Events (MAP D-PHASE), 704–705, growth processes, 349–361
710 impacts on precipitation, 361–363
Mesoscale circulations, 266 Micropulse, 469
Mesoscale models Microwave radiometers, 420
accuracy order, mountain waves, 549–552 Mie-scattering, 427, 428, 469, 470
atmospheric flows prediction, 537–542 Minimum range, 422, 423, 439, 468, 496. See
bottom boundary conditions, 625 also Radars, specific issues
capabilities, 323–325 MISS, 466
error source, 594 Mixed-phase clouds, 352
finite differencing, 543–548 Mixing, 71, 152, 165, 194, 222, 235, 265
grid spacing, 596–597 MM5
high resolution, 160 Cascade Mountains, 391
high-resolution two-dimensional nonlinear, horizontal diffusion scheme, 171
172 MM5-GFS simulation, 658
metric terms consistency, 557–559 5–7 November 2006 Flooding Event, 325,
Newtonian weather forecast, 532–536 327
numerical diffusion and filtering, 562–567 precipitation shadowing, overforecast, 324
orographic precipitation structures, SARJET, 197
335–336, 361 simulations, Cascade Mountains, 20
parameterization, 392, 606 turbulence parameterization schemes, 611
performance, 601 Model
predictability, atmospheric flows, 567–582 evaluation, 275
pressure gradient representation, 559–562 forecasts, 153, 243–245, 275, 325, 386,
RANS approach, 628 391, 397, 538, 574, 576, 681, 697,
realistic convergence bands, 670 705
regional forecasting, 592 intercomparison, 616
simplified version, 173 precipitation capabilities, 323
spatial resolution, 621 precipitation sensitivities, 334
turbulence parameterization, 606–611 validation and verification, 505
vertical coordinates, 552–556 Model output statistics (MOS)
vertical spacing, 597–598 central vs. western United States, 665–667
Mesoscale NWP description, 19
benefits and challenges, 660 foehn forecasting, 245
computing platforms, 660 GMOS, 667
description, 659–660 gridded, 672–674
developments, 663 MM5, 683
finer grid models, 662 post-processing mesoscale NWP
hands-off approach, 662 application, 658, 659, 670
MOGREPS and use, 663 approaches, 673
physical realism, 661–662 DGEX, 671–672
post-processing, MOS (see Model output finer scale model, 671
statistics (MOS)) GFS, 671
predictability, 668–669 gridded, 671–674
swirls and eddies, 661 limitations, 673
verification results and factors, 662 NAEFS, 675
METCRAX, 611, 615 NWS, 671
Meteor Crater, 48, 49, 51, 62, 88, 611, 615 terrain influence, 658
738 Subject Index
Moist convection, 85, 459–462 turbulence, 137

Moist neutral flow, 305–307, 311, 327, 379 vertically propagating, 164, 165, 170
Moist static stability, 301 vertical structure and precipitation, 311
Moisture gradient impacts, 301 vertical wavelength, 311
Momentum balance, 57, 63 Mountain weather forecasting
Momentum equation, 44, 125, 134, 627 human-machine mix, 2, 23–24
Momentum flux, 42–44, 53, 515, 607, 625, 627 Mount Washington Icing Sensors Project
Monin-Obukhov, 612 (MWISP), 457
Monin-Obukhov surface layer scaling, 70 Mt. Pinatubo, 473, 475
Monostatic, 422, 426 Multiple Doppler analysis, 460, 462
Monsoonal circulation, 85
Monte Carlo simulations, 48
Morning transition, 36–37, 81–84, 99, 440, N
442, 623 National Profiler Network (NPN), 464, 466
MOS. See Model output statistics (MOS) National Severe Storms Laboratory, 709
Mountain basins. See Basins NBB rainfall. See Non-brightband rainfall
Mountain meteorology community, 99, 694, Near-wall behavior, 633, 634
697, 701, 705, 707, 709 Nesting
Mountain-plain winds grid stretching and, 547
Alps, 86–87 one-way, 547, 568, 571, 602–634, 641
circulation two-way, 547, 602, 604–605, 637, 642
diurnal pressure oscillation, 84–85 Net radiation, 47, 50, 52, 510
Rocky Mountains, 87 Nighttime phase, 36–37, 80–81
equalizing flow, 86 Nitric oxide (NO), 416, 417
flows, 36, 85, 87 Nitrogen oxides (NOx), 276, 277, 416
system, 37, 84–88 Non-brightband (NBB) rainfall, 350, 452–455
VERA, 86 Non-dimensional mountain height, 122, 126,
VERTIKATOR, 86 127, 133, 148, 164–166, 169, 203,
Mountain thunderstorms, 459, 460, 462 206, 207, 301, 303, 305, 314, 327
Mountain-valley wind, 234, 240, 272 Nonhydrostatic
Mountain venting, 267, 282 Boussinesq equations, 553–555
Mountain waves. See also Lee waves continuous solution, 550
amplification, 166 equations, 536
boundary layer, 148 hydrostatic models, 542
clear-air turbulence, 126 Non-stationary, 44, 53–56, 158, 172, 512
and downslope wind predictability, Nordkette, 45, 235
567–571 November 2006 flooding event (Pacific
high-amplitude, 10 Northwest)
hydrostatic, 131 linear model precipitation, 332, 334
impact, accuracy order, 549–552 MM5 setup, 325–327, 330–334
impact on precipitation, 303 model validation, 327, 329, 333–336
internal gravity waves, 125 precipitation, 327, 329, 330, 332–336
large amplitude, 143, 145, 155, 158, 163, synoptic setup, 13–14, 173
206, 572 Nowcasting, 173, 705
latent heat, 168 foehn, 242
linearization, 127 NOx. See Nitrogen oxides (NOx)
linear mountain wave theory, 320, 321 NPN. See National Profiler Network (NPN)
long horizontal wavelengths, 129 Numerical diffusion, 510, 562–567
precipitation-enhancement effects, 500–504 implicit, 563, 566–567
propagation characteristics, 126 Numerical errors, 555, 558, 562, 564, 595,
rotor, 206 597, 631, 637, 642. See also
rotor clouds and blowing dust, 139 Discretization errors
Taylor-Goldstein equation, 132 Numerical stability, 510, 541, 564–565, 595,
temporal variability, 493 602
Subject Index 739
Numerical weather prediction (NWP) probability density function (PDF), 679,

empirical techniques, 378 688
forecast representativeness, observations
ensemble systems, 18–19, 325, 336, data assimilation method, 676
663, 681, 688, 705, 709 roles and fine grids, 676, 688
MOS, 19, 245, 658, 667, 670, 671, 673, RTMA, 676, 681
674, 698, 704 sensor errors, 676
value added utility, evaluation, 18 tools
horizontal wavelengths, 153 climatologies and survey, 681, 682
interactions, 397–398 conceptual models, 389, 661, 681, 686
measurements, 504–511 ensemble systems, 663, 681
microphysical schemes satellite image, 680–681
bulk, 389–390
explicit, 388–389
objective tools, 18–19 O
orographic precipitation Objective forecast tools
condensation and warm rain processes, lightning, 18, 681
393–394 NWP (see Numerical weather prediction
depositional growth rates, 394–395 (NWP))
ice initiation, 394 radar, 16, 18
particle type issues, 395–396 satellite, 16–18, 173
PSD, 395 surface-based observations, 15
predicted species, 397 Olympic Winter Games
weather forecasting Callaghan Valley, 684, 686
applications and special field programs, circulations and orographic forcing, 683,
682–687 687, 706
atmospheric predictability, 664–670 2D maps and meteograms, 683, 695
complex terrain, 661–663 learning experiences, 687
post-processing mesoscale, 670–675 researchers and forecasters, interaction, 23,
process and complementary forecasting 692–693, 704, 706–710
tools, 679–682 Salt Lake City, 683
representativeness, observations, Vancouver Olympics, 686
675–679 Ora del Garda, 76, 77
NWP. See Numerical weather prediction Order of accuracy, 543–545, 549–552, 557,
NWP and weather forecasting, complex terrain 558, 573, 582, 583
classic patterns, 657 Orographic convection, 307–310
communication and interpretation, 584, 689 Orographic field studies, 292, 294, 305, 334,
computing power and numerical methods, 347, 348. See also Field studies
205, 657 Orographic lifting, 255, 350, 363, 371, 379,
critical decisions, 687, 689 452, 503
degrees of freedom, reduction, 657 Orographic precipitation
impact, role, 656 bands, 303, 310, 320, 331
mesoscale (see Mesoscale NWP) cold-season, 452–459
MOS application, 658 conceptual models, 317, 319, 335
performance metrics, 688 definition, 292
predictability factors for, 292, 300, 346, 371, 377–384,
central vs. western United States, 665, 594
667 future questions, vi, 7, 319, 334
coherent structures, 664 growth
ensemble, 662, 663, 666–667, 679 time scales, 363–365
error growth, 567, 665–666 large scale impacts, 293, 317, 320, 346,
LAM and CMC models, 662, 665 364, 371, 372, 382, 456
mesoscale models, 655–657, 659–664 predictability (see Predictability)
740 Subject Index
Orographic precipitation (cont.) surface property, 97

processes, vi, 292, 296–319, 323, 334, topographic effects, 52
345–399 turbulence, mesoscale models, 606–611
review studies, 296–297, 397 turbulence (see Turbulence)
simple model, 317, 319–323 Particle size distributions (PSD), 388–390,
Orographic storms, 354, 358, 360, 361, 369, 395, 397, 446
373, 382, 387 Particulate matter (PM), 12, 276
frontal precipitation, 374 Passes
occluded front, role in orographic storm Brenner, 179
life cycle, 373–374 coastal range, 309
post-frontal precipitation, 374–375 frontal systems, 196
pre-frontal precipitation, 374, 375 horizontal aspiration theory, 233
upper cold front, role in orographic storm Maloja, 76
life cycle, 374, 375 mountain, 72
Oscillations Snoqualmie, 385
diurnal, 85 Passive sensors, 420
drainage flow, 62 Periodic flow, 74
vertical, 93 Persistent cold-air pools, 93. See also Cold air
Outflow pools
abnormal flow development, 77 Phase speed, 239, 545, 546, 580
basin-to-plain, 89 Phenomenological challenges, complex terrain
canyon, 442, 481, 508, 611 cold-air pools and air quality, 3, 12
characteristics, 491 convective storms, 3, 11
cold pool, 235 description, 2–3
cool-air, 415, 508 droughts, 3, 8
valley, 499–500 floods, flash floods and debris flows, 6–7
valley exit jets, 69 ice-storms, 2–3, 5–6
westerly-wind, 481 outdoor recreation, 2
Owens Valley, 138, 139, 143–145, 155, snowstorms (see Snowstorms)
268–270, 466, 467, 479, 483, wildfire behavior, 3, 9–10
568–570, 615, 617–620, 639–640, windstorms, 3, 10
668 Pitt Lake, 141–416, 418, 419, 499–501
Ozone (O3 ), 12, 223, 246, 247, 264–265, 276, Plain-to-basin winds, 88, 94, 271, 272
277, 281, 282, 410, 415, 470, 471, Plain-to-mountain wind, 84
494, 500 Plain-to-plateau wind, 88, 94
Ozonesonde, 418, 499 Plan-position indicator (PPI), 145, 450
Plateaus
Bolivian Altiplano, 88
P Tibetan Plateau, 88
Parameterizations Plume dispersion, 268, 274
approximation, 514 PM. See Particulate matter (PM)
boundary-layer, 75 Polarization, 16, 445–447, 454, 456, 462, 463,
description, 393 469, 476, 478, 503, 578
land-use datasets, 617–618 Pollutants
limitations of turbulence, 611–615 aged, 416
mesoscale setting, 43 and aerosols, 138
momentum flux, 42 air, 38
radiation models, 620–625 cloudy cold-air pools, 93
sky radiation, 51 cold pools development, 12
soil moisture issues, 618–620 complex terrain, 71
stable boundary-layer, 265 concentration, 261–262
subgrid-scale cumulus, 387 dispersion, 12, 264, 267, 283
subgrid-scale processes, 577 eddies, 269
surface flux, 625–627 ozone, 410
Subject Index 741
recirculation, 265, 276, 278 cellularity, 369

surface concentrations, 271 embedded convection, 369
transport, 85 gravity waves, 294, 303, 305, 367,
Pollution transport, 414, 494 368
Postfrontal flow, 493, 499 mountain height, 305, 366
Potential temperature mountain width, 367
ambient distribution, 42 small-scale topographic variability, 188,
differences, 182 193, 555, 596
Exner function, 555 speed of flow, 365
gap flow layer, 182 turbulence, 303
perturbation, 41, 578 Precipitation, growth of
profile, 41 cold and mixed-phase clouds
shear-induced turbulence, 194 aggregation, 303, 336, 356–357
steps, 190 depositional growth, 352–356, 372,
surface cooling/heating, 41 394–395
tendency equation, 46–47 rime-splintering, 359–360, 369, 372
two-dimensional ridge, 153 riming, 314, 335, 357–360, 366, 372,
vertical velocity, 572 374, 375, 384, 396, 397, 456
volume average, 67 secondary ice production, 358–360,
PPI. See Plan-position indicator (PPI) 369, 373, 374
Prandtl model, 44 warm clouds
Precipitation collision and coalescence, 349–351,
efficiencies, 504 454
enhancement effects, mountain waves, Predictability, 18, 168–169, 171–173,
500–504 319–334, 531–584
error sources, 576, 594 orographic precipitation, 296, 336,
field studies, 294 (see also Field studies) 574–575
heavy upwind, 238 Predictability of orographic precipitation, 296,
loss processes, 361–363 336, 574–575
lower air mass, 224 Pressure-driven channeling, 71, 283. See also
mesoscale, 372 Channeling
orographic (see Orographic precipitation) Pressure fluctuations, 150, 248, 249
processes Pressure gradient
efficiency, 305 differential heating rates, 68
error growth, 324 downslope layer, 63
isotope ratios, 300 geostrophic, 92
pattern, 324 local topography, 10
role, stability, 301 reversal, 65
water vapor flux, terrain-forced sensible heat flux, 65
ascent and vapor condensation, synoptic-scale, 75, 481
296–300 terrain-following coordinates,
and transition regions 559–562
conditions, 377–378 PRF. See Pulse-repetition frequency (PRF)
and orography, 379–382 Probabilistic methods, 241–243
types, 375–376 Profile models, 44, 190
types, transition regions over sloping Prognostic equations, 388, 397, 555, 607,
terrain, 376–377 609–610, 628, 630, 632
U.S. West coast, 292, 295, 305, 317, 319, PSD. See Particle size distributions (PSD)
335 Pulse-repetition frequency (PRF), 424, 444,
windward slope, 227 468, 474
Precipitation growth, effects of mountain and Pyrenees Experiment (PYREX), 126, 140, 157,
flow characteristics on 167
blocking, 227, 302, 303, 305, 320 Pyrocumulus, 476
742 Subject Index
Q boundary conditions, 602

Quantitative precipitation estimation (QPE), diffuse shortwave, 49
16, 350, 445, 463 effect, slope and aspect models, 622–623
Quantitative precipitation forecasts (QPFs), main lobe, 430
319–320, 324, 382, 397, 462, 697 mesoscale models, 606
Quasi-stationary net, 50
foehn, 235–236, 250 parameterization, 661
precipitation bands, 291, 310 physical processes, 410
Rayleigh-scattered, 470
solar, 37, 81, 90
R surface, 47
Radars topographic shading, 623–625 (see also
brightband (BB), 350, 357, 375, 452, 503, Shading, topographic)
504 Radiative
characteristics, 360 flux divergence, 47–49, 90, 593
cold-season orographic precipitation, processes, 39, 65, 79, 99–100, 397–398
452–459 and sensible heat flux, 42, 47, 49, 51, 54,
continuous wave (CW), 421 57, 58, 65, 79, 90
Doppler, 175, 177, 292 transfer (RT), 47, 49, 51, 67, 79, 327, 479,
equation, 425–426 537
ground-based, 428 Radiometers, 48, 50, 410, 420, 502, 503
measureables Rain-snow transition zone, 5, 379
clear-air returns, 448 slush particles, 380, 382
Doppler velocity, 446 wet snow (WS), 375, 376, 380–382
LWC/IWC, 445 Raman scatter, 470
multiple wavelengths, 447 RAMS. See Regional Atmospheric Modeling
polarization, 446–447 System (RAMS)
objective tools, 16 Range-height indicator (RHI), 145, 268–270,
polarimetric and dual-wavelength, 399 450
remote sensor types, 412–414 RANS equations. See Reynolds-averaged
SAR, 175, 196–198, 201 Navier-Stokes (RANS) equations
scanning, 450–451 RASS, 237, 496
scans, 503 Rawinsonde, 87, 278, 411, 413, 464, 466, 503,
sine-wave signal, 518 506, 508–509, 620
SMR, 503–504 Rayleigh scattering, 427, 469, 470
specific issues Real-Time Mesoscale Analysis (RTMA), 667,
maximum range, 448–449 676–679
protective dome, 450 Reduced gravity, 135
siting, 449–450 Redundancy, 439, 464, 517
warm-season orographic precipitation, Reference station, 242, 555, 557, 561
459–462 Reflections
weather, 410 amplification, 163
wind profilers (RWPs) profiler sidelobe signal, 464
applications, 466–467 smaller-scale flow, 601
temperature, 465 wave energy, 130
types, 464–465 wind and stability, 206
WSR-88D, 315, 327 Reflectivity
Radar wind profilers (RWPs), 87, 412–414, airborne dual-Doppler derived vertical
429, 432, 443, 463–468, 506, 508, velocity, 308
514, 516–517 cellular, 456
Radar WSR-88D, 16, 17, 315, 316, 327, 443, depolarization ratio, 446
444, 451, 452, 454, 455 Hovmoller plot, 330–331
Radiation measurements, 466
backscattered, 426 radar, 428, 429, 452
Subject Index 743
sodar backscatter, 513 models, 243–245

SPol, 309 NWP models, 173
S-PROF, 453, 454 range, 424–425, 472
Regional atmospheric modeling system slope temperature structure profiles, 48
(RAMS), 43, 506, 508, 592, 599, spatial, 76, 496, 702
613, 622–623, 638 ultra-fine, 448
Relaxation boundary condition, 605 VERA, 86
Remote-sensing wind speed, 565, 566
active, 420, 421, 424, 432, 437 Retrieval errors, 478
categories, 420–421 Return flows, 63–64, 73, 85, 87, 188–189, 278
characteristics, transmitted signal Reuss Valley, 223, 225, 227, 235, 237, 248
antenna, 430–432 Reynolds-averaged Navier-Stokes (RANS)
basic system properties, 424–425 equations, 41–44, 54, 60, 606,
effects, averaging, 432–435 628–630, 634–638, 640, 641
wavelength œ, 425–430 Reynolds fluxes, 42
description, types, 412–414 wind shear, 42
ground-based instrumentation, 143 RHI. See Range-height indicator (RHI)
lidar, 479 (see also Lidar) Rhine Valley, 11, 225, 227, 229, 237, 246, 257,
measurement system, 410, 412 262, 278–283
meteorological measurements, 273 Richardson number, 165, 194, 608, 612
polarimetric and dual-wavelength radar, Ridge height, 65, 67, 91, 224, 235, 245
399 Riming, 314, 335, 357–360, 366, 372, 374,
pulsed remote sensors, 421–424, 431 375, 384, 396, 397, 456
surface-based meteorological and Riviera Valley, 52, 69, 70, 84, 599, 618, 619,
turbulence sensors, 499 639–641
techniques, classification, 513 Rocky Mountain National Park, 494
uses, 99, 141 Rocky Mountains
Representativeness of observations, 656 Chinook, 255
Research scientists, 696, 698, 701–702, 710 diurnal flow patterns, Central Colorado, 96
Research Support Desk (RSD), 707–708 downslope windstorms, 157
Reservoirs, 26, 136, 178–180, 185–188, 500 eastern slopes, 10
Residence time, 300, 314, 368 front range, 491
Residual layer, 264–265, 498 measurement techniques, 143
Resolution plain wind system, 87
beam, 422 Utah, 451
finer and larger scales, 635–637 WRF model, 637
grid and accuracy, 546–547 Rosalian mountain range, 72
high Rossby number, 122, 127, 203, 310–311
advantages, 393 across-barrier, 122, 203
digital elevation databases, 76 Rossby radius of deformation, 195, 203, 206,
2D nonlinear mesoscale model, 172 310–311
Doppler-lidar scan data, 505–506 Rotational acceleration term, 232–234
measurements, 510 Rotor clouds, 138, 139, 142
mesoscale model, 160, 324, 593 Rotors
numerical models, 155 atmospheric recirculations, 638–639
NWP forecast, 18, 173 gravity waves, 593
prediction system, 683 jump-like rotors (type 2 rotors), 138
visible satellite image, 140, 144 lee waves
horizontal and vertical classification, 138
coordinate system, 599–600 clouds and blowing dust, 138–139
grid aspect ratio, 598–599 models and forecasting, 152–155
terrain and flow representation, 595–597 observations, 140–145
HSRL, 470, 503 structure, trapped, 137, 138
large-scale operational models, 619 theory, 145–152
744 Subject Index
Rotors (cont.) Raman, 470

lidar vertical-slice scans, 483 signal pulse, 421–422
reverse-flow, 466, 467, 481 targets, 421, 425–429, 443, 448, 462–463,
type I, 138, 149–152, 206, 207 468, 495
type II, 138, 153, 206 transmission properties, light, 469
Roughness length, 70–71, 625 transmitted sound pulse, 435
RTMA. See Real-Time Mesoscale Analysis transverse dimension and properties, 425
(RTMA) uniform field, 432
RWPs. See Radar wind profilers (RWPs) volume, 425, 427, 429–430, 470
Science and Nowcasting of Olympic Weather
for Vancouver 2010 (SNOW-V10),
S 704, 707
Saharan dust, 234, 250, 476 Scorer parameter, 132, 137, 138, 145, 149, 163
Salt Lake Valley, 43, 58, 69, 423, 593, 598, Sea breezes, 62, 72, 85, 255, 271, 272, 275,
601, 611, 613, 614. See also Great 277, 278, 415, 485, 486, 490, 491,
Salt Lake 673–674
dead zone, 423 Seasonal variability, 239
lidar observations, 598 Secondary circulations, 84
mesoscale models, 593 Secondary PM, 276
Utah’s, 43, 58, 596 Secondary pollutants, 264–265
VTMX, 611, 614 Seeder-feeder process, 315, 316, 335, 347,
Wasatch Mountains, 69 348, 364, 369–373, 503
San Joaquin valley, 269 Seiches, 60, 62, 93
Santa Ana, 10, 157, 162, 220, 257 Shading, 37, 52, 177, 266, 416, 491, 508, 563,
SAR. See Synthetic aperture radar (SAR) 623–625, 639, 640
Satellite self shading, 623, 624
applications, 480 topographic, 623–625, 639
description, 16–18 Shallow foehn, 227, 238, 250, 251, 253. See
high-resolution visible image, 140, 144, also Foehn
680 Shallow slope, 43, 48, 55, 58, 61
images, lee-wave cloud formations, 137 Shallow water equations, 125, 134, 163, 189
radiometers, 410, 420 Shear
SSMI, 327 cross-barrier, 382
Saturation cross-slope, 58
mixing ratio, 349 cyclonic, 234
point, 392 density strength, 356
surface, 297 different densities, 237
water, 349, 352, 354, 369, 372, 374, 394 directional, 491
Scale interaction, 13–14, 238–239 flow patterns, 486
foehn, 238–239 forward, 170, 570
forecast funnel, 13–14 induced turbulence, 194–195, 370
turbulence, 42, 53, 69, 74, 75, 264 lidar-data, 488–489
Scale-similarity model, 630–634 reverse, 165
Scanning microwave radiometer (SMR), stress, 626, 627
502–504 TKE, horizontal wind, 614
Scattering vertical and destroyed, 57
advantages/limitations, 495 wind, 27, 42, 53
angles, 438 zero and non-uniform stability, 132
atmospheric, 467–468 Shear production, 53, 57, 70, 437, 487, 612,
Bragg, 428–429, 443, 448 613
cross section ¢ C definition, 430 Shelikof strait, 175, 176
Doppler velocity, 446 Shock jets, 201
droplets, particles/ molecules, 426–428 Side-lobe, 430, 431, 439, 443, 445, 449, 456,
linear polarization and spherical, 446–447 457, 462–465, 467, 468
Subject Index 745
Sierra Nevada Snake River Plain, 6, 69

inland, 505 Snow eater, 220
mountain waves and rotors, 142–143 Snow density processes
north-south extent, 568 ageing, 383–384
orographic storm, 367, 373 compaction, 383
Owens Valley, 139, 155, 269 degree of riming, 383–386
role of discretization errors, 571 fragmentation, 383
slopes, 270 habit, 383
T-REX, 667 Snowmelt, 7, 8, 24, 26, 246
vertical cross sections, 479 Snowpack, 3–4, 8, 15, 26, 356, 377, 384, 418,
Sierra Rotors Project (SRP), 143 504, 514
Sigma coordinate, 171, 560, 598, 599, 638 Snow ratio, 382, 384–386. See also Snow
Signal-to-noise ratio (SNR), 412, 432, density
435–437, 442 Snowstorms
Sill, 134, 174, 178, 180, 187, 191 avalanches, 4, 5
Similarity theory, 57, 610, 612, 625, 627, 633, economic losses, 3
634, 637, 640 recreational skiing, 4–6
Sine-wave, 450, 515, 517–519 removal, 3, 4
Sky view, 52, 620 Sodar
SLEVE coordinate, 554, 555, 562 attenuation, 429–430
Slope angle, 44, 46, 52, 55–61, 599, 618, 620, backscatter record, Great Salt Lake basin,
634 440
Slope aspect, 52, 620, 622 beamwidth and sidelobe patterns, acoustic
Slope flows frequencies, 431
analytical models description, 435–437
hydraulic flow, 44–45 Doppler minisodar, 441
profile, 44 high-resolution profiles, 496
atmosphere, 36 instrument types, description, 413–414
channeling, 282 measurements, 278, 442
development and cross-slope wind, 61 operating characteristics, 439
laboratory tank models, 45–46 phased-array, 439
large-scale description, 39 small-scale turbulence, 513–514
layer, 49 temperature fluctuations, 437–438
LES models, 43–44 transmitted frequencies, 435
RANS equations, 41–43 Soil moisture, 26, 37, 47, 52, 76, 97, 616–620,
skin flows, 40–41 639, 643
soil moisture, 52 Solar Differential Optical Absorption
steady-state analytical, 59 Spectroscopy, 420
turbulence role, 53–54 Solar Occultation Flux (SOF), 420
Slope shadowing, 43 Solenoid theory, 232–234
Slope wind system Southerly Buster, 196, 200–201
anabatic, 39 Spatial discretization, 538–540
drainage, 39 Spatial filtering, 628, 629, 638
katabatic, 39 Spatial variability, 47, 51, 52, 90, 276, 451,
Smooth 597, 616, 618, 623
boundary conditions, 627 CBL heights, 266
larger-scale patterns, 661 description, 618–619
lee-wave field, 152 environment, 612
momentum and scalar equations, 605 mountain atmosphere, 411–412
mountains, 130 pollution fields, 273
numerical model solution, 670 soil moisture, 620
terrain, 605 surface-based Doppler lidar scan data, 490,
transition, 555, 605, 638 491
type II rotor, 138, 139 terrain-induced wind shear, 486
746 Subject Index
Spatial variability (cont.) Subcloud layer, 11, 496–497

turbulence properties, 612 Subcritical, 134–136, 160, 163, 174, 188–189,
vertical-slice scan, 488 207, 612
Specialized forecast, mountains Subjective forecast tools
avalanches, 26–27 conceptual models, 22–23
aviation, 27–28 geographic familiarization, 19
description, 24–25 pattern recognition and climatology
fire weather forecasts, 25 mesoscale modeling system,
hydrological runoff forecasts, 26 20, 21
Spectral broadening, 349–350 precipitation distribution, 14, 20
Spill-over precipitation, 363 United States, 20, 21
Spinup, 600 rules of thumb, 19–20
SPol radar, 447, 456 Subrotors, 144, 151, 158
Spray line, 235 Subsidence
SPROF (S-band, profiling radar), 452–454 adiabatic temperature, 192–193
SRP. See Sierra Rotors Project (SRP) air masses, 224
Stability limits, 537, 541–542, 602 cold-air advection, 84
Stable boundary layer (SBL), 43, 52, 53, 126, daytime, 72
264–265, 468, 512, 513, 634. See large-scale, 374
also Boundary layer lee-side, 362–364
Stable core, 82, 83 mechanisms, 68
Stable layer, 36, 54, 80, 82–84, 158, 160, 163, sidewalls and compensating, 67
164, 170, 207 valleys and mountain waves, 305, 311
Stable stratification, 84, 127, 169, 180, 230, Supercooled drizzle, 350
264, 266, 437 Supercooled liquid water, 349, 359, 371–372,
Stable unblocked flow, 303–305 503, 504
Staggering, 3, 547–548, 560, 561, 568, 626, Supercritical, 68, 134–136, 158–160, 163–166,
627 170, 174, 184, 186–189, 207
Stagnant layer, 148, 149, 167, 168, 175–176, Surface drag, 61, 625, 627
180, 182, 186, 189, 207 Surface energy budget
Stagnation, 72, 133, 165, 275, 382 characterization, 50
Standing waves, 235 cloudiness, 92
St. Andrew’s Cross, 580 evening transition phase, 79–80
Static displacement, 235 heat transfer processes, 67
Stationary hydraulic jump, 233–234 imbalance, 50
Statistical analysis, 2, 98, 230 landforms, 36
Statistical methods, 19, 168, 245, 686 latent heat fluxes, 47, 50, 67
Steady state magnitudes, simulated surface heat fluxes,
barrier jet, semi-geostrophic, 202–203 510
buoyancy term and surface friction, 55 morning transition phase, 81–84
characteristics, equilibrium, 39–40 radiative variables, 51
equations, atmospheric water, 321 subsequent influence, 620
equilibrium flow, 57 temporal and spatial variations, 47
horizontal and vertical discrete group, 551 Surface heat flux, 45, 49, 50, 70, 71, 510, 512,
hydraulic theory, 184 593
linear model, orographic precipitation, 320 Swiss Plateau, 235, 252
turbulent flow, 625 Synoptic
Steep slope, 7, 46, 58, 74, 599 pressure field, 223, 641
Stochastic backscatter model, 46, 633 situations, 240, 241
Storm Prediction Center, 709 weather analysis, 696
Storm-storm interactions, 460 weather pattern, 238
Streamflow, 500 Synthetic Aperture Radar (SAR), 174, 175,
Streamflow modeling, 295 196–198, 201, 205
Subject Index 747
T Terrain smoothing, 595, 601, 605

TAF. See Topographic amplification factor Terrain transform metric terms, 557–559, 582
(TAF) Tethersonde, 45, 59, 278, 419, 611
Taku wind, 157, 162 Thermal low, 162, 271, 272
Tangent linear model, 576–577, 580, 581 Thermally driven wind systems, 39, 271, 283
Taylor-Goldstein equation, 132 Himalayas, 36
Telescope, 468, 480 Thermodynamics
Temperature excess, 39, 40, 42, 46, 55 dynamic effects, 509
Temperature gradient energy and continuity equation, 41–42
development, 251 mesoscale models, 607
effects, near-surface, 48 moist ascent, 297
horizontal pressure, 504–505 processes and role, 220–221
reversing, 64 up-and down-valley flows, 66
sensible heat flux, 65 zonal momentum equations, 557
valley stations and Alpine ridge site, 245 Three-dimensional
vertical, 560, 561, 565–567 advection and sedimentation, 387
wind, 547 barrier impacts, 310–313
Temperature inversion foehn, 238–239
air pollution, valley, 262 numerical models, 167
basins, 62, 91–92 NWP model, 172
break-up, surface based, 265 shear production, 613
Colorado, 82 structure, 227
depth, 58 valley topography, 50
downslope flow, 59 vertical flow structure, 143
formation and dissipation, 36 volume scans, 482
ground-based stable layer, 80 wind vector, 460, 464
multi-day, 93 Time differencing, 465, 540–542
nighttime temperature, 90 Time-height cross sections, 440, 475, 491, 493,
nocturnal temperature, 92 499, 505, 508
nocturnal, 78–79 TKE. See Turbulent kinetic energy (TKE)
persistent/multi-day, 6, 7, 12, 84, 85, Top boundary, 67, 600, 602
92–93, 100, 174, 264, 660–661 Topographical map, 220, 221
persistent wintertime, basins, 92–93 Topographic amplification factor (TAF), 66,
plateau, 94 67, 79, 82, 89, 90, 98
stable layers, 163 Topographic effects, 20, 47, 52, 73–75, 283
two-dimensional simulations, 152 Topographic shading, 639. See also Shading
Terminal velocities of precipitating Topography
drops/particles, 446 complex, 36, 506
Terra incognita, 637, 643 DIAL systems sensing ozone, 494
Terrain-following coordinates, 553, 556, fine-scale, 25
559–562, 564, 565, 599, 600 flow types, 18
Terrain-following flow models, 171, 363–365, gap flows, 180, 189
367–368, 562 LES, 634–635
Terrain forced ascent, 296–300, 302 linear Boussinesq flow, 549
Terrain-induced Rotor Experiment (T-REX) MM5 simulations, 205
clouds and blowing dust, 139 mountainous, 443, 565
description, 143 multi-scale, 559
downslope windstorms observation, 172 one-dimensional version, shallow water
NRL COAMPS model, 155 equations, 134
orographic flow, 126 Owens-Valley metric box, 568, 569
Owen’s Valley, 168, 615 radiation, 620
Sierra Nevada mountains, 141, 666–667 radiation field, 51
small-scale intense subrotors, 144 resolution, 191
Terrain-induced wind shear, 485, 486 resolution and flow representation, 595–597
748 Subject Index
Topography (cont.) Turbulent kinetic energy (TKE)

shading, 623–625 ABL simulations, 630
smaller scale, 52, 98 Alpine foehn events, 229
turbulent mixing, 611 Doppler-lidar radial velocity, 515
Tracer plume, 476 profile, 40, 57
Trajectory analysis, 223 prognostic equation, 630
Trajectory calculation, 224, 246 shaded, 571
Transport slope flows, 53–54
air pollution, 414 turbulent length scale, 609
and diffusion, 38 values and accuracy, 510
high-altitude advantageous, 556 vertical, distribution, 70
horizontal, 82 Turbulent mixing, 124, 151, 180, 263, 264,
long-range moisture, 72–73, 85 392, 499, 512, 563, 569–570,
mixing, 71 610–613
rotor clouds and dust, 138–139 Turbulent momentum flux, 42–43, 57, 607
synoptic-scale systems, 192
turbulence and scalar, 604
vertical and horizontal, 493 U
vertical cross section, pollutant, 272 Undersampling, 413, 421
Transportation Unified Model, 157, 158, 592
aviation and ground, 16 Updrafts
cable, 248 ascending air, 502
snowstorms, 3 dendritic growth zone, 354
Trapped lee waves. See Lee waves downdrafts, 369, 460, 462
T-REX. See Terrain-induced Rotor Experiment driven, 357–358
(T-REX) lee-wave, 503–504
Tributary flows, 62, 73 orographic, 372
Tributary valleys, 63, 67, 68, 73–74, 179, 238, speckled area, 318
245–246, 499, 595 terrain-induced, 393
Truncation error, 171, 543–545, 555, 556, 559. troposphere, 368
See also Discretization error Upper Rhine Valley, 11, 278–283
Turbulence Upslope flow, 39, 40, 43, 45–47, 52–56, 66,
first-order closure, 607, 609 82, 83, 90, 94, 96, 99, 199, 263, 265,
length-scale equations, 609 278, 292, 299, 307, 310, 319, 332,
LES (see Large-eddy simulation (LES)) 372, 379, 452, 503, 624
local schemes, 610 correlation with precipitation, 360
Mesoscale Alpine Programme, 69, 639 Up-valley winds, 27, 281, 283, 415, 566, 639,
non-local schemes, 607, 608, 610, 611 640
1.5-order closure, 609 Up-valley flows, 63, 68, 78, 98, 640
1.5-order TKE, 609–611, 630, 632 Kali Gandaki Valley, 67–68
parameterization, 42, 43, 54, 75, 99, 265, San Joaquin Valley, 68, 78, 269
510, 512, 606–615, 625, 635, 637 Wipp Valley, 67–68
(see also Parameterizations)
RANS (see Reynolds-averaged Navier-
Stokes (RANS)) V
role in cellularity and precipitation VAD. See Velocity-azimuth display (VAD)
enhancement, 369, 370 Valley
second-order atmosphere, 36
advection, 546, 549, 550, 558, 559, bend, 81
571–573 channeling effect, 225
closures, 607, 609 deep, 52, 81, 83, 265, 330, 615
diffusion, 563–564, 567 diurnal wind systems, 94
velocity variance, 57, 437, 516 Doppler-derived down valley flow, 314
Turbulent fluxes, 49–52 downslope flows, 58, 61
Subject Index 749
drainage W
ASCOT project, 72 Warm cloud microphysics, 349
Rosalian mountain range, 72 Warm slope zone, 66
exit jet Wasatch Mountains, 20, 22, 69, 258, 294, 303,
Salt Lake Valley, 69 314, 317, 323, 324, 365, 367, 369,
Snake River Plain, 69 384, 392
Wasatch Mountains, 69 Wasatch winds, 157, 324, 384
flows, 69–76 Waterfall theory, 232–234
foehn, 253 Water vapor flux (Fw ), 293, 296–300, 616
geometry, 66, 75, 82 Wavelength
Lower Fraser, 262, 265, 275–278 backscatter measurements, 478
narrow, 564, 565 DIAL advantages, 470–471
nocturnal drainage flow, 81 frequency, 421, 518
outflows, 484, 499–500 grid points, 546, 551
Pitt Lake, 414–416 horizontal, 129, 130, 137, 147–148, 153
Rhine, 227, 229, 237 lidars, 468–469
Rhone, 183 multiple, 445, 447
Riviera (see Riviera Valley) NWP models, 153
rotor formation, Owens Valley, 138, 139 Nyquist, 596
small-scale ridges, 369 polarimetric and dual, 399
three-dimensional, topography, 50–51 streamlines, sinusoidal mountains, 130
Upper Rhine, 278–283 transmitted remote-sensing pulse, 422
Willamette, 327, 332 transmitted signal
wind system, 63–78 attenuation, 429–430
Wipp, 179, 186, 188, 190, 191, 243 backscatter cross section, 426
Valley winds clear-air, 428–429
anomalous, 76, 78 equation, received power, 425–426
cross-valley flows, 72, 639, 640 inertial subrange and structure function,
vs. slope winds, 78–84 429
Vancouver, British Columbia, Canada, 414, Rayleigh scattering, 427
656 scattering, 426
Vegetation, 7, 47, 50, 56, 591–592, 615–618, vertical, 129
625, 642, 643 Waves
Velocity-azimuth display (VAD), 244, 413, amplitude
450, 451, 480, 490, 491, 496, 503, downslope bora flow, 162
515 gravity wave motion, 128
Vertical growth, 128
coordinates, 41, 168, 171, 552–556, 559, lee waves, 145–149
582, 599–600, 605 linear theory, 164
diffusion, 510 mountain, 132
resolution, 54, 154, 435, 439, 491, 510, phase, 155
550, 556, 594–601, 604, 643, 657 trapped, 150
slice scans, 457, 479, 481, 483, 486, 488, breaking
491, 506, 515 asymmetry, 127
velocity variances, 57, 437, 515 flow blocking and gravity, 323
Vertical Transport and Mixing (VTMX), 466, hydraulic jumps, 238
481, 490–492, 506, 593, 598, 611, isentropes, 165
614, 638 low-level, 133, 138, 152
Virtual tower, 491 non-dimensional mountain height,
Visibility 165
in winter orographic storms, 358 self-induced critical layer, 157, 163,
Volcanic ash, 473 170
VTMX. See Vertical Transport and Mixing type II rotor, 206–207
(VTMX) upper-level, 570
750 Subject Index
Waves (cont.) low level shooting flow and wave-induced

equation, 123–125, 128–130, 132, 164, stagnant layer, 167
166, 184, 238, 538, 541, 542, 562, parameterization, 42
578 terrain-induced, 485
Weather radar, 16, 410, 443, 450, 451, 462, turbulence, 466
463, 468, 489, 708 vertical-slice scans, 491
Weather Research and Forecasting (WRF) Windstorms
model amplitude, 667, 668
horizontal grid spacings, 597 Boulder, 254
horizontal resolution, 596 downslope, 155–173, 180, 181, 246
mesoscale atmospheric model, 295 foehn-related, 220
MM5, 391 high-amplitude mountain waves and gap
nested grid configurations, 637 flows, 10
two-way nesting, 604–605 jumplike structure, 487
vertical velocity, 602 structure and dynamics, 456–457
Wegener–Bergeron–Findeisen process, Windward precipitation bias, 392
352 Winter Olympics, NWP. See Olympic Winter
Wet-adiabatically, 221 Games
Widmer Index, 242–244 Wipp Valley, 68, 179, 186, 188, 190, 191, 243,
Wildfire behavior, 9–10 497
Wind energy, 511, 593 WRF model. See Weather Research and
Wind shear Forecasting (WRF) model
aircraft glide paths, 487 WSR-88D. See Radar, WSR-88D
anemometer-based and radar-based,
488–489
flow patterns, 486 Z
horizontal vorticity budget, 152 Z-diffusion, 564–566
lidar-data, 487 Zonda, 10, 157, 257, 258

(De Wekker&snyder) 2013 Mountain Weather Research and Forecasting PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

(De Wekker&snyder) 2013 Mountain Weather Research and Forecasting PDF

Uploaded by

Copyright:

Available Formats

Mountain Weather Research and Forecasting

Springer Atmospheric Sciences

For further volumes:

Mountain Weather Research

ISBN 978-94-007-4097-6 ISBN 978-94-007-4098-3 (eBook)

Library of Congress Control Number: 2012940730

© Springer Science+Business Media B.V. 2013

Printed on acid-free paper

Springer is part of Springer Science+Business Media (www.springer.com)

1 Mountain Weather Prediction: Phenomenological

9 Mesoscale Modeling over Complex Terrain: Numerical

Author Index.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 717

Subject Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 725

Simon Vosper Met Office, Exeter, UK

Michael P. Meyers and W. James Steenburgh

M.P. Meyers ()

F. Chow et al. (eds.), Mountain Weather Research and Forecasting, 1

Contemporary mountain weather forecasting involves the integration of geophysical

1.2 Phenomenological Challenges in Complex Terrain

accessories and vacations (US $6 billion in Canada) (International Snowmobile

Fig. 1.3 Recreational skiing in Colorado (Courtesy of Meyers)

1.2.3 Floods, Flash Floods and Debris Flows

1.2.5 Wildfire Behavior

was burnovers (40%), “a situation where personnel or equipment is caught in an

1.2.6 Severe Windstorms Created by High-Amplitude

1.2.7 Severe Convective Storms

Mountainous terrain can have an indirect impact on tornado development by

1.2.8 Cold-Air Pools and Air Quality

1.3 The Contemporary Forecast Process in Complex Terrain

At its core, weather forecasting is a scientific endeavor involving hypothesis

1.3.1 Scale Interaction: The Forecast Funnel

1.3.2 Objective Tools

1.3.2.1 Surface-Based Observations

Fig. 1.6 WSR-88D coverage

1.3.2.5 Numerical Weather Prediction (NWP)

1.3.3 Subjective Tools: The Human Element

1.3.3.1 Geographic Familiarization

Forecasting in complex terrain requires a familiarization of the geography to provide

1.3.3.2 Rules of Thumb

A “rule of thumb” is an easy-to-remember forecasting method based on experience

1.3.3.3 Pattern Recognition and Climatology

1.3.3.4 Conceptual Models

A conceptual model is a physically-based model detailing a simplified represen-

1.4 Forecasting in Complex Terrain

1.4.1 Complexity of Mountain Weather Forecasting

objective forecasting tools used in mountainous terrain as discussed in the previous

1.4.2 Specialized Forecast Examples

1.4.2.1 Fire Weather Forecasts

1.4.2.2 Hydrological Runoff Forecasts

Hydrologists have a variety of responsibilities over complex terrain. One of their

1.4.2.3 Avalanche Forecasts

Similar to hydrologists, avalanche forecasters must maintain an active weather

1.4.2.4 Aviation Forecasts

Aviation customers in the mountains include commercial industry, private pilots,

This chapter has summarized some of the phenomenological challenges of mountain

Acknowledgments We thank the participants in the 2008 AMS/COMET/MSC Mountain Weather

Ludwig, F. L., D. K. Miller and S. G. Gallaher, 2006: Evaluating a hybrid prognostic/diagnostic

Dino Zardi and C. David Whiteman

F. Chow et al. (eds.), Mountain Weather Research and Forecasting, 35

The scientific study of diurnal mountain winds has important practical

Meteorological Society’s 1988 workshop on mountain meteorology, as later

2.2 The Slope Wind System

Fig. 2.2 Typical wind and a