Robust adaptation decisions amid

climate change uncertainties

Suraje Xembu Rauto Dessai

A thesis submitted for the Degree of

Doctor of Philosophy
in the
School of Environmental Sciences
University of East Anglia
Norwich, UK

November 2005

This copy of the thesis has been supplied on condition that anyone who consults it is understood to
recognise that its copyright rests with the author and that no quotation from the dissertation, nor any
information derived therefrom, may be published without the authors prior, written consent.
Abstract

Projections of future climate change are plagued with uncertainties, which cause
difficulties for planners taking decisions on adaptation measures. This thesis presents an
interdisciplinary framework that allows the identification of adaptation strategies that
are robust (i.e. insensitive) to climate change uncertainties. A range of tools and
methods are used to explicitly quantify uncertainties in climate change projections.
These include scenario analysis, Monte Carlo simulations and the pattern-scaling
technique applied to simple and complex climate models.

The framework is applied to a case study of water resources management in the East of
England, more specifically to the Anglian Water Services 25 year Water Resource Plan
(WRP). The institutional setting where adaptation decisions occur is explained.
Adaptation options are elicited from decision-makers and linked to changes in the
climate. Three sets of analyses are performed on the framework: (1) sensitivity analysis
determines which are the most influential variables of the framework from an adaptation
perspective; (2) uncertainty analysis assesses whether currently planned adaptation
strategies are robust against climate change uncertainties, assuming risks can be
quantified; and (3) robustness analysis assesses adaptation strategies against climate
change and other uncertainties, assuming the uncertainties cannot be quantified using
probabilities.

Water resources are found to be sensitive to uncertainties in regional climate response

from General Circulation Models and dynamical downscaling, in climate sensitivity and
in climate impacts. Aerosol forcing and greenhouse gas emissions uncertainties are also
important, whereas uncertainties from sampling strategies, ocean mixing and the carbon
cycle are not. The uncertainty analysis estimates the conditional probability of
additional water required for two Water Resource Zones (WRZs) in the Anglian region.
This is performed for the 2030s under the SRES A2 scenario and using the most skilled
models. At the 95% confidence level the bounds of additional water required ranged
from -6.67 to 26.34 Megalitres/day (Ml/d) for the East Suffolk & Essex (ES&E) WRZ
and from -24.30 to 111.81 Ml/d for the Ruthamford WRZ. This represents an

1
uncertainty range of 33.01 Ml/d (33% of demand) for the ES&E WRZ and 136.11 Ml/d
(39% of demand) for the Ruthamford WRZ.

Despite these large uncertainties, Anglian Waters WRP remains robust to the climate
change uncertainties sampled because of the adaptation options being considered (e.g.
new reservoirs) and because the climate model used for planning (HadCM3) predicts
drier conditions than other models. According to the robustness analysis, the WRP is
vulnerable to some combinations of uncertain factors (e.g., if very high climate
sensitivity is combined with low aerosol forcing and large decreases of precipitation per
degree of global warming). This suggests that the current WRP needs some adjustment
for it to be robust to uncertainties. This research raises the question of how much
certainty is required in climate change projections to justify investment in adaptation
measures.

2
Acknowledgments

There are so many people I need to thank that I am bound to forget someone and for that
I apologise. Firstly I would like to thank my supervisory panel for all the guidance they
have given me over the past four years. I am eternally grateful to my main supervisor,
Mike Hulme, for the freedom he gave me to pursue many different avenues and for the
trust he had in my completion of this thesis. I would also like to thank my other
supervisory panel members, Sarah Raper for her guidance in using MAGICC, a simple
climate model and Kevin Hiscock for introducing me to the water world, in particular to
groundwater. Special thanks go to Mike Hulme, Ida Wilson, Katharine Vincent and
Kevin Hiscock for proof-reading and commenting on various chapter drafts. Marisa
Goulden is thanked for helping me put this thesis together during the final stages.

Family, friends and colleagues have always supported me throughout the last four years.
My parents and my sisters have always given me unremitting support for whatever I
wanted to do and the PhD was no exception. Ida Wilson has been crucial in keeping me
sane, healthy and happy for over two years. Numerous friends have supported and
inspired me throughout the years, in particular I would like to thank my ex-house mates,
Annika Swindell, Glen Richardson and Dario Muhamudo; my office mates, Marisa
Goulden, Kat Haynes, Ali Colls, Katharine Vincent; my squash partner Esteve Corbera;
my kitesurfing buddy Tom Lowe; my circuit training pals Ben Murray and Kirsty
Southerland; and Bernardo Carmo for encouraging me to come to the UK. Several
colleagues at the Tyndall Centre in UEA have been key to providing the critical mass to
make this PhD happen, in particular I would like to thank Neil Adger, Nick Brooks,
Alex Haxeltine, Jonathan Khler, Xianfu Lu, John Turnpenny, Emma Tompkins and
Rachel Warren. Besides office and UEA colleagues, several other individuals have
contributed to this thesis in a number of ways, such as providing data or influencing
ideas. For that I would like to thank: Robert Willows and Nick Reynard (Environment
Agency and CEH), Myles Allen, Dave Stainforth and Jamie Kettleborough (Oxford
University and RAL), Frans Berkhout and Julia Hertin (formerly at SPRU), Peter
Challenor, Jim Hall and Tony OHagan (Tyndall Uncertainty project), Natalia
Andronova, Chris Forest, Jonathan Gregory, Retto Knutti, James Murphy, Michael

3
Schlesinger, Richard Tol and Tom Wigley (for climate sensitivity data), Roger Jones
(CSIRO) and James Risbey (Monash University).

The case study would not have been possible without the collaboration of Anglian
Water Services. In particular I would like to thank Gerry Spraggs (Anglian Water) and
Geoff Darch (Atkins) for providing me with Water Resource Plan documents and
sensitivity data respectively as well as advice. Tom Wigley and Sarah Raper are thanked
again for allowing me to use their simple climate model, MAGICC. I would like to
thank folks at the Climatic Research Unit, in particular Mike Salmon, Tim Osborn,
Craig Wallace and Marie Ekstrom for computing advice and data. Ole Bssing
Christensen at DMI is thanked for advice on PRUDENCE data, which was funded by
the EU through contract EVK2-CT2001-00132. Ulrich Cubasch is thanked for global
warming data from the IPCC TAR. Interpolated GCM data were obtained from Richard
Jones at the Hadley Centre via Xianfu Lu; both are thanked.

Finally, I would like to extend my deepest gratitude to the Foundation for Science and
Technology (Fundao para a Cincia e a Tecnologia) in Portugal for funding this
research under grant SFRH/BD/4901/2001.

4
Table of contents

LIST OF ACRONYMS .................................................................................................... 8

CHAPTER 1. INTRODUCTION ................................................................................... 10
1.1 Climate change...................................................................................................... 11
1.2 Adaptation to climate change................................................................................ 13
1.3 Sensitivity, uncertainty and robustness ................................................................. 14
1.4 The Anglian region of the UK: climate and water................................................ 16
1.5 Roadmap ............................................................................................................... 17
CHAPTER 2. LITERATURE REVIEW ........................................................................ 19
2.1 Greenhouse gas emissions and its drivers............................................................. 19
2.2 Global climate ....................................................................................................... 20
2.3 Global impacts ...................................................................................................... 22
2.4 Regional climate.................................................................................................... 23
2.5 Regional/local impacts .......................................................................................... 26
2.6 Adaptation and planning ....................................................................................... 27
2.7 Integrated assessment of climate change .............................................................. 29
2.8 Other areas ............................................................................................................ 31
CHAPTER 3. METHODOLOGICAL FRAMEWORK................................................. 33
3.1 Adaptation assessment approaches ....................................................................... 33
3.1.1 The IPCC approach ........................................................................................ 33
3.1.2 The resilience/vulnerability/human development approach .......................... 35
3.1.3 The risk approach........................................................................................... 36
3.1.4 Exploratory modelling and robustness analysis............................................. 39
3.1.5 The approach in this thesis............................................................................. 41
3.2 Managing climate change uncertainties ................................................................ 45
CHAPTER 4. ASSESSMENT OF UNCERTAINTIES AT THE GLOBAL LEVEL ... 50
4.1 World development paths and emissions of greenhouse gases............................. 50
4.1.1. Mitigation and stabilisation scenarios........................................................... 56
4.2 Climate sensitivity................................................................................................. 59
4.3 Using a simple energy balance model to propagate uncertainties at the global
scale............................................................................................................................. 60
4.3.1 Experimental setup......................................................................................... 61
4.3.2 Results at the global level .............................................................................. 65

5
 4.4. Summary .............................................................................................................. 73
CHAPTER 5. ASSESSMENT OF UNCERTAINTIES AT THE REGIONAL LEVEL74
5.1 Description of the climate of the Anglian region.................................................. 74
5.2 The history of climate scenario development for the Anglian region................... 85
5.3 Applying the pattern-scaling technique ................................................................ 93
5.4 General circulation models ................................................................................... 96
5.5 Regional climate models ..................................................................................... 101
5.6 Summary ............................................................................................................. 109
CHAPTER 6. WATER MANAGEMENT IN THE ANGLIAN REGION .................. 110
6.1 A brief history of water policy in the UK and the Anglian region ..................... 110
6.2 Water resources planning.................................................................................... 117
6.3 Climate change and water resources planning .................................................... 120
6.4 Are UK water managers adapting to climate change? ........................................ 124
6.4.1 AMP3 and climate change ........................................................................... 125
6.4.2 AMP4 and climate change ........................................................................... 126
6.4.3 A preliminary assessment ............................................................................ 133
6.4.3.1 Uncertainty............................................................................................ 133
6.4.3.2 Complexity............................................................................................ 134
6.4.3.3 Timescale .............................................................................................. 135
6.4.3.4 Financing............................................................................................... 135
6.5 Adaptation options .............................................................................................. 136
6.5.1 East Suffolk & Essex Water Resource Zone................................................ 145
6.5.2 Ruthamford Water Resource Zone .............................................................. 150
6.6 Linking climate change to water resources decisions......................................... 154
6.7 Summary ............................................................................................................. 158
CHAPTER 7. SENSITIVITY, UNCERTAINTY AND ROBUSTNESS ANALYSIS 160
7.1 Sensitivity analysis.............................................................................................. 160
7.1.1 Natural multi-decadal climate variability .................................................... 162
7.1.2 Greenhouse gas emissions............................................................................ 166
7.1.2.1 Mitigation/stabilisation scenarios ......................................................... 171
7.1.3 Climate sensitivity........................................................................................ 173
7.1.4 Other climate parameters ............................................................................. 175
7.1.5 Regional climate response............................................................................ 176
7.1.5.1 Dynamical downscaling ........................................................................ 180

6
 7.1.6 Climate impacts............................................................................................ 185
7.1.7 Sampling strategy......................................................................................... 186
7.1.8 Summary ...................................................................................................... 187
7.2 Uncertainty analysis ............................................................................................ 189
7.3 Robustness analysis............................................................................................. 197
7.3.1 Summary ...................................................................................................... 203
CHAPTER 8. DISCUSSION AND CONCLUSION ................................................... 204
8.1 Limitations .......................................................................................................... 204
8.1.1 Linear simplicity versus non-linear complexity........................................... 204
8.1.2 Climate impacts interface............................................................................. 207
8.1.3 Adaptation planning ..................................................................................... 209
8.2 Policy and practical implications ........................................................................ 212
8.2.1 Regulatory system: prediction, investment and organisational culture ....... 212
8.2.2 Water industry: searching for guidance ....................................................... 218
8.2.3 Climate prediction and the future of climate scenario construction ............ 222
8.2.4 Adaptation planning ..................................................................................... 226
8.2.5 Post-normal science and interdisciplinarity ................................................. 229
8.3 Future research .................................................................................................... 231
8.4 Concluding remarks ............................................................................................ 232
APPENDIX 1. OPTIONS IN THE WATER RESOURCES PLAN ............................ 234
APPENDIX 2. QUESTIONNAIRE.............................................................................. 243
LIST OF REFERENCES .............................................................................................. 252

7
 LIST OF ACRONYMS

AGCM Atmospheric General Circulation Model

AMP Asset Management Plan
AOGCM Atmosphere/Ocean General Circulation Model
APF Adaptation Policy Framework
AWS Anglian Water Services Ldt.
CCIRG Climate Change Impacts Review Group
CDF Cumulative Distribution Function
CET Central England Temperature
DEFRA Department for the Environment, Food and Rural Affairs
DJF December, January, February
EA Environment Agency
EBM Energy Balance Model
EOETS Ely-Ouse to Essex Transfer Scheme
ES&E East Suffolk and Essex
EWP England and Wales Precipitation
FORWARD Forecasting of Water Resource and Demand
GCM General Circulation Model (or also Global Climate Model)
GHG Greenhouse gases
IAM Integrated Assessment Model
IPCC Intergovernmental Panel on Climate Change
JJA June, July, August
PDF Probability distribution function
PRUDENCE Prediction of Regional Scenarios and Uncertainties for Defining
European Climate change risks and effects
MAGICC Model for the Assessment of Greenhouse-gas Induced Climate Change
MAGLOOP Probabilistic version of MAGICC
MAM March, April, May
NPC Net Present Cost
RCM Regional Climate Model
SON September, October, November
SRES Special Report on Emissions Scenarios

8
SST Sea Surface Temperatures
TAR Third Assessment Report
UKCIP United Kingdom Climate Impacts Programme
UKWIR United Kingdom Water Industry Research
WAFU Water Available For Use
WRP Water Resource Plan
WRZ Water Resource Zone
WTW Water Treatment Works

9
 CHAPTER 1. INTRODUCTION

This thesis concerns itself with climate change and the human response of adaptation to
this uncertain change. The thesis is an interdisciplinary endeavour that combines various
disciplines and approaches from the climate sciences to the decision sciences. Its main
objective is to develop a framework that allows the identification of adaptation
strategies that are robust (i.e. insensitive) to climate change uncertainties. The
motivation for this work originates from the problem faced by planners and decision-
makers when confronted with uncertain climate change information. To date, planners
have ignored, delayed or gambled with decisions about adaptation to climate change
because of the lack of consistent treatment of uncertainty in climate change projections.

This thesis aims to examine this problem by:

? Explicitly quantifying climate change uncertainties ranging from emissions of
greenhouse gases to climate change impacts, using a range of methodologies
? Explaining the institutional setting within which adaptation decisions occur
? Linking options available to decision-makers with changes in the climate
? Running sensitivity analysis to determine which uncertainties matter from an
adaptation perspective
? Running an uncertainty analysis to assess whether currently planned adaptation
strategies are robust against climate change uncertainties under a prediction
paradigm (i.e., assuming risks can be quantified)
? Running a robustness analysis where adaptation strategies are assessed against a
range of deep uncertainties (where risks cannot be quantified using probabilities)

This general framework is tested within a single sector case study on water resources in
the Anglian region of the UK. This chapter introduces the concept of climate change
(section 1.1), the nature of adaptation to climate change (section 1.2), the terms
sensitivity, uncertainty and robustness (section 1.4) and the case study area (section
1.5). The final section (1.6) provides a roadmap to the entire thesis.

10
1.1 Climate change

Climate change is one of the most pressing global problems of our time. There has been
a significant rise in greenhouse gas (GHG) emissions, such as carbon dioxide, methane,
nitrous oxide, etc., since the industrial revolution. This increase in emissions has led to a
rise in the concentration of GHGs in the atmosphere and subsequently a rise in radiative
forcing.1 Over the 20th century, the increase in global average surface temperature has
been of 0.6 0.2C2 according to the Intergovernmental Panel on Climate Change
(IPCC) Third Assessment Report (TAR) (2001b). The IPCC also concluded that most
of the observed warming over the last 50 years is likely3 to have been due to the
increase in GHG concentrations (p. 10). Changes in climate occur as a result of both
internal variability within the climate system and external factors (both natural and
anthropogenic). The influence of external factors on climate can be broadly compared
using the concept of radiative forcing. A positive radiative forcing, such as that
produced by increasing concentrations of GHGs or black carbon, tends to warm the
surface. A negative radiative forcing, which can arise from an increase in some types of
aerosols (e.g., sulphur) tends to cool the surface. Natural factors, such as changes in
solar output or explosive volcanic activity, can also cause radiative forcing.

Using emissions scenarios from the IPCC Special Report on Emissions Scenarios
(SRES, Nakicenovic et al. 2000), the IPCC TAR projected that global mean temperature
will increase between 1.4 and 5.8C over the period 1990 to 2100.4 For the same set of
scenarios, global mean sea level rise was projected to rise by 0.09 to 0.88 meters
between 1990 and 2100. Precipitation is projected to increase during the 21st century,
but there are regional discrepancies. These projections indicate that the warming would
vary by region, and be accompanied by increases and decreases in precipitation. In
addition, there would be changes in the variability of climate, and changes in the
frequency and intensity of some extreme climate phenomena (IPCC 2001a, p. 3).

1
Radiative forcing is a measure of the influence a factor has in altering the balance of incoming and
outgoing energy in the Earth-atmosphere system, and is an index of the importance of the factor as a
potential climate change mechanism. It is expressed in Watts per square metre (W/m2).
2
At the 95% confidence level.
3
Likely in IPCC terminology means 66-90% chance.
4
No confidence levels were given for these projections.

11
The IPCC TAR Working Group Two noted that available observational evidence
indicated that recent regional changes in climate, particularly increases in temperature,
have already affected a diverse set of physical and biological systems in many parts of
the world (IPCC 2001a). Examples of observed changes include shrinkage of glaciers,
thawing of permafrost, later freezing and earlier break-up of ice on rivers and lakes,
lengthening of mid- to high-latitude growing seasons, poleward and altitudinal shifts of
plant and animal ranges, declines of some plant and animal populations, and earlier
flowering of trees, emergence on insects, and egg-laying in birds (p. 3) (see, e.g.,
Parmesan and Yohe 2003, Root et al. 2003). Natural systems can be especially
vulnerable to climate change because of limited adaptive capacity, and some of these
systems may undergo significant and irreversible damage (p. 4). Many human systems
are sensitive to climate change, but the vulnerability of these systems varies with
geographic location, time, and social, economic, and environmental conditions.

Projected adverse impacts based on models and other studies include: reduction in
potential crop yield in most tropical and sub-tropical regions; decreased water
availability for populations in many water-scarce regions, particularly in the sub-tropics;
an increase in the number of people exposed to vector-borne and water-borne diseases,
and an increase in heat stress mortality; widespread increase in the risk of flooding for
many human settlements both from increased heavy precipitation events and sea-level
rise; increased energy demand for space cooling due to high summer temperatures.

Projected beneficial impacts include: increased potential crop yields in some regions at
mid-latitudes; increased water availability for populations in some water-scarce regions;
reduced winter mortality in mid- and high-latitudes; reduced energy demand for space
heating due to higher winter temperatures; increased tourism and leisure in some
regions.

There are two major responses to climate change: mitigation and adaptation. Mitigation
of climate change relates to the reduction of greenhouse gas emissions and the
sequestration of GHGs from the atmosphere (e.g., through trees or underground
storage). Adaptation to climate change refers to the successful reduction of the adverse
effects of climate change and the enhancement of beneficial impacts. This thesis is

12
almost exclusively focused on adaptation responses to climate change, which are
elaborated in further detail next.

1.2 Adaptation to climate change

Adaptation to climate change is one of the major responses to climate change. However,
it was given little attention when climate change emerged as an important issue in the
1980s. Mitigation has always been synonymous with combating climate change and that
is evident in the high prominence it was given in the United Nations Framework
Convention on Climate Change and its Kyoto Protocol. It is only in the last decade that
adaptation has emerged as a credible response to climate change (Burton 1997, Smith
1997b, Yohe and Neumann 1997, Parry et al. 1998, Pielke Jr. 1998, Scheraga and
Grambsch 1998, Fankhauser et al. 1999, Risbey et al. 1999, Smith et al. 1999, Pittock
and Jones 2000, Smith et al. 2000, Yohe 2000, Barnett 2001, Adger et al. 2005). The
IPCC TAR noted that adaptation is a necessary strategy at all scales to complement
climate change mitigation efforts (p. 6). It also noted that adaptation will incur costs
and will not prevent all damages. The ability of human systems to adapt to and cope
with climate change depends on such factors as wealth, technology, education,
information, skills, infrastructure, access to resources, and management capabilities
(p. 8). These are the determinants of adaptive capacity. Adaptation as defined by the
IPCC is adjustment in ecological, social, or economic systems in response to actual or
expected climatic stimuli and their effects or impacts (p. 879). As Adger et al. (2005)
have noted, adaptation can involve both building adaptive capacity thereby increasing
the ability of individuals, groups, or organisations to adapt to changes, and
implementing adaptation decisions, i.e. transforming that capacity into action (p. 78).
This thesis is almost exclusively focused on planning adaptation decisions and the role
of information, more specifically climate information, for adaptation to climate change.

Planning at its most basic is about minimising regret as it goes to the heart of how
humans think about the future and how decisions are made under uncertainty. Planning
is essentially the formalisation of human reasoning. When it comes to planning for
adaptation to climate change, effective reasoning is complicated by the fact that
evidence needs to be gathered from a number of sources to make informed decisions.

13
These could include observational data, modelled results, gut feelings, etc. All of these
types of information have uncertainty, which makes the assessment of climate change
risk a difficult task. Furthermore, it may often be the case that climate change is a
relatively small factor amongst the many pressures considered in real world planning.
This means that most adaptation decisions will only be partially influenced by climate
change (cf. Willows and Connell 2003, Adger et al. 2005), but this depends on the time
horizon of the planning.

1.3 Sensitivity, uncertainty and robustness

The extent to which adaptation will be required depends on the sensitivity of the system
to climate change. According to the IPCC, sensitivity is the degree to which a system is
affected either adversely or beneficially, by climate-related stimuli.5 The central focus
of this thesis is to examine the sensitivity of adaptation decisions to climate change
uncertainties. In order to do this, sensitivity analysis is used. Sensitivity analysis is the
process by which uncertainty in inputs is related to uncertainty in outputs. In a decision
analysis context sensitivity analysis amounts to evaluating the robustness of a
particular decision by repeating computations for varying problem structures,
probabilities and values (van Lenthe et al. 1997, p. 232). This is the type of analysis
this thesis carries out using adaptation decisions and climate change uncertainties.

Saltelli (2000) defines sensitivity analysis as the study of the relationships between
information flowing in and out of the model (p. 4). Saltelli (2000) has grouped
sensitivity methods into three classes: screening methods, local sensitivity analysis
methods and global sensitivity analysis methods. Screening methods allow the
identification of parameters that control most of the output variability in models that are
computationally expensive.6 They tend to provide qualitative sensitivity measures.
Local sensitivity analysis focuses on the local impact of the factors on the model
whereas global sensitivity analysis incorporates the influence of the whole range of

5
Climate-related stimuli encompass all the elements of climate change, including mean climate
characteristics, climate variability, and the frequency and magnitude of extremes. The effect may be
direct (e.g., a change in crop yield in response to a change in the mean, range or variability of
temperature) or indirect (e.g., damages caused by an increase in the frequency of coastal flooding due to
sea-level rise).
6
The central question of screening is: Which factors, among the many potentially important factors, are
really important?

14
variation and the form of the probability density function (PDF) of the input (p. 10-11).
Closely linked to sensitivity analysis, and sometimes done in combination, is
uncertainty analysis. A good way to distinguish between the two is to look at the
different questions they pose:
? What is the uncertainty in y(x) given uncertainty in x?
? How important are the individual elements of x with respect to the uncertainty in
y(x)?
Uncertainty analysis tackles the first question while sensitivity analysis deals with the
second.

Uncertainty is pervasive in the climate change problem. There is uncertainty about how
much climate change we have experienced and even further uncertainty about how
climate will change in the future. In this thesis uncertainty analysis is performed in
order to quantify the uncertainties associated with climate change, relevant to adaptation
decision-making. These include a combination of probabilistic approaches (for climate
system properties) and scenarios analysis (for human system properties) in order to deal
with different types of uncertainties (epistemic, stochastic, human reflexive; see section
3.2). The lack of traceable and consistent treatment of uncertainties in climate change
assessments was noted as a major deficiency in current knowledge by the IPCC TAR
(Carter et al. 2001). The IPCC also noted that a major challenge lies in addressing
uncertainties associated with adaptation (Ahmand et al. 2001). This thesis addresses
both of these shortcomings.

In addition, the IPCC noted that there have been very few cases in which decision
analytic frameworks have been used in evaluating adaptation options (Ahmand et al.
2001). They urged a greater use of methods in support of adaptation decisions in order
to establish their efficacy and identify directions for further research. In this thesis a
method that evaluates adaptation options against the uncertainties introduced by climate
change is elaborated. This is called robustness to uncertainty here. The IPCC defines
robustness as strength; degree to which a system is not given to influence (p. 894).
Lempert and Schlesinger (2000) propose that society should seek strategies that are
robust against a wide range of plausible climate change futures. For these authors robust
strategies are insensitive to uncertainty about the future. The robustness of adaptation
strategies to climate change uncertainties is examined in detail here.

15
1.4 The Anglian region of the UK: climate and water

The Anglian region is the driest in the UK. Extending from the Humber estuary to the
Thames (Figure 1.1), the region is low lying and intensively farmed. The region is also
experiencing one of the fastest growing populations in the UK. Competition over water
for public water supply, agriculture, industry and the environment represents a major
challenge for decision makers and planners in this region. There is uncertainty about the
adequacy of water resources in the region because stakeholders have widely varying
perceptions about the security of their water resources (Hams and Taylor 1999).
Nevertheless, the Environment Agency (EA) regional water resources strategy
concluded that water is a scarce resource in the region and that long-term planning is
required to manage this resource sustainably (EA 2001b).

Figure 1.1 The Anglian region of the UK. Source: Anglian Water and Environment
Agency.

Climate change is expected to increase average temperature in the region over all
seasons, with warming being more pronounced over the summer. Average precipitation
is projected to increase over the winter and decrease over the summer; i.e., wetter
winters and drier summers are expected for the region (Hulme et al. 2002). This will
have implications for water availability, particularly during the summer. Climate change

16
is also likely to cause demand for water to increase in the region (Downing et al. 2003).
During hotter, drier summers it may be increasingly difficult for supplies to keep up
with demand, with implications for water service providers (EERA and SDRT 2004).
According to a stakeholder workshop on water, Taylor and Hams (2001) report that
these changes will require considerable adaptation by economic sectors dependent on
large, regular quantities of water, like agriculture and housing development (p. 6).
Planning for adaptation to climate change will be crucial in this region of the UK. This
makes the Anglian region and water resources a particularly pertinent case study to test
the framework developed in this thesis.

1.5 Roadmap

This thesis is long and sequential so it is important to provide a general overview of the
forthcoming chapters. Much effort in this thesis is put on quantifying uncertainties so
Chapter 2 provides a review of other studies conducted on the various areas where
uncertainty was quantified, such as with respect to global climate, regional climate, etc.
There are very few studies that have looked at adaptation planning and climate change
uncertainties, but these are reviewed in this chapter. Chapter 3 explains the overall
methodological framework of the thesis. Because various tools and approaches are
borrowed from previous work, a review is often conducted. Quantification of
uncertainties at the global level, using a simple climate model, is presented in Chapter 4.
This includes sampling uncertainty in GHG emissions, climate sensitivity and other
climate parameters such as aerosol forcing. Probabilistic estimates of global temperature
change are presented here for a range of uncertainties. Chapter 5 conducts the
assessment of uncertainties at the regional level. This chapter also includes a description
of the climate of the Anglian region (based on observed data) and a review of past
climate change scenarios for the region and the UK. Uncertainties sampled at this level
included different Global and Regional Climate Models and the issue of model skill is
also explored. Chapter 6 goes into greater depth on the institutional setting where
adaptation decision-making could occur in the water sector. A review of how climate
change is currently included in water resources planning is provided. This section also
elaborates on the adaptation options available to a particular decision-maker involved in
long-term planning, Anglian Water Services (AWS), for two Water Resource Zones in

17
the region. Finally, this chapter also explains how adaptation decisions were linked to
changes in the climate. Chapter 7 then brings everything together by conducting
sensitivity, uncertainty and robustness analysis on the entire framework. Chapter 8
discusses the limitations and implications of this work and finishes with some ideas on
further work and concluding remarks.

18
 CHAPTER 2. LITERATURE REVIEW

This chapter reviews previous attempts at dealing with uncertainty in climate change.
Because climate change cuts across many different subject areas and because of the
sequential nature of its assessment, the literature is discussed in subject areas rather than
by individual studies.7

2.1 Greenhouse gas emissions and its drivers

There have been few attempts to quantify uncertainties in future greenhouse gas (GHG)
emissions probabilistically. Exceptions include the early work by Nordhaus and Yohe
(1983) and Edmonds et al. (1986) (as cited by O'Neill 2004) who developed
probabilistic versions of energy-economic growth models to calculate uncertainty
distributions for optimal emissions, given uncertainty in a long list of parameters,
including those affecting population growth, labour productivity, the energy system, and
the carbon cycle.

Uncertainty in certain key drivers of GHG emissions have been explored in probabilistic
terms, namely population growth (Lutz et al. 1997, 2001) and technological change
(Gritsevskyi and Nakicenovic 2000). However, rarely have these been combined to
produce probabilistic greenhouse gas emissions, mainly because the PDFs for a number
of key drivers (e.g., per capita income, hydrocarbon resource use and land-use change)
are unavailable/unknown and the interconnection between drivers is complex. One
exception is a study that developed a consistent set of future emissions with known
probabilities based on a computable general equilibrium model of the world economy
(Webster et al. 2002). They performed a sensitivity analysis to identify the most
important parameters, whose uncertain PDFs were constructed through expert elicitation
(by five in-house economists) and drawing from the literature. The uncertainty of the
eight independent sets of input parameters (e.g., labour productivity growth,
autonomous energy efficiency improvement rate, and several sources of GHGs) was
propagated into the model. Through a Monte Carlo simulation, PDFs of GHG emissions
for each time period were produced.

7
This section is an update of section four of Dessai and Hulme (2003).

19
An earlier study performed something rather similar to this, but went beyond it by
constraining the global energy model according to observations of energy consumption
and carbon emissions through a Bayesian technique (Tsang and Dowlatabadi 1995).8
Another study estimated global CO2 emissions until 2100, using a Monte Carlo method
that draws on PDFs based on historic data (calculated using regression analysis) or
expert assessments (Leggett et al. 2003). From 2500 runs, Leggett et al. (2003)
concluded that it is highly unlikely that CO2 emissions would more than triple over this
century, with 95% probability bounds that ranged from 10-20 Gigatonnes of carbon
(GtC) with a median of 14.1 GtC (CO2 emissions from fossil fuels and land use were
around 7 GtC in 1990 and 8 GtC in 2000). Webster et al. (2002) performed a Monte
Carlo simulation with 10,000 runs, where 5-95% of the results stayed in the range 7-39
GtC, with a median of 20 GtC. Interestingly, neither of these studies span the whole
IPCC SRES range (3.3-36.8 GtC) at the 95% confidence level.

More recently ONeill (2004) developed probabilistic projections of population

conditional on the storylines used in the SRES scenarios. Through simple linear scaling
(with per capita emissions rates derived from the SRES scenarios), ONeill (2004)
developed conditional probabilistic emissions scenarios, using the IPCC SRES
scenarios as a basis. He found a much wider range of uncertainty in emissions compared
with the original SRES scenarios: for example, his 95% uncertainty interval was 14-32
GtC/yr by 2100 compared to the much smaller range of 28-33 GtC/yr for the SRES A2
scenarios.

It is fair to say that the studies above are exceptions in the literature. The most common
approach to quantifying uncertainties in GHG emisisons is scenario analysis (cf.
Nakicenovic et al. 2000). This is further elborated in section 4.1.

2.2 Global climate

Most of the work on quantifying uncertainties has been performed at the global climate
system level, particularly looking at key uncertain parameters such as climate

8
This is similar to what is currently being done with Global Climate Models (see section 2.2), but which
is rarely done for models involing human systems in the environmental field.

20
sensitivity, heat uptake by the oceans or aerosol forcing. One of the earliest studies that
explored the uncertainty of key climate variables was that of Morgan and Keith (1995),
who interviewed a number of US climate experts to elicit subjective PDFs of climate
sensitivity. Their results showed a diversity of expert opinion, which led them to
conclude that the overall uncertainty of climate change is not likely to be reduced
dramatically in the next few decades (a prediction so far borne out). Using a number of
different methods researchers have run their previously deterministic climate models in
a probabilistic manner (Zapert et al. 1998, Visser et al. 2000, Webster and Sokolov
2000, Dessai and Hulme 2001, Wigley and Raper 2001). It is important to note that
within this approach the output likelihood is dependent on the subjective prior PDFs
attached to uncertain model parameters (these are mostly based on expert judgement).

Another strand of research that complements earlier efforts and attempts to reduce
uncertainty is the method of constraining certain climate parameters, in particular
climate sensitivity, by using recent observed changes in the climate system (Tol and de
Vos 1998, Allen et al. 2000, Forest et al. 2000, Andronova and Schlesinger 2001, Forest
et al. 2001, Forest et al. 2002, Gregory et al. 2002, Knutti et al. 2002, Stott and
Kettleborough 2002, Jones et al. 2003, Knutti et al. 2003). This is essentially a Bayesian
approach that will prove most useful as more observed data are gathered in the future.
Uncertainties in climate change detection and attribution have also been articulated and
quantified using a formal probabilistic protocol (Risbey et al. 2000, Risbey and
Kandlikar 2002).

Uncertainty in General Circulation Models (GCMs) has been mainly explored through
means of intercomparison and validation statistics between model results and observed
climatology (Lambert and Boer 2001). There are also a few examples of evaluating
GCM output with impact models (Williams et al. 1998). However, with computational
power on the increase there are a few studies that have started to run large ensembles of
GCMs in order to quantify uncertainty in the climate response (Murphy et al. 2004,
Stainforth et al. 2005). Murphy et al. (2004) performed a local sensitivity analysis for a
selection of parameters of the HadAM3 climate model (an atmospheric general
circulation model coupled to a slab ocean) to generate a PDF for climate sensitivity.
Parameter selection and the range over which parameters were varied was based on

21
expert elicitation.9 In constructing the climate sensitivity PDF, a climate prediction
index that weights ensemble members according to degree of correspondence with
observations was applied. Their 95% confidence range for climate sensitivity was 2.4-
5.4C. The Stainforth et al. (2005) study uses the same model but runs many more
simulations (over two thousand) and does not attempt to constrain the results.10 They
found climate sensitivities that ranged from less than 2C to more than 11C.

2.3 Global impacts

The likelihood of global impacts such as sea level rise (Patwardhan and Small 1992,
Titus and Narayanan 1996), the collapse of the West Antarctic Ice Sheet (Vaughan and
Spouge 2002), the global carbon cycle (Craig and Holmen 1995, Shackley et al. 1998,
Jones et al. 2003), global economic impact (Nordhaus 1994), and the overturning of the
thermohaline circulation (Mastrandrea and Schneider 2001, Vellinga and Wood 2002)
have been the subject of some research and much media attention (the last two studies
do not estimate likelihood, but explore a forced temporary collapse). These
approaches have relied heavily on expert judgement techniques because of the high
impact/low probability nature of the events. There are numerous other studies that deal
with global impacts such as the large-scale eradication of coral reef systems, biome
migration or changes in ENSO (see, e.g., O'Neill and Oppenheimer 2002), but few have
represented uncertainty explicitly through probabilities.

Titus and Narayanan (1996) developed probability-based projections of future sea level
rise using a combination of Delphic and Monte Carlo techniques. The Monte Carlo
experiment was divided into two parts. In part one, they developed a simplified model
(based on Wigley and Raper 1992) for estimating sea level rise as a function of 35
major uncertainties, where the probability distributions for each parameter were derived
from the existing literature, and performed 10,000 simulations. In part two, the results
from part one were circulated to the Delphic panel of approximately two dozen experts
(including climatologists, oceanographers, and glaciologists) who reviewed Titus and
Narayanans assumptions and suggested their own subjective probability distributions

9
To the knowledge of the author no formal protocol of elicitation (as in Morgan and Keith 1995, Risbey
et al. 2000) was followed to, for example, de-bias experts, etc.
10
But that is on-going work.

22
(which were treated as equally likely), with which they re-ran 10,000 simulations. They
estimated that global temperatures are most likely (50% probability) to rise 2C and sea
level rise of 35cm by 2100, which is somewhat lower than previous IPCC assessments.

2.4 Regional climate

At the level of regional climate, probabilistic approaches have been less explored
because the compounding of uncertainty from the global level is large and not well
quantified, and the number of GCM runs is still small (although this is changing). Thus,
validation statistics (e.g., mean, standard deviation, pattern correlation, etc.) have been
traditionally used to explore regional uncertainties in future climate (Kittel et al. 1998,
Giorgi and Francisco 2000, Giorgi et al. 2001b). Giorgi and Mearns (2002) introduced
the Reliability Ensemble Averaging method, a summary measure, which calculates
average, uncertainty range, and reliability of regional climate changes from GCM
simulations.

Some researchers, however, have looked at probabilistic methods for regional climate.
New and Hulme (2000) used a simple climate model to sample uncertainty in the global
climate and then used the pattern-scaling technique to propagate this uncertainty to
the regional level using 14 runs of GCMs as a super-ensemble. Risnen and Palmer
(2001) ignored the upstream uncertainties and considered 17 GCMs a probabilistic
multimodel ensemble projection of future regional climate. Benestad (2004) produced
probabilistic regional temperature scenarios for northern Europe based on spatially
interpolated empirically downscaled trends, derived using a multi-model GCM
ensemble as well as various downscaling options. Goodess et al. (2003) have developed
a sophisticated method to estimate the probability of future extreme weather events.
They applied this to short and long drought frequency in the UK as a function of global
mean temperature change, computed by scaling observed time series means and
distribution shapes according to relationships fitted to a GCM and a regional climate
model. Giorgi and Mearns (2003) extended their Reliability Ensemble Averaging
method to calculate the probability of regional climate change exceeding given
thresholds using nine GCM simulations under SRES A2 and B2. This method goes

23
beyond previous studies by using the reliability factor to estimate the probability of
future regional climate change.

One specific problem of the Giorgi and Mearns (2003) study is that it conflates different
types of uncertainties by giving a likelihood to emissions scenarios (in this case SRES
A2 and B2 had equal weight). More recently, Bayesian approaches have started being
applied to regional probabilities of climate change (Tebaldi et al. 2004, Tebaldi et al.
2005). In these studies, PDFs of temperature and precipitation change are determined
from combining information from a multimodel ensemble of GCMs with observations,
using the criteria of bias and convergence in the relative weights implicitly assigned to
the ensemble members. Dessai et al. (2005a) examined the sensitivity of regional
climate change probabilities to various uncertainties using the pattern-scaling technique,
which combines a simple climate model with results from GCMs. By 2100 temperature
change probabilities are dominated by emissions scenario and climate sensitivity
uncertainty. For precipitation change, GCM weighting scheme uncertainty is also
significant. Regional climate change probabilities have also been estimated using results
from Regional Climate Models (Ekstrm et al. 2005, Hingray et al. 2005).

The most important caveat with these probabilistic approaches is likely to be the limited
number of GCM runs sampled, which is further limited by commonality of code
between GCMs. Hopefully this will be overcome in the future but in the mean time
these studies are important methodological additions to the representation of
uncertainties in regional climate change projections. Allen and Ingram (2002) have
suggested that once the new generation of climate experiments is underway, it might be
possible to constrain regional climate with the combined use of observed global-mean
temperature and the hydrologic cycle, though the latter could be a weak constraint. A
particular regional uncertainty that has been consistently overlooked by assessments,
and thus is far from any type of probabilistic treatment, is the impact of land-use change
and landscape dynamics on regional and global climate (Pielke Sr. 2002).

Another problematic issue in dealing with regional climate uncertainty is the issue of
spatial downscaling. The mismatch between the coarse resolution of GCMs (hundreds
of kilometers) and local or regional impact applications has led to the development of a
plethora of downscaling techniques (Wilby and Wigley 1997, see Appendix 10.4 in

24
Giorgi et al. 2001a for a list of examples). These include dynamical downscaling, which
describe the fine atmospheric processes nested within the GCM outputs, the so-called
Regional Climate Models (RCMs) (e.g., Jones et al. 1995, Giorgi et al. 1998); and
statistical downscaling, including weather typing (e.g., Wilby 1994), stochastic weather
generators (see Wilks and Wilby 1999 for a review) and neural networks (e.g., Trigo
and Palutikof 1999). Both New and Hulme (2000) and Raisanen and Palmer (2001)
used raw GCM data (sometimes called unintelligent or simple downscaling since it
only involves interpolation of the output) in their probabilistic studies whereas Benestad
(2004) and Goodess et al. (2003) applied some downscaling techniques to the GCM
data.

Managing downscaling uncertainties will prove particularly difficult because of the

various techniques being developed and the conflation of scales (global to local), which
makes the appropriateness of techniques highly context dependent. At present,
validation statistics are being used to explore downscaling uncertainty in a number of
settings (e.g., water and agriculture), between statistical downscaling and raw GCM
output (e.g., Wilby et al. 1999), between statistical methods (e.g., Wilby et al. 1998),
between RCMs and statistical downscaling (e.g., Mearns et al. 1999, Wilby et al. 2000).
Two large European projects have recently completed the comparison of results using
different RCM simulations (PRUDENCE 2002) and different statistical downscaling
techniques (STARDEX 2002).

Representation of uncertainty in spatial downscaling in probabilistic terms is

constrained by computational power, but Katz (2002) and Prudhomme et al. (2002)
suggest that statistical downscaling, in terms of conditional stochastic processes, could
be particularly useful for uncertainty analysis, rather than empirical methods.
Downscaling has become a large enterprise within climate change science, but these
techniques are very dependent on the GCM outputs they utilise downscaling cannot
correct for model inaccuracies (Mitchell and Hulme 1999).

25
2.5 Regional/local impacts

There are a plethora of impact studies that have used one or a few more climate change
scenarios to represent uncertainties from climate projections (IPCC 2001a). This is
clearly insufficient (see Katz 2002 for a review of uncertainty techniques in this area),
but few studies have ventured into the probabilistic realm for the same reasons given
earlier, in particular the compounding and management of uncertainty. Similarly,
Schimmelpfennig (1996) noted that uncertainty has been poorly represented in the
economic models of climate change impacts, suggesting that a full probabilistic analysis
be conducted. However, in most studies uncertainty is not comprehensively covered,
especially with respect to climate change scenarios. For example, Woodbury et al.
(1998) used four different GCMs to derive what they call a probabilistic climate
change scenario, Venkatesh and Hobbs (1999) used four climate scenarios, while
Scherm (2000) used a fuzzy scenario. The most comprehensive approach is provided by
Jones (2000a) who has a similar approach to New and Hulme (2000), but extends this to
numerous impact models using critical impact thresholds. In Jones and Page (2001) an
uncertainty analysis was carried out to assess the contribution of global warming vis--
vis other components, as well as a Bayesian analysis to test the sensitivity of the results
to initial assumptions. For the water resources of the Macquarie river catchment, 25% of
the uncertainty originates from global warming whereas precipitation changes
contribute 64%. The Bayesian analysis showed that the risk of threshold exceedance is
rather insensitive to changes in the input assumptions for rainfall or global warming.
Morgan et al. (2001) determined subjective probability distributions of the impact of a
doubling CO2 climate change on forest ecosystems through the elicitation of experts.
For one single climate change scenario various non-climate key factors and processes of
forest ecosystems were explored, for example, human land use, physical disturbance,
soil properties, etc.

Another study combined the results of New and Hulme (2000; i.e., 25,000 climate
scenarios randomly generated by a Monte Carlo simulation using several GCMs, SRES-
98 emissions scenarios and climate sensitivities) with a hydrological model to quantify
uncertainties of climate change impact on the flood regime of five small catchments in
Great Britain (Prudhomme et al. 2003). The analysis showed a large variation of results

26
(varying by a factor of 10), but most scenarios showed an increase in both the
magnitude and frequency of flood events, generally not greater than natural variability
(which in this study constituted 95%-confidence intervals of historical data). The largest
uncertainty was attributed to the GCM used rather than emissions scenarios or climate
sensitivity, though the former starts playing a larger role by the 2080s. Uncertainties in
the hydrological model itself or downscaling were not explored so it is not possible to
make definitive recommendations on where further research should be targeted.
Schumann and Antl (2001) used the same 25,000 climate scenarios to examine the
hydrological impacts in the Danube river basin in Germany. Their results showed a high
variety of future runoff for this basin. This led Schumann and Antl (2001) to conclude
that water managers should accept these restrictions and be able to consider a wide
range of uncertainties in planning.

2.6 Adaptation and planning

It is perhaps not surprising that there is so little work done on adaptation and planning
for climate change considering the level of uncertainty involved. One of the earliest
studies in this area is that of Yohe (1991) who examined the efficacy of adaptation
options to protect developed property in Long Beach Island, New Jersey from uncertain
greenhouse gas-induced sea level rise. In a subsequent paper, Yohe and Neumann
(1997) argue that current procedures by coastal planners to evaluate shoreline protection
projects can deal with climate change uncertainty, except in situations where planners
try to achieve maximum economic efficiency.

Because quantification of uncertainty in climate change assessments is problematic,

Risbey (1998) performed a qualitative sensitivity analysis that showed that water-
planning decisions were sensitive to uncertainty in the range of GCMs simulated for the
Sacramento basin in California. Though only a few GCMs were used and a simple
scenario matrix approach taken for adaptation decisions, this study is nonetheless
ground-breaking because it links future climate with planning decisions of today under a
range of plausible scenarios. Most of the other local impact studies in the literature lack
this important component the sensitivity of adaptation decisions to upstream
uncertainties even though uncertainty is sometimes quantified in terms of probability.

27
In the Great Lakes region, Hobbs et al. (1997) used decision trees and Bayesian analysis
to assess climate change risk to two specific investment decisions. They found that
beliefs about climate change can affect optimal decision. Hobbs (1997) provides a
review of the application of Bayesian analysis of climate change and water resource
uncertainties. Using Bayesian Monte Carlo-based decision analysis, Venkatesh and
Hobbs (1999) showed that taking climate change uncertainty into account made a
difference in investments for managing Lake Erie. In particular, beliefs of prior
distributions over climate scenarios affected the optimal strategy. However, this is not
always the case. Wood et al. (1997) showed that reservoir reallocatoin decisions on the
Green River (WA, United States) would not differ significantly if climate change
information were incorporated in the planning process. This conclusion, however, is
limited by the the fact that only three GCM outputs were considered to represent climate
change uncertainty.

Without using probabilities of climate change, Risbey et al. (1999) examined farming
decision-making under the uncertainty of climate variability and change for Australia.
At the micro-scale (farm level) they constructed a simple model to compare the
performance of hypothetical farmers operating under a variety of possible behaviours
for example, a hedge farmer who is risk averse or a predictive farmer who uses forecast
information with the performance of a clairvoyant farmer. This study showed that
using climate forecast information has substantial benefits over decision rules based on
historical trends, but this cannot be generalised because of the simplicity of the model.
The study also showed that the model is sensitive to the shape of the crop response
surface used.

More recently, Whitehead et al. (2005) performed an appraisal of adaptation strategies11

to cope with the impacts of climate change on nitrogen in a lowland chalk stream in the
UK (River Kennet). They demonstrated that the most effective strategy is to change
land use or reduce fertiliser use, followed by water meadow creation, and atmospheric

11
Reduced fertiliser application rates/land use change; reduced atmospheric deposition rates for nitrate
and ammonia; river water quality management via the creation of water meadows or connected wetland
areas; and multiple strategies.

28
pollution controls. The timing of adaptation was influenced by GCM uncertainty, which
was sampled using three GCMs.

2.7 Integrated assessment of climate change

While this thesis is focused on adaptation decisions and all the climate change
uncertainties upstream of that, it is worth examining integrated assessments of climate
change because much uncertainty analysis has been conducted here. Some of the studies
reviewed in previous sections (e.g., Webster et al. 2002, Leggett et al. 2003) are
modules of larger Integrated Assessment Models (IAMs). Various IAMs of climate
change have represented uncertainty explicitly (e.g., Dowlatabadi and Morgan 1993,
Hope et al. 1993, Chao 1995, Peck and Teisberg 1996, Nordhaus and Popp 1997, Scott
et al. 1999), which Weyant et al. (1996) have called uncertainty-based models in their
overview of the literature. They counted ten of these models in 1996, some of which are
reviewed next.

Nordhaus and Popp (1997) estimated that the value of early information (assuming
decision-makers act optimally once the information is known) is between one and two
billion dollars for each year that resolution of uncertainty is moved forward in time.
They estimate that damages of climate change and the costs of reducing greenhouse gas
emissions are the most important uncertain variables when estimating the value of
scientific knowledge. On the other hand, Dowlatabadi and Morgan (1993) found that
uncertainties in climatic/geophysical parameters dominate policy outcomes. Hope et al.
(1993) found that adaptation to climate change might be cheaper than mitigation, but the
results were sensitive to various parameters including climate sensitivity, mitigation
costs, etc. Within probabilistic-based methods, the treatment of uncertainty in IAMs has
taken a number of methodological approaches including Latin hypercube sampling
(e.g., Nordhaus and Popp 1997, Scott et al. 1999), method of probabilistic collocation
(e.g., Webster et al. 2003) or Monte Carlo simulation (e.g., Roughgarden and Schneider
1999). However, there are other approaches to treat uncertainty in IA. van der Sluijs
(1996) has explored the problems of uncertainty management in IAMs of climate
change and concluded that a methodology based on the Numerical Unit Spread

29
Assessment Pedigree (NUSAP) system (Funtowicz and Ravetz 1990, van der Sluijs et
al. 2005) is most appropriate to disentangle the uncertainty problem in IAMs.

Though some results might be probabilistic, the treatment of adaptation in IAMs is still
very simplistic, if at all existent (Toth 2000). This work implicitly uses damage
functions that remain considerably uncertain mainly because they use global mean
temperature as an indicator of impacts (see e.g., Smith and Hitz 2003). In order to
explore this particular uncertainty, Roughgarden and Schneider (1999) used an IAM
(Nordhaus and Popp 1997) to explore several alternative published estimates (e.g.,
Fankhauser 1995) and opinions (Nordhaus 1994) of the possible damages from climate
change. Roughgarden and Schneider (1999) showed that incorporating alternate damage
estimates results in a significantly more aggressive optimal policy than using a single
damage function, thus emphasising the need to consider estimates of uncertainty of the
various parameters of IAMs.

Scott et al. (1999) performed sensitivity and uncertainty analysis in their previously
deterministic IAM. The sensitivity analysis showed that market damages are much more
sensitive to variables relating to economic activity or emissions than climate sensitivity.
For example, an increase of the index of labour productivity in the developing world by
50% would increase global temperature by 1.24C by the year 2100 and market
damages by 290%. A 50% increase of income elasticity of demand for energy in the
developing world would increase temperature by over 2.54C. Climate sensitivity is
ranked fourth out of 74 variables of importance for market damages. The uncertainty
analysis showed that the income elasticity of demand of energy in the developing world
is the single largest uncertainty for either emissions, temperature change, market
damage or non-market damage in the year 2100.

Toth et al. (2003) performed uncertainty analysis in four decision variables to explore
the effects of policy/target-related uncertainties on the shape of the emission corridors.
Impact constraints, climatic constraints (such as the magnitude and rate of the global
mean temperature increase) and aerosol emission scenarios had an important influence
in the emission corridors. In a similar effort and without using probabilities, Caldeira et
al. (2003) showed how uncertainty in climate sensitivity propagates to uncertainty in
allowable carbon emissions for a specified climate change scenario, arbitrarily chosen to

30
be a stabilisation to 2C global mean warming after the year 2150. Estimated allowable
carbon emissions later this century could be less than 0 GtC or 13 GtC per year
depending on whether climate sensitivity is 4.5C or 1.5C respectively. These authors
conclude that even if climate sensitivity is low, a considerable transition to CO2
emission-free technologies will be required (see Hoffert et al. 2002).

Webster et al. (2003) have recently combined probabilistic GHG emissions (Webster et
al. 2002) with constrains on climate parameters from observations (Forest et al. 2002).
They found that in the absence of policy intervention their 95% probability range of
global mean temperature change for 2100 is 1.0 to 4.9C, with a mean of 2.4C. The
policy scenario reduces (mean of 1.7C in 2100), but doesnt eliminate the chance of
substantial warming. While this approach is certainly pioneering in terms of climate
change uncertainty analysis, there remain important caveats surrounding the issue of
policy versus no-policy scenarios. Where does reflexivity fit with these scenarios? This
study used arbitrary climate policy and no-climate policy scenarios (Reilly et al. 1999),
however, unlike the SRES scenarios little information is given about the worlds that
underlie these scenarios, which limits the usefulness of the results. Also, these results
might be useful to evaluate a certain mitigation strategy, but in terms of adaptation it is
not because one could envisage millions of policy scenarios; adaptation researchers
want to know which is the most likely to occur.

2.8 Other areas

It is also important not to forget that other areas of inquiry in the environmental field
(and beyond) have examined the issue of decision-making under severe uncertainty. For
example, Lempert et al. (2003, 2005) demonstrated the application of robust decision
methods to the problem of global sustainable development. From thousands of scenario
simulations that describe plausible futures, they identified robust, adaptive near-term
pollution control strategies to help ensure economic growth and environmental quality
throughout the 21st century. Regan et al. (2005) used information-gap theory (see Ben-
Haim 2001) to propagate uncertainty and to rank conservation management decisions
about an endangered species, both under severe uncertainty. They showed that different

31
management decisions may result when uncertainty in utilities and probabilities are
considered in decision-making problems.

32
 CHAPTER 3. METHODOLOGICAL FRAMEWORK

This chapter presents the overall methodological framework of this thesis. Due to the
interdisciplinary, modular and sequential nature of the work it is not possible to describe
the entirety of the methodology in this chapter. Instead detailed methodology is
provided in subsequent chapters where it is more relevant, while this chapter provides
an overall picture. The chapter is divided into two main parts. Section one reviews
different adaptation assessment approaches and then elaborates on the methods applied
in this thesis. Section two reviews how uncertainties have been managed in the past and
how they are tackled in this thesis.

3.1 Adaptation assessment approaches

Several approaches have been proposed to deal with adaptation to climate change. This
section briefly reviews these approaches as they are relevant to the methodology
developed in this thesis.12

3.1.1 The IPCC approach

One of the earliest guidelines for impact and adaptation assessment was developed by
the IPCC in the early 1990s (Carter et al. 1994, Parry and Carter 1998). This approach
followed seven steps, which included:

1. Define problem (including study area, its sectors, etc.);

2. Select method of assessment most appropriate to the problems;
3. Test methods/conduct sensitivity analysis;
4. Select and apply climate change scenarios;
5. Assess biophysical and socio-economic impacts;
6. Assess autonomous adjustments;
7. Evaluate adaptation strategies.

12
This section is partially based on Dessai et al. (2005b).

33
This approach relies heavily on climate change scenarios (step 4) as this is the main
driver of the impacts (step 5), from which adaptation strategies are devised in the last
step. This framework was later expanded to include socio-economic scenarios (Feenstra
et al. 1998). Because this approach is so scenario-driven, further guidance was provided
on the use of scenario data (both climate and non-climate scenarios) for climate impact
and adaptation assessment (IPCC-TGCIA 1999) and more recently on climate scenarios
developed from regional climate model experiments (Mearns et al. 2003) and from
statistical downscaling methods (Wilby et al. 2004). This approach gives great
significance to climate scenarios in informing adaptation planning.

The IPCC approach has become increasingly sophisticated in the literature with the
inclusion of socio-economic and land use scenarios (see, e.g., Arnell 2004a, Arnell et al.
2004, Schrter and others 2005), better representation of uncertainties in climate change
projections (see, e.g., Jones 2000a, Dessai 2003, Dubrovsky et al. 2005) and increased
spatial resolution (through statistical or dynamical downscaling, see ,e.g., Wilby et al.
2000). Nevertheless, few studies, if any, have been able to provide robust information
for decision-makers and risk managers. According to Burton et al. (2002) this occurs
because of: 1) the wide range of potential impacts (issue of uncertainty); 2) the
mismatch of resolution between global climate models and adaptation measures (usually
local or site specific; issue of scale); 3) impact assessments not designed to consider a
range of adaptation options; 4) adaptation is incorporated as an assumption rather than
explored as a process; 5) impact and adaptation assessments were initially developed for
the scientific purpose of understanding impacts not for informing decisions. Despite
these shortcomings, the IPCC approach is probably the most widely used in the
literature. Also known as the standard approach or top-down approach, Figure 3.1
shows the common elements of the IPCC approach towards impacts and adaptation.

34
 (physical)

Figure 3.1 Two complementary approaches to impacts, adaptation and

vulnerability assessments. Modified after Dessai and Hulme (2004).

3.1.2 The resilience/vulnerability/human development approach

The limitations of the IPCC approach have led Burton et al. (2002) to encourage a
paradigm shift from adaptation for impacts/mitigation (what they call first generation
impacts and adaptation research) to adaptation to reduce vulnerability in a development
context. For these scholars, scenarios and probabilities of climate change are much less
relevant. Instead, a strategy of resilience and adaptive environmental management that
enhances coping capacity is preferred (Pielke Jr. 1998, Adger 1999, Handmer et al.
1999, Kelly and Adger 2000, Barnett 2001, Burton et al. 2002, Clark and Pulwarty
2003, Tompkins and Adger 2003). These authors argue that in the face of the
considerable uncertainty over climate change projections and their impacts, one is better
off adapting to the present day (or recent historic) climate variability as this is assumed
a good proxy for near-term climate change. These so-called bottom-up approaches
(see Figure 3.1) have been very useful for understanding society's vulnerability to
present day climate and the underlying causes of vulnerability. In particular, social
vulnerability scholars are concerned with the capacity of individuals or social groups to
respond to (i.e., to cope with, recover from or adapt to) any external stress placed on

35
their livelihoods and well-being; this method of analysis emerged from the work of Sen
(1981), Blaikie et al. (1994) and others. Social vulnerability has been measured by
developing a set of indicators of relative vulnerability, for example, for a small rural
community this could include poverty, the use of resources, the distribution of assets
and income, institutional effectiveness, etc. For an excellent case study of social
vulnerability to extreme events, see OHare (2001).

Other approaches use so-called analogues to learn from past climate adaptation
experiences (e.g., Pulwarty and Melis 2001) because their basis in actual experience is
viewed as an advantage over modelled quantitative scenarios (Meyer et al. 1998).
However, history may not be the best guide in viewing how adaptation would unfold
with future changes in hazard exposure, though it is certainly an important one (Yohe
and Dowlatabadi 1999). This is the case because there are two fundamental limitations
to the use of analogues in climate-society research: analogues between cases are never
perfect and analogues can say little about long term climate change (Meyer et al. 1998).
Climate change has a unique global forcing that is likely to produce non-analogous
impacts. Therefore, while analogues can be extremely useful to calibrate our
understanding of how the system works, they are limited by the unique and dynamic
nature of future climate change.

3.1.3 The risk approach

Risk assessment is the process of identifying, evaluating, selecting, and implementing

actions to reduce risk to human health and to ecosystems (USPCC RARM as cited in
Jones 2001). Central to risk assessment is the management of uncertainties (see section
3.2), which allows the risk (i.e., probability times consequence) of some activity or
phenomenon to be determined. Risk assessment and risk management have been widely
applied to a number of environmental problems, but only very recently has this thinking
been introduced into climate change (see, e.g., Jones 2000b, Jones 2000a, Willows and
Connell 2003). Reasons for this include the high level of uncertainty associated with
climate change projections, the global nature of the problem and the difficulty of
attaching probabilities to different world development paths (Dessai and Hulme 2004).
Jones (2001) developed an environmental risk assessment/risk management framework

36
to assess the impacts of climate change on individual exposure units identified by
stakeholders as potentially vulnerable to climate change. This work was based on Carter
et al. (1994, i.e., the IPCC approach), but it incorporated features of risk assessment
frameworks. While the framework is conceptually simple, Jones acknowledges that it is
difficult to implement due to the complexity of climate change. The climate impact risk
assessment proposed by Jones (2001) has seven steps:

1. Identify the key climatic variables affecting the exposure units being assessed;
2. Create scenarios and/or projected ranges for key climatic variables;
3. Carry out a sensitivity analysis to assess the relationship between climate change
and impacts;
4. Identify the impact thresholds to be analysed for risk with stakeholders;
5. Carry out risk analysis;
6. Evaluate risk and identify feedbacks likely to result in autonomous adaptations;
7. Consult with stakeholders, analyse proposed adaptations and recommend
planned adaptation options.

Climate scenarios play an important role in this framework (step 2), like the IPCC
approach, but this particular risk approach is not totally scenario driven. It is more
dependent on stakeholder involvement and their definition of critical impact thresholds.
The analysis phase of the framework involves the linking of key climatic variables with
impact thresholds. The conditional probabilities of exceeding those thresholds are then
assessed within the context of projected ranges for key climatic variables under climate
change (Jones 2001).

The United Kingdoms Climate Impact Programme (UKCIP) published a guidance

report that helps decision-makers take account of the risk and uncertainty associated
with climate variability and change and identify good adaptation options (Willows
and Connell 2003). This generic framework is focused on institutional decision-making.
It is composed of eight stages:

1. Identify problem and objectives;

2. Establish decision-making criteria, receptors, exposure units and risk assessment
endpoints;

37
3. Assess risk (tiered);
4. Identify options;
5. Appraise options;
6. Make decision;
7. Implement decision;
8. Monitor, evaluate and review.

Unlike the previous examples, this framework does not have climate scenarios as a
specific stage of the framework. Instead, the use of climate scenarios is suggested as a
tool/technique amongst many others in stage 3 (risk assessment). They suggest that the
decision-maker adopt a different level of analysis (tier) depending on: the level of
decision; the level of understanding they have about how climate change will affect
his/her decision; and whether they are making a climate adaptation decision or a
climate-influenced decision. Tier 1 would focus on risk screening, tier 2 on qualitative
and generic quantitative risk assessment and tier 3 in specific quantitative risk
assessment. It is not clear if climate scenarios play different roles in each tier, but it
seems likely the higher the number of the tier (e.g., 3) the more climate scenarios will
be needed and the more quantification of uncertainty will be required. Within this
framework climate scenarios are just tools to help the risk assessment of climate change
on particular decision-making problems.

As the literature moved from impacts assessment to adaptation priorities the need for an
Adaptation Policy Framework (APF) emerged (Burton et al. 2002, Lim et al. 2005). The
APF, prepared by the United Nations Development Programme, aims to provide
guidance to developing countries for formulating national policy options for adaptation
to climate change. It builds on the IPCC approach, but presupposes adaptation to short-
term climate variability and extreme events will reduce vulnerability to longer-term
climate change (Lim et al. 2005): the essential starting point is the present (Burton et
al. 2002, p. 154). The APF is composed of five basic components, where engaging
stakeholders and enhancing adaptive capacity are crosscutting components:

1. Defining project scope;

2. Assessing current vulnerability;
3. Characterising future climate risks;

38
4. Developing an adaptation strategy;
5. Continuing the adaptation process.

Within the third component, one of the usual outputs is a set of future climate scenarios
and an analysis of associated risk. Climate scenarios certainly play a role in the APF,
but the uncertainty in predicting future climate has led the APF to anchor adaptation
assessments firmly with an understanding of current climate risk. According to the APF
this helps to provide a road map from known territory into uncertain futures. The APF is
perhaps the most mixed approach since it builds on the IPCC approach, it contains risk-
based approaches (step 3) and human development approaches (steps 2, 4 and 5).

3.1.4 Exploratory modelling and robustness analysis

While exploratory modelling has not yet been used in the context of adaptation to
climate change it is a particularly relevant approach (as shown in section 7.3).
Exploratory modelling was first introduced by Bankes (1993) and has since been used
in a number of climate change studies (Lempert et al. 1996, Lempert et al. 2000),
military studies, private sector planning, etc.13 Exploratory modelling (also known as
Computer-Assisted Reasoning) uses the capabilities of modern computing to combine
the best features of traditional quantitative decision analysis with those of narrative
scenario-based planning (Lempert 2002). Its supporters argue that it is completely
distinct from consolidative modelling: the use of models to consolidate knowledge into
a package that is used to predict system behaviour (similar to the IPCC approach and
advocates of climate prediction).

Bankes (1993) provides numerous examples and reasons why prediction using models
is doomed to fail, in particular in complex and uncertain systems. These are not repeated
here. Exploratory modelling, on the other hand, uses computational experiments to
explore the implications of varying assumptions and hypothesis. According to Bankes
(1993), even when a model is not validated it can serve as an inference engine,
showing where innocuous-appearing assumptions lead to predicted behaviours different
from initial expectations (p. 441). One of the problems of exploratory modelling is

13
See http://www.evolvinglogic.com for more examples.

39
how to select the limited number of experiments that can be run practically to best
inform the question of interest. The answer, according to Bankes (1993), is that there is
no mathematically rigorous strategy for sampling, instead human judgment has to be
used to prioritise the investigation of the uncertainties involved. Question-driven
exploratory modelling is particularly relevant for adaptation to climate change because
this type of analysis searches amongst an ensemble of plausible models to answer a
question of interest (such as: do climate change uncertainties matter for this decision?).
Bankes argues that exploratory modelling can provide a bridge for moving from
deductive analysis of closed systems, to interactive analytic support for inductive
reasoning about open systems where the contextual pragmatic knowledge possessed by
users can be integrated with quantitative data residing in the computer (2002, p.
7264).

In this thesis exploratory modelling is conducted under the name of robustness analysis.
Since definitions are so important, the definition of robustness analysis from Willows
and Connell (2003) is given below:

Robustness analysis may be used to help determine the robustness of the answers
within an options appraisal to possible uncertainties as to the values of key sensitive
variables and parameters (as identified from the sensitivity analysis). It identifies the
extent to which the decision-maker might be exposed to potential costs and errors if
some uncertain eventualities regarding these parameters should arise in future.
Robustness analysis is sometimes used to investigate the impact on the decision of a
reasonable range of input values for the key parameters identified by the sensitivity
analysis, or a range of values that considered plausible (p. 134-135).

The last characterisation of robustness analysis by Willows and Connell (2003) is

perhaps the closest to the application in this thesis.

40
3.1.5 The approach in this thesis

The methodology applied in this thesis combines several aspects of the frameworks
reviewed above. In this section a general overview of the methodology applied is
provided, but more detail is given in each individual chapter.

It should be clear by now that decisions (to adapt to climate change) are at the core of
this thesis. As mentioned in the introduction (section 1.2) these decisions are planned,
anticipatory and positive in their characteristics (cf. de Lo et al. 2001). A distinction
has been made between decision-makers and stakeholders for the sake of consistency.
Decision-makers are agents14 (could be single or multiple) who make judgments about
what actions need to be taken in order to adapt to climate change, and implement these
actions now or in the future. For example, in the context of public water supply the main
decision-makers are the water companies because they are the ones who have to do the
adapting. However, these adaptation decisions have to be agreed by the environmental
and economic regulators so they are also important decision-makers in this context.
Stakeholders include all those parties interested in the adaptation process and affected
by the decisions. So, for the public water supply example, depending on what adaptation
decision is being considered, stakeholders could include the small population of a
village affected by the extension of a reservoir and extend to all the customers of a
water company if the costs of adapting are borne by the customers. This distinction
between decision-makers and stakeholders is conceptually important because in this
thesis the focus is almost exclusively on decision-makers options (as shown in Chapter
6). However, in the real world stakeholders are very important if adaptation decisions
are to be accepted by society as a whole.15

The research described in this thesis focuses on the options available to decisions-
makers to adapt to climate change. A number of methods were used to research these
options. These included extensive literature reviews (based on peer-review and grey
literature), elite interviews with decision-makers and in-depth examination of planning
documents (see section 6.5). The adaptation options compiled included characteristics
14
Adger et al. (2005) mention individuals, firms, civil society, public bodies and governments (at local,
regional and national scales), and international agencies.
15
Furthermore, it could be argued that customers are in fact decision-makers since they have the power to
reduce demand to a certain extent, for example, by saving water, so the distinction is not straightforward.

41
of how much demand savings or supply increases could be achieved if the option was
implemented (in Millions of litres per day; Ml/d) so that the options could be linked
with changes in a climate variable or variables (e.g., temperature or precipitation). This
is perhaps one of the most difficult tasks of the assessment because agents adapt to a
number of different pressures simultaneously, not just to climate change.16 This task
was accomplished using planning documents that already incorporated climate change
and which included sensitivity analysis (AWS 2003, 2004b).

Climate change has been incorporated in water resources planning by combining

climate change scenarios with observed climate records which have served as input to
an impact model, which outputs changes in the variable of interest, for example, yield
change (more details given in Chapter 6). Using these results it is possible to link
changes in climate variables (e.g., summer temperature or precipitation change) with
impacts and subsequently with adaptation options required (to maintain a certain level
of service, for example). Part of this work has been conducted by the decision-maker so
this avoids running further simulations with impact models. Instead a relationship
(transfer function) is established (by linear regression) between changes in the climate
and water resources loss (or additional water required). Considering these are in-house
models they should be trusted by the decision-makers. Ideally one would run a number
of impact models to sample uncertainty in the models and also within models by
changing parameterisations and calibrations periods (see, e.g., Wilby 2005)

In order to link adaptation decisions with climate change uncertainties, regional climate
change probabilities were estimated. These focused essentially on seasonal temperature,
precipitation and potential evaporatranspiration change. There is a plethora of different
methods to estimate these climate variables as shown in section 2.4 (GCMs, RCMs,
statistical downscaling, etc.). The approach taken here tries to quantify as much of the
uncertainty as feasible without jeopardising the main objective of the thesis. This has
led to the use of the pattern-scaling technique (originally developed by Santer et al.
1990). This allows the analysis to be neatly divided into a global and regional scale. At
the global scale a simple energy-balance model (Wigley and Raper 1992, 1995, 2001)
was used to calculate future global mean temperature change. This model, known as

16
Adger et al. (2005) note that it is difficult to separate climate change adaptation decisions or actions
from actions triggered by other social or economic events.

42
MAGICC (Model for the Assessment of Greenhouse-gas Induced Climate Change),
allows the performance of uncertainty analysis on certain key parameters such as
emissions of GHGs, climate sensitivity, carbon cycle, ocean diffusivity and aerosol
forcing. A combination of literature review and expert advice was used to define
uncertainty ranges for these parameters. For example, for the emissions of GHGs,
scenarios from the IPCC SRES (Nakicenovic et al. 2000) were used, whereas for
climate sensitivity PDFs available in the literature (e.g., Andronova and Schlesinger
2001, Forest et al. 2002) were used. The connection between the simple climate model
and regional climate change is made using the pattern-scaling technique and results
from global and regional climate models. The GCM results originated from the IPCC-
Data Distribution Centre.17 The RCM results originated from the PRUDENCE project.18
A number of different forcings (either from different emissions scenarios or different
driving GCMs) and resolutions were sampled. Uncertainties were explored by testing a
number of different combinations including weighting models by skill (more details in
Chapter 5 and section 7.1.5). Figure 3.2 summarises some of the main features of the
framework described.

17
http://ipcc-ddc.cru.uea.ac.uk/
18
http://prudence.dmi.dk/

43
 Adaptation options
Are these strategies robust to all these climate change uncertainties?

Impact model
Relationship between regional climate change and impact

Regional climate change PDF (temperature, precipitation, evaporation)

GCM database RCM database

Two SRES scenarios (A2 and B2) Two SRES scenarios (A2 and B2)
Several models (9) Three driving GCMs (HadAM3, ECHAM4, ARPEGE)
Different spatial resolutions (12, 22, 50km)
Different initial conditions

Scaled by global temperature change PDF

MAGICC (Wigley and Raper, 2001)

Emissions of GHG: uses scenarios (either 6 marker scenarios or 4
scenario families)
Climate sensitivity: uses PDF (from the literature)
Carbon cycle: uses log- normal PDF based on expert judgement
Ocean diffusivity: uses uniform PDF based on expert judgement
Aerosol forcing: uses uniform PDF based on expert judgement

Figure 3.2. Schematic of the overall methodological framework used in this thesis.

It will become clear in the chapters ahead how the concepts reviewed in this section
relate to the methodology applied. For example, Jones (2001) mentions the
identification of key climatic variables affecting the exposure unit being assessed; this is
conducted in section 6.6 of this thesis. Climate change scenarios are constructed in
Chapters 4 and 5, while sensitivity analysis is performed in section 7.1. Risk analysis as
defined by Jones (2001) is what is referred here as uncertainty analysis (section 7.2).
Exploratory modelling is carried out under the name of robustness analysis in section
7.3.

44
3.2 Managing climate change uncertainties

uncertainty management is the raison detre of risk assessment, extreme care must
be exercised through out an assessment, so that uncertainties are identified, the nature
of their propagation throughout the assessment is understood and that they are
communicated as part of the results (Jones 2001, p. 207).

Since the management of climate change uncertainties is so critical to reach the aims of
this thesis, this section reviews previous efforts and describes how the issue is
approached here. The quote above by Jones (2001) neatly describes why uncertainty
management is so important. Jones (2000b) also noted that there is little evidence to
show which methods are best for managing uncertainty (for climate change), so a
number need to be tested under a range of conditions and for different purposes. This is
certainly one of the aims of this thesis.

Much has been written about uncertainty management (uncertainty assessment and
communication are included in this definition). The uncertainty literature is vast and
still growing (Knight 1922, Kahneman et al. 1982, Funtowicz and Ravetz 1990, Morgan
and Henrion 1990, French 1995, Helton and Burmaster 1996, Winkler 1996, Casman et
al. 1999, Stewart 2000, van Asselt 2000, Goodman 2002, Helton and Davis 2002, van
der Sluijs et al. 2003, Walker et al. 2003, van der Sluijs et al. 2004). For each proposed
taxonomy there is probably an equal number of different methodologies to handle
uncertainties. The basic premise in this thesis (which comes out of exploratory
modelling, risk assessment and the operational research literature, see e.g., French 1995)
is that there is uncertainty (sometimes deep uncertainty) about almost every aspect of a
climate risk assessment. There is uncertainty about what values various parameters
should have, uncertainty about the structure of the models, uncertainty about the options
available to decision-makers, uncertainty about the methodology, etc. The approach
taken here tries to quantify as much of the uncertainty as possible using knowledge from
the literature and expert judgement in the framework explained in the previous section.
A combination of methods is used to achieve this including: scenario analysis where
there is deep uncertainty (e.g., in emissions of GHGs), exhaustive fractile sampling (in

45
MAGICC) and Monte Carlo analysis using Latin Hypercube sampling for the pattern-
scaling and climate impacts (these methods are explained in more detail in the chapters
where they are applied). Each facet of the framework presented in Figure 3.2 has a
different uncertainty treatment because they represent different types of uncertainties
(but also sometimes for pragmatic reasons). It is important to discuss these different
types of uncertainties (in climate change) so that managing and communicating them
can be done appropriately. 19

On the nature of uncertainty, Hulme and Carter (1999) note that uncertainty about
climate change originates from both incomplete and unknowable knowledge. In a
general context these have been classified as epistemic (or subjective, type B,
reducible, and state of knowledge) and stochastic uncertainty (or aleatory, type A,
irreducible, ontic and variability) (Helton and Burmaster 1996, Helton and Davis 2002).
Epistemic uncertainty originates from incomplete knowledge of processes that influence
events. In relation to climate change, this type of uncertainty includes unknown values
for the climate sensitivity, the rate of heat uptake by the deep ocean or the
parameterisation of an impact model. Unknowable knowledge derives from the
indeterminacy of human systems and the unpredictability of the climate system.
Because global GHG emissions depend on human behaviour, they are inherently
uncertain; for example, the uncertainty of the future fertility rate or technology choices.
The climate system is also stochastically unpredictable to a certain extent because of its
chaotic nature, i.e., small differences in the initial conditions of a global climate model
can yield very different results (Lorenz 1993, Smith 2002). In the context of climate
change, incomplete knowledge matches perfectly the concept of epistemic uncertainty
in that by collecting more information, this type of uncertainty can be reduced (Parry
1996), although it is also possible that uncertainty increases with more research (as
shown in Chapter 4 with climate sensitivity). For complex systems, like climate change,
it is more likely than for simple well-constrained systems that this type of uncertainty
grows at first with more research and then starts converging as the science matures.

Representing epistemic uncertainty in a probabilistic fashion has been relatively widely

accepted, either through probability distributions based on scientific evidence or expert

19
The discussion that follows is largely based on Dessai and Hulme (2004).

46
judgements, although the aggregation of expert opinions is still controversial (see
Clemen and Winkler 1999 for a review). Nonetheless, in the past, scientists have been
uncomfortable with the notion that there was a subjective element to their analysis
(Moss 2000, for an example see Vaughan and Spouge 2002). This has led these
uncertainties to be sometimes ignored by overlooking available techniques for
improving subjective assessments of probability and confidence levels (Moss 2000)
and sometimes under-reported, especially in public policy studies of controversial or
politically sensitive issues, such as climate change (Pat-Cornell 1996).

In the case of climate change, unknowable knowledge does not translate solely into
stochastic uncertainty. This is where a split between natural and social scientists is most
noticeable. Stochastic (or aleatory) uncertainty stems from variability in known (or
observable) populations and, therefore, represents randomness in samples (Pat-Cornell
1996). This type of uncertainty arises in weather and climate prediction, which climate
scientists are overcoming with so-called ensemble simulations (Palmer 2000, 2001).
Dealing with stochastic uncertainty in the social sciences has been more problematic
and so attaching probabilities to world development paths and emissions of greenhouse
gases has been hotly debated (Grubler and Nakicenovic 2001). These authors argue that
probabilities in the natural sciences are different from probabilities in the social
sciences. Schneider (2002) provides a rebuttal to this argument, which is supported by
the author, but not repeated here.

Instead, the notion of reflexivity is emphasised, which some scientists call human
volition or feedback. Humans are capable of reflecting critically on the implications
of their behaviour and making adjustments in the light of experience (Berkhout et al.
2002). In a climate change context, if scientists state that global temperature will
increase between 1.4 and 5.8C by 2100 (with or without a probability distribution),
society will surely react. By critically reflecting upon this information, society will
create a perception of the problem (Is it good? Is it bad? Will it affect me or my
children?) and act upon it (even if it means doing nothing or continuing with business as
usual). This reaction in a collective form can conventionally take two forms: mitigating
the problem by reducing greenhouse gas emissions and enhancing sinks, and/or
adapting to the problem by devising and enhancing coping strategies to deal with the
impacts of a changing climate. By mitigating (or adapting to) the problem, people are

47
changing the future, which would lend the scientists' original statement incorrect had he
or she attached an estimate of likelihood. Thus, within the unknowable knowledge
sphere it is appropriate to introduce a new category of reflexive uncertainty, which
together with stochastic uncertainty provides a comprehensive picture of unknowable
knowledge in the context of climate change. Reflexive uncertainty only applies to
human systems because natural systems are not reflexive to information about the future
(predictions).20

In the case of the climate change problem, the fact that humans are part of the system
being researched makes the uncertainty irreducible in the context of prediction; it makes
all probabilities provisional. According to Slaughter (1994) predictions (with explicit
or implicit probability evaluations) are useless in the context of social systems where
qualitative phenomena relating to human choice are dominant. In the case of climate
change Sarewitz et al. (2003) note that the process of prediction for decision is hindered
by the fact that relationships that inform expert probabilities are themselves highly
non-stationary and perhaps influenced by the predictions themselves (p. 806).

Table 3.1. Characteristics of different types of uncertainty in the context of climate

change (Dessai and Hulme 2004).

Type of knowledge Type of uncertainty Possible to represent with

probabilities
Incomplete Epistemic Yes, but limited by knowledge
Incomplete-Unknowable Natural stochastic Yes, but with limits
Unknowable Human reflexive No, scenarios required

Table 3.1 tries to summarise the different types of uncertainty discussed here. Great care
is taken in this thesis to make sure the appropriate representation of uncertainty is used
for the type of knowledge being considered. For issues where human choice has a large
role to play for example, future emissions of GHGs or future water demand scenario
analysis has been applied. For parameters pertaining to the climate system, probabilities
(PDFs) have often been used. When there is little knowledge of some element, uniform

20
For example, it could be argued that the publications of the four SRES storylines have already made a
B1-type world more likely than an A2-type world, in which the impacts of climate change would be much
more pronounced. However, if one perceived to be following a B1 world one may become complacent
with regard to policy and thus move towards a more A2-like world. This crude example shows the
significant reflexive uncertainty social systems exhibit. For an explanation of the SRES scenarios see Box
4.1 in section 4.1.

48
distributions have often been used. In this thesis, human reflexive uncertainty is never
compounded with epistemic or natural stochastic uncertainty as is sometimes done in
the literature (see, e.g., Giorgi and Mearns 2003, Tebaldi et al. 2004). When both types
of uncertainty are present, uncertainty is propagated mathematically using probability
theory, but within a certain scenario. So probabilities are conditional on a particular
scenario or scenario family. This is the approach taken for the sensitivity and
uncertainty analysis (sections 7.1 and 7.2). Under the robustness analysis (section 7.3),
which follows the exploratory modelling approach, it is assumed that climate change
uncertainties (of any type) cannot be represented with probabilities because of the
presence of deep uncertainty. This leads to an entirely different thinking compared to
risk analysis as shown in section 7.3.

It is pertinent to conclude this section with a cautionary remark: probabilities and

scenarios are but two methods available to manage uncertainties in climate change,
perhaps the most widely used. However, there are other methods available to represent
uncertainty such as: Dempster-Shafer evidence theory, random sets, imprecise
probabilities, possibility theory, fuzzy sets, polynomial chaos expansions, info-gap
models, etc. While some of these methods have been applied to climate change (Bass et
al. 1997, Ha Duong 2003, Kriegler and Held 2005) and water resources (Luo and
Caselton 1997), others remain untested. It is also important to acknowledge that the
framework applied in this thesis addresses conditional probabilities, i.e., probabilities
that are conditional on the assumptions taken (based on best available information) and
the scenarios applied. With the advancement of knowledge these probabilities are likely
to change, hence their characterisation as conditional or provisional probabilities
(Dessai and Hulme, 2004).

49
 CHAPTER 4. ASSESSMENT OF UNCERTAINTIES AT
THE GLOBAL LEVEL

This chapter explores and assesses uncertainties at the global climate system level.
Uncertainty in global mean temperature is quantified in order to scale multiple climate
model response patterns in the next chapter. This climate information serves as the basis
to assess whether adaptation decisions are sensitive and robust to uncertainty in Chapter
7.

In this chapter, section 4.1 explains how uncertainties in future development pathways
and emissions of GHGs have been quantified. Section 4.2 looks at one of the most
critical parameters in climate change assessments: climate sensitivity. Section 4.3
elaborates on how a simple climate model was used to propagate uncertainties and
estimate future global mean temperature. Section 4.4 presents some results and section
4.5 summarises the findings of this chapter.

4.1 World development paths and emissions of greenhouse

gases

The IPCCs Second Assessment Report concluded that the balance of evidence suggest
a discernible human influence on global climate. The Third Assessment Report
concluded that most of the warming observed over the last 50 years is attributable to
human activities (IPCC 2001b, p. 10). These human activities to which the IPCC refers
are essentially related to emissions of GHG. How these emissions of GHGs will change
in the future is the starting point of almost all climate change assessments.

Estimating future emissions of GHGs is difficult because of the complex interactions

between driving forces such as demographic development, socio-economic
development, and technological change (Nakicenovic et al. 2000). Their future
evolution is characterised by deep uncertainty, so scenario analysis has been the
dominant approach to quantifying uncertainties in future GHG emissions (although
there are exceptions as section 2.1 showed). Scenarios are stories of how the future
might unfold which explore what if type questions. They have been used in post-

50
World War II military planning, business strategy development for major corporations
(e.g., Shell) and more recently to planning for sustainable development (e.g., the IPCC
SRES and the recent Millennium Ecosystem Assessment) (Schwartz 1991, van der
Heijden 1996). According to Nakicenovic et al. (2000) scenarios are an appropriate tool
with which to analyse how driving forces may influence future emission outcomes and
to assess the associated uncertainties. An important feature of scenarios is that they
should be internally consistent, i.e., for example, a scenario with very high population
growth could not have been driven by low fertility rates.

There are a number of global future scenarios in the literature (see, e.g., Raskin et al.
1998), but not all include quantitative estimates of the evolution of GHG and other
aerosols. This is a pre-requisite to run the simple climate model. Results from the IPCC
SRES (Nakicenovic et al. 2000) are used in this thesis because they provide a mix of
qualitative scenarios in the form of narrative storylines with quantitative estimates
performed by six different modelling approaches. Besides being considered the state-of-
the-art in the field, these scenarios were used because they have correspondence with
scenarios used in the UK, such as the Foresight Environmental Futures (DTI 1999,
OST 2002), UKCIP socio-economic scenarios (Berkhout and Hertin 2000, UKCIP
2000, Berkhout et al. 2002, Hulme et al. 2002) and Environment Agency water demand
scenarios (EA 2001a). Furthermore, a recent critical assessment of the IPCC SRES
scenarios concluded that while the scenarios leave much to be desired, the range of
future GHG emissions remains undisputed (Tol et al. 2005).

The IPCC SRES writing team developed a range of individual scenarios following on
extensive literature assessment, six alternative modelling approaches, and an open
process that solicited wide participation and feedback. They cover a wide range of the
main demographic, technological and economic driving forces for GHG and sulphur
emission. These scenarios do not include explicit mitigation measures or policies
(additional climate policy initiatives), although they necessarily encompass various
policies of other types, some of which have the effect of reducing emissions.

Each scenario links one of four narrative storylines with one particular quantitative
model interpretation. All the scenarios based on a specific storyline constitute a scenario
family. Box 4.1 briefly describes the four narrative storylines. Each storyline

51
represent the playing out of different social, economic, technological and environmental
developments (or paradigms), which may be viewed positively by some people and
negatively by others. Possible surprise and disaster scenarios were excluded
(Nakicenovic et al. 2000).

Box 4.1 The main characteristics of the four SRES storylines

The A1 storyline describes a future world of very rapid economic growth, low
population growth and rapid introduction of new and more efficient technology.
Major underlying themes are convergence among regions, capacity building and
increased cultural and social interaction, with a substantial reduction in regional
differences in per capita income.

The A2 storyline describes a very heterogeneous world. The underlying theme is self-
reliance and preservation of local identities. Fertility patterns across regions converge
very slowly, resulting in high population growth. Economic development is primarily
regionally-oriented, and per capita economic growth and technological change are
more fragmented and slow compared to other storylines.

The B1 storyline describes a convergent world with rapid change in economic

structures toward a service and information economy, reduction in material intensity
and the introduction of clean and resource-efficient technologies. The emphasis is on
global solutions to economic, social and environmental sustainability, including
improved equity, but without additional climate initiatives.

The B2 storyline describes a world in which the emphasis is on local solutions to

economic, social, and environmental sustainability. It is a world with less rapid, and
more diverse technological change, but with a strong emphasis on community
initiative and social innovation to find local and regional solutions. While policies are
also oriented towards environmental protection and social equity, they are focused on
local and regional levels.

Source: Nakicenovic et al. (2000)

In addition, scenarios in the A1 family were categorised into three groups according to
their technological emphasis: on coal, oil and gas (A1FI), non-fossil energy sources
(A1T) or a balance of all three (A1B). The SRES team used six different integrated
assessment models (IAMs): AIM, ASF, IMAGE, MARIA, MESSAGE and MiniCAM
that are representative of different modelling approaches and different integrated
assessment frameworks in the literature (Table 4.1).

52
Table 4.1 Models used to generate the SRES scenarios.

Model Acronym Source Reference

Asian Pacific Integrated Model AIM National (Morita et al. 1994)
(AIM) Institute of
Environmental
Studies in Japan
Atmospheric Stabilization ASF ICF Consulting (Sankovski et al.
Framework Model (ASF) in the USA 2000)
Integrated Model to Assess the IMA RIVM in The (Alcamo 1994,
Greenhouse Effect (IMAGE) Netherlands Alcamo et al. 1998,
de Vries et al. 1999)
Multiregional Approach for MI Science (Mori and Takahashi
Resource and Industry University of 1998)
Allocation (MARIA) Tokyo in Japan
Model for Energy Supply MES IIASA in (Riahi and Roehrl
Strategy Alternatives and their Austria 2000)
General Environmental Impact
(MESSAGE)
The Mini Climate Assessment MIN PNNL in the (Edmonds et al.
Model (MiniCAM) USA 1996)

Figure 4.1 shows how different integrated assessment models (AIM, MI, MES, ASF,
IMA) have converted the six illustrative SRES scenarios (A1B, A1FI, A1T, A2, B1 and
B2) into CO2 emissions up to 2100. The illustrative marker scenarios (A1B, A2, B1 and
B2) are no more or less likely than any of the other scenarios, but are considered by the
SRES writing team as illustrative of a particular storyline (Nakicenovic et al. 2000). By
2100, emissions of CO2 could range between halving to quadrupling compared to 1990
levels. This means that by 2030 the uncertainty range is in the order of 6 Gt C whereas
in 2100 it is 25 Gt C.

53
 30
A1B-AIM
25 A1FI-MI
CO2 emissions (Gt C)
A1T-MES
20 A2-ASF
B1-IMA
15 B2-MES

10

0
1990 2010 2030 2050 2070 2090
Year

Figure 4.1 Total global anthropogenic emissions of carbon dioxide (Gt C) from
fossil fuel burning and deforestation, from 1990 to 2100, for the six illustrative
SRES scenarios (A1B, A1FI, A1T, A2, B1 and B2) using different integrated
assessment models (AIM, MI, MES, ASF, IMA).

Figure 4.2 shows the evolution of CO2 emissions for all the scenarios explored within
the A2 storyline with different IAMs. Figure 4.3 shows scenarios run under the B2
storyline. Both figures show a substantial amount of uncertainty in translating storylines
into emissions by different energy-economy-environment IAMs. Generally, however,
only the SRES marker scenarios (black dotted lines in Figures 4.2 and 4.3) are used as
input to climate model simulations. In the A2 scenario family the SRES marker scenario
(A2-ASF) is in the top range of realisations (Figure 4.2), whereas in the B2 family the
SRES marker scenario (B2-MES) is in the bottom range of realisations (Figure 4.3).
Table 4.2 shows how the uncertainty range (measured as the difference in emissions
between the highest and lowest scenario) changes with time.

54
 35
A2-ASF
30 A2A1MI
CO2 emissions (Gt C) A2AIM
25
A2MES
20 A2MIN
A2GIM
15

10

0
1990 2010 2030 2050 2070 2090
Year

Figure 4.2. Total global anthropogenic emissions of carbon dioxide (Gt C) from
fossil fuel burning and deforestation, from 1990 to 2100, for the SRES A2 scenario
family and all its scenarios (using different IAMs). The marker scenario (A2-ASF)
is shown in black with dots.

25
B2-MES
B2AIM
20
CO2 emissions (Gt C)

B2ASF
B2IMA
15 B2MIN
B2HIMI
10

0
1990 2010 2030 2050 2070 2090
Year

Figure 4.3. Total global anthropogenic emissions of carbon dioxide (Gt C) from
fossil fuel burning and deforestation, from 1990 to 2100, for the SRES B2 scenario
family and all its scenarios (using different IAMs). The marker scenario (A2-ASF)
is shown in black with dots.

55
Table 4.2. Uncertainty range (in Gt C) between the highest and lowest scenarios
run by different integrated assessment models under the A2 and B2 SRES scenario
family.

1990 2000 2010 2020 2030 2040 2050 2060 2070 2080 2090 2100
A2 family 0.00 0.00 1.83 3.75 6.22 7.45 7.96 6.94 7.20 7.30 10.1114.91
B2 family 0.00 0.00 1.19 3.68 4.85 4.82 6.01 7.16 8.19 8.95 9.83 10.78

The uncertainty is large in the near future (e.g. 2020s) and it gets bigger as time
progresses. Uncertainty in the A2 family peaks at 2050 and then it decreases for 30
years before increasing again. The uncertainty range for the B2 family in 2100 (10.78
Gt C) is almost as large as emissions for the B2-MES SRES marker scenario (13.32 Gt
C). It is also worth noting that this uncertainty range is much larger than some of the
differences between storylines as shown in Figure 4.1.

This uncertainty of how the IAMs convert storylines into emissions is commonly
forgotten in climate modelling studies as only the SRES marker scenarios (usually A2-
ASF and B2-MES, but sometimes including others as well) have been used as input
forcings in GCM simulations. This has led Stott and Kettleborough (2002) to conclude
that global mean temperature change is insensitive to uncertainty in emissions of GHGs
over the next 40 years. This is partially true because near term temperature changes are
mostly dependent on past emissions due to the long lifetime of CO2 and other GHGs.
However, if all the modelled SRES scenarios of each scenario family had been included
in such analysis, uncertainty in GHG emissions might not be entirely negligible. In this
thesis, uncertainty associated with different storylines and uncertainty associated with
the conversion of storylines into GHG emissions by IAMs are both included in the
assessment. To the knowledge of the author, the latter uncertainties have never been
considered explicitly in climate modelling or impact and adaptation research.

4.1.1. Mitigation and stabilisation scenarios

The scenarios described above are no-climate policy scenarios, i.e., they have no
explicit policies to tackle climate change. Since mitigation policies are being enacted
today (e.g., the Kyoto Protocol or the UK carbon levy) and are likely to continue, these
scenarios could be deemed unrealistic. Therefore, mitigation and stabilisation scenarios

56
are also analysed here. This will allow the examination of whether pursuing climate
mitigation policies make a difference to climate change impacts and adaptation
decisions (section 7.1.2.1).

Following the publication of the SRES scenarios, nine modelling teams participated in a
special comparison programme to quantify SRES-based mitigation scenarios. The so-
called Post-SRES mitigation scenarios have the original SRES as baseline scenarios.
These mitigation scenarios were quantified on the basis of storylines for each of the six
SRES scenarios that describe the relationship between the kind of future world and the
capacity for mitigation. Over seventy stabilisation scenarios were produced for several
concentrations of atmospheric CO2, including 450, 550, 650, and 750ppmv. Swart et al.
(2002) analysed 18 of these scenarios on the basis of their comprehensiveness in terms
of GHGs, an assessment of their scientific soundness, and a preference for modelling
groups who also participated in the SRES exercise (to enhance the comparability with
the SRES reference cases). In this thesis, 13 post-SRES scenarios are examined (Figure
4.4 and Table 4.3) and compared with their corresponding baseline SRES scenario. As
one would expect, Figure 4.4 shows that the lower the stabilisation level the lower the
emissions of carbon dioxide. It also makes sense that scenarios with low emissions of
GHGs such as B1-IMA and AIT-MES, do not have stabilisation runs above 550 ppm,
because they are already below that level (Table 4.3).

M a r k e r ' s t o t a l C O 2 emission scenarios

18 A2-750 ASF
A1FI-750 MiniCAM
16
A1B-650 AIM
14
A1FI-650 MiniCAM
12 A1B-550 AIM
A2-550 ASF
10
GtC

B2-550 MESSAGE
8
A1T-550 MESSAGE
6 A1FI-550 MiniCAM
A1B-450 AIM
4
B1-450 IMAGE
2
A1T-450 MESSAGE
0 A1F1-450 MiniCAM
1990 2010 2030 2050 2070 2090

Figure 4.4 CO2 emissions for Post-SRES stabilisation scenarios (at different
stabilisation levels: 450 ppm in green; 550 ppm in blue; 650 ppm in black and 750

57
ppm in red) for each SRES illustrative scenario (A1B, A2, B1 and B2, A1FI and
A1T). Figure supplied by Morita, 2001.

Table 4.3. Post-SRES stabilisation scenarios used in this thesis (arranged by

stabilisation level: 450, 550, 650 and 750 ppm) for each SRES illustrative scenario.
Scenarios in grey are illustrated in Figure 4.5.

450 ppm 550 ppm 650 ppm 750 ppm

A1B-AIM X X X
A1FI-MI X X X X
AIT-MES X X
A2-ASF X X
B1-IMA X
B2-MES X

Figure 4.5 shows how one of the illustrative SRES scenario (A1B) baseline compares
with the stabilisation scenarios (450, 550, 650 ppm). It is clear that considerable
emissions reductions are achieved if mitigation policies aimed at stabilising
concentrations are pursued.

18
16
CO2 emissions (Gt C)

14
12
10
8
6 A1B-AIM
4 A1B-650-AIM
A1B-550-AIM
2
A1B-450-AIM
0
1990 2010 2030 2050 2070 2090
Year

Figure 4.5. Global CO2 emissions under the illustrative baseline scenario A1B-AIM
and its respective stabilisation scenarios (650, 500, 450 ppm)

58
4.2 Climate sensitivity

Climate sensitivity is an important parameter in climate change research and one that
can be prescribed in MAGICC. It is examined in this thesis in order to determine
whether adaptation decisions are sensitive to uncertainty in this parameter. Climate
sensitivity is the equilibrium change in global surface temperature resulting from a
doubling of atmospheric equivalent CO2 concentration. As the name suggests if climate
sensitivity is high then climate change may be an extremely serious problem, but if
climate sensitivity is low then it might be a manageable problem.

The concept of climate sensitivity has more than a century of history. Andronova et al.
(2005) provide an historic account of the development of climate sensitivity from
Arrhenius to the present day (see also, van der Sluijs et al. 1998). Recent studies that
incorporate uncertainty in climate sensitivity into important societal decision-making
options (e.g., Caldeira et al. 2003, Yohe et al. 2004) demonstrate the need for adequate
quantification of this uncertainty, a goal that remains elusive. As mentioned in the
literature review (section 2.2) and reviewed in Andronova et al. (2005) there are a
number of studies that have estimated the uncertainty associated with climate sensitivity
by constraining models with observations (Tol and de Vos 1998, Andronova and
Schlesinger 2001, Forest et al. 2002, Gregory et al. 2002, Knutti et al. 2002, Murphy et
al. 2004). The author of this thesis contacted the lead author of each of these
publications to obtain climate sensitivity PDFs from each study.

Figure 4.6 shows the assembled estimates of climate sensitivity as cumulative

distribution functions (CDFs) as well as estimates from GCMs used in the IPCC TAR
(Cubasch et al. 2001). These independent estimates indicate that it is exceptionally
unlikely (i.e, < 1% probability in IPCC terminology) that climate sensitivity is below
0.5C; this is a particularly robust finding. The central value (50th percentile) ranges
from 2 to 6C while the upper bound range (for instance the 95th percentile) is even
wider. The studies represented in Figure 4.6 use a wide range of models (simple climate
models, statistical models, intermediate complexity models and even a GCM), apply a
number of different techniques (optimal fingerprinting, bootstrapping or Bayesian
techniques) and constrain parameters using many different variables (sulphate aerosols,

59
solar forcing, surface temperatures or ocean temperatures). Until recently, GCMs had
not been used to characterize uncertainty in climate sensitivity, mainly due to high
computational costs, but as Murphy et al (2004) have shown, this is changing. Figure
4.6 also shows the IPCC range of climate sensitivity (1.5 to 4.5C) in the horizontal
axis, which has remained unchanged since the publication of the First Assessment
Report of IPCC in 1990.

Figure 4.6 Cumulative distribution functions of climate sensitivity (the equilibrium

response in global surface temperature change to a doubling of equivalent CO2
concentration) for a number of studies. On the bottom axis the climate sensitivity
of individual GCMs used in the IPCC TAR (diamonds) is indicated and also the
IPCC consensus range (thick black line in the horizontal axis). Modified after
Dessai and Hulme (2004).

4.3 Using a simple energy balance model to propagate

uncertainties at the global scale

This section uses a simple climate model to propagate uncertainty at the global level to
serve as input to the next chapter, for the calculation of regional climate change
probabilities. One simple climate model that has been widely used in IPCC reports since
1990 and in various other applications (such as IAMs) is that developed by Wigley and
Raper (1992, 1995, 2001). MAGICC, as the model is known, is a synchronously
coupled carbon cycle/gas cycle/upwelling-diffusion energy-balance climate model
(Wigley and Raper 1992, 1995). In MAGICC, emissions are converted to atmospheric

60
composition changes (or directly to radiative forcing for SO2 emissions) using a set of
gas-cycle models. In the IPCC TAR, MAGICC was tuned to match results from a
number of GCMs in order to produce projections of future global temperature change
and sea level rise.

This section uses the probabilistic version of the Wigley and Raper (2001) model
(known as MAGLOOP) to estimate global mean temperature change and its associated
uncertainties. MAGLOOP considers, in a probabilistic way, uncertainties arising from:
future emissions of GHGs and other aerosols, the climate sensitivity, ocean mixing
(through the ocean models vertical diffusivity), the carbon cycle (by considering
uncertainties in the CO2 fertilization factor and the air-sea CO2 flux, and in various
temperature feedbacks) and aerosol forcing (direct and indirect). These parameters were
chosen by Wigley and Raper (2001) as they were identified as contributing the most to
uncertainties in output. Having identified the key uncertain parameters in their model,
Wigley and Raper (2001) defined PDFs for these uncertain parameters based on
information in the literature and expert judgement. In terms of sampling strategy,
MAGLOOP uses exhaustive fractile sampling, which considers all possible
combinations of results for the division of the input distribution into prespecified
fractiles (Wigley and Raper 2001, p. 453). This method is similar to Latin Hypercube
Sampling which divides the PDFs into equally probable fractiles, and randomly samples
the fractiles (using median values within the fractiles to represent the fractile value). In
MAGLOOP there are 25 fractiles for climate sensitivity and 5 fractiles for each of the
other three parameters. For 35 emissions scenarios that makes a total of 109,375
simulations (3525555).21 Probabilistic projections are based on the assumption
that the model structure is sufficiently realistic that it can be used outside its calibration
zone.

4.3.1 Experimental setup

This section describes and justifies the PDFs (or scenarios) used for the five key
uncertain parameters in MAGLOOP. Future emissions of GHGs and other aerosols are
deeply uncertain because of their dependency on human choice over the next 100 years

21
This takes roughly five hours to run in an average 2001 computer.

61
(see section 4.1). For this reason PDFs are not used to describe emissions of greenhouse
gases (as has been done by some authors in the past; see section 2.1 for examples),
instead scenario analysis is used (see section 2.2 for the reasoning on this). Wigley and
Raper (2001) also used the SRES scenarios, but they combined all 35 complete SRES
scenarios (i.e., they assumed they were all equally plausible) to construct one PDF of
global mean temperature change. This is not done here as it conflates uncertainty from
different storylines. Therefore, experiments are conducted using the four illustrative
marker scenarios (A1B-AIM, A2-ASF, B1-IMA and B2-MES) and two additional
illustrative scenarios (A1FI-MI and AIT-MES) separately or using all scenarios from
each of the four scenario families (A1, A2, B1 and B2). The latter approach has the
advantage of representing uncertainty on how different IAMs convert storylines into
GHG emissions (and in the A1 case it also includes uncertainty about technology).
Table 4.4 shows the 35 SRES scenarios used in this analysis.

Table 4.4. Scenarios used in the analysis ordered by storyline families. Illustrative
scenarios are distinguished by being in bold.22

SRES A1 SRES A2 SRES B1 SRES B2

A1B-AIM A2A1-MIN B1-AIM B2-AIM
A1B-ASF A2-AIM B1-ASF B2-ASF
A1B-IMA A2-ASF B1-IMA B2-IMA
A1B-MES A2-MES B1-MES B2-MES
A1B-MIN A2-MIN B1-MIN B2-MIN
A1C-AIM A2G-IMA B1HI-MES B2HI-MIN
A1C-MES B1HI-MIN
A1C-MIN B1T-MES
A1G-AIM
A1G-MES
A1FI-MIN
A1T-AIM
A1T-MES
A1V1-MIN
A1V2-MIN

22
In the case of the A1 scenario family several scenarios groups were considered (A1C: large scale
synfuel economy based on coal; A1G: massive development of unconventional oil and gas resources;
A1FI: the combination of A1C and A1G into one fossil intensive group; A1T: technology focus on
renewable energy sources ; A1B: balanced mix of technologies and supply sources). V1 and V2 relate to
alternative interpretations of the A1 scenario storyline with different demographic, economic, and energy
development patterns. A2A1 refers to an alternative interpretation of the A2 scenario storyline in the form
of a delayed development or transitional scenario between the A2 and A1 scenario families. HI
(High) refers to alternative interpretations on rates and potentials of future dematerialisation and energy-
intensity improvements.

62
There are two main reasons to keep storylines separate. The first and most important
one is not to conflate different types of uncertainties. Scenario analysis was used in the
IPCC SRES because of the presence of deep uncertainty. According to the typology of
uncertainties introduced in Chapter 2, future emissions of GHG fall under the category
of human reflexive uncertainty, which can only be represented using scenarios. Thus, it
is important to keep the scenario storylines separate so that further down the assessment,
robust adaptation decisions can be identified, i.e., decisions that work reasonably well
under all scenario storylines. Second, by sampling each scenario storyline independently
one is actually sampling the inter model uncertainty that exists when different IAMs of
energy, economy and environment translate the storyline into quantified GHG
emissions. Simulations using the illustrative scenarios, which have been the main inputs
into Global Climate Models, are compared to results using all the scenarios from each
storyline.

In order to examine the impact of mitigation/stabilisation initiatives on adaptation

decisions, mitigation/stabilisation runs have also been conducted using the simple
climate model. This uses the 13 Post-SRES scenarios shown in Table 4.3 and Figure 4.4
in Section 4.1.1. Compared to their SRES scenario baselines, the Post-SRES scenarios
only have different CO2 emissions, i.e., the other GHGs and aerosols remain the same
(this simplification is a limitation).

Climate sensitivity is an important uncertain parameter as section 4.2 has shown. In this
analysis a number of PDFs from the literature (Andronova and Schlesinger 2001, Forest
et al. 2002, Knutti et al. 2002, Murphy et al. 2004, Stainforth et al. 2005) and expert
judgement (Wigley and Raper 2001) are used. The Wigley and Raper (2001) PDF
originates from a log-normal interpretation of the IPCCs 1.5 to 4.5C range. The results
from Stainforth et al. (2005) are here interpreted as a uniform distribution from 1-11
degrees C (called Uniform cp.net) because these results have not been constrained by
any sort of observed data; it is only the raw data from the uncertainty analysis. Figure
4.7 shows the 25 sensitivity values used in MAGLOOP for each study.

63
Figure 4.7 Climate sensitivity CDFs used in this analysis. The dots in each line
represent the 25 values used to define climate sensitivity in MAGLOOP.

Wigley and Raper (2001) determined carbon cycle uncertainties offline because of the
large number of parameters that might potentially contribute to this uncertainty. They
arrived at five quintile values for three uncertain input parameters associated with
carbon cycle. For ocean mixing five quintiles were used for different values of the
vertical diffusivity: 1.47, 1.92, 2.3, 2.76 and 3.61 cm2/sec. Note that the medium value
(2.3) is the IPCC TAR best guess. Aerosol forcing uncertainty is also represented using
quintile values for four uncertain input parameters (1990 value of direct aerosol forcing,
indirect aerosol forcing, biomass aerosol forcing and fossil organic + black carbon),
which when added have this distribution: -0.58, -1.051, -1.3, -1.6 and 1.819 W/m2. The
IPCC TAR best guess for aerosol forcing is the medium value used again. Figure 4.8
shows the uniform distributions assumed for ocean mixing and aerosol forcing in this
analysis.

64
 0.25 0.25

0.20 0.20

Probability

Probability
0.15 0.15

0.10 0.10

0.05 0.05

0.00 0.00
1.47 1.92 2.3 2.76 3.61 -0.58 -1.05 -1.30 -1.60 1.819
Vertical diffusivity (cm2/sec) Aerosol forcing (W/m2)

Figure 4.8 Uniform PDFs used to represent uncertainty in ocean mixing (left
panel) and aerosol forcing (right panel).

4.3.2 Results at the global level

Using the probabilistic simple climate model with the PDFs and scenarios described
produces a large amount of output. Since this thesis is mostly concerned with the
sensitivity of adaptation decisions it is not important to present all the results, but to get
an overall view of the uncertainties at the global level.

Figure 4.9 shows the probability of global mean temperature change for the end of the
century (compared to 1990) under the six illustrative SRES scenarios. Under these
scenarios and climate sensitivity PDF (Wigley and Raper 2001), global temperature
could increase from less than 1C to almost 7.5C, which means there is a large range of
uncertainty. The magnitude of global temperature change by 2100 appears to be heavily
dependent on SRES scenarios, with B1-IMA producing the least global warming and
A1FI-MI the maximum in the range. Figure 4.10 shows the evolution of global
temperature change with time under the six illustrative SRES scenarios. The SRES
scenarios introduce less uncertainty in the near term compared to the end of the century.
It is interesting to note that illustrative scenario A1T-MES is the warmest in 2030 the
second warmest in 2050 and the second least warmest in 2100 (compared to the other
scenarios). This change in ranking occurs because illustrative scenario A1T-MES has
the lowest emissions of sulphur dioxide (SO2) throughout the century whereas most
other scenarios increase in the near future and then decrease towards the end of the
century. Sulphur dioxide has a negative radiative forcing, which means it cools the
atmosphere, but by how much is still not well known, hence the PDF assumed for
aerosol forcing in the simple climate model. This is also observable with A2-ASF:

65
lowest warming in 2030, second lowest warming in 2050 and second highest in 2100.
This occurs because A2-ASF has the highest emissions of sulphur dioxide compared to
the other scenarios.

0.08
0.07 A1B-AIM
A1FI-MI
0.06
A1T-MES
Probability

0.05
A2-ASF
0.04
B1-IMA
0.03
B2-MES
0.02
0.01
0
0 1 2 3 4 5 6 7 8
Global temperature change (C)

Figure 4.9 Probability of global temperature change for 2100 compared to 1990 for
the six illustrative SRES scenarios using the Wigley and Raper (2001) PDF for
climate sensitivity and PDFs of other parameters as described in section 4.3.1. 3125
simulations were carried out for each scenario.23 Bins of 0.1C.

0.3

0.25 2030 A1B-AIM

A1FI-MI
A1T-MES
A2-ASF
0.2
Probability

B1-IMA
B2-MES
A1B-AIM
0.15 2050

0.1
2100
0.05

0
0 1 2 3 4 5 6 7 8
Global temperature change (C)

Figure 4.10 Same as Figure 4.9 but including different time horizons (2030, 2050
and 2100).

23
1 scenario x 25 climate sensitivities x 5 carbon cycle x 5 ocean mixing x 5 aerosol forcing.

66
Figure 4.11 shows the probability of global temperature change by 2100 for the
different SRES scenario families (A1, A2, B1 and B2), assuming each scenario within
the family is equally plausible. This is a reasonable assumption, that takes a multi-
model approach to how IAMs convert storylines into GHG emissions Because more
scenarios are used than in Figure 4.9 the PDFs are smoother in Figure 4.11. Also
evident is the distinctively flatter distribution of the A1 family compared to the other
scenario families. This occurs because the A1 family includes scenarios with alternative
developments of energy technologies (coal, oil and gas; non-fossil energy sources; and a
balanced approach), which under this approach are all equally plausible.

0.06

0.05 A1
Probability

0.04 A2
B1
0.03
B2
0.02

0.01

0
0 1 2 3 4 5 6 7 8
Global temperature change (C)

Figure 4.11 Probability of global temperature change for 2100 compared to 1990
for each SRES scenario family using the Wigley and Raper (2001) PDF for climate
sensitivity and PDFs of other parameters as described in section 4.3.1. Bins of
0.1C.

In Figures 4.2 and 4.3 it was shown that the illustrative SRES marker scenarios are
either in the upper (for the A2 scenario family) or lower (for the B2 scenario family)
range of carbon dioxide emissions of the scenarios considered in each family. Figure
4.12 compares global mean temperature change calculated using the illustrative SRES
marker scenarios (A2-ASF and B2-MES) and all the other scenarios within the A2 and
B2 scenario family. Using the marker scenarios skews the distribution considerably
compared to using the entire scenario family.

67
 0.07
0.06
A2-ASF
0.05

Probability
B2-MES
0.04 A2
0.03 B2
0.02
0.01
0
0 1 2 3 4 5 6 7
Global temperature change (C)

Figure 4.12 Probability of global temperature change for 2100 (compared to 1990)
for SRES family A2 and B2 and their respective marker scenario (A2-ASF and B2-
MES) using the Wigley and Raper (2001) PDF for climate sensitivity and PDFs of
other parameters as described in section 4.3.1.

Figure 4.13 shows the impact of mitigation initiatives in the baseline illustrative
scenarios in 2100. In almost all stabilisation scenarios global temperature change is
reduced by mitigation (the only exception is A1T-550 which remains essentially the
same as the baseline24). High global temperature change baseline scenarios (such as
A1FI-MI) have the largest reductions. Figure 4.14 shows how mitigation/stabilisation
initiatives impact on global temperature change over time for one particular marker
scenario (A1B). The impact of mitigation is clearly more discernable in the long term
than in the short term as one would expect.

24
This probably occurs because the A1T baseline concentration level is close to 550ppm.

68
 0.08 0.08
A1FI-MI
0.07 A1B-AIM 0.07
A1FI-750
0.06 A1B-650 0.06
A1FI-650

Probability

Probability
0.05 A1B-550 0.05
A1FI-550
0.04 A1B-450 0.04 A1F1-450
0.03 0.03
0.02 0.02
0.01 0.01
0 0
0 1 2 3 4 5 6 7 8 0 1 2 3 4 5 6 7 8
Global Temperature Change (C) Global Temperature Change (C)

0.08 0.08
0.07 0.07
A1T-MES A2-ASF
0.06 0.06
A1T-550 A2-750
Probability

Probability
0.05 0.05
A1T-450 A2-550
0.04 0.04
0.03 0.03
0.02 0.02
0.01 0.01
0 0
0 1 2 3 4 5 6 7 8 0 1 2 3 4 5 6 7 8
Global Temperature Change (C) Global Temperature Change (C)

0.08 0.08
0.07 0.07
0.06 B1-IMA 0.06 B2-MES
Probability

Probability

0.05 B1-450 0.05 B2-550

0.04 0.04
0.03 0.03
0.02 0.02
0.01 0.01
0 0
0 1 2 3 4 5 6 7 8 0 1 2 3 4 5 6 7 8
Global Temperature Change (C) Global Temperature Change (C)

Figure 4.13 Probability of global temperature change by 2100 compared to 1990

for each marker scenario (A1B-AIM, A1FI-MI, A1T-MES, A2-ASF, B1-IMA and
B2-MES) and their respective stabilisation scenarios using the Wigley and Raper
(2001) PDF for climate sensitivity and PDFs of other parameters as described in
section 4.3.1.

0.3
2030
0.25
A1B-AIM
A1B-650
0.2 A1B-550
Probability

A1B-450
2050 A1B-AIM
0.15

0.1
2100
0.05

0
0 1 2 3 4 5 6
Global Temperature Change (C)

Figure 4.14 Probability of global temperature change for different time horizons
(2030, 2050 and 2100) for the A1B-AIM baseline marker scenario and its

69
respective stabilisation scenarios (650, 550 and 450 ppmv) using the Wigley and
Raper (2001) PDF for climate sensitivity and PDFs of other parameters as
described in section 4.3.1.

Up to now all the results shown have used the Wigley and Raper (2001) climate
sensitivity PDF. Figure 4.15 shows global temperature change by 2100 for two scenario
families (A2 and B2) using a number of different climate sensitivity PDFs (shown in
Figure 4.7). For one climate sensitivity PDF (Andronova and Schlesinger 2001) global
temperature change ranges from 1C to beyond 9C for the A2 scenario family. The
mode of the various distributions (the most likely value) ranges from about 2 to 5C for
the A2 scenario family. It is noticeable that the distributions under the B2 scenario
family produce less warming than in the A2 scenario family; this is consistent with
Figure 4.11.

70
 0.06
W&R (2001)

0.05 A&S (2001)

Knut t i et al. (2002)

0.04
Probability
Unweight ed Murphy et al. (2004)

Weight ed Murphy et al. (2004)

0.03 Uni f orm cp.net ( 2005)

Uni f orm Forest et al. (2002)

0.02
Exper t Forest et al. (2002)

0.01

0
0 1 2 3 4 5 6 7 8 9 10
Global Temperature Change (C)

0.06
W&R (2001)

0.05 A&S (2001)

Knut t i et al. (2002)

Probability

0.04 Unweight ed Murphy et al. (2004)

Weight ed Murphy et al. (2004

0.03 Uni f orm cp.net ( 2005)

Uni f orm Forest et al. (2002)

0.02
Exper t Forest et al. (2002)

0.01

0
0 1 2 3 4 5 6 7 8 9 10
Global Temperature Change (C)

Figure 4.15 Global temperature change for 2100 compared to 1990 using different
climate sensitivity PDFs for the A2 (top) and B2 (bottom) SRES scenario family,
and PDFs of other parameters as described in section 4.3.1.

The impact of sulphur dioxide emissions on global temperature change has been
observed in some scenarios. This issue has been examined in more detail by considering
only the lowest and highest values of the uniform PDF applied up to now. Figure 4.16
shows how global temperature change evolves over time for high and low aerosol

71
forcing under two contrasting scenarios: SRES family A2, which is high on SO2
emissions and illustrative scenario A1T-MES which is low on SO2 emissions. When
SO2 emissions are high (as in the left panel) the impact of aerosol forcing uncertainty on
global temperature change is considerable, especially in the early decades of the century
when emissions are highest in the A2 scenario family (left panel). In illustrative
scenario A1T-MES this is not observable because SO2 emissions are low (right panel).

0.25 0.25
Low-2030 Low-2030
0.2 High-2030 0.2 High-2030
Probability

Probability
Low-2050 Low-2050
0.15 0.15
High-2050 High-2050
0.1 Low-2100 0.1 Low-2100

High-2100 High-2100
0.05 0.05

0 0
0 1 2 3 4 5 6 7 8 0 1 2 3 4 5 6 7 8
Global temperature change () Global temperature change ()

Figure 4.16 Probability of global temperature change for different time horizons
(2030, 2050 and 2100) and different aerosol forcing parameters (low and high) for
the A2 SRES scenario family (left panel) and illustrative scenario A1T-MES (right
panel), using the Wigley and Raper (2001) PDF for climate sensitivity and PDFs of
other parameters as described in section 4.3.1.

Uncertainties associated with ocean mixing and carbon cycle were also tested by using a
high and low value of the PDFs described in section 4.3.1 (Figure 4.17). Carbon cycle
appears to have a significant impact on global temperature change, especially at the end
of the century, whereas ocean mixing seems to have a small impact.

0.25
0.25
Low-2030 Low-2030
0.2 High-2030 0.2 High-2030
Probability

Low-2050
Probability

0.15 Low-2050
0.15
High-2050 High-2050
0.1 Low-2100 0.1 Low-2100
High-2100 High-2100
0.05 0.05

0 0
0 1 2 3 4 5 6 7 8 0 1 2 3 4 5 6 7 8
Global temperature change () Global temperature change ()

Figure 4.17 Probability of global temperature change for different time horizons
(2030, 2050 and 2100) for the A2 SRES scenario family using the Wigley and
Raper (2001) PDF for climate sensitivity, comparing low and high value
parameters for carbon cycle (left panel) and ocean diffusivity (right panel) and
PDFs of other parameters as described in section 4.3.1.

72
4.4. Summary

This chapter has propagated several uncertainties through a simple climate model in
order to estimate the uncertainty associated with future global temperature change. This
work is different from that performed by Wigley and Raper (2001) because
uncertainties are treated differently (scenario uncertainty is not conflated with model
uncertainty)25 and because more objective estimates of climate sensitivity were
sampled. It was shown that future global temperature change remains uncertain,
especially as the estimates go further into the future. Both emissions scenarios and
climate sensitivity were shown to change the shape and size of global temperature
change PDFs. It was also shown that the illustrative marker scenarios can underestimate
or overestimate global temperature change compared to estimates that use the entire
scenario family (uncertainty due to the IAMs). Aerosol forcing uncertainty was shown
to be important for some scenarios (where sulphur dioxide emissions are high) as was
carbon cycle and ocean mixing uncertainties, in particular by the end of the century. The
relevance of these experiments will become more apparent when they are applied to a
decision context in Chapter 7. One uncertainty that was not tackled in this chapter was
that related to choice of energy balance model. It is the authors opinion that using a
different energy balance model would not change the results significantly (because
climate sensitivity and greenhouse gas emissions remain the most important
uncertainties) but this remains to be tested.

25
Or as the language introduced in section 2.2 would say: Human reflexive uncertainty is not merged
with epistemic uncertainty.

73
 CHAPTER 5. ASSESSMENT OF UNCERTAINTIES AT
THE REGIONAL LEVEL

This chapter builds upon the previous chapter by propagating uncertainty in climate
projections from the global level to the regional level. It starts however, by describing
the climate of the region under study, the Anglian region of the UK (section 5.1). It is
crucial to have a good understanding of the past and present climate of a region in order
to make an assessment of future climates for the region. Section 5.2 reviews the history
of previous climate change scenarios for the region. This is essential to distinguish
robust from non-robust information (as changed through time) and also to get an
overview of what sort of climate information has previously been given to decision-
makers. Section 5.3 describes the pattern-scaling technique, originally pioneered by
Santer et al. (1990). Section 5.4 describes GCM results for the Anglian region and
section 5.5 describes results from dynamically downscaled RCMs. Finally, section 5.6
summarises the chapter.

5.1 Description of the climate of the Anglian region

The climate of the British Isles can be generally characterised as temperate and mild.
This is the result of the UKs maritime location, its position within the main flow of the
mid-latitude westerlies and its proximity to the mild waters of the north-east Atlantic
Ocean (Barrow and Hulme 1997). This contributes to a climate of few extremes when
compared to regions of equivalent latitude (e.g., Moscow or Hudson Bay). Although the
British Isles makes up only a relatively small geographical area of the Earth's surface,
there are regional differences within the country. The west is dominated by maritime
air, whereas the climate in the east is more influenced by continental regimes.26

The Anglian region is the driest in the UK. According to the Environment Agency
(2001b), annual average rainfall is 595 mm, compared with a national average for
England and Wales of 897 mm. Effective rainfall (precipitation minus evaporation) is
only a quarter of average rainfall at 147 mm. Long dry summers during which
evaporation exceeds rainfall are a normal part of the climate in the region (EA 2001b).
26
As Figure 1.1 has shown the Anglian region is situated in the East of England.

74
Met. Office data, which perhaps is calculated differently from EA data,27 estimates
average rainfall for England and Wales at 905.6 (919.7) mm and 601.1 (605.6) mm for
East Anglia, for the period 1961-1990 (1971-2000 in brackets). Met. Office long-term
annual average station data for the Anglian region are shown in Table 5.1 for the two
most recent 30 year periods, 1961-1990 and 1971-2000. It is noticeable that both
temperature and precipitation have increased between these periods. Temperature has
increased by between 0.2 and 0.4 C during this last decade, whereas precipitation has
increased by between 6 and 21 mm (1-4%) except in Lowestoft where there was a
decrease of 8 mm (1%). These changes are within the bounds of natural variability for
this region (see section 5.1.1). Figure 5.1 shows mapped averages of summer and winter
precipitation for the entire UK. Particularly noticeable is the northwest to southeast
gradient in rainfall and the driest areas found in the Anglian region.

Table 5.1 Annual mean maximum and minimum temperature and precipitation
for selected locations in the Anglian region for the period 1961-90 (and 1971-200 in
brackets). Source: http://www.metoffice.com/climate/uk/averages/index.html

Bedford Cambridge Lowestoft Marham Wattisham

(85 m (26 m (25 m (21 m (89 m
AMSL) AMSL) AMSL) AMSL) AMSL)
Average 13.1 (13.5) 13.8 (14.1) 12.8 (13.0) 13.4 13.0 (13.4)
maximum (13.8)
temperature (C)
Average 5.3 (5.6) 5.8 (6.1) 6.6 (7.0) 5.5 (5.7) 5.5 (5.8)
minimum
temperature (C)
Rainfall (mm) 563 (584) 548 (554) 584 (576) 610 (621) 568 (574)
AMSL = Above Mean Sea Level

27
Furthermore, the EAs Anglian region boarder is different from the East Anglians boarder.

75
Figure 5.1 Summer and winter mean precipitation for the UK covering the 1961-90
(top panels) and 1971-2000 (bottom panels) periods. Source:
http://www.metoffice.com/climate/uk/averages/index.html. For method see Perry
and Hollis (2005b, 2005a)

76
Climate data for the Cambridge station are examined in more detail.28 One could argue
that this station is representative of the Anglian region because it is roughly in its centre
and because of the low topography throughout the region. Figures 5.2 and 5.3 show a
time series of monthly mean temperature and precipitation for this station. There is an
increasing trend for both variables. Because there is seasonality in these variables it is
important to look into each season.

25

20
.

15
Temperature (C)

10

0
1959

1961

1963

1965

1967

1969

1971

1973

1975

1977

1979

1981

1983

1985

1987

1989

1991

1993

1995

1997

1999

2001

2003

2005
-5
Year

Figure 5.2 Time series of monthly mean temperature (C) from January 1959 to
February 2005 for the Cambridge station. The black line represents a linear trend,
estimated as 0.26C/decade.

160

140
.

120
Precipitation (mm)

100

80

60

40

20

0
1961

1963

1965

1967

1969

1971

1973

1975

1977

1979

1981

1983

1985

1987

1989

1991

1993

1995

1997

1999

2001

2003

2005

Year

Figure 5.3 Time series of monthly mean precipitation (mm) from January 1961 to
February 2005 for the Cambridge station. The black line represents a linear trend,
estimated as 0.86mm/decade.

28
This was available on-line at: http://www.metoffice.com/climate/uk/stationdata/index.html

77
Figures 5.4 and 5.5 show changes compared to the long-term seasonal average
(calculated over the period 1961-1990), sometimes called relative variability (Linacre
1992); this is particularly useful for precipitation. It is interesting to note that since 1998
there have been no negative anomalies for temperature in any season. JJA and SON
have not had any negative anomalies since 1994 and MAM since 1997. This certainly
emphasises the point that the climate is getting warmer in the Anglian region. For
precipitation there is a large inter-annual variability (Figure 5.5).

4
f

3
2
Temperature change (C)

1
0
1959
1961
1963
1965
1967
1969
1971
1973
1975
1977
1979
1981
1983
1985
1987
1989
1991
1993
1995
1997
1999
2001
2003
2005
-1
-2
-3 DJF
MAM
-4
JJA
-5 SON
-6
Year

Figure 5.4 Time series of seasonal (DJF, MAM, JJA, SON) temperature anomalies
(C) compared to the long-term average (1961-1990) for the Cambridge station.

120
v

DJF
100
MAM
80
Precipitation change (%)

JJA
60
SON
40
20
0
1961
1963
1965
1967
1969
1971
1973
1975
1977
1979
1981
1983
1985
1987
1989
1991
1993
1995
1997
1999
2001
2003
2005

-20
-40
-60
-80
Year

Figure 5.5 Time series of seasonal (DJF, MAM, JJA, SON) precipitation anomalies
(expressed as %) compared to the long-term average (1961-1990) for the
Cambridge station.

78
5.1.1 Estimating natural climate variability for the Anglian region

Climate is known to vary naturally on various time scales (e.g. multi-decadal) for
reasons that have nothing to do with anthropogenic forcing (Hulme and Carter 1999) or
natural external forcing such as the sun or volcanic eruptions. This part of natural
variability is induced by interactions between various components of the climate
system, i.e., the atmosphere, ocean, cryosphere, soil, vegetation, etc (Ruosteenoja et al.
2003). Estimating natural variability is important because it allows the comparison of
the noise caused by internal variability with the climate change signal caused by
anthropogenic forcing in order to determine the significance of the model-simulated
response to anthropogenic forcing. This was demonstrated by Hulme et al. (1999) in the
context of regional impacts in Europe, who concluded that there is considerable
variability in unforced 30-year climates and this natural climate variability can induce
non-trivial responses in impact indicators (p. 690).

There are essentially two methods to estimate natural climate variability. One approach
is to use long series of observed data, but this is usually limited by how far back the
observation record goes. The second approach is to use multi-century unforced GCM
simulations. In this analysis both approaches are used. The following observational data
have been used:
? Central England Temperature (CET) from 1659-2005 (Manley 1974, Parker et
al. 1992, Parker and Horton 2005).29 The monthly mean surface air
temperatures, for a region representative of the English Midlands, are expressed
in degrees Celsius for the period from 1659 to the present.
? England and Wales Precipitation (EWP) from 1766-2005 (Wigley and Jones
1987, Gregory et al. 1991, Jones and Conway 1997, Alexander and Jones
2001).30 The monthly precipitation totals are representative of the land area of
England and Wales and are derived from measurements at 35 gauges across the
region. The series runs from 1766 to the present and are expressed in tenths of a
millimetre.

29
Available from: http://www.met-office.gov.uk/research/hadleycentre/CR_data/Daily/HadCET_act.txt
30
Available from: http://www.met-
office.gov.uk/research/hadleycentre/CR_data/Monthly/HadEWP_act.txt

79
 ? Ely Ouse to Denver Complex and Wensum to Costessey Mill catchment rainfall
from 1865-2002 (Jones 1984, Jones and Lister 1998, Jones et al. 2004).31

In terms of modelled data the following GCMs have been used:

? HadCM2 1400-year control data (Johns et al. 1997)32
? CGCM1 1001-year control data (Flato et al. 2000)33

Figure 5.6 shows the approximate grid boxes for the models used and the catchment
rainfall areas considered (numbers 10 and 8).

Figure 5.6 Grid box used for the East of England in CGCM1 (larger box) and
HadCM2 (smaller box) and outline of the Ely Ouse to Denver Complex catchment
(10) and the Wensum to Costessey Mill catchment (8) in the Anglian region.
Modified after Jones et al. (2004).

31
Available from: http://www.cru.uea.ac.uk/cru/data/riverflow/
32
This data was provided by Dr. Tim Osborn at the Climatic Research Unit
33
Available from: http://www.cccma.bc.ec.gc.ca/data/data.shtml. Used coordinates (I=1; J=38). This is a
land gridbox.

80
A series of 30-year overlapping means34 were calculated for all the data sets for summer
and winter. Results are shown in Figure 5.7 as a function of anomalies for a certain
specified period. The model and observation clouds overlap better for summer than for
winter. It shows that the models are reasonable in estimating natural variability,
although they are better for summer than winter. It is interesting to note that for winter
the Anglian catchment anomalies are almost exclusively in the warm and wet or cool
and dry quadrants. This is consistent with synoptic patterns in this region. Air flowing
over the North Atlantic to the UK (westerlies) is relatively warm and moist, whereas air
flowing out of Siberia (easterlies) is cold and dry (see, e.g., Davies et al. 1997). It is
very rare for the region to experience a cold/wet or mild/dry winter.

34
There is evidence that results using non-overlapping means do not systematically deviate from those
inferred from overlapping means (Ruosteenoja et al. 2003).

81
Figure 5.7 Winter (top) and summer (bottom) temperature and precipitation
anomalies in individual overlapping 30-year periods for observed (EWP, Ely Ouse
and Wensum) and modelled (CGCM1 and HadCM2) data. All observational data
used the CET for temperature. All anomalies are calculated with respect to the
mean for the period shown in the legend or for the modelled data, the entire
period.

82
From these sets of tridecadal averages, means and standard deviations of temperature
(C) and precipitation (mm/day) were calculated. Table 5.2 shows modelled, observed
and contaminated estimates of natural climate variability. The contaminated
estimates differ from the observed estimates in that they include periods after 1950 and
hence could already show an anthropogenic signal. The models seem to do a better job
at capturing the mean and variability of temperature in summer compared with winter.
For precipitation the results are rather similar, with summer variability being better
captured by the models than winter variability. A comparison between the
contaminated (1659-2004) and observed (1659-1950) estimates show evidence of
increased mean temperature and a slight increase in variability for both seasons in
recent 30-year periods. For precipitation, there is evidence of drier summers and wetter
winters for the EWP and the Anglian catchments in recent 30-year periods. This is
consistent with other studies that have examined observed precipitation trends (Osborn
et al. 2000, Osborn and Hulme 2002). Precipitation variability increased for the EWP
for both seasons, but decreased for the Anglian catchments over winter.

Table 5.2 Mean and standard deviation (and standard deviation as percentage
change) of natural climate variability for summer and winter temperature (top)
and precipitation (bottom) using modelled (CGCM1 and HadCM2), observed
(CET, EWP, Ely Ouse and Wensum with periods before 1950) and
contaminated (CET, EWP, Ely Ouse and Wensum including periods after 1950)
data.

Temperature Summer (JJA) Winter (DJF)

Mean St. Dev. Mean St. Dev.
(C) (C) (C) (C)
Modelled CGCM1 17.03 0.13 5.11 0.17
HadCM2 14.49 0.19 4.56 0.23
Observed CET (1659-1950) 15.24 0.19 3.62 0.39
CET (1766-1950) 15.24 0.17 3.70 0.37
to match EWP
CET (1894-1950) 15.13 0.11 4.08 0.29
to match Anglian
catchments
Contaminated CET (1659-2004) 15.27 0.20 3.69 0.40

83
 Precipitation Summer (JJA) Winter (DJF)
Mean St. Dev. St. Dev. as Mean St. Dev. St. Dev. as
(mm/day) (mm/day) % change (mm/day) (mm/day) % change
Modelled CGCM1 2.37 0.09 3.89 4.21 0.13 3.03
HadCM2 2.25 0.10 4.58 2.24 0.06 2.86
Observed EWP 2.52 0.10 3.77 2.48 0.20 8.23
(1766-1950)
Ely Ouse 1.82 0.09 4.75 1.55 0.12 7.54
(1894-1950)
Wensum 2.03 0.12 5.87 1.80 0.16 8.63
(1894-1950)
Contaminated EWP 2.47 0.13 5.32 2.56 0.22 8.49
(1766-2004)
Ely Ouse 1.81 0.07 4.04 1.81 0.07 4.04
(1894-2002)
Wensum 1.96 0.12 5.97 1.83 0.12 6.61
(1894-2002)

The normality of these sets of tridecadal averages were checked by plotting histograms.
Figure 5.8 shows the frequency distribution of 30-years of summer precipitation from
observed (periods before 1950) and modelled estimates.35 Two of the longer sets of data
(CGCM1 and EWP) resemble a normal (Gaussian) distribution, but the shorter sets of
data (Ely Ouse and Wensum catchments) do not. The HadCM2 curve is less like a
normal distribution, but this could be due to the existence of many missing numbers
from its data set, which reduced the number of tridecadal averages analysed. For the
purpose of this thesis (in particular Chapter 7) it was assumed that all five sets of
tridecadal averages are normally distributed (data from Table 5.2 were used to define a
mean and standard deviation). Figure 5.8 matches well with Figure 5.1 as the Ely Ouse
is probably one of the driest places in the UK, the Wensum is slightly wetter and the
EWP is considerably wetter than both of these. Both modelled data (CGCM1 and
HadCM2) sit in between.

35
Only results for summer precipitation are shown because this is the variable of most interest to
decision-makers, as shown in Chapter 6. The thesis focuses on summer precipitation for two main
reasons. Firstly, because water managers identified summer precipitation as the main climatic driver of
future investment (see Appendix 2; answer to question 6a). Secondly, because this thesis only examines
probabilistic changes in seasonal temperature and precipitation as it assumes higher uncertainty at higher
temporal scales (monthly or daily data). Given that all simulations show increases in winter precipitation
focusing only on summer precipitation is in fact a conservative approach.

84
 0.3
CGCM1 (972) EWP (156)
0.25 Ely Ouse (57) Wensum (57)
HadCM2 (1022)
0.2
Frequency
0.15

0.1

0.05

0
1.65 1.8 1.95 2.1 2.25 2.4 2.55 2.7 2.85
30-year JJA precipitation (mm/day)

Figure 5.8 Frequency distribution of 30-year summer precipitation for observed

(EWP, Ely Ouse and Wensum; same periods as in Figure 5.7) and modelled
(CGCM1 and HadCM2) data. In brackets is the number of 30-year sets available
for each distribution. Bins of 0.05 mm/day.

5.2 The history of climate scenario development for the Anglian

region

Climate scenario construction worldwide has been comprehensively reviewed by

Mearns et al. (2001) so this is not attempted in this section. Instead, the focus is
exclusively on how climate change scenarios have been constructed for the UK. Hulme
and Turnpenny (2004) provide a review of UK climate change research which describes
its institutional history, but does not go into sufficient depth regarding climate change
scenarios. There are various European climate scenario studies that include the UK, but
these will not be reviewed in this section; the focus is exclusively on the UK, with a
particular focus on results for the Anglian region.

The history of UK regional climate change projections spans over 20 years so it is worth
discussing. The first official UK climate change scenarios were published in 1991 by
the Climate Change Impacts Review Group (CCIRG 1991). This was mostly based on
previous work by Hulme and Jones (1989), which is further elaborated in Warrick and
Barrow (1991). Hulme and Jones (1989) examined five GCMs under single and double

85
atmospheric CO2 concentrations to construct regional climate scenarios for the UK.
Seasonal and annual changes for south-central England were estimated and are shown in
Table 5.3.

Table 5.3 Seasonal and annual temperature (C) and precipitation (mm/day)
anomalies for a gridpoint representation of the UK from five GCM experiments
(UKMO, OSU, GFDL, NCAR and GISS) for a doubling of CO2 concentration;
defined as (2xCO2-1xCO2). All points lie on the Greenwich meridian. Adapted
from Hulme and Jones (1989).36

Temperature Latitude DJF MAM JJA SON Annual Global mean

temperature
change
UKMO 52.5N 5.2 6.2 5 7.8 6 5.2
OSU 52.0N 2.8 2.8 3.1 2.9 2.9 2.8
GFDL 51.1N 5.8 4.8 5.9 5.4 5.5 4
NCAR 51.1N 4.6 3.2 3.3 3.5 3.6 3.5
GISS 50.87N 4.6 3.3 3.9 3.4 3.8 4.2
Precipitation Latitude DJF MAM JJA SON Annual Global
precipitation
change (%)
UKMO 52.5N 0.20 0.70 0.86 0.58 0.58 15
OSU 52.0N 0.20 0.01 -0.10 0.06 0.05 11
GFDL 51.1N 0.50 0.26 -0.4 0.00 0.09 7.1
NCAR 51.1N 0.80 1.37 0.22 0.75 0.51 8.7
GISS 50.87N 0.50 0.09 -0.30 0.28 0.15 Not known,
but positive

These results suggested a generally warmer and wetter UK with the exception of
summer where models diverge on precipitation sign. Warrick and Barrow (1991)
elaborated on these results by making them time-dependent (or transient) by combining
normalised GCM patterns with results from a simple upwelling-diffusion energy-
balance climate model. The simple climate model (Wigley and Raper 1987) was tuned
to a Business-As-Usual scenario of increasing GHG concentrations (1.5% per annum)
and a best estimate of climate sensitivity of 3C. Under these assumptions Warrick
and Barrow (1991) estimated the world would be warmer by 0.7, 1.4 and 2.1C than
1990 in the years 2010, 2030 and 2050 respectively. These authors estimated that UK
summer mean temperature change is comparable with global mean values, whereas
winter temperature changes were slightly higher. Summer precipitation change was

36
For UKMO and OSU an average of two grid points was used whereas for the other three models a
single grid point representation was used.

86
estimated to be 05%, 011% and 016% for 2010, 2030 and 2050 respectively. Winter
precipitation changes were 33%, 55% and 88% for the years 2010, 2030 and 2050
respectively.

The second set of official UK climate change scenarios came with the publication of
the second CCIRG report in 1996 (CCIRG 1996). These scenarios were more detailed
than those provided in the first CCIRG report as they included estimates of raindays,
precipitation intensity, potential evapotranspiration, etc, and greater spatial (four grid
boxes used for the UK) and temporal (daily) resolution. However, the scenarios were
mostly based on the results of one UK model, the transient climate change experiment
completed by the Hadley Centre in 1992 (known as UKTR or HadCM1, Murphy 1995,
Murphy and Mitchell 1995). Like the 1991 CCIRG scenario, pattern-scaling was
applied using a simple climate model (Wigley and Raper 1992, Raper and Cubasch
1996). This time the IPCC IS92a greenhouse gas emission scenario and a climate
sensitivity of 2.5C were used.

CCIRG (1996) observed a gradient of decreasing warming, from southeast to northwest

for both seasons. By 2050 winter warming was in the range 0.8-2C, while summer
warming was in the range 1.2-1.8C. This pattern of warming was the opposite of that
reported in the 1991 CCIRG scenario and is believed to be due to the explicit inclusion
of a dynamic ocean in the GCMs. The 1996 CCIRG scenario showed wetter winters
throughout the UK, up to 10% in south eastern England by the 2050s. For the summer
the scenario showed drying in the south and wetting in the north of the UK. Summer
precipitation was reduced by up to 9% in the extreme south of England by 2050.
Potential evapotranspiration was calculated using the Penman formula. Large increases
were estimated for winter (but in absolute terms these remain rather small) and
increased by up to 30% in the summer over southern England. While only one model
was used to create the 1996 CCIRG scenarios there was appreciation of the fact that
different climate models with similar forcing yielded different patterns of climate
change (see Table 5.4).

87
Table 5.4 Percentage changes in mean precipitation by the 2050s with respect to
1961-90 derived from 11 global climate model experiments for a region
representing central England. The acronyms refer to the respective climate
modelling centres. The effect of different model climate sensitivities has been
eliminated (from CCIRG 1996).

GCM Precipitation change (%)

experiment Annual Winter Summer
UKHI 14.7 24.6 6.0
UKTR 9.3 11.9 -0.5
UKLO 8.3 4.4 4.4
CSIRO 7.0 8.1 1.8
LLNL 6 11.1 8.5
CCC 4.9 11.1 -4.0
GFDL 4.5 10.1 7.3
GISS 1.5 3.3 1.6
OSU 1.1 4.7 5.2
ECHAM 0.7 4.0 -5.6
BMRC -0.7 7.3 -5.6

The UK Climate Impacts Programme (UKCIP) was established in 1996. One of its first
major products was the publication of a new comprehensive set of climate change
scenarios called the UKCIP98 scenarios (Hulme and Jenkins 1998). UKCIP98
described four possible climate futures for the UK. The Low scenario used IPCC IS92d
emissions scenario with a low (1.5C) climate sensitivity scaled from HadCM2 GGd
(Johns et al. 1997). The Medium-Low scenario used the HadCM2 GGd forced
experiment (0.5% increase in equivalent carbon dioxide concentrations per annum). The
Medium-High scenario used the HadCM2 GGa forced experiment (1% increase per
annum). The High scenario used IPCC IS92a emissions scenario with a high (4.5C)
climate sensitivity scaled from HadCM2 GGa. The results for annual and seasonal mean
temperature were similar to those of the 1996 CCIRG scenario; i.e., northwest to
southeast gradient in the magnitude of the climate warming and UK warming rate
similar to global mean warming. Precipitation changes were also similar to the 1996
CCIRG scenario. Annual and winter precipitation increased for all periods and
scenarios. For summer, there was a general tendency for drying in the south of the UK
and wetting in the north. More detail was provided for the Medium-High scenario such
as diurnal temperature range, relative humidity, potential evapotranspiration, etc.
Besides seasonal means, Hulme and Jenkins (1998) also analysed interannual
variability. For temperature all seasons become more variable except winter, which
becomes less variable. Precipitation variability increased in all seasons. In addition,

88
Hulme and Jenkins (1998) put the UKCIP98 scenarios in the wider context of other
climate models (CGCM1, ECHAM4 and GFDL-R15) and natural variability. The other
three models all show more rapid annual warming over the UK compared to the
Medium-High scenario, by almost 1C in the case of ECHAM4. For precipitation,
ECHAM4 shows a strong pattern of drying over southern England, while GFDL-R15
shows slight drying annually over the southwest. For the Medium-High scenario,
seasonal temperature changes are largely due to human-induced climate change rather
than natural climate variability. However, a large proportion of the seasonal mean
precipitation changes are due to natural climate variability.

Besides Warrick and Barrow (1991) the only peer-reviewed article dedicated
exclusively to future climate change scenarios for Great Britain is that by Conway
(1998). Conway (1998) produced scenarios for four regions in Great Britain (Northern
Scotland, Northern England, Wales and Central England) using results from HadCM2
under two different forcings: greenhouse gas only (GHG) following IPCC IS92a and
greenhouse gas plus sulphate aerosols (GHG + SUL). Four experiments were performed
under each forcing. Conway (1998) showed a warming trend throughout the country,
slightly warmer over central England than northern Scotland. Annual temperatures in
the GHG + SUL ensemble were roughly 0.5C lower than the GHG ensemble by 2100.
The precipitation signal was much less pronounced than the temperature signal under
both forcings. There was no major change in annual precipitation, but increases
occurred in winter over the whole country and slight decreases in the summer in the
south.

In 2002, UKCIP published the UKCIP02 scenarios (Hulme et al. 2002). These included
four scenarios of future climate change for the UK at a higher spatial resolution than
previous efforts. The regional climate model used was HadRM3 (50 km resolution).
This model was run for the periods 1961-90 and 2071-2100 with boundary conditions of
an atmospheric GCM (HadAM3H). HadAM3H was run for a reference period (1961-
90) driven by observations of sea-surface temperatures (SST) and sea-ice for that
period. A second run (2071-2100) was driven by changes in SST and sea-ice from
HadCM3, added to the observations. This double-nesting approach improves the
quality of the simulated European climate (Hulme et al. 2002, Appendix 2). Due to the
computational expense of this double-nesting technique, only four RCM experiments

89
were conducted (three with the SRES A2 forcing scenario and one with B2). To derive
climate change scenarios for other time periods and emissions scenarios, Hulme et al.
(2002) used the pattern-scaling technique. The master pattern was derived by averaging
the results from the three experiments with HadRM3 under SRES A2. Then a scaler
(global warming from HadCM3 simulations) was used to obtain climate change for
other time periods and emissions scenarios. The same nomenclature as UKCIP98 was
used, but this time each scenario was linked to a different emission scenario. So the
Low scenario related to SRES B1, Medium-Low scenario to SRES B2, Medium-High
scenario to SRES A2 (the original master pattern) and High scenario to SRES A1FI.
Figure 5.9 compares the results of UKCIP02 (HadCM3) with other GCMs for winter
and summer, temperature and precipitation change in the 2080s.

90
.

80 CGCM2
70 CSIRO Mk2
Precipitation change (%)

60 CSM 1.3
ECHAM4
50
GFDL
40
MRI2
30 NIES2
20 PCM
10 HadCM3
0
0 1 2 3 4 5 6 7
Temperature change (C)

10
.

5 CGCM2
0
Precipitation change (%)

CSIRO Mk2
-5 0 1 2 3 4 5 6 CSM 1.3
-10 ECHAM4
-15 GFDL
-20 MRI2
-25 NIES2
-30 PCM
-35 HadCM3
-40
Temperature change (C)

Figure 5.9 Changes in average winter (top) and summer (bottom) temperature and
precipitation in the 2080s over the East of England with respect to the average
1961-90 climate, for the SRES A2 scenario. Each coloured circle represents the

90
result from a different global climate model. The red circle represents the model
(HadCM3) used in UKCIP02.

For summer precipitation, HadCM3 was amongst the driest of the models, whereas for
winter precipitation and overall temperature HadCM3 lay in the middle of the model
range. Based on these results, Hulme et al. (2002) provided expert-based uncertainty
margins for seasonal temperature and precipitation changes for each emission scenario
for the 2080s. Like in previous reports, the UK warming rate is very similar to the
global rate and a northwest to southeast gradient is again present. The warming in the
southeast is particularly pronounced in summer. Winter precipitation increased for all
periods and scenarios, with increases larger than natural variability for almost the whole
country (except Scotland). Summer was expected to become drier especially for the
Anglian region. Over most of the country, winter and spring temperatures became less
variable, whereas summer and autumn became more variable. Precipitation variability
tended to increase in the winter and decrease in the summer. The UKCIP02 report
presents results for a number of other climate variables such as: diurnal temperature
range, cloud cover, relative humidity, etc, and also daily climate variables which allows
the examination of heating and cooling degree days, daily precipitation, etc.

Figure 5.10 tries to summarise all the climate change scenarios reviewed in this section,
using a climate variable that water managers in the Anglian region are interested in:
summer precipitation change (see section 6.6). Until the 2020s most modelled results
are within the bounds of natural multi-decadal climate variability (as represented by a
95% confidence interval for the Ely Ouse catchment), expect beyond UKCIP02
Medium-Low. Even by the 2050s, many results are within these bounds, except
UKCIP98 Medium-High and High and all UKCIP02 results. By the 2080s, only
UKCIP98 Low could be within the bounds of natural variability. As time has progressed
(from CCIRG 1991 to UKCIP02) estimates of summer precipitation change seem to
have decreased in general (i.e., showing further drying). The Conway (1998) study
shows that the inclusion of sulphur emissions reduces summer drying by a small
amount.

91
 2010s 2020s 2030s 2050s 2080s

Conway_HadCM2_GHG+SUL
Natural climate variability

1996_CCIRG_High Sens
1996_CCIRG_Low Sens

1996_CCIRG_11 GCMs

Conway HadCM2_GHG
1996_CCIRG_UKTR

1996_CCIRG_UKTR
1991_CCIRG

1991_CCIRG
1991_CCIRG
UKCIP98 MH

UKCIP02 MH

UKCIP98 MH

UKCIP02 MH

UKCIP98 MH

UKCIP02 MH
UKCIP98 ML

UKCIP02 ML

UKCIP98 ML

UKCIP02 ML

UKCIP98 ML

UKCIP02 ML
UKCIP98 H

UKCIP02 H

UKCIP98 H

UKCIP02 H

UKCIP98 H

UKCIP02 H
UKCIP98 L

UKCIP02 L

UKCIP98 L

UKCIP02 L

UKCIP98 L

UKCIP02 L
20
15
10
.

5
Precipitation change (%)

0
-5
-10
-15
-20
-25
-30
-35
-40
-45

Figure 5.10 Chronology of summer precipitation change (%) for the East of
England, for different time horizons (2010s, 2020s, 2030s, 2050s and 2080s), for the
studies reviewed in this section and natural variability (as estimated in section 5.1.1
for the Ely Ouse catchment; 95% confidence level). The interval ranges are based
on the judgment of the authors of the studies based on the GCM results (1991
CCIRG), other GCM results (1996 CCIRG and UKCIP02 2080s) or ensemble runs
(UKCIP98 MH).

A large number of impact studies originated from the climate change scenarios
described here. Some of them were published together with the climate scenarios
(CCIRG 1991, 1996), others elsewhere in the peer-review literature (e.g., Arnell and
Reynard 1996, Pepin 1997, Arnell 1998, Pilling and Jones 2002, Arnell 2003b, 2004b,
Diaz-Nieto and Wilby 2005 all on water resources and, Holman et al. 2005a, Holman et
al. 2005b on a multi-sectoral integrated assessment that included water resources), and
more recently in the grey literature that is more accessible to stakeholders (UKWIR
2003, EERA and SDRT 2004, CIRIA 2005).

92
This review and in particular Figure 5.10 have shown that uncertainty has received a
poor treatment in previous climate change scenarios efforts in the UK. Perhaps the best
treatment of uncertainty was given by the first set of scenarios (CCIRG 1991) as
uncertainty ranges were provided for climate variables such as temperature and
precipitation change. The ranges were subjective, but they explained the contradictory
results given by GCMs and showed that uncertainty expanded with time. As the degree
of sophistication and complexity increased with the years, uncertainty description
decreased. This is probably due to the difficultly of characterising uncertainty in
complex systems (such as GCMs) and the reluctance to provide subjective uncertainty
ranges. In UKCIP02, uncertainty margins were given for the 2080s (for average winter
and summer temperature and precipitation change) but these margins have rarely (if
ever) been used in impact and adaptation assessments. The robust finding that comes
out of this review (through time) is that summers are expected to get drier in the East of
England although its magnitude remains uncertain.

5.3 Applying the pattern-scaling technique

Having established the basic characteristics of the climate of the Anglian region, in
particular natural climate variability, and reviewed the history of climate scenario
construction for the region, this section starts to explain how regional climate change
projections were constructed in this thesis. In particular, this section introduces the
pattern-scaling technique and how it is applied here.

The pattern-scaling technique has been widely used to create climate change projections
for emissions scenarios and/or time spans not simulated directly by GCMs. It was
originally developed by Santer et al. (1990) to create transient runs from equilibrium
GCM simulations. The technique normalises GCM patterns by global average
temperature change at the end of the century to create a spatial response pattern (this is
now in temperature or precipitation units per degree of global warming). The scaler,
annual global mean temperature change, is calculated using a simple climate model (or
as in UKCIP02, it is derived from an A/OGCM). Using time series from a simple
climate model the response pattern and the scaler are multiplied to produce regional
climate change scenarios. The critical assumption is that there is a linear relationship

93
between the scaler and the response pattern. Mitchell et al. (1999) found this assumption
to be broadly valid for annual temperature and Mitchell (2003) extended this to
precipitation and seasonal climate (at least for one model, HadCM2). Several different
types of pattern-scaling have been applied. Mitchell et al. (1999) used the entire time
series of a transient GCM run to estimate patterns using linear least-squares regression
on a sequence of non-overlapping periods. Schlesinger et al. (2000) conducted two-
pattern scalings: one for GHG forcing and another for aerosol forcing, and combined
them arithmetically.

Here the more commonly used time-slice method was used because the regional climate
model simulations were only available for two discrete 30-year periods (1961-90 and
2071-2100; see section 5.5). The traditional time-slice method is calculated by:

? T Scenario, time, season = ? T_EBM Scenario, time X (? T_GCM Gridbox, season / ? T_GCM Global)

where ? T is temperature change, EBM refers to the Energy Balance Model (in this case
MAGICC) and GCM relates to Global Climate Model. Global temperature change from
the simple climate model is usually known as the scaler. The expression in parenthesis
is essentially the normalised master pattern, here called response pattern. This
expression is usually calculated for the strongest available forcing scenario at the end of
the century (e.g., 2071-2100) in order to maximise signal-to-noise ratio. In order to
increase the sample size of GCMs and RCMs both SRES A2 and B2 forced simulations
were used. To calculate regional precipitation changes one simply has to replace
? T_GCM Gridbox, season by ? P_GCM Gridbox, season, where precipitation change (? P) is
calculated by: (Pfuture / Pcontrol) 1. When using RCM results the ? T_GCM Global is
estimated from the driving AOGCM. The geographical pattern of temperature and
precipitation change is assumed to be independent of the forcing, while the amplitude at
every location is linearly proportional to the global mean change in surface air
temperature.

A probabilistic interpretation of the time-slice patter-scaling method is not entirely

straightforward so it is explained in some detail. In a probabilistic setting, ? T_EBM is
replaced by the MAGLOOP results from Chapter 3. These results are conditional not
only on which emissions scenario and time period is being examined, but which climate

94
sensitivity, aerosol forcing, carbon cycle and ocean mixing PDF has been used (see
section 4.3.1). The response patterns (gridbox temperature or precipitation change per
degree of global warming) are deterministic results from individual GCM and RCM
simulations. There are several different plausible approaches to converting these
deterministic results into probabilities.

Perhaps the simplest approach is to use the deterministic results, which once multiplied
by probabilistic global temperature change (from MAGLOOP) result in probabilistic
regional temperature change. This approach weighs each model pattern equally, i.e.,
they have the same probability of occurrence. It could be argued, however, that some
models are better than others. Dessai et al (2005a) calculated skill scores for several
GCMs based on model performance (how well the model simulates the observed
climate) and model convergence (how well models converge to the multi-model
ensemble average in the future) at the regional level for seasonal precipitation and
temperature change. It was shown that using a probabilistic approach was more
informative than using the ensemble mean, but skill scores do not make a substantial
difference from not using them. Wilby and Harris (2006) also found that skill scores
make relatively little difference. In Chapter 7, some tests with skill scores are conducted
to see if this is the case from an adaptation decision point of view.

The other plausible approach is to construct a PDF for the response pattern. Jones
(2000a), for example, assumes the response pattern has a uniform probability
distribution (this author also assumes a uniform distribution for global temperature
change). Hingray et al. (2005) apply an ANalysis Of VAriance (ANOVA) to estimate
the uncertainty associated with the response pattern (they call it the scaling ratio).
Ekstrm et al. (2005) compared these two methods for changes in annual temperature
and precipitation in five different locations in Europe. They found that the directions of
change are similar in both methods but the magnitude of the expected changes differs.
In Chapter 7, uniform probability distributions have also been used to define the
response pattern based on the highest and lowest model results.

In this thesis a number of approaches to PDF construction of response patterns have

been considered. In general, for GCM results a uniform distribution has been assumed
because of the small number of simulations available. For RCM results each simulation

95
was assumed to be equally plausible when constructing the PDF. These assumptions are
varied in Chapter 7 to test the sensitivity of adaptation decisions to different
constructions (e.g., using only the best performing models).

5.4 General circulation models

Now that the pattern-scaling technique has been explained in more detail, it becomes
clear that Chapter 4 essentially dealt with calculating the scaler (global temperature
change from a simple climate model) while this chapter is mostly concerned with
estimating the response patterns from more complex atmosphere and ocean models.
This section examines the results from Atmosphere-Ocean General Circulation Models
(AOGCMs; here abbreviated to GCMs) for the region under study. Climate scenario
construction has relied heavily on GCMs because they are currently the most advanced
tools available for simulating the response of the global climate system to changing
atmospheric composition (Mearns et al. 2001).

The same GCM data as used by the IPCC TAR and UKCIP02 are used in this thesis.37
Table 5.5 describes the features of the GCMs used. The nine GCMs used come from a
number of different countries and possess different grid resolutions. Depending on the
model used simulated global warming by the end of the century ranges between 1.2-
4.5C for emissions scenario SRES A2 and between 0.9-3.4C for emissions scenario
SRES B2. All these values are within the ranges estimated in Chapter 4 using a simple
climate model. Temperature and precipitation data were extracted from each GCM for
the relevant grid box above the Anglian region in the East of the UK.

37
The original GCM simulation outputs were interpolated onto a 2.5 by 2.5 degree grid by Dr. Richard
Jones at the Hadley Centre.

96
Table 5.5 Characteristics of GCMs used in this analysis. Modified after Table 8.1
of McAvaney et al. (2001).

Global
Atmospheric
warming by
resolution Ocean
GCM Country Centre the 2080s Reference
(longitude x resolution
under SRES
latitude)
A2 and B238
CSM 1.3 2.8o?2.8o L18 2.0o?2.4o L45 USA NCAR 2.29 1.71 (Boville et al.
2001)
CGCM2 3.8o?3.8o L10 1.8o?1.8o L29 Canada CCCma 3.54 2.46 (Flato and Boer
2001)
CSIRO Mk2 3.2o?5.6o L9 3.2o?5.6 o L21 Australia CSIRO 3.46 2.70 (Gordon and
O'Farrell 1997)
ECHAM4/OPYC3 2.8 ?2.8 L19
o o
2.8 ?2.8 L11
o o
Germany DKRZ 3.05 2.32 (Roeckner et al.
1996)
GFDL_R15_b 4.5o?7.5 o L9 4.5o?3.7 o L12 USA GFDL 2.87 2.20 (Dixon and
Lanzante 1999)
DOE PCM 2.8o?2.8o L18 0.67o? 0.67o L32 USA NCAR 2.34 1.81 (Washington et
al. 2000)
CCSR/NIES2 5.6 ?5.6 L20
o o
2.8 ?3.8 L17
o o
Japan CCSR/ 4.50 3.36 (Nozawa et al.
NIES 2000)

HadCM3 2.5o?3.75 o L19 1.25o? 1.25o L20 UK UKMO 3.24 2.38 (Gordon et al.
2000)
MRI2 2.8o?2.8o L30 2.0o?2.5o L23 Japan MRI 1.24 0.92 (Yukimoto et
al. 2000)

Figure 5.11 shows changes in summer and winter temperature and precipitation at the
end of the century (2080s) compared to the present (1961-90) for the Anglian region
under emissions scenarios SRES A2 and B2.39 In general these GCMs corroborate the
findings of previous climate change scenarios (section 5.2) that winters will get wetter,
summers will get drier and temperature will increase over all seasons. These projections
are consistent with recent observations (section 5.1). The higher forcing scenario (SRES
A2) produces larger temperature and precipitation changes than the lower forcing
scenario (SRES B2). Also evident is a trend towards warmer and drier summers, and
warmer and wetter winters.

38
These values were calculated as 30-year averages over the period 2070-2099 compared to the period
1961-1990. This data was supplied by Prof. Ulrich Cubasch and originates from the IPCCs TAR WGI
Figure 9.6. (Cubasch et al. 2001, p. 542)
39
The autumn and spring seasons were not examined in GCM simulations because they were not
available in the interpolated grid results provided and because summer was the season of most interest to
decision-makers (as shown in section 6.6).

97
 .
80

Precipitation change (%)

60
40 Summer A2
Winter A2
20
Summer B2
0 Winter B2
0 1 2 3 4 5 6 7
-20

-40
Temperature change (C)

Figure 5.11 Changes in average winter and summer temperature and precipitation
for the Anglian region in the 2080s (2070-2099) relative to 1961-90 for nine GCMs
under two SRES emissions scenarios (A2 and B2).

The values from Figure 5.11 were subsequently normalised by the respective global
temperature change over the period 2070-2099 from each GCM. These normalised
values assume the response pattern is independent of the forcing from the two different
emissions scenarios. Figure 5.12 shows summer and winter temperature and
precipitation change per degree of global warming. Perhaps not surprisingly Figure 5.12
is not too dissimilar to Figure 5.11. Figure 5.12 shows that some models estimate that
the Anglian region will warm up more slowly than the global average, whereas other
models estimate larger warming than the global average with one model (MRI2)
estimating more than double the warming of the global average in summer.

98
 20

15
Summer
10 Winter

Precipitation 5
change (%)
0
per degree of
global warming-5 0 0.5 1 1.5 2 2.5

-10
-15

-20
Temperature change (C) per degree of global warming

Figure 5.12 Summer and winter temperature and precipitation change per degree
of global warming for the Anglian region (based on 18 GCM simulations forced by
SRES A2 and B2; see Table 5.5).

Using data from Figure 5.12, frequency distributions of temperature and precipitation
change per degree of global warming were constructed for the Anglian region. Because
there were only 18 simulations (nine for each SRES scenario) large bin sizes had to be
used to construct a continuous distribution. Figure 5.13 shows the results when each
GCM simulation is assumed equally likely, as part of a super ensemble. The right panel
shows that most models show a decrease in summer precipitation while winter
precipitation is almost certain to increase. The left panel shows that the magnitude of
summer temperature increases could be larger than winter increases.

0.35 0.5
0.45
0.3 Summer
0.4
Winter Summer
.

0.25 0.35
.

Winter
Frequency

0.3
Frequency

0.2
0.25
0.15 0.2

0.1 0.15
0.1
0.05 0.05

0 0
0.5 0.75 1 1.25 1.5 1.75 2 2.25 2.5 -20 -16 -12 -8 -4 0 4 8 12 16 20 24
Temperature change (C) per degree of global warming Precipitation change (%) per degree of global warming

Figure 5.13 Frequency of temperature change (C) (left panel) and precipitation
change (%) (right panel) per degree of global warming for summer and winter
using nine GCMs under two emission scenarios (i.e., totalling 18 simulations) . Bins
of 0.25 C and 4%.

99
In Chapter 7 regional response patterns of summer precipitation change (per degree of
global warming) are assumed to be uniform - due to the small number of simulations
available - which is not dissimilar from the distribution in Figure 5.13 (red curve in the
right panel). Also examined in Chapter 7 is the issue of model skill. Figure 5.14
compares GCM control simulations (for the 1961-90 period) with observed records for
England and the Anglian region. It shows that some models are better than others at
representing past mean climatic conditions. The ellipses encircle the best performing
models for precipitation and temperature, which are used as representative of skilled
models in Chapter 7.40 There are dangers associated with evaluating model skill using
output from a single grid-box. For example, while a model could be estimating
temperature and precipitation correctly (due to compensating errors) it could be getting
the large-scale features of the atmosphere wrong. More detailed model evaluation would
require the estimation of model skill at different scales, for example: locally (single
grid-box), regionally/hemispheric (some/many grid-boxes) and globally (all grid-
boxes).

4
.

3.5 CET/EWP
Cambridge/Ely Ouse
3 Cambridge/Wensum
Precipitation (mm/day)

CGCM2
2.5 CSIRO Mk2
2 CSM 1.3
ECHAM4
1.5 GFDL
MRI2
1 NIES2
PCM
0.5 HadCM3
0
0 5 10 15 20 25
Temperature (C)

Figure 5.14 Mean summer temperature and precipitation for the 1961-90 period as
modelled by nine GCMs over central and eastern England (in circles; see Table
5.5) and as observed by a combination of stations and catchments (represented by
the symbols +, -, ; see section 5.1). The ellipses surround the best performing
models for temperature and precipitation for CET/EWP.

40
Natural variability ranges could have been used to objectively define the dimensions of the ellipses
(and hence the best performing models). This would have led to the selection of the same GCMs except
CSM 1.3 and NIES2 for temperature.

100
5.5 Regional climate models

The coarse resolution of GCMs, in the order of hundreds of kilometres, has been the
major impetus for the development of downscaling techniques (Mearns et al. 2001).
Downscaling can be broadly divided into two approaches: statistical and dynamical.
Statistical downscaling relates local climate with large-scale features of the climate
system (e.g., circulation patterns, weather types, etc.) using techniques such as linear
regression, neural networks, etc. This is first done using observational or quasi-
observational (e.g., NCEP reanalysis) data in order to build the statistical model;
subsequently, this model is applied to GCM data. One of the major assumptions of
statistical downscaling is that this relationship remains constant in the future. Dynamical
downscaling on the other hand uses Regional Climate Models driven by boundary
conditions provided by GCMs. In comparison with statistical downscaling, dynamical
downscaling is more physically based, but also more computationally expensive.

Both statistical and dynamical downscaling techniques have been developing since the
1990s, but it is only recently that efforts have been made to compare different
techniques (Wilby and Wigley 1997, Mearns et al. 1999, Wilby et al. 2000). One such
effort is the informal cluster of three European projects: PRUDENCE (Prediction of
Regional scenarios and Uncertainties for Defining EuropeaN Climate change risks and
Effects), STARDEX (STAtistical and Regional dynamical Downscaling of EXtremes
for European regions) and MICE (Modelling the Impact of Climate Extremes).41 MICE
used information from global and regional climate models (developed by the Hadley
Centre) to explore future changes in extreme events across Europe in response to global
warming. STARDEX was concerned with improving statistical downscaling
methodologies for the construction of scenarios of changes in the frequency and
intensity of extreme events. PRUDENCE provided high-resolution climate change
scenarios for 2071-2100 for Europe using Regional Climate Models.

This section increases the number of response patterns sampled in this thesis by
examining results from Regional Climate Models from the PRUDENCE project.42
Unfortunately, there was not enough time to delve into statistical downscaling, the other

41
http://www.cru.uea.ac.uk/projects/mps/
42
http://prudence.dmi.dk/

101
major form of downscaling, so this limitation must be acknowledged. The PRUDENCE
project constructed high-resolution climate change scenarios for Europe using four
high-resolution Atmospheric General Circulation Models (AGCMs) (see Table 5.6) and
11 RCMs (see Table 5.7).

Table 5.6 Characteristics of atmospheric GCMs used in the PRUDENCE project,

including global warming (?T) by the end of the century (2080s) compared to the
present (1961-90) under emissions scenarios SRES A2 and SRES B2.

AGCMs Resolution Institution Country ? T-A2 ? T-B2 Reference

(latitude by (C) (C)
longitude)
HadAM3H 1.24 o?1.88 o Hadley Centre for United 3.09 2.28 (Pope et al.
Climate Prediction Kingdom 2000)
and Research
ECHAM4/OPYC3 2.8 o? 2.8 o Max-Planck-Institut Germany 3.56 2.76 (Roeckner
fr Meteorologie et al. 1999)
ARPEGE/OPA-SST Variable* Centre National de France 3.02 2.35 (Gibelin
Recherches and Deque
Mtorologiques 2003)
ARPEGE/HadCM3- Variable* Centre National de France 3.07 2.30 (Gibelin
SST Recherches and Deque
Mtorologiques 2003)
* - from 50 km in the centre of the Mediterranean sea to 450 km in the southern Pacific
ocean; resolution of 50-70 km over Europe

The atmospheric GCMs used observed and modelled sea-surface conditions (from
AOGCMs) and radiative forcing specified according to SRES A2 and B2. HadAM3H is
a high-resolution version of the atmospheric component of HadCM3 (Gordon et al.
2000). The baseline period extends from 1961-90 and is driven by observed SSTs and
sea ice extent. The second run (2070-99) is driven by changes in the SST and sea ice
extent predicted by HadCM3 and added to the observations. Similarly, the atmospheric
component of ECHAM4 was run at high resolution. ARPEGE (Action de Recherche
Petite Echelle Grande Echelle) is a variable resolution atmospheric model with a
resolution of 50-70 km over Europe (due to grid stretching), which makes it comparable
with RCMs (hence it appears in Tables 5.6 and 5.7). The 1960-99 period was forced by
monthly mean observed SSTs and historical GHG and sulphate aerosol concentrations.
Changes in mean SST were derived from transient simulations with the ARPEGE-
Ocean Paralllis (OPA) (Royer et al. 2002) and HadCM3 AOGCMs.

102
The various RCMs listed in Table 5.7 (except ARPEGE) were run with a combination
of different boundary conditions (provided by HadAM3H43 or ECHAM4/OPYC),
different emission scenarios (SRES A2 and B2), different spatial resolutions and
different initial conditions. Each RCM run has a control period (1961-90) and a future
period (2070-99). All RCMs were run at horizontal spatial scales of approximately 50
km, though a few were also run at higher spatial resolution (12 and 25 km). Monthly
and seasonal data from the RCMs were interpolated to a common grid (the CRU grid)
of 0.5 ? 0.5 by the PRUDENCE team. The seasonal data were used in this analysis.

43
There are also some runs with an improved version of HadAM3H called HadAM3P.

103
Table 5.7 Characteristics of the RCMs used in the PRUDENCE project and in this
analysis.

RCMs Acronym Institution Country Grid Reference

distance ()
ARPEGE CNRM Centre National de France Variable (Gibelin and
Recherches Deque 2003)
Mtorologiques
CHRM ETH Institute for Switzerland 0.50 (Vidale et al.
Atmospheric and 2003)
Climate Science
HadRM3H HC Hadley Centre for UK 0.44 (Hulme et al.
Climate Prediction 2002)
and Research
HIRHAM DMI Danish Denmark 0.44 (Christensen
Meteorological et al. 2001)
Institute
REMO MPI Max-Planck-Institut Germany 0.5 (Jacob 2001)
fr Meteorologie
ICTP- ICTP International Centre Italy N/A (Giorgi et al.
RegCM for Theoretical 1993a, Giorgi
Physics et al. 1993b,
Giorgi et al.
1999, Pal et
al. 2000)
PROMES UCM Universidad Spain N/A (Arribas et al.
Complutense de 2003)
Madrid
CLM GKSS Institute of Coastal Germany 0.5 (Steppeler et
Research al. 2003)
RCAO SMHI Swedish SMHI 0.44 (Risnen et
Meteorological and al. 2004)
Hydrological
Institute
RACMO KNMI Royal Netherlands The 0.44 (Lenderink et
Meteorological Netherlands al. 2003)
Institute
HIRHAM METNO Norwegian Norway 0.5 (Christensen
Meteorological et al. 2001)
Institute

Tables 5.8 and 5.9 show the different combination of RCM (or atmospheric GCM),
Atmospheric GCM (or AOGCM), emission scenario and spatial resolution used here.
Every RCM has been run with at least HadAM3H under SRES A2. There are a total of
21 runs under A2, seven runs under B2, 20 runs under HadCM3H, four runs under
ECHAM4/OPYC3 and four runs under ARPEGE. There are a number of gaps from the
possible number of combinations.

104
Table 5.8 Number of simulations for each combination of RCM, driving
atmospheric GCM (HadAM3H, ECHAM4/OPYC3 and HadAM3P) and emission
scenario (SREAS A2 and B2). For RCMs with spatial resolutions other than 50kms
are shown in parenthesis (e.g., 12, 25 klms).

HadAM3H ECHAM4/ OPYC3 HadAM3P

RCM A2 B2 A2 B2 A2 B2
3+2(12,25
DMI km) 1 1
HC 1 3 1
GKSS 2
MPI 1
SMHI 1+1(22 km) 1 1 1
ICTP 1 1
METNO 1 1
KNMI 1
ETH 1
UCM 1 1

Table 5.9 Number of simulations for each combination of atmospheric GCM

(HadAM3H, ARPEGE), driving AOGCM (HadCM3, ARPEGE.OPA) and
emission scenario (SRES A2 and B2) used in the analysis.

HadCM3 ARPEGE/OPA
A/GCM A2 B2 A2 B2
HadAM3H 2 1
ARPEGE 3 3

Based on this data, 30-year seasonal means of temperature and precipitation change
were extracted for grid boxes above the Anglian region, specifically to match the
observed catchments examined in section 5.1.1 (Ely Ouse and Wensum) and the water
resource zones introduced in Chapter 6 (East Suffolk & Essex, Ruthamford and
Norwich & the Broads). For each catchment the average over all the relevant grid boxes
was calculated. As applied in section 5.4 with GCMs, all RCM runs were normalised by

105
the global mean temperature from the driving GCM (see table 5.6; this follows the
methodology applied by Hulme et al. 2002).

Figure 5.15 shows temperature and precipitation change per degree of global warming
for all high-resolution AGCMs and RCMs simulations for the East Suffolk and Essex
Water Resource Zone (an area within the Anglian region introduced in the next chapter)
for all seasons. Summers are expected to get drier and winters wetter, while spring and
autumn could go either way or remain the same. The high-resolution AGCM and RCM
results are fairly consistent with the results using AOGCMs (see Figure 5.12). AGCM
and RCM results show less uncertainty than AOGCMs results, but this is partially
because they have only been driven by two AOGCMs.

20
Summer
15 Winter
10 Spring
Autumn
Precipitation 5
change (%)
per degree 0
of global 0 0.5 1 1.5 2
warming -5

-10

-15

-20
Temperature change (C) per degree of global warming

Figure 5.15 Mean temperature and precipitation change per degree of global
warming for all high-resolution AGCMs and RCMs for the East Suffolk and Essex
Water Resource Zone for different seasons.

Figure 5.16 shows all the red circles of Figure 5.15 (representing mean summer
temperature and precipitation change) sorted by driving model (either AGCM or
AOGCM).44 The AGCM ARPEGE (driven by ARPEGE/OPA under emissions scenario
A2) produces the lowest change in precipitation and temperature while the AGCM
HadAM3H (driven by HadCM3 under emissions scenarios A2 and B2) produces some

44
Results for the AOGCM ARPEGE/OPA are not shown in Figure 5.16 under Driving AOGCMs.

106
of the largest precipitation and temperature changes. The RCM results (driven by
HadAM3H and ECHAM4/OPYC3 under emissions scenarios A2 and B2) seem to lie
somewhere in between. It is also noticeable that most B2 forced simulations are at the
drier and warmer range of results. This occurs because RCM and AGCM response
patterns are normalised by global mean temperature from the driving AOGCM, which is
lower for B2 forced simulations than for A2 forced simulations.

0
0 0.5 1 1.5 2
-2

-4

-6 A2 ARPEGE/OPA
Precipitation A2 ECHAM4/OPYC3
change (%) -8
per degree B2 ECHAM4/OPYC3
of global -10 A2 HadCM3
warming
B2 HadCM3
-12
A2 HadAM3H
-14 B2 HadAM3H
-16 Driving AOGCMs

-18
Temperature change (C) per degree of global warming

Figure 5.16 Summer mean temperature and precipitation change per degree of
global warming for the East Suffolk and Essex Water Resource Zone, for all high-
resolution AGCMs as a function of driving AOGCM (AREGEPE/OPA or
HadCM3 under emissions scenarios A2 and B2) and all RCMs as a function of
driving AGCM (ECHAM4/OPYC3 or HadAM3H under emissions scenarios A2
and B2). The black circles represent the driving AOGCMs ECHAM4/OPYC3 and
HadAM3H estimates for the East of England under emissions scenarios A2 and B2

Due to the larger number of simulations with RCMs and AGCMs, compared to the
AOGCMs in section 5.4, it was possible to use the deterministic results shown here and
assume each model was equally likely.45 Skill was also investigated by comparing RCM

45
This is also the case because none of the simulations crossed zero (or were very close to zero) for
summer precipitation change, which was not the case with AOGCMs (in section 5.4). In probabilistic
pattern-scaling using deterministic estimates that cross zero to represent the response pattern produces
non-continuous PDFs of regional precipitation change. In order to overcome this, the AOGCM response
pattern is assumed to have a uniform PDF (see Chapter 7). For RCMs it is feasible to assume that all
simulations are equally likely (as a super-ensemble) because all the estimates are to one side of zero, thus
producing a continuous PDF of regional precipitation change once multiplied by the scaler.

107
and AGCM control simulations with observed data. Figures 5.17 and 5.18 shows how
all RCM and AGCM control simulations match with comparable catchment data for the
Ruthamford and Norwich & the Broads Water Resource Zones. A number of models are
estimating mean summer precipitation and variance accurately for the Ruthamford
Water Resource Zone. For the Norwich & the Broads Water Resource Zone this number
is perhaps smaller. These results are elaborated further in Chapter 7 to establish the
most skilled models to use in the analysis.

0.75
JJA St. Dev. Precipitation

0.7
0.65
0.6
(mm/day)

0.55
0.5
0.45
RCM Ruthamford
0.4
Observed Ely Ouse
0.35
0.3
1 1.2 1.4 1.6 1.8 2 2.2 2.4
JJA Mean Precipitation (mm/day)

Figure 5.17 Comparison between mean summer precipitation and variance for the
1961-90 period as modelled by RCMs and AGCMs over the Ruthamford Water
Resource Zone and as observed for the Ely Ouse catchment.

0.75
JJA St. Dev. Precipitation

0.7 RCM Norwich

0.65 Observed Wensum
0.6
(mm/day)

0.55
0.5
0.45
0.4
0.35
0.3
1 1.2 1.4 1.6 1.8 2 2.2 2.4
JJA Mean Precipitation (mm/day)

Figure 5.18 Comparison between mean summer precipitation and variance for the
1961-90 period as modelled by RCMs and AGCMs over the Norwich & the Broads
Water Resource Zone and as observed for the Wensum catchment.

108
5.6 Summary

This chapter has performed a number of tasks. First it characterised the climate of the
Anglian region. In particular it estimated natural multi-decadal climate variability for
several areas in the Anglian region. This is important as it allows the comparison of the
magnitude of anthropogenic changes with natural variations. Secondly, it reviewed
previous climate change scenarios for the region and determined that summer
precipitation decreases have remained a robust result throughout assessments. The
pattern-scaling technique, also used in previous climate change scenarios, was then
explained in more detail. Sections four and five examined the responses of complex
climate models to anthropogenic forcing in the Anglian region, with a particular focus
on summer precipitation change. It was shown that most modelled simulations estimate
drier and hotter summers for the future in the Anglian region, but the magnitude of the
change is considerably uncertain. Response patterns were normalised by global mean
temperature and control simulations were compared to observations. The combination
of these results with results from Chapter 4 allows the description of future climates for
the Anglian region in a probabilistic way (i.e., with estimates of uncertainty associated
with it). Together with information on climate impacts, which is introduced in the next
chapter, this will serve as the basis for the analysis conducted in Chapter 7.

109
 CHAPTER 6. WATER MANAGEMENT IN THE ANGLIAN
REGION

This chapter applies the scenario development framework to a specific adaptation

decision-making context. The context is water policy in the UK, more precisely long-
term water resources planning in the Anglian region. Water is essential to life on the
planet, one of its most precious resources. Water is also part of the climate system,
which prima facie makes it particularly sensitive to changes in the climate. Water
resources was the sector chosen to illustrate the decision-making context because of its
sensitivity to climate change and its societal (and environmental) importance. The
Anglian region was the area chosen because water is already a scarce resource and
climate change is expected to make it scarcer, thus suggesting that adaptation decisions
will have to be taken in the near future in order to manage this resource sustainably.

The chapter is divided into six sections. Section 6.1 provides a background to water
policy in the UK and the Anglian region. Section 6.2 describes how water resources
planning is conducted in England and Wales. Section 6.3 explains how climate change
has been incorporated in water resources planning in the past. Section 6.4 elaborates on
why water managers are still not adapting to climate change. Section 6.5 describes how
adaptation options were elicited and collected and section 6.6 demonstrates how the
options can be linked to changes in the climate. Finally, section 6.7 provides a summary
of the whole chapter.

6.1 A brief history of water policy in the UK and the Anglian

region

In the UK, there are a number of European and national legislations that regulate water
or have a major impact on water. At the European level various EU directives are
relevant such as the: Nitrates Directive, Habitats Directive, Freshwater Fish Directive,
Shellfish Waters Directive, Dangerous Substances Directive, Groundwater Directive,
Urban Wastewater Treatment Directive, Drinking Water Directive, Bathing Water
Directive and Surface Water Abstraction Directive. The most relevant directive in terms
of European Water Policy is the Water Framework Directive which entered into force in

110
December 2000. Its main aim is to protect all types of water (rivers, lakes, coastal
waters, and groundwaters) with an ambitious objective to ensure that all European
waters meet good status by 2015 (WFD 2000). It achieves this through integrated
river basin management plans which require cross border cooperation and active
participation of all stakeholders in water management activities. The Water Framework
Directive does not explicitly mention climate change but it is clear that the good
status of water could be impacted by climate change.

At a national level, the Water Resources Act of 1991 has been the major piece of
legislation regulating the water resources sector since privatisation in 1989. However,
over the last half a century the water resources sector in the UK has witnessed
considerable organisational and institutional change (Hassan 1998). In the 19th century
water supply was largely private sector activity, regulated by local governance
structures and accountable to parliament. In the late 19th and early 20th centuries a
large proportion of these private companies were taken over by local-level
municipalities with water supply and sewerage being brought under public control
through the 1963 Water Resources Act (Johnson and Handmer 2002, p. 349).

Table 6.1 provides a summary of the institutional changes, the meteorological events
and the concerns and responses of the largest water company in the Anglian region
(Anglian Water), during the last 40 years as experienced by a senior water resource
manager. The first decade of the 1963 Water Resources Act (1964-1974) was a decade
of considerable resource development. Three reservoirs, Grafham Water, Pitsford and
Covenham reservoirs, were developed in the Anglian region (see Figure 6.1 for
locations) with a combined yield of 354 Ml/d and two inter-basin transfers with a
potential yield in excess of 400 Ml/d. The 1975-76 drought highlighted the fragility of
the water resource situation, particularly in eastern and southern England, despite
considerable resource development of the previous decade. The drought combined with
concerns over long-term supply of water versus increasing demand, led to further
development of water resources from 1977-87. Two new reservoirs, Rutland Water and
Alton Water, provided a combined yield of 355 Ml/d (Smith 1997a). This period also
saw the creation of the Regional Water Authorities with responsibility for water supply
and distribution, regulatory functions and sewerage functions. Some 29 private water

111
supply companies were retained as agents of the Water Authorities (Johnson and
Handmer 2002).

Figure 6.1 Rivers, lakes and reservoirs of the Anglian region (in grey). Source:
Environment Agency.

The period 1988-92 was a rather troubled period for the water industry with a prolonged
drought (Marsh et al. 1994) and the privatisation of the water authorities. Because of
concerns about the environmental impacts of unconstrained demand the response to this
drought was mainly focused on demand management rather than resource development.
Subsequently, the 1995 drought recorded only 40% of average rainfall for the Anglian
region between April and September, and exceptionally high demand. Nonetheless,
water resources were not unduly stretched, due in part to previous investments in
drought-related schemes and the reservoir refill strategy adopted over the previous

112
winter (Smith 1997a, p. 12). This was not the case for all water companies in England
and Wales. Several struggled to cope, with half of the major water companies restricting
water use during the 1995/1996 period (Subak 2000). Yorkshire Water Services resorted
to trucking water to customers (using over 700 tanker trucks around-the-clock for five
months) at great expense after failing to induce the public voluntarily to reduce their
consumption sufficiently (Haughton 1998, Bakker 1999, Bakker 2000). Soon after, co-
ordinated by the government, a group made up of representatives of all the water
companies met and put together an action plan. The Agenda for Action called for
water companies, in consultation with the Environment Agency, to estimate water
supply and water demand over the next 25 years (Subak 2000). This is how long-term
resource planning was institutionalised in the UK (section 6.2 explains water resources
planning in more detail).

The early decades of the last 40 years were characterised by resource development and
little concern for the environment whereas the most recent decades have tended to focus
more on demand management and environmental monitoring. It is noteworthy that
concerns over rising demand and extreme meteorological events, such as droughts, have
been the main drivers for developments in Anglian Water.

More recently, the Water Bill of 2003 has amended the Water Resources Act 1991,46
but it is the latter that has regulated the sector for over a decade. It is noteworthy that
privatisation has not deregulated the industry entirely; instead it has been re-regulated in
what remains a curious public-private hybrid (Bakker 1999, p. 368). According to
Barrauqu (2003), England and Wales offers an extreme example in Europe of
liberalism with central state regulation. Water companies remain firmly established as
monopolies, presiding over networks that give consumers no option in their choice of
service provider (Bakker 1999, p. 368). The Government has carefully considered
whether to introduce competition for household customers. It concluded that the
drawbacks of increasing competition in the supply of water for household customers are
likely to outweigh the potential benefits (DEFRA 2002).

46
The Bill moves from licensing based on purpose of use to one based on volume consumed.

113
In response to housing shortages and rising house prices, the Government set out its
proposals to build an additional 1.1 million homes between 2002 and 2016 in the wider
South East in the Sustainable Communities Plan The areas in the plan include the
already established Thames Gateway and the three new growth areas of the Milton
Keynes/ South Midlands, London-Stansted-Cambridge and Ashford. These
developments are likely to be accompanied by increases in water demand across the
region.

Since privatisation in 1989 the main players in the UK water supply sector have been:
? Private water companies: they have the statutory responsibility to supply
water to customers.
? DEFRA (formerly DETR): responsible for policy, standard setting and the
regulatory framework.
? Drinking Water Inspectorate: monitors and enforces drinking water
standards.
? The Environment Agency: has a statutory responsibility to secure the proper
use of water resources and has to pay particular attention to the statutory
requirements of water companies; it also has a duty to protect and enhance
the environment, and to implement some aspects of European legislation at
designated nature conservation sites (EA 2003a).
? Ofwat: is the economic regulator; it has a duty to ensure that water
companies statutory functions are properly carried out and that they are
able to finance these functions.
? English Nature and Countryside Council for Wales: are statutory advisers to
government on nature conservation.

Other players include:

? Consumer Council For Water (formerly WaterVoice): represents customers

interests.
? Water UK: is the industry association that represents all UK water and
sewerage companies.

114
 ? UKWIR (United Kingdom Water Industry Research Limited): provides a
framework for the procurement of a common research programme for UK
water operators on 'one voice' issues.

Every five years since privatisation, the economic regulator has conducted a periodic
review of water company prices (maximum increases companies may charge to
consumers). Therefore, water companies investment is cyclical and spread out over a
five-year period. Ofwat reviewed water company prices in 1994, 1999 and most
recently in 2004. This process is also know as the Asset Management Plan (AMP)
periodic review. The 2004 periodic review was the fourth of its kind so it is known as
AMP4. As part of the cyclical AMP process both the Environment Agency and Ofwat
scrutinise water companies Water Resource Plans every five years. Water resources
planning focuses on the need to ensure that companies maintain security of supply to
customers in a way that is economically, socially and environmentally sound (EA
2003a). It shows how the water industry intends to maintain the balance between water
supply and demand for the next 25 years.

115
 Table 6.1 Development of water resources in the Anglian region from 1963-1995 (adapted from Smith 1997a, Johnson and Handmer 2002).

Period Institutional Characterisation Developments in Anglian Water Concerns

framework
1963-74 1963 Water Decade of quantity; introduction Construction of Grafham Water, Pitsford and Flooding of agricultural land by
Resources Act of abstraction licensing; era of Covenham reservoirs; Inter-basin transfer such landowners and their local rural
Public control major resource development as the Ely Ouse Essex scheme and the Trent- communities
with localised justified by simple extrapolation of Witham-Ancholme scheme
decision making existing demands
1975-76 Drought; two warm summers Widespread introduction of demand Maintaining water supplies
(1975 and 1976) and one dry management through hosepipe bans and
winter (1975-76) selective banning of non-essential use;
emergency boreholes were developed
1977-87 1973 Water Act Decade of plenty; regional Rutland Water was commissioned and Alton Over the availability of water
Public control and planning started with the creation Water was built; Great Ouse Groundwater resources in the longer term and the
centralised of Regional Water Authorities in Scheme developed continuing upward trend of
decision making 1974; water plans were developed demands; awareness of the impact
of leakage
1988-92 1989 Water Act Privatisation of the water Hosepipe bans and a concerted advertising Environmental impacts of
Privatised companies and creation of the programmer to control demand; restriction on unconstrained demand
framework National Rivers Authority; irrigation water use
Drought (dry, mild winters, with
limited replenishment of resources,
and dry, hot summers with high
demand)
1992-94 1991 Water Act Post-drought optimism Undertakes baseline environmental scooping Securing flexibility and security of
exercises; further promotion of domestic supplies while maximising the
metering and continued leakage reduction utilisation of existing licensed
resources within an acceptable
environmental regime
1995 Drought Coped without major problems

116
6.2 Water resources planning

Water resources planning has gone through considerable development and

sophistication over the last decade in the UK.47 Figure 6.2 shows the conceptual
framework for the current guidelines. The unit the Water Resources Plan (WRP) is
compiled in is the water resource zone which is defined as: The largest possible zone
in which all resources, including external transfers, can be shared and hence the zone
in which all customers experience the same risk of supply failure from a resource
shortfall (EA 2003c, p. 18). For example, Anglian Water has 12 Water Resource Zones
(WRZs), although these were aggregated from 89 Planning Zones. The basic
framework of a water company plan follows a structure that considers all of the
components of a companys water balance for each year until the planning horizon of
2030. In broad terms this involves evaluating the amount of water available for supply
as well as all of the demands on this water (EA 2003a, p. 8). This is done for dry
year annual average unrestricted daily demand as the norm, but normal year and dry
year critical period are also explored to test sensitivities. This is first carried out for
baseline conditions (2002/03) and then for final planning scenarios. Changes to the
availability of water from existing sources must be forecast; such changes could arise
from planned environmental improvements, the impact of climate change, or the effect
of gradual pollution. A forecast of demand must also be made, taking into account
current company policies on such issues as leakage control, promoting the efficient use
of water and household metering (EA 2003a, p. 8).

Central to the water resources plan is the concept of headroom that combines
numerically the forecasts of supply and demand. Prior to AMP3 there was no agreed
methodology for calculating headroom, which resulted in inconsistencies in regulatory
submissions. Carnell et al. (1999) developed a pragmatic methodology for converting
uncertainty in the supply/demand balance into headroom. The following concepts are
important to understand headroom:

Target headroom: the minimum buffer that a prudent water company should allow
between supply (including raw-water imports and excluding raw water exports) and

47
For a review of how US Federal water resources planning has developed see Major and Frederick
(1997).

117
demand to cater for specified uncertainties in supply and demand (except for those due
to outages).

Available headroom: the difference between Water Available For Use (WAFU;
including raw-water imports less raw water exports) and demand.

Supply: water available for use.

Demand: dry year annual average unrestricted demand (or critical period demand).

There is a range of uncertainties that affect headroom, climate change being one of them
both on the supply and demand side (see Table 6.2).

Table 6.2 Key uncertainties affecting headroom (Carnell et al. 1999).

Factor Characteristic
Supply Vulnerable surface-water licences Expected loss
related Vulnerable groundwater licences Expected loss
Time-limited licences Expected loss
Bulk transfers Expected loss
Gradual pollution causing a Expected loss, increases with time
reduction in abstraction
Accuracy of supply-side data Random variability
Single-source dominance and Random variability, possible
critical periods demand increase
Uncertainty of climate change on Unexpected changes, increases
yield with time
Demand Accuracy of sub-component data Random variability
related Demand forecast variation Unexpected changes, increases
with time
Uncertainty of climate change on Uncertainty of climate change
demand

The headroom methodology provides a risk envelope against which companies should
derive a balanced portfolio of supply and demand management options for each
resource zone. The company will have to identify the costs, benefits and risks for each
option. The final stage of the planning process it to produce a final planning
scenario which shows the results of the chosen options for the zone in question. This
should confirm that that the companys plan will maintain security of supply in a way
that is environmentally, socially and economically acceptable (EA 2003a, p. 10).

118
 Determine resource zones and agree with
Agency.

Assemble base year data

Actual out-turn (annual review) base year

Assemble supply-demand
forecasts: determine planning
STAGE 1 scenarios
ASSEMBLE
BASE YEAR Supply-side: assemble deployable Demand-side: assemble information on
DATA & output figures and consider impact of raw water use from point of abstraction
SUPPLY/ climate change
DEMAND
Supply-side: identify & specify Demand -side: assemble information on
FORECASTS Sustainability reduction and assess potable water to point of delivery
outage

Supply-side: existing water available Demand-side: establish normal year (base year)
for use consumption and consider impact of climate change

Demand-side: set out assumptions for

unrestricted dry year demand for annual average
or critical period

Prepare initial SupplyDemand

STAGE 2 Determine available and target headroom

IDENTIFY
PLANNING
PROBLEM Establish if target headroom is available within available headroom

Amend as necessary through consultation with the Agency

Define set of relevant total water management options & decide
modelling framework

Undertake economic analysis & calculate average incremental social cost

Test sensitivity of initial list of total water management options and

consider risks and uncertainties

STAGE 3
OPTION Refine initial list of total water management options and determine solution
SELECTION

Plan sequence of development of total water management options

Compile planned SupplyDemand Balance

WATER RESOURCES PLAN

Figure 6.2 Conceptual framework for the preparation of a water resources plan

119
(EA 2003c).

6.3 Climate change and water resources planning

The previous section explained concisely how water companies put together water
resource plans. In this section, the treatment of climate change in the water resource
planning is examined in more detail. The first requirement to incorporate climate
change scenarios in water resources planning was part of the Third Asset Management
Plan (AMP3), which was agreed in 1999 and ran from 2000-2005. Water companies
were required to produce water resources plans for the period 2000-2005, which looked
out until 2025 and included climate change in the form of the UKCIP98 scenarios (see
Hulme and Jenkins 1998). Using these climate change scenarios through a fast track
approach, factors of change in monthly streamflow and annual groundwater recharge
were estimated for various regions of England and Wales (Arnell et al. 1997). These
factors were then combined with observed data to produce projected conditions under a
changed climate in the 2020s.

Following the AMP3 process there was an outcry for higher resolution climate change
scenarios from the UK water industry, as the whole UK was covered by only three grid
boxes. For example, the Welsh Water company believed that the scenarios were too
aggregated for a region as large as Wales (Subak 2000, p. 152). There was certainly a
mis-match between the spatial scale of the climate scenarios (250 km) and the areas the
water managers were interested in (catchment scale). This user demand and the
advancement of regional climate modelling led to the construction of the UKCIP02
scenarios (Hulme et al. 2002) now at a 50km resolution, but only constructed using a
single Regional Climate Model driven by a single Global Climate Model (both from the
Hadley Centre) (cf. section 5.2).

The AMP4 process was slightly different from AMP3 in particular in its degree of
modelling sophistication. Instead of using a deterministic approach to calculating
headroom, a probabilistic approach was adopted. The uncertainties of each headroom
component are defined as probability distributions and combined using probabilistic

120
simulation techniques (sometimes called Monte Carlo simulation) (EA 2003c).
UKCIP02 scenarios were translated into streamflow and groundwater changes by Arnell
(2003a) and verified by Reynard and Young to be an appropriate approach for the initial
assessment of the impact of climate change (UKWIR 2003, p. 63). However, the
Environment Agency noted that it does not provide sufficient certainty for a major
investment. Box 6.1 provides a verbatim of the supplementary guidance provided by
the Agency on the treatment of climate change.

121
 Box 6.1 Supplementary Guidance Note 1: Treatment of Climate
Change
Supply

? Impact should be considered on a resource zone level.

? Arnells report provides an indication of the amount of change that should be applied for
different decades between the 2000s and the 2020s (section 6 table 12), together with
guidance for using these factors.
? Ideally, a resource zone simulation model should be used to look at impacts:
? River flow either use Arnells scaling factors or a simulation model using rainfall
and evaporation scaling;
? Reservoir yield use river flows as above;
? Groundwater sources use aquifer models where possible. Where this is not possible,
examine the existing vulnerability of yield to drought. This should help to identify
those groundwater sources that are vulnerable to climate change. For these sources,
estimate the effect of reduced recharge in line with Arnells estimates. Where sources
without good models appear to be sensitive to climate change, in the long term it may
be appropriate to carry out further modelling work.
? Where no resource zone model exists, estimates of the impact on individual sources
should be made. An estimate of the cumulative impact for the resource zone should be
made, but care must be taken to ensure that this does not lead to the inappropriate
accumulation of impacts.
? In most cases the four core UKCIP scenarios will give very similar results to the 2030s.
Therefore for the purpose of establishing the consequences of climate change, results need
only be presented for Arnells medium scenario, which reflects both the medium-high and
medium-low UKCIP scenarios. This does not preclude the use of other UKCIP scenarios
for sensitivity purposes.
? Arnells approach is suitable only for flow and rainfall records before 1990. If the
deployable output of sources is dependent on droughts that occurred in the 1990s, it is
possible that these climatic events already include some impact of climate change. An
allowance for this must be made and justified.

Demand

? Impact should be considered on a resource zone level.

? Household demand - % factors from CC:DEW should be used for average demand. These
refer to 2030 and should be scaled linearly between the base year and 2030.
? It is not anticipated that companies will factor peak demands but where peak demands are
significant to the supply demand balance and investment then a well argued case will be
considered. This will need to consider the link between current peak demand and weather
conditions and show that it is reasonable to assume that peak demand would change with a
different climate.
? Industrial demand: CC:DEW provides some guidance for different sectors. This can be
used but assumptions must be described clearly. Alternatively, an assumption of no impact
for industrial demand may be made.
? Agricultural demand: for most water companies this is small. CC:DEW provides estimates
of unconstrained future irrigation demand. The benefits of meeting such demand must be
considered carefully to produce an estimate of the total agricultural demand for which the
company should plan. Only if water companies can demonstrate the proportion of their
non-household water use that is used within agriculture for irrigation purposes should an
estimate of the impact of climate change on agriculture be considered. Assumptions must
be stated clearly.

122
 Box 6.1 (continued)
Impact on supply-demand balance

The effect of climate change on the supply-demand balance should be assessed for each
resource zone.

? If it is clear that the impact of climate change makes little or no difference to activities
before 2030, the company may state this and does not need to change its water resources
plan;
? If the impact of climate change makes little difference before 2020 but could to 2030
consideration must be given to the timing of the necessary investigations;
? If the impact of climate change is great enough to require changes to the water resources
plan before 2020, the company should consider the further investigations and analysis that
will be needed.

An allowance for climate change may be made explicitly or in headroom depending on the
outcomes referenced above. Where included in headroom, this should reflect only the
uncertainty associated with climate change. Where climate change is incorporated explicitly
in forecast supply or demand, only the uncertainty associated with this should be included in
headroom. Where no explicit allowance for climate change is incorporated in supply or
demand, a greater allowance may be incorporated in headroom. In all cases assumptions must
be stated clearly.

For each resource zone a statement of the impact of climate change should be included in the
supply-demand balance submission.

The analysis described here will not be sufficient to justify significant expenditure. However,
for zones where it is identified that the impact of climate change may be a significant driver
for investment in the next 10 years, it should provide sufficient information to allow the
scoping of such action to begin. Options in these zones are:
? To decide to carry out further investigations between 2005 and 2010;
? To decide to carry out further investigations after 2010.

If this climate change analysis indicates that there may be a need to start work on
implementing new options before 2010, companies will be expected to finish investigations
as early as possible. The nature of such investigations will depend on the scale and type of
problem but they could include:
? Developing resource zone models where existing models are weak;
? Carrying out further work to understand source outputs and the risks associated with
climate change;
? More detailed hydrological modelling using the UKCIP work;
? Further modelling using results from other global circulation models (GCMs).

The effort and cost involved in using other GCMs could be significant, and it will be
necessary to consider this when scoping further work.
The outcome of this analysis might identify the need to take further action. If it is agreed that
this requires significant expenditure before 2010, Ofwat has indicated that it could be dealt
with through the usual process for logging up.

Source: Environment Agency (2003c)

123
Furthermore, the Agency guidelines note:

The results arising from the application of climate change scenarios must be used
sensibly, pragmatically and consistently, accepting that continued research effort will
no doubt go on to refine and develop better and more reliable climate predictions. It is
important that companies think holistically about climate change when considering
options. Overall the Agency encourages a low regrets approach, built around the
robustness of options and flexibility of companies plans in adaptation to alternative
scenarios. (EA 2003c, p. 84)

In broad terms, climate variability and change represent a threat to water supply in
England and Wales insofar as they increase the magnitude and duration of dry periods,
as well as the frequency of occurrence of such events. Changes in extreme events such
as droughts are what water managers would like to know. However, the scenarios used
are usually of mean changes not changes in extremes; the latter are usually more
uncertain than the former. The Agency (EA 2003c) notes that even if significant
changes in water availability are to occur, it is not expected that these will be sudden or
dramatic (p. 84).

6.4 Are UK water managers adapting to climate change?48

The previous sections described how water companies put together water resources
plans and how climate change has been incorporated in them. This section examines the
resulting water resource plans and their treatment of climate change. Based on the
literature, the AMP3 process is briefly reviewed but the majority of the analysis focuses
on AMP4 based on available literature and elite interviews.49 A preliminary assessment
is made on whether water companies are indeed adapting to climate change.

48
This sections title is inspired by a talk given by Ian Holman at Cranfield University on 15 March 2004
entitled Climate change and water resources why arent water managers adapting for climate change?.
49
The author of this thesis interviewed Paul Hart from the Environment Agency Anglian Region (12
August 2003), Paul Sayer and colleagues at Essex and Suffolk Water (10 September 2003), and Ed Smith
and Gerry Spraggs from Anglian Water (12 February 2004). The interviews followed a semi-structured
protocol and where aimed at understanding how climate change was incorporated into water resources
planning, what difficulties they had in doing this, and what adaptation options were available to them. On
average interviews lasted about two hours. As will become evident in subsequent sections, the author only
managed to collaborate with one of these decision-makers (Anglian Water) due to time restrictions.

124
6.4.1 AMP3 and climate change

The Third Asset Management Plan (AMP3) in 1999 was the first one to include climate
change. Arnell et al. (1997) examined the effect of four UKCIP98 climate change
scenarios on river flows and groundwater recharge. The Environment Agency
summarised its results and implications as:

The impact varies according to location and underlying rock type. For rivers
dominated by groundwater, average flows decrease in late summer and increase
through the rest of the year. The decrease in late summer flows is greatest in the south
and east. Average recharge to aquifers is expected to increase in all scenarios for all
aquifer types. Rivers not dominated by groundwater show a similar pattern, with lower
flows in July, August and September throughout England and Wales. However, the
decreases are small in the north. Flows increase in the rest of the year in all scenarios,
with the increases being lowest in the south. The implications of this for water users is
mixed. As recharge increase everywhere in all scenarios, groundwater abstractions
should be at least as reliable as they are now. Abstractions that need summer water will
become less reliable through much of England and Wales, as licence conditions that
protect low flows become effective more often. This will be a particular problem in the
south, and will apply also to rivers dominated by groundwater in this area. However,
higher winter flow mean that other abstractions should continue at current levels of
reliability. The storage of winter water should be more reliable, with more potential for
reservoir filling later into the spring (EA 2001c, p. 77).

Subak (2000) investigated UK water managers perceptions of past variability and

future scenarios. She noticed that the new requirement to incorporate regional climate
change scenarios in future supply assumptions appears to have had little impact on
planning in the region to date (p. 137). This was because many water resource
planners believe that the scenarios generated are too aggregated and do not encourage
a precautionary approach to planning (p. 137). Subak also noted that two companies
have been notable in their sustained interest in climate change: Anglian Water and
Severn Trent. Their concern for climate change has led them to apply different
strategies: Anglian has focused primarily on demand management whereas Severn Trent

125
is pursuing a strategy to try to increase water supplies through new construction.
Generally, however, most water companies do not view climate change to be of
significant concern in their region (p.150). For the year as a whole, even the driest
scenario projects only a few percentage reduction in annual streamflows in the more
arid regions of England and Wales. Accordingly, it is not surprising that many
companies do not believe climate change should elicit a change in water resource
planning (p. 153). Subak (2000) concludes that most water companies implemented
their supply adaptation to climate change during the late 1970s/early 1980s and their
demand adaptations during the mid-1990s. This author also concludes that:

Worries over providing sufficient future water supply appear to be a concern for only a
few water companies, and for reasons largely to do with the regions rapidly expanding
populations. Unless climate change scenarios offer very different results in the future,
it would appear that their implications for planning will remain minor. Indeed,
concern over leaving natural ecosystems with sufficient water may be a more important
driver for water resources planning than the prospect of climate change (p. 154).

6.4.2 AMP4 and climate change

AMP4 was different from AMP3 on a number of issues. It was more sophisticated (it
included probabilistic headroom) and it used different climate change scenarios. One
would not expect identical results, but how do they compare? Starting with the climate
change scenarios, UKCIP02 were more severe than UKCIP98, i.e., they showed more
rainfall in the winter and less rainfall in the summer compared to UKCIP98 results, at
least for the Anglian region. This consequently led to projections of higher streamflow
in winter and even lower streamflow in the summer. The result for water companies of
England and Wales was a mixed picture regarding the impacts of climate change. Box
6.2 provides the view of the Environment Agency.

126
 Box 6.2 Climate Change
Most companies have made an allowance for climate change either explicitly or in
headroom. The range of impacts calculated by the companies is enormous despite
common guidance and industry-wide methods. A few companies tell us that present
estimates of climate change will make little difference to their long-term strategy. Most
companies have told us that they can accommodate the impact of climate change by
making small changes to their plans in the decade between 2020 and 2030. We urge
companies that have calculated the impact of climate change to be negligible or small to
keep their assessments under active review as more information becomes available.

A few companies have identified serious impacts from climate change. We welcome the
preliminary analysis that these companies have carried out. We have agreed with the
water industry that these analyses are not on their own sufficient to justify
significant changes to investment programmes. We expect all companies that have
identified significant impacts from climate change to carry out more detailed analyses
over the next few years. These include detailed modelling studies using alternative
global climate models (GCMs). We will expect to see timetables for this work from
these companies. For many companies this is a new area of work and they may need to
seek advice and expertise from climate change specialists. Joint work through Water UK
and UKWIR may be a good way to carry out some of this analysis. We will work with
the water industry to help to scope further investigations into the impact of climate
change.

Most companies that have identified significant impact from climate change seem to
intend to deal with this by developing new water resources and, in particular, reservoirs.
We understand that these are preliminary assessment of need in the late 2020s but we
expect companies to follow Government guidelines and consider a full range of possible
responses to climate change rather than only the more obvious engineering options.

From Environment Agency (2004a, p. 20).

It is hardly surprising that the range of impacts was enormous. If all water companies
used Arnells cool and wet and warm and dry scenarios then they would find
opposing precipitation signs (decrease or increase) for certain seasons, which would
result in a large range of impacts. Also, if each company used their own locally-
calibrated and validated catchment or aquifer simulation model then the potential for
wide-ranging results is enormous. As the Environment Agency (EA 2001c) notes
individual catchments respond to rainfall and evaporation in different ways (p. 77).
The range of impacts is also explained by the geographical location of the water
company and its main source of supply (surface or groundwater).

127
These two factors can be explained by what Price (1998) calls the rainfall gradient and
the storage gradient. Price (1998) divides Great Britain into two along a line that
extends roughly from Teeside, down the east of the Pennines, though the Midlands and
south to Dorset. The area to the south-east of the line is densely populated and
characterised by low effective rainfall and high demand for water for public,
commercial and agricultural use. North-west of the line the country is less densely
populated, enjoys higher rainfall and sees lower demand for water. These factors are
controlled largely by the geology (Price 1998). The storage gradient is the opposite of
the rainfall gradient. In the north-western area there is little natural storage in aquifers
for water except in superficial deposits. Discharges increase rapidly in response to
effective rainfall, but also decline quickly once rainfall has ceased. In the south-eastern
area many of the rocks are permeable, so much of the effective rainfall infiltrates to
provide large supplies of groundwater; baseflows components are much larger than in
the north-west and river flows are more evenly distributed throughout the year. Due to
this diversity, it is not astonishing to find such an enormous range of impacts due to
climate change for water companies of England and Wales. Finally, different water
supply systems will respond to changes in streamflow and groundwater in different
ways, according both to their physical characteristics and to the way in which they are
operated.

The second paragraph in Box 6.2 seems to suggest that the Agency is not yet prepared
to give the go ahead for adaptation to climate change, in particular as proposed by some
water companies through new reservoirs. This view is shared by the Government, who
believes that, where major new resources are planned, further investigations are
necessary before further commitment is given. What this means is that while climate
change is being incorporated into water resources planning, adaptation to this change is
not being considered at present (if developing new water resources is the strategy
proposed by the water company) until more detailed analysis are carried out.

If water companies are not adapting to climate change then what are they adapting to? It
is important not to forget all the drivers and pressures that are considered in the WRP
other than climate change: changes in demand, changes in licences, environmental
improvements, changes in leakage, pollution, raw water quality, etc. These are in
essence the major drivers of the water sector.

128
Based on the Environment Agencys (2004a, 2004b) review of water companies water
resources plans, Table 6.3 provides an overview of the situation for selected water
companies in the Anglian region, companies bordering the region and companies that
found a significant impact from climate change. Water-only companies such as
Cambridge Water and Tendring Hundred Water suggest climate change will have no
impact upon their business. In fact, the latter company did not include climate change in
their final plan because they were suspicious of contradictory scenarios! Major water
companies such as Anglian Water and Severn Trent Water show a range of impacts
from minor to major. Interestingly, their strategies to cope with climate change have not
changed at all since Subak (2000) reported them. Three other water companies reported
significant climate change impacts: Mid Kent Water, South West Water and Southern
Water, all in the south of the UK. Their proposed adaptation strategies includes a mix of
new resource development and demand management.

It is rather surprising that companies that are currently ranked low in Ofwats security of
supply index (i.e., they have a significant deficit against target headroom), such as
Essex and Suffolk Water or Thames Water, are not flagging climate change as an added
long-term problem. This is astonishing because almost all other water companies
surrounding these two companies flag climate change as a serious issue. However, this
could also reflect their priority to deal with the current deficits rather than long-term
problems. Finally, it is worth noting Three Valleys Waters situation, who do not flag
climate change as an issue, but who argue that the company has limited opportunities to
develop additional resources (i.e., supply-side measures) within its supply area. Could
this reveal a limit to adaptation for this company?

129
 Table 6.3 Characteristics of selected water companies in the Anglian region, companies bordering the region and companies that found a significant
impact from climate change (compiled from EA 2004a, b).

Company Population Security of supply Baseline supply- Supply Impact of climate change Options
supplied index for planned demand balance
service
assumption
Anglian 4 million B (Marginal deficit Currently has a deficit Mixture of surface and The scenarios suggest that Demand management;
Water against target in two of its 12 resource groundwater, with a by 2030 there could be a Extend the water treatment
headroom) zones; forecasts that a large integrated significant impact on works at Rutland Water by
further four resource reservoir system surface water sources with 2009; develop strategic
zones will have a deficit operating in the west of a loss of some 40 Ml/d. mains within and between
between now and 2012 the companys area The impact on resource zones; expects that
groundwater abstractions this will deliver 100 Ml/d of
are insignificant. Did not additional water by 2010 and
include the impact of a further 40 Ml/d before
climate change on demand 2030
Cambridge 300,000 A (No deficit Forecasts that it has Integrated supply No impact N/A
Water against target enough water to supply system based entirely
headroom in any demand until beyond on groundwater sources
zone) 2030
Essex and 1.7 million C (Significant Currently has a deficit Mixture of surface and Demand management,
Suffolk deficit against in the Essex resource groundwater, but leakage reduction and
Water target headroom) zone; forecasts that a predominately surface resource development
deficit in the Northern
and Central zone in
2025

130
Tendring 150,000 A (No deficit Forecasts that it has Groundwater and The numbers forecast are N/A
Hundred against target enough water to supply Ardleigh reservoir not significant and can be
Water headroom in any demand until beyond accommodated within
zone) 2030 future headroom.
Furthermore, the potential
scenarios are contradictory
and for this reason the
Company has made no
specific provisions in its
Final Business Plan
relating to the effects of
climate change
Three 3 million B (Marginal deficit Currently has deficits in Groundwater (60%), Limited opportunities to
Valleys against target the Northern and River Thames develop additional resources
Water headroom) Southern resource abstraction and transfer
zones; forecasts a from Grafham reservoir
deficit in the Central
resource zone by 2007
Other
Severn 7.2 million C (Significant Currently has a deficit River, groundwater and Estimates that by 2030 Leakage reduction, new
Trent deficit against in 3 of the 6 resource reservoir sources deployable output will be groundwater resources,
Water target headroom) zones; a fourth is reduced by nearly 180 reallocation of existing
forecast to have a deficit Ml/d due to climate change surface water abstractions,
from 2005 and a fifth etc. After 2010 they propose
from 2014 an additional 275 Ml/d of
resource development
(includes a major new
reservoir)

131
Thames 8 million D (Large deficit Currently has a deficit in
Water against target the London resource
headroom) zone; forecasts a deficit
in 2 resource zones
before 2010 and another
by 2018
Mid Kent 580,000 Forecasts that it will
Water have a deficit in all seven
resource zones between
now and 2030
South West 1.6 million Currently has a deficit in
Water (increases 2 resource zones ; the
by over third is forecast to have a
half a deficit by 2009
million
during the
summer
period
Southern 2.3 million 3 of 9 resource zones
Water have a deficit at times of
both average and peak
demand. Forecast to have
a deficit in one resource
zone by 2010, another by
2013, two more by 2018
and an eight by 2030
River Thames and
groundwater
Predominately It believes that the volume
dependent on of water available from its
groundwater sources groundwater sources is
more likely to reduce due
to climate change than it
previously thought
Believes climate change is
partially responsible for
about 2% less water
available than reported in
AMP3
Relies on groundwater It has made a moderate
for 70% of its supply allowance for climate
(many of its zones are change, but has assessed
sensitive to peak that the worst case impact
demands) could be significant
132
Reducing leakage, further
development of groundwater,
a large desalination plant (in
2018) and a major new
reservoir by 2021
Enhancing outputs from
groundwater sources;
increasing transfers between
resource zones; and
developing a new strategic
transfer; expects to deliver a
total of 35 Ml/d of resource
development by 2020
Combination of water
efficiency measures and
infrastructure enhancement
Further leakage control;
infrastructure improvements;
increased transfer capacities;
and demand management
6.4.3 A preliminary assessment

The UK is perceived to be at the forefront of adaptation planning, but is this really the
case? This case study has shown that the UK is indeed at the forefront of planning for
climate change with its inclusion on water resources planning guidelines. However,
enacting adaptation decisions to climate change based on water resource plans is still
not a reality. There are several reasons for this, which are discussed next in turn.

6.4.3.1 Uncertainty

One would have thought that water managers would be experts in dealing with
uncertainty from previous history of dealing with increased demand, droughts (climate
variability) and increased environmental concern (cf. section 6.1). However, it seems
that may not apply to climate change uncertainties. Headroom methodology is a
technique to manage uncertainty in the supply/demand balance that allows the
calculation of a safety margin where water companies can achieve their levels of
service. While various uncertainties are included in headroom calculations (see Table
6.2), water managers are much more familiar with water-related uncertainties (e.g., bulk
transfers or time-limited licences) than with climate change uncertainties. Therefore, the
degree of awareness, familiarity and trust in uncertain climate information is an
important factor in determining whether adaptation goes ahead or not (cf. views of
Tendring Hundred Water).

As mentioned in previous chapters, regional climate change projections are very

uncertain, for a range of reasons. At present, this seems to be the major deterrent for the
Agency to allow adaptation to go ahead (if the planned strategies are resource
development). Their advice is for water companies to do more detailed analysis that
uses more than one global climate model, which thus tries to sample more of the
uncertainty in future climate than did UKCIP02 (which uses just one regional climate
model driven by just one global climate model). The uncertainties associated with
regional climate change projections are substantial and will not decrease in the near
future. It will be interesting to see how long actors in the water sector can use
uncertainty to postpone taking decisions on adaptation. Conversely, the question that

133
should be asked is how much evidence is necessary to justify taking adaptation
decisions? Of course this is down to the decision-makers, but this will be explored in
Chapter 7 by quantifying uncertainties in regional climate change projections and
associating them with potential adaptation options.

6.4.3.2 Complexity

, the management of the risks associated with water supply security are immensely
complex. (Johnson and Handmer 2002, p. 357)
Water-resources planning is complex and requires the understanding of uncertainty
associated with physical, social, economic, and environmental factors in order to
produce a sustainable and practicable supply/demand balance. (Carnell et al. 1999, p.
414)
It is clear that the assessment of the future adequacy of water resources is a complex,
multi-criteria problem. Furthermore, each criterion is beset by uncertainty. (McIntyre
et al. 2003, p. 1)

Section 6.2 has shown the institutional and technical complexity of water resources
management. The three quotes above simply confirm this. Climate change brings its
own complexities and uncertainties to water resources planning, but even without
climate change the (water resource) system is incredibly complex. This complexity is
demonstrated in two ways: technically and institutionally. Technical complexity
emerges from putting the water resources plans together. The institutional complexity
comes from the fact that the water resources plans are almost negotiated between the
main stakeholders involved (private water companies, Ofwat and the Environment
Agency).

From a technical perspective, there are many important drivers, pressures and
uncertainties that dominate water resources planning other than climate change. For
example, increased demand or impeding directives from Europe can have a significant
impact on headroom. This emphasises the importance of sensitivity analysis in
determining how important each component is. However, if the system/model is not

134
integrated50 (which is probably the norm) then the sensitivity analysis will not pick up
any feedbacks. For example, climate change might lead to more droughts and the Water
Framework Directive requires certain minimum flows to preserve a certain species.
These two issues combined could have significant implications for the water company,
whereas individually these issues might not represent a threat to security of supply. The
complexity of the system also makes it difficult to pick out adaptation decisions to
climate change from adaptation decisions to any other changes.

6.4.3.3 Timescale

Intimately linked with complexity and uncertainty is timescale or planning horizon. The
further in the future one looks, the more uncertain and complex the planning becomes.
While the water resources plans have a planning horizon of 25 years, Ofwats price caps
only look into the next five years. There seems to be a mismatch between investment in
the next five years and the impacts in 25 years. Discounting might play a part in this. It
seems that Ofwat and the Environment Agency give much prominence, and rightly so,
to the short term problems of getting water companies out of deficit. However, in doing
so they are forgetting the longer-term.

6.4.3.4 Financing

Financing is extremely important if adaptation decisions are to be implemented.

Arguably, investment to adapt to climate change needs to start now or in the near future
if water systems are to cope with the gradual changes expected over the next 25 years.
But how can investment be justified in adapting to an uncertain future when there are
considerable deficits in headroom (in certain companies) today? This is perhaps one
reason why adaptation actions are not being funded yet. Another reason is the fact that
currently there is not enough certainty in climate change projections to justify
investment (or at least there is that perception; this is discussed further in Chapter 8).

50
By integrated the author means models that include sub-components (climate change, demand, etc.)
that interact with each other and are not analysed individually. This is prevalent in integrated assessment
modelling, but absent in many fields.

135
Finally, the issue of who will pay for adaptation is also important. Will the costs be
simply externalised to customers as is normal practice in the sector? This raises issues
of equity and water affordability. The Water Act 2003 means that the next set of
company water resource plans will have a statutory basis and that companies will have
to carry out wide public consultation on their draft plans. One hopes that this will lead
to an open debate on some of these thorny issues that will face the public eventually.

6.5 Adaptation options

The previous sections have described the institutional dimensions of water resources
decisions-making and how climate change has been dealt with. This section elaborates
on the options available to water managers in coping with pressures such as climate
change, increased demand, point source pollution, borehole deterioration, environmental
improvements, etc.51 They are based on elite interviews (with water manages from
Anglia Water, Essex and Suffolk Water and the Anglian Region Environment Agency),
water companies WRPs and other literature. It focuses specifically on the Anglian
region of the UK.

As previous sections have shown, the balance between supply and demand of water is
the core issue in water resources planning. Therefore, it is no surprise that adaptation
options can be divided into supply or demand related. Supply options mainly relate to
resource development, i.e., the augmentation of supply. Demand options are related to
demand management with a view to decreasing demand. Table 6.4 provides examples
of the options considered by the Environment Agency in their Water resources for the
future strategy for England and Wales.

51
See Frederick (1997) for a general overview of adaptation options in the water sector, with a particular
focus in the United States and demand management.

136
Table 6.4 Types of options considered in Water Resources for the Future: A
strategy for England and Wales (EA 2001c).

Resource Development Demand Management

New reservoir Improved leakage control
Reservoir raising Rainwater use (new development, non-
potable)
Winter storage reservoirs (single farm) Greywater use (new development, non-
potable)
Winter storage reservoirs (farm Waste minimisation of
consortium: 10-15) industrial/commercial
Surface water abstraction (unsupported White goods subsidies
for pws)
Groundwater abstraction or enhancement Retrofit of toilets to dual flush/interruptible
flush
Desalination Increased household metering
Wastewater re-use (direct - pws) Tariffs for measured charges
Wastewater re-use (direct - non-pws)
Wastewater re-use (indirect - pws)
Wastewater re-use (indirect - non-pws)
Aquifer artificial recharge and recovery
Canal transfer
River transfer
Pipeline transfer
Operational improvements

Anglia Water uses a software package called FORWARD (FORecasting of WAter

Resource and Demand) to forecast annual water demand for a 30 year period.
FORWARD then determines whether the current supply is sufficient and, if it is not,
calculates the lowest cost set of actions to maintain a supply and demand balance for a
given return level of service. FORWARD considers a number of different types of
schemes, which can be split into three main categories of demand, leakage and resource
management (Table 6.5).

137
Table 6.5 Types of schemes evaluated by FORWARD (AWS 2004a).

Type of Scheme Capital Investment Operational

Required Expenditure Required
DEMAND MANAGEMENT
Commercial Water Audit Low None
Household Water Audit Low None
Cistern Displacement Devices Low None
Additional Metering Medium Medium

LEAKAGE MANAGEMENT
Active Leakage Control Low High
Pressure Management Medium Low
Supply Pipe Replacement High None
Distribution Mains Rehabilitation High None

RESOURCE MANAGEMENT
Bulk Transfer of Water Medium Medium
Resource Development High Medium

Sheriff et al. (1996) have described in some detail the strategic resource development
options for England and Wales which were considered in an older resource strategy by
the National Rivers Authority. Figure 6.3 and Table 6.6 summarise the key strategic
options.

138
Figure 6.3 Strategic options considered by Sheriff et al. (1996).

139
Table 6.6 Key strategic resource options for England and Wales: yields and costs
(Sheriff et al. 1996). The shaded options are the most relevant to the Anglian
region.

Option Additional Indicative Indicative Time to

potential yield capital pumping cost promote
(Ml/d) cost ( (000s/Ml/day
million) per annum)
Severn-Thames transfer Supported 249 57 31.5 Medium
capacity 200 Ml/d
Severn-Thames transfer Supported 425 92 31.5 Long
capacity 400 Ml/d
Enlarge Craig Goch Up to 775 105 0 Long
reservoir regulating River
Severn
Partial redeployment 152 37 25.9 Short
Vyrwy reservoir (74 Ml/d
redeployed)
South West Oxfordshire 350 400 23.5 Medium
reservoir
Severn-Trent transfer Capacity 100 26 16.3 Medium
Canal transfer from River Capacity 100 49 20.3 Medium
Severn to River Thames
(initial transfer to canal
system via Severn-Trent
transfer)
Trent-Anglian transfer (to Capacity 200 108 26.6 Medium
Essex reservoirs via
Fossdyke Navigation, R
Witham and existing Ely
Ouse-Essex transfer)
East Anglian reservoir 174 69 0 High
figures only available for
Great Bradley reservoir
Birmingham groundwater 50 4 6 Medium

The strategy described by Sheriff et al. (1996) revolves around supplying more water to
the southeast by developing more storage to use winter river flows and transferring
water from the north of England or Wales. For the Anglian region, the strategy proposes
a transfer from the River Trent to the Anglian region and a new pumped storage
reservoir in East Anglia (see Figure 6.4). The transfer corridor would use the lower
River Witham to Boston, an aqueduct across the Fens into the cut-off Channel at
Denver, and then into the existing Ely Ouse-Essex Scheme. An alternative longer route
using the Greater Ouse system could also supply Rutland and Grafham Waters. At the
time, two principal sites were being considered for a new reservoir in the Anglian

140
region. These were Great Bradley near Newmarket on the headwaters of the River Stour
and a fenland site on the South Level fen between Feltwell and Ely (cf. de Lande Long
and Scott 2000 for a feasibility study of a reservoir at Feltwell). Both sites could be
supplied with winter water pumped from the Ely Ouse and would be used in
conjunction with the existing Ely Ouse-Essex transfer system to augment the Stour and
other Essex rivers.

Figure 6.4 Trent-Anglian transfer and/or East Anglian Reservoir (Sheriff et al.
1996).

The Environment Agencys water resources strategy for the Anglian region considered
a wide range of options (EA 2001b). The Environment Agency used three approaches to
assess these options: 1) a costing exercise, looking at the broad financial costs of each

141
option; 2) a risk and uncertainty framework, looking at the risks, uncertainties and
opportunities of options and strategies; 3) a sustainability appraisal, looking at
sustainability in its widest sense. The risk and uncertainty framework considers:
? uncertainty in the technology, investigation, time, cost and resource value of an
option;
? opportunities to meet wider objectives, including: the ability to be flexible in
implementation, opportunities for environmental enhancement, resilience to
climate change, and providing amenity and recreational benefits;
? constraints that may limit the success of the option, including attitudes and
aspirations and legal or institutional barriers.

Table 6.7 shows the application of this framework to different resource and demand
management options. The option of building new reservoirs or raising old ones provide
some of the highest resource value and are also resilient to climate change. On the other
hand, they take a long time to be implemented. Other schemes such as a desalination
plant takes less time to implement, but has a high cost. While this list of options might
seem long, it is important to remember that particular zones in the Anglian region only
have a handful of options available (as shown ahead).

Table 6.7 Options risk and uncertainty for the Anglian Region. Source:
Environment Agency (2001b).

Resource Time to Estimated Renewal Resilience

Type Option value implement cost period to climate
(Ml/d) (yr) (m/Ml/d) (yr) change
New reservoir
Feltwell reservoir 80-150 15-20 Medium 100 High
Great Bradley reservoir 50 15-20 Medium 100 High
Lower Witham Storage 100-200 15 Medium 100 High
Wash Storage 180 15 Medium 50 Low
Reservoir raising
Resource Development

Abberton Reservoir 50 10 Medium 100 High

Reservoir extension
Ardleigh Reservoir 5-15 10-20 Low 100 High
Winter storage reservoirs (single farm) <1 1-2 Low 15-30 High
Winter storage reservoirs (farm
Reservoir

consortium:10-15) <5 3-5 Low 15-30 High

Greater utilisation of existing reservoir
Eye Brook Reservoir 11 1-2 Low 75 High

142
 Groundwater abstraction or

Groundwater
enhancement
Greater utilisation of groundwater localised 2-3 High 30 Medium
Major groundwater development for
export 50-100 5-10 Medium 30 Medium
Dewatering water locally (inc Thurrock) 5-10 1-3 Medium 20-30 Medium
Desalination
Brackish Suffolk/Thames estuaries 5-10 2-5 High 25 High
Effluent re-use (direct - pws)
Newton-le-Marsh - Covenham 15 3-5 High 30 High
Effluent re-use (indirect - pws)
Ipswich - Alton Reservoir 30 3-5 High 30 High
Peteborough - Rutland 30 3-5 High 30 High
Other schemes

Witham - Essex reservoirs 10 3-5 High 30 High

Norwich - River Wensum/Costessey
intake 30-40 3-5 High 30 Low
Aquifer Storage Recovery
Essex Chalk 10-20 5-10 High 15-30 High
Canal transfer
Grand Union transfer to Anglian (Nene) 50-100 1-3 Medium 50+ Low
River transfer
Increased TWA transfer (summer) 30 1-2 Low 30 Low
Increased TWA transfer (winter) 300 2-3 Low 30 Medium
Transfer across Fens from TWA (rivers) 150 5-10 Medium 30 Low
Pipeline transfer
Trent to Rutland raw water transfer
(winter) 200 5-10 Medium 30 High
Enhanced links Rutland/Grafham (inc
Wing) 50 2-3 High 30 Medium
Bulk transfer TWUL (Chigwell) - Anglian 20 2-3 Low 30 Medium
Bulk transfer TVW to Essex 25 2-3 Low 30 Medium
Bulk transfer TWUL (Abingdon) - Anglian 20-50 15-20 Low 30 High
Increased bulk transfer Severn Trent/AWS 20 2-3 Low 30 Medium
Dewatering water Kent - Essex (rail
Transfers

tunnel) 15-20 10-15 Low 30 Medium

Transfer across Fens from TWA (pipes) 150 5-10 High 30 Medium
Transfer across Fens from TWA (tunnel) 150 10-15 High 50 Medium
Operational improvements 5 1-3 High 30 Medium
Ops. Conjuctive use 25 1-5 Medium 15-30 High
Leakage reduction 1-5 Low-High High
Rainwater use (new development, non-
potable) 1-3 High Low
Greywater use (new development, non-
potable) 1-3 High High
Demand Management

Waste minimisation of
industrial/commercial 1 Low-High High
White goods subsidies 1-3 High High
Retrofit of toilets to dual
flush/interruptible flush Medium High
Metering (domestic) 10 High 10 High
Tariffs for measured charges 10+ High
Household Water Audits 1-2 High High

143
In order to examine in more detail the options available to water managers in the region,
Anglian Water Services (AWS) Water Resource Plan (WRP) was scrutinised in detail
(see AWS 2004b). The WRP is an extensive document of which the following is mainly
a synthesis. Anglian Water has identified as a number of major schemes in order to
maintain service levels (see Table 6.8).

Table 6.8 Major schemes for each Water Resource Zone (AWS 2004b). In shades
of grey are the WRZs that are examined in more detail.

Water Resource Zone Option Name

1 South Humberside - Irby to Elsham Booster
2 Lincoln - Irby Link to Caistor Reservoir
- Waddingham Booster
3 Lincolnshire Coastal - Manby to Louth Booster
2 Lincoln - Saltersford Water Treatment Works
Membrane Plant
4 Lincolnshire Fens - Wing / Etton Booster
5 Fenland - Peterborough to March Link
- Rezoning Sutton to March Planning Zone
5 Fenland - Stoke Ferry Water Treatment Works
Extension and
- Stoke Ferry to Downham Market Link
6 North Norfolk Coast - Metton to Matlaske for Blend & Treatment
7 Norfolk Rural - Carbrooke Recommissioning and Booster
- Bradenham to Dereham Link
- Bradenham to Beetley Link
7 Norfolk Rural - Old Buckenham Recommissioning
8 Norwich & The Broads - Yare Valley Groundwater Development
8 Norwich & The Broads - Witton Park Booster
9 Cambridgeshire & West - Ixworth [Stanton] additional borehole
Suffolk
9 Cambridgeshire & West - Long Melford Booster
Suffolk
10 East Suffolk & Essex - Alton Water Treatment Works extensions
- Cliff Quay Effluent Re-use
- Alton to Horkesley Link
- Coggeshall to Parkfield Tower
10 East Suffolk & Essex - Ardleigh Reservoir Increased Take and
Lexden Blending
11 Ruthamford - Wing Water Treatment Works Extensions
- Wing Strategic Mains
11 Ruthamford - Clapham Water Treatment Work Extensions

144
The WRP also identified the need to plan for the development of new water resources
towards the middle and end of the 25 year planning period. This is due to the potential
for loss of deployable output due to climate change, deterioration in raw water quality
due to point source pollution, and an increase in demand due to population and
potential economic growth. The Final Business Plan therefore includes allowance for
investigations for a raw water storage reservoir in the lower river Witham, supported
by transfers from the river Trent, and investigation of the potential for water resources
to be transferred into and within the Anglian Region through development and potential
linking of the existing Trent-Witham and Ely-Ouse Essex transfer schemes (AWS
2004b, p. 8).

Like other water companies, Anglian Water included climate change in their WRP
following Environment Agency guidelines (see section 6.3). Their modelling, using
UKCIP02 results, suggests that by 2030 there could be a significant impact on surface
water sources with a loss of 40 Ml/d. The impact on groundwater abstractions is thought
to be negligible. As mentioned before, these results have not led to investment
decisions, but this is currently being explored.

It became clear that it was not possible to do an analysis, such as one suggested by this
thesis, for the entire company as it was necessary to work in the same unit as the Water
Resource Plan: the Water Resource Zone (cf. section 6.2).52 In order to make it climate
change relevant, two WRZs where climate change was a major driver of headroom were
chosen: East Suffolk & Essex and Ruthamford. These two WRZs are described in great
detail next, as contextual information is very important for adaptation.

6.5.1 East Suffolk & Essex Water Resource Zone53

The East Suffolk & Essex (ES&E) WRZ is predominantly supplied by groundwater
with some contribution from surface water. Surface water is developed through raw
water storage reservoirs at Alton and at Ardleigh. The latter is jointly operated with
Tendring Hundred Water Services through the Ardleigh Reservoir Committee. The

52
This highlights the difficulty of translating best-available science into practical methodologies.
53
This section and the next (6.5.2) are heavily based on verbatim passages from AWS WRP (2004b, p.
67-80)

145
reservoirs have individual yield characteristics. Alton Water reservoir is filled from the
Gipping catchment that is characterised by low baseflow from groundwater and limited
input from the return of treated effluents upstream. Ardleigh reservoir has a very short
retention period and relies upon the return of treated effluent from towns along the
Colne valley and baseflow from the glacial sand and gravel aquifer that overlies the
Chalk and Tertiary clays, but is itself confined by boulder clay. Both Alton and
Ardleigh have augmentation schemes to secure their yield during low flow conditions.

Demands are centred on the large towns of Ipswich, Colchester and Braintree, all of
which are served by major road and rail links to London. They have attracted fringe
growth both from commuters and from company relocation. Total domestic demand for
the zone for the base year is assessed at 75.6 Ml/d and the commercial demand is
assessed at 25.6 Ml/d.

The zone is adjacent to the M11 corridor with development included in the strategic
plans published by the ODPM (Office of the Deputy Prime Minister) earlier in 2003. It
is anticipated that growth in the M11 corridor and around Stansted airport would spill
over into the AWS supply area using the A120 and A12 road links. Detailed figures are
not available about location or timing of development and so no allowance has been
made for growth for the proposed housing developments at this stage. Domestic growth
is forecast to take place at the rate of 1350 properties per annum. New property
occupancy rate is expected to be just under 1.99 persons per household. In spite of this
Anglian Water expects the total domestic demand to decline to 70.5 Ml/d by the year
2029/30. This is due in part to the gradual increase in metering penetration from 53% in
2002/03 to 93% in 2029/30. Other factors influencing domestic demand are the
expected overall fall in household occupancy size and the net effect of water efficiency
and demand management initiatives on the this demand. Commercial demand is
expected to remain steady over the planning period at 25 Ml/d for the year 2029/30.

In the ES&E WRZ target headroom is affected by the risk of climate change on Alton
Water reservoir. There are also fairly large risks associated with the anticipated
deterioration of borehole conditions and loss of supply due to groundwater pollution.
To this is added the uncertainty over forecast demand and leakage. These risks and

146
uncertainties combine to give the ES&E WRZ the highest target headroom in the
region.

The cornerstone for the investment plans is the development of the trunk main system
by linking the Ipswich and Colchester storage reservoirs of Wherstead and Great
Horkesley. This would facilitate the development of surface resources based on Alton
Water to meet the water future needs of economic growth and climate change. This
trunk main system is identified for construction late in the AMP4 period in anticipation
of the upgrading of Alton Water Treatment Works (WTW) during AMP5.

In the short term the AWS has identified the need for increased supplies to the
Colchester Planning Zone. This can be achieved by upgrading Lexden sourceworks to
utilise the current licence by improved blending for fluoride and by a temporary re-
allocation of the proportion of the deployable output of Ardleigh WTW. This together
with the construction of a new treated water storage reservoir at Ardleigh WTW forms
the optimal solution for the Colchester Planning Zone. The Ardleigh Reservoir
Committee has discussed the potential for an amendment to the Ardleigh Agreement
with no objection in principle. AWS has submitted a proposal to Tendring Hundred
Water Services for reallocation of the output from Ardleigh WTW in line with the
provisions of the Ardleigh Reservoir Order; discussions are ongoing. The scheme for
an additional 5 Ml/d at average and 7.5 Ml/d at peak for the AMP4 period only is
included in the WRP.

The development of Alton WTW will require the augmentation of the reservoir yield to
cater for increased output and the potential decline in yield identified through the
analysis of climate change impact. The main options for augmentation are effluent re-
use, transfer through the Ely Ouse-Essex Transfer Scheme or augmentation from the
confined chalk aquifer. Analysis shows that augmentation is required during the AMP4
period to avoid a deficit of target headroom in a dry year. However the risk will be
mitigated by the connection of the Ipswich and Colchester systems. AWS has included
increasing the yield of the Bucklesham raw water transfer through the development and
commissioning of chalk augmentation boreholes in its Drought Plan. This could either
be at Bucklesham itself or by use of the existing Waddling Duck borehole to augment
the flow of the Mill River for re-abstraction downstream. The Business Plan does not

147
include a scheme for upgrading the Bucklesham augmentation scheme, although this
would be considered as part of a Drought Action Plan. The confined chalk aquifer
under the Felixstowe peninsula has potential for future development using Aquifer
Storage Recovery techniques, although this has not been included in plans at present.

In the medium term (AMP5) the proposed increase in the Alton WTW peak output from
42 Ml/d to 60 Ml/d, requires additional augmentation during drier than average years.
Additional water resources are available from Cliff Quay Waste WTW, which serves
most of the Ipswich area, but there may be options for augmentation through
enhancements to the Ely Ouse-Essex Transfer Scheme. AWS has discussed Essex and
Suffolk Water Company proposals for water resources development through the
Abberton Trilogy to secure additional abstraction at Denver, increase the transfer
capacity down the River Stour and raise the dam at Abberton reservoir. In the years
immediately following the completion of the proposed Abberton Trilogy there would be
surplus capacity in the Ely Ouse-Essex Transfer Scheme that could be used to augment
the yield of Alton Water from a new intake at Cattawade at the tidal limit of the River
Stour. A temporary increase for AWS would need support in the longer term towards
the end of the 25 year planning period by investment in the transfer of water from the
Trent described above, i.e., the development of single season critical raw water storage
reservoirs in the Lower Witham and adjacent to the Cut-off-Channel with a new transfer
link. Clearly this strategic inter catchment transfer has potential to provide additional
water resources to other water users including agriculture, navigation and other water
companies.

Figure 6.5 tries to capture the essence of the WRP for this WRZ as summarised above.
It is noticeable that demand management will reduce water demand by very little (0.45
Ml/d) in the next 25 years. Leakage reductions are more substantial in the order of 3
Ml/d. Changes in Water Available For Use (WAFU), essentially resource development,
is AWSs main strategy to cope with increasing target headroom. The first increase in
WAFU in 2006-07 is associated with a temporary increased take from Ardleigh
Reservoir and by upgrading Lexden sourceworks (8 Ml/d). The second increase in
2012-13 (AMP5) is connected to the extension of Alton WTW. The subsequent
decrease in 2014-15 is linked to the temporary take from Ardleigh. This plan for the

148
ES&E WRZ leaves the supply/demand balance always in the positive, but only barely at
the start of AMP4 in 2005-06. By 2030 there is a surplus of 11 Ml/d.

30
Upgrading Lexden
25 sourceworks +
temporary take Change in demand
20 from Ardleigh
Reservoir (8 Ml/d)
Change in leakage
15

Change in WAFU
Ml/d 10
Alton WTW 11 Ml/d
5 upgrading Target headroom
(20 Ml/d)
0 Planned supply/demand
balance
20 17

20 27
20 03

20 07

20 13
20 05

20 09

20 15

20 19

20 23

20 25

9
20 1

20 21
-5

-2
-1

-
-

-
-

-
-
16

26
02

06

12
04

08

14

18

22

24

28
10

20
20

-10
Year

Figure 6.5 Changes in demand, leakage, WAFU (Water Available For Use), target
headroom and supply/demand balance from 2002-03 to 2029-30 for the East
Suffolk and Essex WRZ. Also shown are the adaptation strategies included in
AWS WRP and the supply/demand balance available in 2030 (in red).

Table 6.9 lists the adaptation options collated for the ES&E WRZ. The shaded options
are included in the WRP and in Figure 6.5. The other options are based on plausible
interpretations of the WRP and the options reviewed earlier in the section. Appendix 1
provides a full list of options for this WRZ from AWSs WRP.

149
Table 6.9 Adaptation options for the East Suffolk and Essex Water Resource Zone.
In grey are the decisions already included in Anglian Waters Water Resources
Plan.

Adaptation options Planned Planned Total net

gains in implementation present cost
WAFU or date (M)
savings in
demand by
2029-30
(Ml/d)
Ardleigh Reservoir increased take 0a 2006-07 4.71-1.98d
Upgrading Lexden sourceworks 3 2006-07 3.32-3.15d
Alton WTW extension and effluent re-use 20 2012-13 42.85-33.87d
resource scheme at Cliff Quay
Aquifer Storage Recovery in the confined 10-20 5-10 years to High
chalk aquifer under the Felixstowe implement
peninsulac
Desalination plant (Brackish Suffolk) 5-10 2-5 years to High
implement
Trent-Anglian transfer (to Essex Capacity Medium time to 108
b
reservoirs via Fossdyke Navigation, R 200 promote
Witham and existing Ely Ouse-Essex
transfer)
East Anglian reservoir figures only 174b Long time to 69
available for Great Bradley reservoir promote
a: Temporary increased take of 5 Ml/d between 2006-07 and 2013-14
b: Only a portion of this water would reach the ES&S WRZ
c: based on Environment Agency figure for Essex Chalk
d: based on figures from the WRP and supply-demand worksheet

6.5.2 Ruthamford Water Resource Zone

According to AWSs Water Resource Plan the Ruthamford WRZ has the highest
increase in target headroom due to climate change compared to all other WRZs. In other
words, Anglian Water is expecting the Ruthamford WRZ to be the worst hit by climate
change in terms of water supply. The Ruthamford WRZ is named after the integrated
water resources and supply system formed by the use of RUTland Water, GrafHAM
Water and PitsFORD reservoirs. The zone also includes the smaller surface water
sourceworks at Ravensthorpe and Clapham and the groundwater sources abstracting
from the Woburn Sands aquifer. The supply system in the zone is characterised by long
strategic trunk mains connecting large treated water storage reservoirs. The Ruthamford

150
system is a net exporter of water with bulk supplies from Grafham WTW to Three
Valleys Water Services (91 Ml/d) and Wing WTW to Severn Trent Water (18 Ml/d).54
The environmental concerns in the zone arise from the management of surface water
resources in the large Special Protected Area wetlands of The Wash, Nene Washes,
Ouse Washes and Rutland Water.

The Ruthamford WRZ is by far the largest and includes the major towns in the west of
the region of Peterborough, Huntingdon, Corby, Kettering, Bedford, Wellingborough,
Northampton, Milton Keynes and Daventry.55 The zone has shown steady growth in
parallel with development of Grafham Water in the 1960s and Rutland Water in the
1970s (cf. section 6.1). Demand management has been progressed through the control
of leakage and pressure management in the distribution system, household metering and
the more efficient use of water by customers. This, together with the extension of
Grafham WTW in the mid 1990s, has maintained secure supplies to date, due in part to
the bulk supplies taken being below the reservation quantities. However the continued
growth of the East Midlands region will require further development of water resources.
Both in 2001, and again in 2003, local problems have been experienced in maintaining a
secure level in treated water storage reservoirs in the south-west corner of the region, at
the extremes of the Ruthamford distribution system. Although the disruption to water
supplies was short-lived these incidents confirm the need to invest in both trunk mains
and water treatment works to meet the economic development of the region. Potential
growth comes from both housing and light industrial development in the existing towns
of Milton Keynes, Northampton and Peterborough, which are on the major north-south
transport links and within commuting distance of London. Total domestic demand for
the zone for the base year is assessed at 229.4 Ml/d and the commercial demand is
assessed at 106.1 Ml/d.

AWS expects total domestic growth to increase from 229.4 to 243.9 Ml/d by the year
2029/30. This increase is moderated in part by the gradual increase in metering

54
It has been agreed with both companies that these arrangements will remain as at present for the AMP4
period.
55
The Ruthamford system also indirectly supplies parts of the towns of Luton and Stevenage in the Three
Valleys Water Services supply area and the market towns of Oakham and Market Harborough in the
Severn Trent Water supply area through bulk supplies.

151
penetration from 55% in 2002/03 to 94% in 2029/30. Commercial demand is expected
to fall slightly over the planning period from 106.1 to 104.4 Ml/d.

Target headroom is driven by demand uncertainty, including an initial increase due to

uncertainty over leakage. In addition there is a large risk on the supply side from the
impacts of climate change on the yield of surface water resources. The WRZ has a
current deficit of available against target headroom at peak and at average from the start
of the AMP4 period.

AWSs WRP provides a long list of options available to maintain target headroom.
These include generic demand management options for additional customer metering
and targeted domestic water audits and cistern displacement devices as well as the use
of additional leakage control to reduce leakage to below the current economic level of
leakage. The cornerstone for maintaining a secure supply-demand balance for the
Ruthamford WRZ, which supplies a third of AWS customers and those of adjacent
water companies, is the extension of Wing WTW with a 90 Ml/d peak treatment works.
The construction of the works is dependent upon planning permission being granted by
Rutland County Council, which in turn needs the support of English Nature and other
conservation bodies to the proposed mitigation package to protect the reservoir as a
Special Protected Area for birds. AWSs FORWARD model shows that the Wing
extensions are needed at the start of the AMP4 period in order to maintain target
headroom. However the construction of major engineering works, including
reinforcement of raw water and treated water mains, means that the additional
deployable output will not be available until later in the five year period. Although the
deployable output could be increased earlier by recommissioning of Foxcote WTW and
the upgrading of Clapham WTW, each of which would provide an additional 10 Ml/d,
these are not included in AMP4 as they result in a more expensive plan in the short-term
due to the need for the larger Wing WTW extensions now. Both remain options in the
longer-term.

The FORWARD model also shows that a further increase in supplies is needed during
the 25 year planning period. This could be met by the strategic development of the
River Trent through a new single season winter storage reservoir in the Lincolnshire
Fens, south of the lower River Witham. A multifunctional reservoir could supply a new

152
surface water treatment works for Peterborough that would replace the existing supply
from Wing WTW. The reservoir could be a part of a strategic transfer link to Stoke
Ferry WTW in Norfolk and could provide water for the proposed Fens Waterways and
agricultural abstractors. AWSs supply-demand programme includes provision for
detailed investigation during the AMP4 period into engineering and environmental
issues.

Figure 6.6 tries to capture the essence of water management planning for the
Ruthamford WRZ. Demand management is expected to reduce water demand by 1.5
Ml/d by 2030. Leakage is reduced to a maximum of almost 6 Ml/d, but by 2030 it is
actually expected to increase by 1.5 Ml/d. Changes in WAFU is AWSs main strategy
to cope with increasing target headroom. The first increase in WAFU occurs in 2009-10
and is associated with the extension of Wing water treatment works (50 Ml/d). The
subsequent increase in 2011-12 is linked to the extension of Clapham WTW. The
further increases that take place from 2020 onwards relate to a potential new reservoir in
the Lower Witham, which would be developed in conjunction with a strategic
development of the River Trent (67-81 Ml/d). Most of the AMP4 period is under
negative supply/demand balance due to a rising target headroom. By 2030 there is a 77
Ml/d surplus. Appendix 1 provides a full list of options for this WRZ from AWSs
WRP.

160
Winter storage
Upgrading reservoir Change in
140 Clapham in the demand
Wing Water Water lower
120 Treatment Treatment Witham Change in
Works Works (7 (81 Ml/d)
leakage
100 extension Ml/d)
(51Ml/d)
80 Change in
Ml/d WAFU
60
Target
40 headroom
77 Ml/d
20 Planned
supply/demand
0 balance
20 3

20 5

20 7

20 9

20 1

20 3

20 5

20 7

20 1

20 3

20 5

20 7
20 9

-29

-20
-0

-0

-0

-0

-1

-1

-1

-1

-2

-2

-2

-2
-1
02

04

06

08

10

12

14

16

20

22

24

26
18

28
20

Year

Figure 6.6 Changes in demand, leakage, WAFU (Water Available For Use), target
headroom and supply-demand balance from 2002-03 to 2029-30 for the

153
Ruthamford WRZ. Also shown are the adaptation strategies included in AWS
WRP and the supply/demand balance available in 2030 (in red).

6.6 Linking climate change to water resources decisions

Now that a list of adaptation options has been compiled, the final requirement is to link
these options with changes in the climate. This is perhaps one of the most difficult tasks
of the assessment because water companies adapt to a multitude of pressures
simultaneously, not just climate change. This is clearly evident in the approach taken in
water resources planning and the concept of headroom (see section 6.2), which includes
a number of risks and uncertainties, climate change being one of them. Due to the
difficulty of this task a number of approaches were pursued to build up the evidence
base.

The first approach was based on expert judgement and was conducted by sending a
questionnaire to AWSs water managers (see Appendix 2 for full questionnaire and
answers). To the question of how much extra water (in terms of WAFU) is needed due
to climate change, the water manager answered:

As I pointed out previously it is very difficult to unravel the degree to which

change in the climate variable (principally rainfall) contributes to the strategy because
it went into the 'melting pot' (HOURUS, the Monte Carlo simulation method - see WR
Plan Vol 3) with a number of other variables to determine target headroom. It is
possible to get a rough idea of the climate change component by comparing the
Headroom in Yr28 for Suffolk & Essex (25%) with WRZs where climate change is not
included (7 zones average ~12% [exceptions being Linc Fens, Ruthamford and S
Humberside]) i.e. ~ 50% headroom due to climate change. Given that WAFU would
rise by 23 Ml/d in the ES&E WRZ, 0.5 x 23 =11.5 Ml/d through strategy due to climate
change.

He also noted that the above related to a 10% decrease in summer precipitation based
upon Arnell medium scenario, relative to 1961-90 average (UKCIP02) since the
ES&E WRZ had a significant surface water component, which was much more affected

154
by climate change than groundwater. Thus according to this expert, in the ES&E WRZ
a -10% precipitation change creates an additional need of 11.5 Ml/d. The expert gave
medium confidence to this link.

The second approach was to examine the contribution of climate change to target
headroom (the cyan lines in Figures 6.5 and 6.6). The data were supplied by Atkins
Water,56 who conducted the headroom calculations for Anglian Water. Since headroom
is calculated using Monte Carlo sampling it is relatively straightforward to calculate the
sensitivity of the output to the input drivers. Table 6.10 shows results for the ES&E
WRZ. Climate change and demand are the major uncertainties that impact target
headroom at the end of the planning period. This amounts to climate change producing
a loss of resources of around 9.2 Ml/d by 2030.

Table 6.10 Sensitivity analysis of the key drivers that contribute to target
headroom uncertainty for the ES&E WRZ. Total headroom shown as a percentage
of demand. In grey is the calculated climate change contribution to target
headroom in Ml/d. Source: Atkins Water.

Year 1 5 10 15 20 25 30
Total Headroom 4.0% 10.0% 15.1% 17.7% 21.0% 23.8% 25.7%
%age from base year 100.0% 56.0% 33.8% 33.2% 33.3% 33.9% 36.0%
and demand
uncertainties
%age from climate 0.0% 17.1% 20.7% 27.8% 32.9% 36.5% 36.1%
change
%age from point 0.0% 0.0% 16.7% 14.2% 12.2% 10.5% 9.8%
source pollution risk
%age from borehole 0.0% 26.9% 28.8% 24.9% 21.7% 19.1% 18.1%
(yield) deterioration
Climate change 0.00 1.79 3.19 4.98 6.94 8.70 9.19
contribution to
headroom (Ml/d)

Further interaction with Atkins Water allowed the identification of the precise PDFs
used in the Monte Carlo simulation to describe losses in water resources due to climate
change for both WRZs considered here. The PDFs used are shown in Figure 6.7. The

56
This has been supplied by Geoff Darch at Atkins Water on 2 June 2005.

155
results for the ES&E WRZ are consistent with Table 6.10 and the expert assessment
given by the water manager above. It is the authors understanding that these PDFs were
constructed based on modelling that AWS had performed (on the impact of climate
change on surface water and groundwater AWS 2003) and human judgement.57 This
proves Schneiders (2001, 2002) point that in the absence of experts providing estimates
of likelihood for climate change, users will create their own frequency distributions.
These triangular distributions are essentially how AWS included climate change in their
WRP plan. They are used in Chapter 7 as a benchmark to assess the robustness of the
WRP to wider climate change uncertainties as sampled in this thesis.

P(x) P(x)

Min Max Min Max

Most Likely Most Likely
0 9.6 17.2 0 36.2 52.7
Reduction (Ml/d) Reduction (Ml/d)

Figure 6.7 Probability density function used to describe losses in water resources
by 2030 for the East Suffolk and Essey WRZ (left) and the Ruthamford WRZ
(right).

The third and final approach examined the work AWS conducted on the impact of
climate change on surface water sources (AWS 2003). AWS perturbed past climate data
(rainfall and evaporation) with UKCIP02 climate change scenarios to simulate river
flows and determine reservoir yield. All of AWS reservoirs were run with the medium
climate change scenario (from Arnell 2003a, which is equivalent to UKCIP02 Medium-
Low and Medium-High) and three reservoirs (Grafham, Alton and Covenham) were run
with the cool and wet and warm and dry scenarios as well. Estimates for the other
reservoirs in the ES&E WRZ (Ardleigh) and the Ruthamford WRZ (Pitsford and
Rav/Hollowell)58 were interpolated for the cool and wet and warm and dry scenarios
based on the modelled results for Grafham, Alton and Covenham. The Ruthamford
reservoirs were assumed to behave like Grafham (because it is in the same WRZ; thus
the same changes were applied to these reservoirs), whereas Ardleigh was applied with
the average changes from Grafham, Alton and Covenham. Because summer

57
Personal communication with Geoff Darch, at Atkins Water.
58
Rutland was not included because deployable output is not affected by reductions in yield.

156
precipitation was the main climatic driver identified by the AWS water manager (see
above) losses in water resources were linked to this variable. Figure 6.8 shows the
estimates of deployable output loss for the UKWIR scenarios. The numbers are
consistent with the PDFs devised by Atkins (see Figure 6.7), showing that Atkins
assumed the medium scenario would provide the most likely value for the PDF, while
the other two estimates (warm and dry and cool and wet) would provide roughly for the
minimum and maximum values used. In this thesis, the linear fit (the black line in
Figure 6.8) is used as representative of climate change impacts on water resources in
this particular WRZ. Since there is a considerable uncertainty in arriving at this transfer
function (due to various assumptions, limited sensitivity analysis, etc.) it is assumed to
be uncertain when applied in Chapter 7.

18
Warm and dry 16
15.29
14
12

9.61 10
Loss in Medium 8
deployable
output( Ml/d) 6
4
2
Cool and wet
0 -0.47
-12 -10 -8 -6 -4 -2 -2 0 2

-4
Summer precipitation change (%)

Figure 6.8 Estimates of loss in deployable output in the ES&E WRZ for three
UKWIR scenarios (Medium, warm and dry, cool and wet) as a function of summer
precipitation change. The black line represents a linear fit to the results.

Figure 6.9 shows results for the Ruthamford WRZ. Estimates are larger than in the
ES&E WRZ simply because Ruthamford is a larger WRZ with more reservoirs and
more water available for use. These results are also consistent with the PDF estimated
by Atkins (cf. Figure 6.7 right panel). Table 7.11 compares regression coefficients, their
standard errors and explained variance for both WRZs.

157
 Warm and dry 60
54.44
50

40
33.97
30
Medium
Loss in
deployable 20
output( Ml/d)
10

0
-12 -10 -8 -6 -4 -2 0 2
-8.05
-10
Cool and wet

-20
Summer precipitation change (%)

Figure 6.9 Estimates of loss in deployable output in the Ruthamford WRZ for
three UKWIR scenarios (Medium, warm and dry, cool and wet) as a function of
summer precipitation change. The black line represents a linear fit to the results.

Table 6.11 Table of regression model coefficients, their standard errors and
explained variance about the relationship between summer precipitation change
and loss in deployable output.
Water Coefficients Standard Explained
Resource Zone error variance (R2)
ES&E -1.19 0.15 0.92
Ruthamford -4.26 0.57 0.93

6.7 Summary

This chapter has introduced the case study to which the framework developed in this
thesis will be applied in Chapter 7. Section 6.1 showed how the water supply sector has
undergone considerable restructuring over the last decades, going from the private
sector to the public sector, and then back to a highly regulated private sector (cf. Bakker
2003). Water resources planning was shown to be a complex undertaking. Numerous
uncertainties are included in the assessment, including climate change, in order to
determine the supply-demand balance. However, it is still difficult (perhaps impossible)

158
to determine the true level of risk a water company takes when it chooses a certain
target headroom for a particular level of service (cf. EA 2004a). With guidance from the
Environment Agency, the inclusion of climate change has been relatively
straightforward, but even so, the range of impacts has been vast. A preliminary
assessment on whether adaptation to climate change is currently taking place in the
water supply sector was conducted in section 6.4. While many efforts are increasing
adaptive capacity (e.g., research on the impacts of climate change, mainstreaming
climate change into planning, etc.) adaptation decisions per se are not being
implemented yet. This occurs for a number of reasons, including: uncertainty (in
particular in climate change projections), complexity (from the technical side as well as
the institutional side), timescale (long-term versus short-term planning) and financing
(too uncertain for investment). This issue is discussed further in the context of results
from Chapter 7 in section 8.1.3.

Section 6.5 elaborated on the various adaptation options available to decision-makers in

the water supply sector. It focused particularly on two Water Resource Zones of AWS:
East Suffolk & Essex and Ruthamford. Much detail was supplied on these two WRZs
because of the importance of context in adaptation.59 Finally, section 6.6 linked these
adaptation options with changes in the climate (in particular summer precipitation
change) using a number of approaches. The most convincing approach is the last
approach, which uses AWS hydrological modelling as a basis. This is the approach
used to link adaptation options with changes in the climate in the next chapter.

59
At its simplest and most obvious, one would not expect to build a desalination plant in the Ruthamford
WRZ, for example.

159
 CHAPTER 7. SENSITIVITY, UNCERTAINTY AND
ROBUSTNESS ANALYSIS

This chapter brings together the science of climate change (climate projections and their
uncertainties; Chapters 4 and 5) and decision-making (water supply planning; Chapter
6). It uses sensitivity analysis, uncertainty analysis and robustness analysis to examine
the issue of adaptation to climate change in water resources planning in the Anglian
region. Sensitivity analysis (section 7.1) is used to investigate how sensitive adaptation
decisions are to uncertainties in the various elements of the assessment and whether
Anglian Waters WRP can cope with these uncertainties. Uncertainty analysis (section
7.2) is used to quantify the uncertainty in the most important elements of the framework
(informed by the sensitivity analysis) and propagate it numerically throughout the
assessment. This allows the determination of the risk that climate change poses to
AWS WRP based on best available information (this could also be called the prediction
approach). Finally, robustness analysis (section 7.3) is used to explore an alternative to
the prediction paradigm. Here robust adaptation strategies are sought amongst a wide
spectrum of climate change uncertainties without judging how likely they are. Demand
uncertainties are also examined using the Environment Agency water demand scenarios.

7.1 Sensitivity analysis

This section performs sensitivity analysis on the various elements of the modelling
framework in order to determine whether or not a decision to adapt to climate change is
sensitive to uncertainty in those elements. Elements that can easily change a decision
(high sensitivity) should be included in an uncertainty analysis (risk assessment) while
elements that dont change a decision (low sensitivity) need not.

The modelling framework used in this thesis is a sequential one. Therefore, in order to
examine the sensitivity of adaptation decisions to the various elements of this sequence,
each element is examined in turn to keep uncertainties manageable.60 This allows the
examination of how much uncertainty each parameter contributes while the other
parameters are kept constant. This methodology is consistent with Visser et al. (2000)
60
There are limits to how many uncertainties humans can keep in the mind at one time.

160
and Jones and Page (2001) and is known in the sensitivity analysis literature (see, e.g.,
Campolongo et al. 2000) as one-at-a-time experiments, where the impact of changing
the values of each factor is evaluated in turn. The experiment that uses the standard
values is here called default scheme61 and is shown in Table 7.1. This is summarised for
the purpose of presentation and to facilitate the understanding of the sensitivity of
adaptation decisions to particular uncertainties. Latin Hypercube Sampling was applied
in these experiments.

Table 7.1 Scenarios, PDFs or transfer functions used in the default scheme for the
sensitivity analysis.

Uncertain parameter Default combination

Natural variability Not included
GHG emissions SRES A2-ASF
Climate sensitivity Log-normal PDF (Wigley and Raper 2001)
Carbon cycle Log-normal PDF (Wigley and Raper 2001)
Ocean diffusivity Uniform PDF (Wigley and Raper 2001)
Aerosol forcing Uniform PDF (Wigley and Raper 2001)
Regional climate response Uniform distribution (for GCMs) and
equally likely frequency distribution (for
RCMs)
Climate impacts Transfer function estimated from AWS
hydrological modelling

In the various sections that follow much attention is given to the 95th percentile. This
happens because Anglian Water wants to maintain the level of service at a 95%
confidence level. The level of service is defined by AWS as the:
? Restriction on the use of hosepipes not more than 1 in 10 years
? Use of Drought Orders to enforce restriction on non-essential uses and secure
raw water resources not more than 1 in 40 years
? Imposition of the use of standpipes not more than 1 in 100 years

Results are presented as uncertainty ranges between the maximum and minimum values
that a certain parameter yields at the 95th percentile. The shortfall between AWSs
triangular distribution, assumed to represent climate change impacts (see Figure 6.7),
and the maximum value for the parameter being sampled here is also reported at the 95th
percentile. The WRP is robust if the shortfall is smaller than the supply/demand balance

61
Other authors call this the control experiment or control scenario (Campolongo et al. 2000).

161
in 2030 and weak if the shortfall is larger than the supply/demand balance in 2030.
Therefore, it is important to keep in mind the estimated supply/demand balance values
from the current WRP for the ES&E WRZ (11 Ml/d) and the Ruthamford WRZ (77
Ml/d) in 2030 (see Chapter 6, in particular Figures 6.5 and 6.6).

The list of uncertain parameters examined in the sensitivity analysis includes:

? natural mutidecadal climate variability
? greenhouse gas emissions
o illustrative SRES scenarios
o scenarios within a storyline
o mitigation/stabilisation scenarios
? Climate sensitivity
? Other climate parameters
o Carbon cycle
o Aerosol forcing
o Ocean mixing
? Regional climate response
o Global climate models
? Skill
o Regional climate models (dynamical downscaling)
? Forcing scenario
? Driving Atmospheric GCM
? Skill
? Climate impacts
? Sampling strategy

7.1.1 Natural multi-decadal climate variability

It is important to look at the impact of natural muti-decadal climate variability on water

resources in the Anglian region before examining the additional risks introduced by
anthropogenic climate change. This is crucial because the main driver of water resource
supply precipitation varies considerably at the decadal scale (and other scales),
making it a major source of uncertainty for planning. Therefore, the natural climate

162
variability signal for precipitation is expected to be larger than the anthropogenic signal
in the near term (see, e.g., Hulme et al. 1999, Arnell 2003b).

0.9
0.8
0.7
0.6
AWS
Probability 0.5
Natural variability
0.4

0.3
0.2

0.1
0
-20 -15 -10 -5 0 5 10 15 20 25
Additional water required (Ml/d)

Figure 7.1 Additional water required (Ml/d) by 2030 in the East Suffolk & Essex
WRZ due to natural tri-decadal climate variability (as represented by EWP 1795-
195062) and Anglian Water Services (AWS) strategy to cope with climate change
for the next 25 years.

Figure 7.1 shows the additional water required by the water company solely due to
natural multi-decadal climate variability and AWSs strategy for the next 25 years in the
ES&E WRZ. Natural multi-decadal climate variability introduces large uncertainties to
water resources planning as demonstrated by the black line. If AWS wanted to be 95%
certain to be able to maintain the level of service and cope with drought solely due to
natural variability then an extra 7.28 Ml/d would have to be included in any planning
effort.63 It is promising to observe that the AWS WRP for this WRZ can cope with the
entire range of natural variability if drought is the main concern. It is curious, however,
that no allowances for decreases in additional water required (due to increased summer
precipitation) have been taken into account by AWS. This might occur because in water
resources planning, the main concern is with drought, which leads water managers to
take a conservative approach and ignore part of the uncertainty (i.e., water surplus)
introduced by natural climate variability. Concerns about climate change shifting the

62
Because catchment rainfall was not available for this WRZ data from the England and Wales
Precipitation series was used. See Figure 5.12 (from chapter 5)
63
This value is obtained by reading 0.95 probability in the black line of Figure 7.1.

163
distribution towards the drought side might also influence this conservative approach.
While this is a sensible approach if one is preparing for drought it does overemphasise
the probability of additional water being required. That is, it biases the PDF to one side
of the distribution, when in fact it should be much wider (as shown later).

The other uncertainty worth examining at this stage relates to climate impacts. Figure
7.2 shows the effect of climate impacts uncertainty ( 50% of the relationship between
climate and reservoir yield64) on natural climate variability in the ES&E WRZ. AWS
WRP remains robust to the combination of hydrological modelling uncertainty and
natural variability in the ES&E WRZ if drought is the main concern. Only less than
0.5% of the natural variability (+50%) CDF remains outside AWS strategy. The same
does not occur for the Ruthamford WRZ (Figure 7.3), where the natural variability
(+50%) CDF has nearly 5% of its distribution beyond the AWS plan.

0.9
AWS
0.8 Natural variability
0.7 Natural variability (-50%)

0.6 Natural variability (+50%)

Probability 0.5

0.4

0.3
0.2

0.1

0
-25 -20 -15 -10 -5 0 5 10 15 20 25
Additional water required (Ml/d)

Figure 7.2 Additional water required (Ml/d) in the East Suffolk & Essex WRZ due
to natural tri-decadal climate variability (as represented by EWP 1795-1950) and
Anglian Water Services strategy for the next 25 years. The black natural
variability curve is the same as in Figure 7.1 and represents AWS best estimate for
the relationship between climate and reservoir yield. This relationship is varied by
-50% (red) and +50% (blue).

64
This is applied to the gradient of Figures 6.8 and 6.9 (i.e, to the regression coefficients of Table 6.11).

164
 1
AWS
0.9
Natural variability
0.8
Natural variability (-50%)
0.7 Natural variability (+50%)
0.6

Probability 0.5

0.4

0.3

0.2

0.1

0
-100 -75 -50 -25 0 25 50 75 100
Additional water required (Ml/d)

Figure 7.3 Same as Figure 7.2 but for the Ruthamford WRZ. Natural variability
was estimated using data from the Ely Ouse catchment (1865-1950).

The UK is fortunate to have long historic time series of climate data. This is a rarity
rather than the norm in other locations around the world. Thus, in order to examine
whether this methodology would be reproducible in data poor locations, natural
variability estimates from GCMs (CGCM1 and HadCM2) are compared to observed
data (Figure 7.4).

1
AWS
0.9
Observed
0.8
CGCM1
0.7 HadCM2
0.6

Probability 0.5

0.4

0.3
0.2

0.1

0
-20 -15 -10 -5 0 5 10 15 20
Additional w ater required (Ml/d)

Figure 7.4 Additional water required (Ml/d) in the East Suffolk & Essex WRZ due
to natural tri-decadal climate variability as represented by observed data (EWP
1795-1950) and modelled data (CGCM1 and HadCM2). Also shown is the AWS
strategy for the next 25 years.

165
It is promising that the observed and modelled estimates match closely. At the 95th
percentile, the gap between the observed and modelled is only 1.6 Ml/d. This bodes well
for the application of this methodology in locations with sparse historic data, but it is
certainly not a guarantee as it depends on the performance of climate models in different
regions.

7.1.2 Greenhouse gas emissions

Future GHG emissions are often the first input into any climate change assessment and
are an important source of uncertainty. This section adopts the default scheme (Table
7.1) for the various parameters of the framework and explores uncertainty within the
GHG emissions parameter. Figure 7.5 compares the additional water required due to
natural climate variability (cf. Figure 7.1), climate change due to different GHG
scenarios and the AWS strategy. The climate change CDFs show the impacts of the six
illustrative SRES scenarios (A1B-AIM, A1FI-MI, A1T-MES, A2-ASF, B1-IMA, B2-
MES) on water resources planning when GCM results are used. The range of outcomes
is rather large, ranging from having slightly more water available to having substantially
less water available. This occurs because a minority of GCMs show increases in
summer rainfall while the majority shows a decrease, and because of the application of
a uniform distribution. It is interesting to note that when zero water is required, there is
no uncertainty with respect to emissions scenarios. This is an artefact created by the
pattern-scaling technique when a number is multiplied by zero (or close to zero)
precipitation. It occurs when the regional climate response crosses from one sign to
another (if each individual GCM result is used or if natural variability is added this does
not occur; as shown later).

166
 1

0.9

0.8
AWS
0.7 Natural variability

0.6 A1B-AIM
A1FI-MI
Probability 0.5
A1T-MES
0.4 A2-ASF
B1-IMA
0.3
B2-MES
0.2

0.1

0
-20 -15 -10 -5 0 5 10 15 20 25 30
Additional water required (Ml/d)

Figure 7.5 Additional water required by the 2030s for the ES&E WRZ due to
climate change under the six illustrative SRES marker emissions scenarios (A1B-
AIM, A1FI-MI, A1T-MES, A2-ASF, B1-IMA, B2-MES) and the default scheme
using GCMs. Also shown is AWS plan for this WRZ and the additional water
required solely due to natural multi-decadal climate variability.

At the 95th percentile the uncertainty introduced by different SRES marker scenarios is
in the order of 5.55 Ml/d. The discrepancy between AWS WRP and SRES A1T-MES at
this level of confidence is of 7.35 Ml/d. It is intereting to note that what could be
considered one of the greenest worlds (A1T-MES; has a strong emphasis on
renewable energy) will require more adaptation than a less green world (A2-ASF).65
This occurs due to the impact of sulphate aerosols (examined more detail in section
7.1.4).

65
Compatible with a World Markets scenario and a Provincial Enterprise scenario respectively, or Beta
and Alpha scenarios in Environment Agency terminology.

167
 1

0.9

0.8
AWS
0.7 Natural variability

0.6 A1B-AIM
A1FI-MI
Probability 0.5
A1T-MES
0.4 A2-ASF
B1-IMA
0.3
B2-MES
0.2

0.1

0
-15 -10 -5 0 5 10 15 20 25
Additional water required (Ml/d)

Figure 7.6 Same as Figure 7.5, but only using one GCM (HadCM3).

Figure 7.6 shows the risk of climate change to water resources using only one GCM
(HadCM3). It is promising that results using HadCM3 are close to the AWS strategy.
This shows that the modelling framework applied in this thesis can mimic UKCIP02
results (which used the HadCM3 model). At the 95th percentile, the uncertainty
introduced by emissions scenarios is of 3.2 Ml/d, whereas the discrepancy between
AWS and A1T-MES is 1.45 Ml/d.

0.9

0.8
AWS
0.7 Natural variability

0.6 A1B-AIM
A1FI-MI
Probability 0.5
A1T-MES
0.4 A2-ASF
B1-IMA
0.3
B2-MES
0.2

0.1

0
-20 -15 -10 -5 0 5 10 15 20 25 30
Additional water required (Ml/d)

Figure 7.7 Same as Figure 7.5, but only using RCMs.

Even more promising for the modelling framework used here are the results shown in
Figure 7.7 which use RCM rather than GCM results. The AWS curve matches very well

168
A2-ASF, which is essentially equivalent to UKCIP02 medium-high. Almost all the
RCMs simulations included in the CDFs were driven by the same A/GCM as in
UKCIP02 (HadAM3H). The uncertainty introduced by emissions scenarios at the 95th
percentile is 4.57 ml/d and the AWS strategy is short of 6.73 Ml/d compared to A1T-
MES.

In summary, the uncertainty introduced by the six illustrative SRES scenarios to water
resources planning is significant (in the order of 4.5-5.55 Ml/d), but does not represent a
threat to the ES&E water resource zone because of positive supply/demand balance in
2030 (11 Ml/d). This component will grow in significance the further into the future the
planning horizon is.

0.9

0.8
AWS
0.7 Natural variability

0.6 A1B-AIM
A1FI-MI
Probability 0.5
A1T-MES
0.4 A2-ASF
B1-IMA
0.3
B2-MES
0.2

0.1

0
-60 -40 -20 0 20 40 60 80 100 120
Additional water required (Ml/d)

Figure 7.8 Additional water required by the 2030s for the Ruthamford WRZ
under different SRES marker emissions scenarios (A1B-AIM, A1FI-MI, A1T-
MES, A2-ASF, B1-IMA, B2-MES) and the default scheme using RCMs. Also
shown is AWS plan for this WRZ and the additional water required solely due to
natural multi-decadal climate variability (Ely Ouse catchment data 1865 to 1950).

For the Ruthamford WRZ the uncertainties introduced by emissions scenarios are much
larger than for the ES&E WRZ. This occurs because of the larger size of the WRZ, the
larger drying in this WRZ (using RCM results) and also the higher sensitivity of the
reservoirs in this WRZ to changes in the climate. Thus, at the 95th percentile, emissions
scenarios introduce almost 20 Ml/d of uncertainty. The shortfall for AWS WRP,
compared to A1T-MES is just over 40 Ml/d, which is AWSs best estimate for the
impact of climate change on their entire business (AWS 2004b). AWSs WRP remains

169
robust to emissions scenario uncertainty for the Ruthamford WRZ under the default
scheme considered.

One further sensitivity experiment that was conducted for GHG emissions compared the
default scheme SRES marker scenario (A2-ASF) with the other members of the same
scenario family (A2A1M, A2AIM, A2MES, A2MIN and A2GIM).

0.9

0.8
AWS
0.7 Natural variability

0.6 A2-ASF
A2A1MI
Probability 0.5
A2AIM
0.4 A2MES
A2MIN
0.3
A2GIM
0.2

0.1

0
-15 -10 -5 0 5 10 15 20 25
Additional water required (Ml/d)

Figure 7.9 Additional water required by the 2030s for the ES&E WRZ under
different emissions scenarios for the SRES A2 scenario family (A2-ASF, A2A1M,
A2AIM, A2MES, A2MIN and A2GIM) and the default scheme using GCMs. Also
shown is AWS plan for this WRZ and the additional water required solely due to
natural multi-decadal climate variability.

Figure 7.9 shows the impact of using all the scenarios of the A2 storyline. It is
noteworthy that even uncertainties relating to how Integrated Assessment Models
convert storylines into greenhouse emissions can have an impact on adaptation
decisions. The uncertainty range at the 95th percentile is 3.29 Ml/d and the AWS
shortfall is 3.44 Ml/d. SRES A2-ASF, used in the default scheme, lies roughly in the
middle of the other scenarios. This has important implications for climate modelling as
no GCMs have ever been run with scenarios other than the six illustrative SRES
scenarios. It also shows that the uncertainty introduced by Integrated Assessment
Models when converting storylines into GHG emissions is not negligible from an
adaptation perspective.

170
7.1.2.1 Mitigation/stabilisation scenarios

Mitigation and stabilisation scenarios are important for a number of reasons. In the
real world, climate policies are being enacted (e.g., Kyoto Protocol, EU Emissions
Trading Scheme, UK Climate Change Programme, etc.) and are expected to be
reinforced in the future (but this is extremely uncertain because it depends on human
choice; in particular how much mitigation is going to happen). The SRES scenarios do
not include explicit climate policies (but they do include environmental policies) so one
could argue that they are unrealistic or that they will never come to pass. Including
climate mitigation policies in the SRES scenarios is therefore important (see section
4.1.1) and for this reason the Post-SRES mitigation/stabilisation scenarios are used here
(Morita and Robinson 2001). The inclusion of mitigation adds further uncertainty to the
problem since it is impossible to know at which level atmospheric concentrations of
GHG will be stabilised over the next 100 years. Hence, different stabilisation levels are
examined for a few SRES scenarios. This allows the comparison of mitigation efforts
with adaptation responses in the water sector and their respective trade-offs.

The analysis shows that by the 2030s mitigation policies will have a small influence on
water resources in the ES&E WRZ under the A2-ASF scenario. Thus, stabilising GHG
concentrations at 550ppm (which is essentially the EU/UK position/vision, see
Comission 2005) would lead to a reduction in additional water required by 2030 of
between 1.63-1.93 Ml/d (GCM-RCM range) at the 95th percentile.

0.9

0.8
AWS (ES&E
0.7 Natural variability

0.6 GCMBaseline
GCM750
Probability 0.5
GCM550
0.4 RCMBaseline
RCM750
0.3
RCM550
0.2

0.1

0
-15 -10 -5 0 5 10 15 20 25
Additional water required (Ml/d)

Figure caption overleaf

171
Figure 7.10 Additional water required by the 2030s for the ES&E WRZ under the
A2-ASF baseline scenario and stabilisation scenarios (750 and 550 ppm) for the
default scheme using GCMs and RCMs. Also shown is AWS plan for this WRZ
and the additional water required solely due to natural multi-decadal climate
variability (Ely Ouse catchment data 1865 to 1950).

Table 7.2 shows that greater reductions in additional water required are possible under
the A1FI-MI scenario, reaching a maximum of between 3.34-3.57 Ml/d (GCM-RCM
range) with a 450ppm stabilisation scenario. In this scenario, mitigation policies are not
entirely negligible for adaptation purposes. The higher the emissions in the baseline
scenario, the greater the impact of mitigation policies will be on additional water
resources.

Table 7.2 Additional water required by the 2030s for the ES&E WRZ under A1FI-
MI baseline scenario and different stabilisation scenarios (750, 650, 550 and
450ppm) for the default scheme using GCM and RCM results.

Percentiles Baseline 750 650 550 450

0.0% -5.27 -5.57 -5.66 -5.31 -4.66
2.5% -2.96 -2.67 -2.62 -2.58 -2.31
5.0% -2.25 -2.06 -2.03 -1.99 -1.77
GCM

50.0% 6.84 6.41 6.29 6.10 5.32

95.0% 18.27 17.19 17.21 16.41 14.70
97.5% 20.06 18.93 18.91 18.29 16.42
100.0% 30.58 30.20 29.13 32.17 32.08
Percentiles Baseline 750 650 550 450
0.0% 0.73 0.68 0.69 0.45 0.58
2.5% 2.18 2.03 1.98 1.94 1.74
5.0% 3.76 3.48 3.44 3.28 2.87
RCM

50.0% 10.58 9.82 9.68 9.28 8.23

95.0% 18.09 17.00 16.81 16.24 14.76
97.5% 19.43 18.58 18.36 17.69 16.15
100.0% 30.73 26.97 30.06 25.80 24.33

This analysis shows that adaptation should be of higher priority to AWS than mitigation
because the increases in water required (around 18 Ml/d under A1FI-MI) far outweigh
the reductions due to mitigation (between 1-3.57 Ml/d). This is, of course, not that
simple since AWS does not mitigate worldwide GHG emissions on its own (as
portrayed in the simple climate model); millions of other agents do as well. The analysis
shows, in a crude way, the plausible trade-offs between mitigation and adaptation (from
a water resources perspective), an emerging topic in the forthcoming IPCC assessment
report (cf. Klein et al. 2005).

172
7.1.3 Climate sensitivity

Climate sensitivity is an important parameter to be assessed in any climate change

assessment (see section 4.2). Uncertainty about climate sensitivity has been shown to
influence important mitigation decisions (see, e.g., Caldeira et al. 2003).

Table 7.3 shows the impact of using different climate sensitivity PDFs on the risk that
climate change poses on water resources. Under the default scheme uncertainty about
the shape of the climate sensitivity PDF introduces large uncertainties into water
resources planning. At the 95th percentile, the range of uncertainty for the ES&E WRZ
is 8.27 Ml/d and almost 30 Ml/d for Ruthamford WRZ. For the default scheme, AWS
WRP remains robust to climate sensitivity uncertainty, although for the ES&E the
inclusion of this uncertainty would leave only 1.61 Ml/d in the supply/demand balance
by 2030. It is worth noting that in the case of the Forest et al. (2002) CDF and the
Murphy et al. (2004) CDF using respectively expert judgement and a Climate Prediction
Index to judge simulations (weighted) reduces the range of uncertainty. In the case of
Forest et al. (Expert) it reduces the water required whereas in Murphy et al. (Weighted)
it increases it.

173
Table 7.3 Additional water required by the 2030s for the ES&E and Ruthamford WRZs under the default scheme using GCMs with changing
climate sensitivity PDFs. Also shown is AWS plan and the additional water required solely due to natural multi-decadal climate variability.

Percentile Wigley Andronova Uniform Expert Knutti Unweighted Weighted Uniform AWS Natural
and Raper and Forest et Forest et al. Murphy et Murphy et cp.net variability
(2001) Schlesinger al. (2002) et al. (2002) al. (2004) al. (2004) (2005)
(2001) (2002)
0.0% -5.53 -7.95 -9.09 -5.19 -6.85 -6.24 -5.99 -6.96 0.21 -17.24
2.5% -2.54 -2.49 -2.94 -2.40 -3.40 -2.67 -2.94 -3.54 2.04 -8.81
5.0% -1.98 -1.74 -2.11 -1.88 -2.61 -2.11 -2.37 -2.67 2.87 -7.40
ES&E WRZ

50.0% 5.81 5.05 6.15 5.53 8.07 6.42 7.17 7.83 9.09 0.00
95.0% 16.56 18.36 20.83 15.76 22.43 17.56 18.89 24.03 14.64 7.39
97.5% 18.66 21.90 24.81 17.64 25.03 19.70 20.86 27.08 15.39 8.78
100.0% 34.56 59.49 56.07 29.68 43.58 35.03 33.53 41.20 17.13 17.38
0.0% -19.79 -28.41 -32.50 -18.55 -24.50 -22.32 -21.41 -24.91 1.32 -71.47
Ruthamford WRZ

2.5% -9.10 -8.91 -10.52 -8.59 -12.14 -9.54 -10.51 -12.66 7.64 -39.72
5.0% -7.08 -6.23 -7.54 -6.73 -9.32 -7.54 -8.47 -9.55 10.46 -33.35
50.0% 20.79 18.06 21.98 19.79 28.87 22.97 25.64 27.99 31.14 -0.01
95.0% 59.22 65.67 74.49 56.34 80.21 62.79 67.55 85.91 46.16 33.33
97.5% 66.72 78.32 88.70 63.09 89.49 70.44 74.59 96.85 48.08 39.72
100.0% 123.59 212.74 200.50 106.12 155.83 125.28 119.89 147.32 52.59 71.06

174
7.1.4 Other climate parameters

With the modelling framework applied here it is possible to determine whether the
climate parameters defined in the simple climate model (section 4.3.1) have an impact
on adaptation decisions. The parameters examined included: aerosol forcing, ocean
diffusivity and carbon cycle. One by one, the highest and lowest values of one
parameter was investigated, with the default uniform distribution (or log-normal in the
case of carbon cycle) applied to the other two parameters. Figure 7.11 and Table 7.4
show the impact of these three parameters on water resources of the ES&E WRZ.
Ocean diffusivity and carbon cycle uncertainties have a small influence on water
resources (between 2.69-2.91 Ml/d at the 95th percentile), but aerosol forcing introduces
a larger uncertainty for planning (7.53 Ml/d). This latter uncertainty is large because in
the A2-ASF scenario there are large emissions of sulphur dioxide in the first couple of
decades of the century. Under low aerosol forcing much more water is required than
under high aerosol forcing. This uncertainty is much more scenario dependent than the
other two; i.e., in a scenario with lower sulphur dioxide emissions than A2-ASF, aerosol
forcing uncertainty will be smaller than in A2-ASF. In sum, these are not entirely
negligible parameters when it comes to adapting to climate change, especially the
magnitude of aerosol forcing.

0.9

0.8
AWS
0.7 Natural variability
Default scheme
0.6
Low aerosol
Probability 0.5 High aerosol
Low ocean
0.4
High ocean
0.3 Low carbon
High carbon
0.2

0.1

0
-20 -15 -10 -5 0 5 10 15 20 25 30
Additional water required (Ml/d)

Figure 7.11 Additional water required by the 2030s for the ES&E WRZ under the
default scheme using GCMs and under changing climate parameters (low and high
aerosol forcing, ocean diffusivity and carbon cycle). Also shown is AWS plan for

175
 this WRZ and the additional water required solely due to natural multi-decadal
climate variability (Ely Ouse catchment data 1865 to 1950).

Table 7.4 Additional water required by the 2030s for the ES&E WRZ under the
default scheme using GCMs and under changing climate parameters (low and high
aerosol forcing, ocean diffusivity and carbon cycle). Also shown is AWS plan and
the additional water required solely due to natural multi-decadal climate
variability.

Percentiles Default Low High Low High Low High AWS Natural
scheme aerosol aerosol ocean ocean carbon carbon variability
forcing forcing diffusivity diffusivity cycle cycle
0.0% -5.53 -5.58 -3.35 -5.55 -4.76 -4.34 -5.39 0.21 -17.24
2.5% -2.54 -3.27 -2.01 -2.92 -2.56 -2.55 -2.95 2.04 -8.81
5.0% -1.98 -2.52 -1.61 -2.27 -2.01 -2.00 -2.31 2.87 -7.40
50.0% 5.81 7.71 4.82 6.93 6.13 5.98 7.12 9.09 0.00
95.0% 16.56 20.28 12.75 18.38 15.69 15.82 18.74 14.64 7.39
97.5% 18.66 22.02 13.91 20.02 16.97 17.31 20.22 15.39 8.78
100.0% 34.56 30.38 19.24 29.22 24.59 25.97 28.61 17.13 17.38

7.1.5 Regional climate response

The literature seems to suggest that regional climate response is where some of the
largest uncertainties lie in climate prediction (see, e.g., Giorgi and Francisco 2000). The
default scheme used in the sensitivity analysis assumes a uniform distribution across
available GCM results to represent uncertainty in regional climate response. This is a
major assumption, which is examined next. Figure 7.12 and Table 7.5 display results for
additional water required due to climate change for individual GCM simulations (i.e.,
without assuming any PDF for their distribution). The uncertainty range between the
highest and lowest GCMs at the 95th percentile is 22.66 Ml/d, which is the greatest
uncertainty assessed up to now. The HadCM3 curve is the closest to the AWS plan,
which bodes confidence in the modelling framework applied here. The default scheme
CDF, which uses a uniform distribution across GCM results, goes through every
individual GCM CDF. Due to the central limit theorem it makes the GCMs in the
middle (GFDL and NIES) the most likely.

The AWS WRP remains robust to GCM uncertainty under the default scheme, mainly
because it was planned using one of the driest models (HadCM3). In this sensitivity test

176
a single GCM grid box value (a deterministic number) is being multiplied by a positive
PDF of global mean temperature change. This explains why none of the curves cross the
zero line and why the GCMs closest to zero have narrower distributions than the GCMs
further away from zero. Another peculiarity that emerges from using pattern-scaling
probabilistically is that MRI2 emerges as the driest model but in the original data (see
section 5.4 and Figure 5.9) HadCM3 came out as the driest model. This discrepancy
arises because MRI2 has a low global warming figure, which creates the highest rate of
precipitation decrease per degree of global warming.

0.9 AWS
Natural variability
0.8
Default schem e
0.7 CGCM2

0.6 CSIRO Mk2

CSM 1.3
Probability 0.5
ECHAM4/OPYC3
0.4 GFDL_R15_b
MRI2
0.3
CCSR/NIES2
0.2 DOE PCM
HadCM3
0.1

0
-15 -10 -5 0 5 10 15 20 25 30
Additional w ater required (Ml/d)

Figure 7.12 Additional water required by the 2030s for the ES&E WRZ under the
default scheme using GCMs (uniform distribution) and using different GCMs
forced by SRES A2. Also shown is AWS plan for this WRZ and the additional
water required solely due to natural multi-decadal climate variability.

177
Table 7.5 Additional water required by the 2030s for the ES&E WRZ under the default scheme using GCMs (uniform distribution) and using
different GCMs forced by SRES A2. Also shown is AWS plan and the additional water required solely due to natural multi-decadal climate
variability.

Percentiles CGCM2 CSIRO CSM ECHAM4 GFDL_ MRI2 CCSR/NI DOE HadCM3 Default AWS Natural
Mk2 1.3 /OPYC3 R15_b ES2 PCM scheme variability
0.0% 0.67 -2.85 0.66 2.42 1.78 4.44 1.56 0.44 2.48 -5.00 0.10 -17.98
2.5% 1.23 -1.95 1.22 4.49 3.30 8.23 2.88 0.81 4.59 -2.56 2.03 -8.82
5.0% 1.35 -1.81 1.34 4.91 3.61 9.00 3.16 0.89 5.02 -1.98 2.87 -7.40
50.0% 2.15 -1.18 2.13 7.81 5.74 14.31 5.02 1.41 7.98 5.82 9.09 0.00
95.0% 3.29 -0.74 3.26 11.96 8.79 21.91 7.68 2.17 12.22 16.50 14.64 7.39
97.5% 3.55 -0.68 3.52 12.92 9.49 23.67 8.29 2.34 13.20 18.56 15.40 8.82
100.0% 5.19 -0.37 5.15 18.89 13.87 34.60 12.13 3.42 19.30 34.77 17.00 15.92

178
In order to get a more realistic representation of future climate and to overcome the
pattern-scaling deficiency of dealing with signs, tri-decadal natural climate variability
has been added to Figure 7.13 below. Figure 7.13 shows how the inclusion of natural
variability allows the CDFs to cross between signs.

0.9
AWS
0.8 Natural variability
Default schem e
0.7 CGCM2
CSIRO Mk2
0.6
CSM 1.3
Probability 0.5 ECHAM4/OPYC3
GFDL_R15_b
0.4
MRI2

0.3 CCSR/NIES2
DOE PCM
0.2 HadCM3

0.1

0
-15 -10 -5 0 5 10 15 20 25 30 35
Additional w ater required (Ml/d)

Figure 7.13 Additional water required by the 2030s for the ES&E WRZ under the
default scheme using GCMs (uniform distribution) and using individual GCMs
forced by SRES A2 with added natural variability. Also shown is AWS plan for
this WRZ and the additional water required solely due to natural multi-decadal
climate variability.

The issue of model skill is investigated next. There are a plethora of studies that are
evaluating and weighting model simulations based on how well they represent past or
present climate. In Chapter 5, GCM control simulations were compared with observed
climatology for the Anglian region (cf. Figure 5.14). Here the analysis was redone using
only the best performing models for precipitation (HadCM3, CSM, CCCma and GFDL
weighted equally) and the best performing models for precipitation and temperature
(HadCM3, CSM, PCM and NIES weighted equally) and results were compared to the
default scheme (uniform distribution across all GCM results; Figure 7.14).

Using the most skilful models gives a narrower (less uncertain) PDF of additional water
required, but this is severely limited by the fact that only nine GCM simulations were
examined. Nonetheless, this simple experiment shows that uncertainty can be narrowed
by constraining exercises, but on what grounds model skill should be defined will be

179
hotly contested. The issue of verification, validation and confirmation of earth science
models has been debated for many years (see e.g., Oreskes et al. 1994).

0.10

0.09

0.08

0.07

0.06 AWS

Probability 0.05 Natural variability

Default scheme
0.04
Best GCMP
0.03 Best GCMP&T
0.02

0.01

0.00
-15 -10 -5 0 5 10 15 20 25
Additional water required (M/d)

Figure 7.14 Additional water required by the 2030s for the ES&E WRZ under the
default scheme using GCMs (uniform distribution) and using the best performing
GCMs for precipitation (P) and precipitation & temperature (P&T). Also shown is
AWS plan for this WRZ and the additional water required solely due to natural
multi-decadal climate variability.

7.1.5.1 Dynamical downscaling

Using a small subset of GCMs, mainly HadCM3 (but also ECHAM4 and ARPEGE), a
European project called PRUDENCE dynamically downscaled GCM results into finer
resolutions using 11 different Regional climate models. This section runs the analysis
using the PRUDENCE results.

The default scheme for RCMs did not assign a subjective PDF to their distribution, but
instead assumed they were all equally likely (as a superensemble). Here the uncertainty
introduced by using RCMs is examined in more detail. Figure 7.15 shows the additional
water required for the ES&E WRZ using 37 simulations from 11 RCMs with added
natural variability. Dynamical downscaling introduces large uncertainties to water
resources planning, although not as large as the uncertainty due to GCMs. At the 95th
percentile, the uncertainty is almost 15 Ml/d.

180
 1

0.9 Default scheme

AWS
0.8
Natural variability
0.7

0.6

Probability 0.5

0.4

0.3

0.2

0.1

0
-15 -10 -5 0 5 10 15 20 25 30 35
Additional water required (Ml/d)

Figure 7.15 Additional water required by the 2030s for the ES&E WRZ under all
available RCMs (37 simulations using 11 RCMS) with added natural variability.
The thick black line represents AWS strategy for this WRZ, the thick blue line
represents the default scheme (without natural variability) and the thick brown
line represents the additional water required solely due to natural multi-decadal
climate variability.

Each of the simulations displayed in Figure 7.15 has been run with a specific
combination of forcing scenario (SRES A2 or B2), driving A/GCM (HadCM3,
ECHAM4 and ARPEGE), spatial resolutions and initial conditions. Some of these
combinations are examined in more detail to see if adaptation decisions are sensitive to
them. Different forcing scenarios seem to introduce some uncertainty but it is less than
2 Ml/d at the 95th percentile (Figure 7.16).66 It is not surprising that the default scheme
matches well the A2 forced simulations since the default scheme is comprised of 29 A2
forced simulations and eight B2 forced simulations. These results also match well
Figures 7.5-7.7, where a B2 world is slightly drier than an A2 world. Earlier it was
explained that this was due to the impact of sulphate aerosols on the modelling
framework applied. In the case of RCMs it is difficult to say why this happens without
going into more depth with RCM modelling. It could be due to aerosol forcing, but it
could also just be due to the high variability of precipitation or due to the application of
the pattern-scaling technique (i.e., RCM results are divided by a scaler, which in the
case of B2 is smaller than A2).

66
The uncertainty is larger at the 50th and 5th percentile though.

181
 1

0.9

0.8

0.7
Default scheme
0.6
AWS
Probability 0.5
SRESA2
0.4
SRES B2
0.3

0.2

0.1

0
0 5 10 15 20 25
Additional water required (Ml/d)

Figure 7.16 Additional water required by the 2030s for the ES&E WRZ under the
default scheme using RCMs, the AWS strategy and using RCM simulations forced
by SRES A2 and SRES B2.

Which A/GCM drives the RCMs makes a substantial impact on the amount of
additional water required (Figure 7.17). The difference between HadCM3 and
ECHAM4 is almost negligible from an adaptation perspective; the uncertainty
introduced is only 0.5 Ml/d at the 95th percentile. This is not surprising since both
models show very similar amounts of drying for the East of England (see how close
curves are in Figures 7.12 and 7.13). Large uncertainties are introduced, however, when
the ARPEGE model is taken into account; almost 12 Ml/d at the 95th percentile. The
default scheme is closest to the HadCM3 curve because there are many more RCM
simulations driven by this model than the other two.

182
 1

0.9

0.8

0.7
Default schem e
0.6
AWS
Probability 0.5 HadCM3
ECHAM4
0.4
ARPEGE
0.3

0.2

0.1

0
0 5 10 15 20 25
Additional w ater required (Ml/d)

Figure 7.17 Additional water required by the 2030s for the ES&E WRZ under the
default scheme using RCMs, the AWS strategy and using RCM simulations driven
by different Atmospheric GCMs (HadCM3, ECHAM4 and ARPEGE).

The issue of model skill is here revisited for the case of RCMs. The criteria used to
judge model skill was the degree of proximity of model simulation results to observed
values of summer mean precipitation and variance. Figure 7.18 shows how the
modelled values compare with observations. Models inside the inner circle in Figure
7.18 are deemed to have high skill whereas models between the circle lines are given
medium skill and models outside the outer circle are regarded as having low skill.
Models from each category (high, medium and low skill) were given equal probability
and the analysis was redone using each category separately. An extra analysis, named
combination, was also performed where the high skilled models were three times as
likely as the low skilled models and the medium models were twice as likely as the low
skill models.

183
 JJA St. Dev. Precipitation (mm/day)
0.75
0.7
0.65
0.6
0.55
0.5
0.45
RCM Ruthamford
0.4
Observed Ely Ouse
0.35
0.3
1 1.2 1.4 1.6 1.8 2 2.2 2.4
JJA Mean Precipitation (mm/day)

Figure 7.18 Comparison between summer mean precipitation and variance for the
1961-90 period as modelled by RCMs over the Ruthamford WRZ and as observed
for the Ely Ouse catchment. The two circles enclose the models with high skill
(inner circle) and medium skill (outer circle). All other models have low skill.

Figure 7.19 shows that using or weighting models according to their skill can change
the shape of the PDFs. It is noticeable that the highly skilled models can constrain the
lower bound of the PDF, but only slightly. The biggest difference seems to emerge
between the medium and the low skilled models. At the 95th percentile the uncertainty
range between these categories is in the order of 12 Ml/d. The combined curve, which
includes all three categories, but weighted, is perhaps the most comprehensive
approach, but is not too dissimilar to the default scheme (all simulations equally likely).
At the 95th percentile the difference between the combined approach and the default
scheme is about 3 Ml/d. For a WRZ as large as Ruthamford and for the adaptations
being planned, introducing a measure of skill for RCM performance makes very little
difference.

184
 0.1

0.09

0.08

0.07
AWS
0.06
Default scheme
Probability 0.05 High skill

0.04 Mediumskill
Low skill
0.03
Combination
0.02

0.01

0
0 10 20 30 40 50 60 70 80 90 100
Additional water required (Ml/d)

Figure 7.19 Additional water required by the 2030s for the Ruthamford WRZ
under the default scheme using RCMs, the AWS strategy, and using RCM
simulations with different amounts of skill (high, medium, low) and in
combination.

7.1.6 Climate impacts

The relationship between changes in the climate and additional water required is based
on hydrological modelling performed by AWS (see section 6.6). No uncertainty
analysis was performed on this model so it is assumed the relationship is uncertain by
25%.67 Figure 7.20 shows that the uncertainty in hydrological modelling is not
negligible from an adaptation perspective. The uncertainty range for GCMs and RCMs
is almost the same at around 8.3 Ml/d. AWS WRP remains robust nonetheless.

67
The uncertainty of the regression coefficients (Table 6.11) could have been used to sample this
uncertainty.

185
 1
AWS
0.9
Natural variability
0.8
GCMdefault
0.7 GCMHigh
GCMLow
0.6
RCMdefault
Probability 0.5
RCMHigh
0.4 RCMLow

0.3

0.2

0.1

0
-20 -15 -10 -5 0 5 10 15 20 25 30
Additional water required (Ml/d)

Figure 7.20 Additional water required by the 2030s for the ES&E WRZ under the
default scheme using GCMs and RCMs and with changing climate impact
functions (high and low; 25%.). Also shown is AWS plan for this WRZ and the
additional water required solely due to natural multi-decadal climate variability.

7.1.7 Sampling strategy

In section 7.1.3 it was shown that the shape of the climate sensitivity PDF can have an
impact on adaptation decisions. In this section different sampling strategies are
compared. By default Latin Hypercube Sampling was used because it is an efficient
sampling technique (see McKay et al. 1979, or Helton and Davis 2003 for a recent
review). Latin Hypercube Sampling divides each parameter distribution into a number
of non-overlapping intervals (in this case 500), each having equal probability. Then,
from each interval, a value is selected at random (from 10,000 simulations) according to
the probability distribution within the interval. The other approach, which is tested here,
is Monte Carlo sampling where random numbers are used to sample the parameter
distribution. Latin Hypercube Sampling is generally more precise for producing random
samples than conventional Monte Carlo Sampling because the full range of the
distribution is sampled in a more even and consistent manner (cf. Helton and Davis
2002). Figure 7.21 shows that adaptation decisions are fairly insensitive to sampling
strategy. At the 95th percentile, the uncertainty range between sampling strategies is a
mere 0.07 Ml/d.

186
 0.06

Monte carlo sampling

0.05
Latin hypercube sampling

0.04

Probability 0.03

0.02

0.01

0
0 5 10 15 20 25 30
Additional water required (Ml/d)

Figure 7.21 Additional water required by the 2030s for the ES&E WRZ under the
default scheme using RCMs with two different sampling strategies (Monte Carlo
and Latin hypecube).

7.1.8 Summary

This section distils and summarises the sensitivity experiments/simulations conducted in

the previous sections. Table 7.6 shows the sensitivity of adaptation decisions (measured
as additional water required) to the various parameters analysed for both WRZs. The
largest uncertainty introduced into adaptation planning comes from the regional climate
response, mainly GCMs, but closely followed by dynamical downscaling (RCMs). Both
hydrological modelling (climate impacts) and climate sensitivity uncertainties are level
at third place in the sensitivity ranking. Next comes aerosol forcing followed by GHG
emission scenario uncertainty. Ocean diffusivity and carbon cycle uncertainties are
amongst the parameters to which the decision is the least sensitive.

187
Table 7.6 Quantification of the uncertainty introduced by the parameters sampled
in assessment using the default scheme (with GCMs) in terms of additional water
required (Ml/d) for the East Suffolk & Essex and Ruthamford WRZs (at the 95th
percentile). Also shown is AWSs WRP shortfall.

Parameter East Suffolk & Essex Ruthamford WRZ

WRZ
Uncertainty AWS Uncertainty AWS
range shortfall range shortfall
(Ml/d) (Ml/d) (Ml/d) (Ml/d)
GHG emissions scenario 5.55 7.35 18.56 32.11
Climate sensitivity 8.27 9.39 30.63 40.7
Aerosol forcing 7.53 5.63 26.70 25.92
Ocean diffusivity 2.69 3.74 9.23 19.68
Carbon cycle 2.91 4.01 10.52 20.93
Regional climate response 22.66 7.28 81.05 32.29
GCMs (RCMs) (14.88) (9.10) (58.96) (55.68)
Climate impacts 8.28 6.06 29.64 28.00

AWSs WRP shortfall has been measured as the difference between the maximum value
of additional water required for that particular uncertainty with the AWS strategy at the
95th percentile. In other words, how much extra target headroom would be necessary to
cope with all sampled uncertainty in that parameter. For some uncertainties, AWSs
shortfall is larger than the actual uncertainty range sample; this occurs when AWSs
strategy is not even within the uncertainty range. Perhaps this is the best evidence that a
triangular distribution might not be the most appropriate PDF to represent climate
change uncertainties, which often have long tails (e.g., climate sensitivity, etc.).

Surprisingly, AWSs WRP remains robust to the uncertainties sampled in the default
scheme. This does not mean that the WRP is immune to all climate change
uncertainties, as shown further ahead, but it is immune to the combinations examined in
the default scheme of the sensitivity analysis on a one-to-one basis. However, even
under the default scheme both WRZs were close to eliminating their positive
supply/demand balance with some parameters. It is difficult to compare the WRZs with
Table 7.6 so Table 7.7 shows results as a percentage of demand.

188
Table 7.7 Same as Table 7.6, but as a percentage of demand.

Parameter East Suffolk & Essex Ruthamford WRZ

WRZ
Uncertainty AWS Uncertainty AWS
range (%) shortfall range (%) shortfall
(%) (%)
GHG emissions scenario 5.57 7.38 5.32 9.21
Climate sensitivity 8.30 9.42 8.79 11.68
Aerosol forcing 7.56 5.65 7.66 7.44
Ocean diffusivity 2.70 3.75 2.65 5.65
Carbon cycle 2.92 4.02 3.02 6.00
Regional climate 22.74 7.30 23.25 9.26
response GCMs (RCMs) (14.93) (9.13) (16.91) (15.97)
Climate impacts 8.31 6.08 8.50 8.03

Table 7.7 shows that results are fairly similar between WRZs. The AWS shortfall is
consistently higher for the Ruthamford WRZ compared to ES&E WRZ. This seems to
suggest that Anglian Water is better prepared to cope with future drought in the ES&E
WRZ than the Ruthamford WRZ. The differences between WRZs in terms of
uncertainty ranges are not greater than 0.5% except for RCMs. This occurs because the
RCM simulations suggest that the Ruthamford WRZ will get drier than the ES&E
WRZ.

It is important to note that the relative significance of the uncertainties reported here is
impact variable specific and time-dependent. For example, if one was interested in
temperature change in the 2050s (for the purpose of building design) one could
speculate that climate sensitivity and GHG emissions uncertainties would be more
significant than GCM uncertainty. It should also be noted that the omission of certain
uncertainties such as choice of energy balance model, hydrological modelling (that
explores structural and parameter uncertainty) and downscaling (in particular statistical
downscaling) skews the relative importance of different sources of uncertainty.

7.2 Uncertainty analysis

The previous section on sensitivity analysis identified the most critical uncertain
parameters in a climate change assessment from an adaptation perspective. Essentially,
all the tested parameters introduce uncertainties into adaptation planning, but some are

189
more significant than others (e.g., regional climate response). The only element that was
insignificant for adaptation decisions was sampling strategy, so Latin Hypercube
Sampling is used henceforth.

In this section an uncertainty analysis is performed on the entire framework. Where

possible uncertainty was quantified using PDFs, but where human choice is involved
scenarios were used. The uncertainty analysis is not too different from the sensitivity
analysis. The major difference resides in the fact that where the sensitivity analysis
defines an upper and lower bound for a certain variable (high and low aerosol forcing or
x and y GCM) the uncertainty analysis defines a PDF (or a scenario). It is important to
recognise that these are subjective probabilities (see also Mearns et al. 2001, p. 759).
Table 7.8 illustrates the PDFs or scenarios used in the uncertainty analysis, which in
many cases are not different from the default scheme of the sensitivity analysis.

Table 7.8 Representation of uncertainty in the parameters of the framework.

Parameter Uncertainty ES&E WRZ Ruthamford Reference

represented WRZ
by
Natural Normal PDF Mean: 0 Mean: 0 (Jones et al. 2004)
variability Std: 3.77 Std: 4.75
World Scenarios A1, A2, B1, A1, A2, B1, (Nakicenovic et al.
development B2 B2 2000)
Conversion of Assuming all (Nakicenovic et al.
storylines into models 2000)
emissions of equally
GHGs by plausible
Integrated
Assessment
Models
Climate The most Weighted Weighted (Murphy et al. 2004)
sensitivity recent PDF in Murphy et al. Murphy et al.
the literature (2004) (2004)
Carbon cycle Log-normal (1.5, 2.6, 4.5) (1.5, 2.6, 4.5)(Wigley and Raper
PDF 2001)
Ocean Uniform PDF (1.47, 3.6) (1.47, 3.6) (Wigley and Raper
diffusivity 2001)
Aerosol Uniform PDF (-0.58, -1.819) (-0.58, - (Wigley and Raper
forcing 1.819) 2001)
Regional Assuming all Uniform: Uniform: http://ipcc-
response highly skilled (-18.05, 3.80)a (-21.49, 3.88) ddc.cru.uea.ac.uk/
a
(GCMs and models are http://prudence.dmi.dk/
RCMs) equally likely

190
 plausible or
assuming a
uniform
distribution
Climate Uniform PDF (-1.79, -0.60) (-7.15, -2.38) (AWS 2003)
impacts (with bounds
of 50%)
a: precipitation change (%) per degree of global warming

Some justification for the PDFs (or scenarios) applied is required here. Natural
variability was assumed to be normally distributed because that was the evidence
arising from past observations (see Figure 5.12). That was the case for the ES&E WRZ,
which used data from the England and Wales Precipitation series. For the Ruthamford
WRZ , whose data originated from the Ely Ouse catchment, that was not the case. This
could be due to the fact that less 30-year periods were available (than in EWP), but a
normal distribution was assumed as well.

The evolution of world development in the next 25 years is highly uncertain and
strongly dependent on human choice so scenarios are used to portray this factor. Four
different narrative storylines from the IPCC SRES are used. Each storyline was
converted into future emissions of greenhouse gases and aerosols by at least six IAMs.
Within each storyline, model runs were assumed to be equally likely. This is one
plausible approach to represent uncertainty arising from different characteristics of these
models. Uncertainty in the direction of technological change in the energy system was
also represented in the A1 scenario family and given equal likelihood again.

Chapter four showed that there are a number of climate sensitivity estimates in the
literature. The number of estimates is growing as new models are used or when the
same model is updated with more or different observed data. The sensitivity analysis
(see section 7.1.3) showed that the choice of climate sensitivity PDF does have a certain
impact on adaptation decisions so it is important to select a PDF carefully. Here the
Weighted Murphy et al. (2004) PDF is used on the basis that it is the most recent
estimate from the literature (gathered in this thesis), that it has been constrained using a
number of observational data and that it used a GCM to make the estimate.

191
The other climate parameters (carbon cycle, ocean mixing and aerosol forcing) PDFs
are all based on expert judgement from Wigley and Raper (2001). A uniform
distribution means there is a substantial uncertainty associated with this variable. For
aerosol forcing this is corroborated by a recent expert elicitation conducted by Morgan
et al. (2005). This study elicited subjective probability distributions of aerosol forcing
from a group of 24 leading atmospheric and climate scientists. The range of uncertainty
assessed by a number of experts was significantly larger than values suggested by the
IPCC TAR, ranging between -0.25 W/m2 and -2.1 W/m2. This range is only slightly
larger than the one used in this thesis (-0.58 to -1.82 W/m2; see Figure 4.8).

Uncertainty in regional climate response was represented in two ways. The first
approach assumes that all highly skilled models (skilled at representing summer
precipitation in the East of England) are equally plausible. These comprise of four
GCMs (HadCM3, CGCM2, CSM 1.3 and GFDL; see Figure 5.14) and three RCMs
(HadRM3H, PROMES and HIRHAM; see Figure 7.18) for the Ruthamford WRZ.68
The second approach takes the view that the above models come from a small
population of models (sometimes called ensembles of opportunity) and that the
uncertainty range could be much wider. This approach assumes a uniform distribution
for the regional response factor.

Finally, the uniform distribution assumed for climate impacts is rather arbitrary as there
was no information available on the hydrological model used by Anglian Water.

Figure 7.22 shows the two variants of the uncertainty analysis (skill and uniform) for
the Ruthamford WRZ. The uniform distribution is slightly flatter and more uncertain
than the skilled distribution, which shows that using the models that perform best for
todays climate can constrain the results slightly. The 95% confidence range is reduced
from 150.08 Ml/d (under the uniform CDF) to 136.11 Ml/d (under the skilled CDF). At
the 95th percentile the gap between these two distributions is 6.67 Ml/d and the AWS
shortfall to the uniform distribution is 58.93 Ml/d.

68
For the ES&E WRZ no catchment rainfall data was publicly available so rainfall data (only mean
summer precipitation) for Wattisham (1961-90) was compared to modelled results. That lead to the use of
three GCMs (PCM, CSM and ECHAM4) and six RCMs (CHRM, HadRM3H, HIRHAM, REMO,
PROMES and CLM).

192
 0.12

AWS
0.1
Natural variability
Skill
0.08
Uniform

Probability 0.06

0.04

0.02

0
-75 -50 -25 0 25 50 75 100 125 150 175
Additional water required (Ml/d)

0.9

0.8

0.7

0.6

Probability 0.5
AWS
0.4
Natural variability
0.3
Skill
0.2 Uniform

0.1

0
-75 -50 -25 0 25 50 75 100 125 150 175
Additional water required (Ml/d)

Figure 7.22 Uncertainty analysis of additional water required due to climate

change by the 2030s for the Ruthamford WRZ under SRES A2 using models with
skill and a uniform distribution for regional response. Also shown is AWS strategy
and natural variability.

With Latin Hypercube Sampling it is possible to calculate the percentage of

contribution to the variance of additional water required by the various factors modelled
in the framework. Contribution to variance was calculated by squaring the rank
correlation coefficients and normalizing them to 100%. A statistical test often applied to
determine the relationship between input and output variables is the calculation of
Spearmans rank correlation coefficient. This is a test for the presence of a monotone
relationship between the input and output variables. During the simulation rank

193
correlation coefficients are calculated between every factor of the framework and the
output (additional water required).

Table 7.9 presents the contribution of each parameters uncertainty to the overall
variance of the results. Global temperature change combines the variance from climate
sensitivity, aerosol forcing, ocean diffusivity, carbon cycle and different integrated
assessment models. The sensitivity analysis in section 7.1 showed that the major
contributors to the variance of this parameter are climate sensitivity, closely followed by
aerosol forcing. Table 7.9 should come as no surprise after reading the section on
sensitivity analysis. However, instead of looking at each parameter in turn, Table 7.9
examines the entire sequence of the modelling framework. Furthermore, it adds the
impact of natural variability, which has not been examined before. It is perhaps not
surprising that natural variability plays such an important role in the outcome. In fact, if
a different climate sensitivity PDF is used one that produces less global warming
natural variability comes out as the top contributor to the variance of water needed (not
shown). Table 7.9 shows that using a skilled distribution constrains the uncertainty in
regional response compared to using a uniform distribution.

Table 7.9 Contributions of each parameter to the variance of additional water

required in the Ruthamford WRZ by the 2030s under SRES A2.

Parameter Contribution to variance (%)

Skill Uniform
Regional response 48.4 63.5
Natural variability 36.2 27.3
Climate impacts 10.2 6.3
Global temperature change 4.6 2.5

To the knowledge of the author the only other study that has performed an uncertainty
analysis similar to this one is Jones and Page (2001). They examined the contributions
to uncertainty from global warming (same as global temperature change), local changes
in precipitation (same as regional response) and potential evaporation when assessing
the risk of climate change on the water resources of the Macquarie river catchment.
Their risk assessment is most similar in design to the uniform distribution simulation
used here, although the uncertainties examined are different. The regional response
(local change in precipitation) contribution to variance was 64%, with global warming

194
taking 25% and potential evaporation 12%. The results shown here and those by Jones
and Page (2001) demonstrate, analytically, that almost two-thirds of the uncertainty in
water resources originates from the regional response (local change in precipitation as a
function of global warming).

The only uncertainty that was represented using scenarios was world development over
the next 100 years. Figure 7.23 shows the impact of using different SRES storylines
with the skilled distribution. The gap between storylines at the 95th percentile is 14.33
Ml/d with SRES A2 producing the least water requirements and SRES B1 the highest.
As mentioned before, this gap is largely due to aerosol forcing uncertainty.

0.9
0.8

0.7
0.6
AWS
Probability 0.5
Natural variability
0.4
A1
0.3 A2
0.2 B1
0.1 B2

0
-75 -50 -25 0 25 50 75 100 125 150 175
Additional w ater required (Ml/d)

Figure 7.23 Uncertainty analysis of additional water required due to climate

change by the 2030s for the Ruthamford WRZ under different SRES storylines
(A1, A2, B1, B2) using a skilled distribution for regional response. Also shown is
AWS strategy and natural variability.

Figure 7.24 and Table 7.10 shown results for the ES&E WRZ. Here the skilled
distribution is much poorer at constraining uncertainty in the results. At the 95%
confidence range, uncertainty is reduced from 32.80 Ml/d (under the uniform CDF) to
32.31 Ml/d (under the skilled CDF). What the constraining exercise does is it shifts the
distribution to the right, at the 50th percentile by 2 Ml/d.69 The gap between the skilled
and the uniform distribution at the 95th percentile is negligible at 0.23 Ml/d and the
AWS shortfall compared to the skilled distribution is 7.77 Ml/d.

69
This is similar to what happens to the Murphy et al. (2004) climate sensitivity distribution once it is
weighted by the Climate Prediction Index.

195
 1

0.9 AWS
Natural variability
0.8
Skill
0.7
Uniform
0.6

Probability 0.5

0.4

0.3

0.2

0.1

0
-20 -15 -10 -5 0 5 10 15 20 25 30 35
Additional water required (Ml/d)

Figure 7.24 Uncertainty analysis of additional water required due to climate

change by the 2030s for the ES&E WRZ under SRES A2 using models with skill
and a uniform distribution for regional response. Also shown is AWS strategy and
natural variability.

In terms of contributions to the variance of additional water required the results are
fairly similar to the Ruthamford WRZ.

Table 7.10 Contributions of each parameter to the variance of additional water

required in the ES&E WRZ by the 2030s under SRES A2.

Parameter Contribution to variance (%)

Skill Uniform
Regional response 53.6 61.8
Natural variability 31.1 29.2
Climate impacts 10.2 5.0
Global temperature change 4.3 3.0

Using best available knowledge, this uncertainty analysis has showed analytically that
AWSs WRP remains robust to the climate change uncertainties considered. This means
that if the aim is to maintain current levels of service over the next 25 years, the
adaptation measures contained in the WRP will suffice. This conclusion is, of course,
conditional on the subjective assumptions, PDFs, scenarios and modelling framework
used. However, the sensitivity analysis (section 7.1) showed that even under different
assumptions, PDFs and scenarios, the WRP remained robust.

196
This occurs for two main reasons. AWS planned for climate change using the UKCIP02
scenarios, which were constructed using one of the driest models (HadCM3) amongst
the range examined here. This makes the WRP robust to future droughts. Also
important is the fact that AWSs adaptation strategies involve major increases in supply
(reservoir extension, water treatment works extension, a potential new reservoir, etc.)
which make them robust to climate change uncertainties.

There are, however, certain combinations of climate change uncertainties that make
AWSs WRP insufficient to maintain levels of service. For example, if very high
climate sensitivity is combined with low aerosol forcing and large decreases of
precipitation per degree of global warming the current WRP would be insufficient to
maintain levels of service. Under the uncertainty analysis this sort of combination is
given a very low probability because of the assumptions taken. Under the sensitivity
analysis, uncertainty about climate sensitivity and regional climate response (in
particular GCM uncertainty) almost render the WRP insufficient under the default
scheme, but it still remains robust. The next section examines the possibility of this sort
of combinations using a robust decision-making framework.

7.3 Robustness analysis

In this section an entirely different paradigm to the previous two sections is

investigated. Instead of trying to asses how likely climate change will be and what
adaptation actions should be pursued (the predict-then-act approach, cf. Lempert et al.
2004), it examines the robustness of adaptation strategies without having to first assess
how likely they are.70 That is, it seeks robust strategies amongst deep uncertainty,
without using probabilities. The key factor here is the presence of deep uncertainty,
sometimes called Knightian uncertainty, after Knight (1922), who distinguished risk
from uncertainty. For Knight (1922) risk is a realm where factors can be quantified
whereas uncertainty denotes unknown factors poorly-described by quantifiable
probabilities. Lempert et al. (2003, 2005) have defined deep uncertainty as the condition
where analysts do not know or the parties to a decision cannot agree upon: (1) the
70
Notice that this is different from the concept of robustness of water resources systems as applied by
Hashimoto et al. (1982): a measure of the likelihood that the actual cost of a proposed project will not
exceed some fraction of the minimum possible cost of a system designed for the actual conditions that
occur in the future.

197
appropriate models to describe interactions among a system's variables, (2) the
probability distributions to represent uncertainty about key parameters in the models, or
(3) how to value the desirability of alternative outcomes. Climate change has all these
characteristics and they are all brought to the foreground when it needs to be
incorporated into water resource planning for the next 25 years. The other reason to go
for robustness over prediction is a matter of belief and world view.

Here the view (belief) is taken that both climate change and demand are much more
uncertain than that portrayed in previous sections. Climate change uncertainties are
believed to be extremely difficult to quantify and the previous sections have only
explored a fraction of the plausible uncertainty space. Similarly, it is believed that
AWS demand projections are in fact much more uncertain than what the water
company thinks. These circumstances (beliefs) make it very difficult (perhaps
impossible) for Anglian Water to perform a risk assessment on the security of public
water supply over the next 25 years. This is where a robust decision-making framework
becomes more useful.

The application of a robustness framework is demonstrated for the ES&E WRZ. In

order to characterise uncertainty in demand the Environment Agency scenarios for the
Anglian region were used (EA 2001b).71 Demand changes under these scenarios are
shown in Table 7.11.72 Under the Alpha scenario (Provincial Enterprise and SRES A2)
water demand rises substantially over the next 25 years. In the Beta scenario (World
Markets and SRES A1) the rise is smaller. Under both the Gamma (Global
Sustainability and SRES B1) and Delta (Local Stewardship and SRES B2) scenarios
water demand decreases over then next 25 years. These estimates show there is
substantial uncertainty surrounding future changes in demand in the Anglian region.73

71
These scenarios are based on the Foresight scenarios (DTI 1999, OST 2002) and therefore have
correspondence with the UKCIP socio-economic scenarios (UKCIP 2000) and with the SRES scenarios
(Nakicenovic et al. 2000).
72
Values of percent changes were shifted for 5 years (from 2010 and 2025 to 2015 and 2030) to match
the analysis performed here.
73
For other examples on the presence of substancial uncertainty in future water demand estimates see
Groves et al. (2005) for a case study on California, US.

198
Table 7.11 Demand changes (%) in the future according to the Environment
Agency scenarios for the Anglian region (EA 2001b, a).

2015 2030
Alpha 8% 86%
Beta 10% 22%
Gamma 0% -26%
Delta 0% -25%

In order to keep climate change uncertainties manageable for display only two variables
were used: uncertainty in summer precipitation change and uncertainty in climate
impacts. The first is composed of regional climate response uncertainty, natural
variability and global temperature change so it encompasses over 90% of the variance in
outcome according to the uncertainty analysis (see section 7.2). Climate impacts
represents uncertainty in the hydrological modelling, which contributes between 5-10%
of the variance in additional water required.

For regional response the plausible range extends from an increase in summer
precipitation by 25% to a decrease of 50%. Natural variability alone can explain 10%.
If one combines very high climate sensitivity (e.g., 10.8C) with a GCM that shows
wetter summers (e.g., PCM under SRES B2), values above 20% increase in
precipitation can be found. The same occurs for the other direction (drying) using, e.g.,
HadRM3. The climate impacts range of uncertainty (50%) cannot be justified as
scientifically because detailed modelling was not performed. In part this uncertainty
range represents the authors scepticism of modelling studies that have not performed
sensitivity analysis.

Figure 7.25 shows the performance of AWS WRP for the ES&E WRZ against climate
change uncertainties under AWSs demand projections. The critical threshold for this
WRZ is when additional water required goes beyond 25 Ml/d.74 That means that the
WRP is robust over 65% of the uncertainty space. The plan remains vulnerable to large
precipitation decreases (over 25%) and to hydrological modelling uncertainties. If the
current WRP would be supplemented with an Aquifer Storage Recovery (under the
Felixstowe peninsula) and a desalination plant, AWS robustness would rise to 90% of

74
That includes all the uncertainties taken into account by AWS in their WRP (demand, borehole
pollution, etc.), the uncertainty introduced by climate change and the available headroom in 2030 (11
Ml/d).

199
the uncertainty space. Full elimination of risk would only occur through transfers from
the river Trent or an East Anglian reservoir under the AWS demand projections.

Under the Alpha scenario there are large increases in demand, which render AWSs
WRP insufficient to maintain levels of service under the entire uncertainty spectrum.
Transfers from the river Trent or an East Anglian reservoir would be required to keep
the plan robust over most of the uncertainty space. Another option would be to decrease
the levels of service.

The Beta scenario displays a moderate increase in demand which leaves AWSs WRP
vulnerable to almost 90% of the uncertainty space. An Aquifer Storage Recovery and a
desalination plant would increase robustness to almost 50% of the uncertainty space.
Transfers from the river Trent or an East Anglian reservoir would eliminate any risk
from the uncertainties considered under this scenario.

Finally, the Gamma and Delta scenarios see similar decreases in demand which render
AWSs WRP robust across much of the uncertainty space. A supplement to the WRP
such as an Aquifer Storage Recovery and a desalination plant would almost eliminate
the risk entirely.

200
 AWS

-50

-25

Climate impacts
0
uncertainty (%)

25

50
25 20 15 10 5 0 -5 -10 -15 -20 -25 -30 -35 -40 -45 -50
Summer precipitation change (%)

Figure caption
-75--50 -50--25 -25-0 0-25 25-50 50-75 75-100
on page 203
Alpha

-50

-25

Climate impacts
0
uncertainty (%)

25

50
25 20 15 10 5 0 -5 -10 -15 -20 -25 -30 -35 -40 -45 -50
Summer precipitation change (%)

-75--50 -50--25 -25-0 0-25 25-50 50-75 75-100 100-125 125-150 150-175 175-200 200-225 225-250 250-275

201
 Beta

-50

-25

Climate impacts
0
uncertainty (%)

25

50
25 20 15 10 5 0 -5 -10 -15 -20 -25 -30 -35 -40 -45 -50
Summer precipitation change (%)
Figure caption
-75--50 -50--25 -25-0 0-25 25-50 50-75 75-100 100-125 125-150
on page 203

Gamma/Delta

-50

-25

Climate impacts
0
uncertainty (%)

25

50
25 20 15 10 5 0 -5 -10 -15 -20 -25 -30 -35 -40 -45 -50
Summer precipitation change (%)

-75--50 -50--25 -25-0 0-25 25-50 50-75

202
Figure 7.25 Additional water required (in intervals in Ml/d) in order to maintain
levels of service by the 2030s in the ES&E WRZ under different demand scenarios
(AWS projections and Environment Agency scenarios: Alpha, Beta, Gamma and
Delta) as a function of regional climate response uncertainty (represented by
summer precipitation change in the horizontal from a 25% increase to a 50%
decrease) and climate impacts uncertainty (represented by hydrological modelling
uncertainty in the vertical from -50% to 50%).

7.3.1 Summary

This robustness analysis has shown that the adaptation decisions being planned by
Anglian Water are not entirely immune to climate change uncertainties, as the
predictive/probabilistic approach in section 7.2 would lead one to believe. This
conclusion also applies to AWSs projections of demand. If demand is considered more
uncertain than assumed by AWS (e.g., by using the Environment Agencys demand
scenarios) then the robustness of the WRP changes dramatically. Under the Alpha and
Beta scenarios the WRP is weak at maintaining the level of service. Even without
climate change considerable more measures would be required. In a Gamma or Delta
scenario the WRP is slightly more robust than with AWSs projections of demand.

With this approach to decision-making it is possible to identify the vulnerabilities of the

current WRP and suggest further strategies that will overcome these vulnerabilities. If
water managers at AWS are adamant that their demand projections are correct and are
worried about potential large summer drying (<-25%) then further adaptation strategies
would be required. An Aquifer Storage Recovery, a desalination plant and some
additional demand management measures would make the WRP robust across a large
range of climate change uncertainties. If water managers have doubts about their
demand projections things get more complicated. In two scenarios the WRP is fairly
robust to uncertainties, while in the other two the WRP is very weak. If robustness to
uncertainty is the sole indicator being used to judge successful adaptation then it is clear
that further measures such as transfers from the river Trent or an East Anglian reservoir
would be required. If issues such as costs, social acceptability, environmental impacts,
and many others are used to define successful adaptation, then a trade-off between
robustness to uncertainty and these various issues would be required.

203
 CHAPTER 8. DISCUSSION AND CONCLUSION

The previous chapters have demonstrated the application of a climate risk assessment
framework to an adaptation decision context. In this chapter the limitations and
implications of this work are discussed in more detail. The first section examines the
limitations of the modelling framework used in this thesis and elaborates on some of the
current constraints in water planning imposed by the regulatory environment on some of
the results. Section 8.2 examines the implications of this work for the regulators, water
companies, future climate scenario construction and adaptation planning. It also
includes some reflections of the work conducted in the context of post-normal science
and interdisciplinarity. Section 8.3 describes future research questions and section 8.4
provides some concluding remarks.

8.1 Limitations

This section examines some of the limitations of the work conducted here. The first
section discusses some of the linear assumptions made throughout the thesis. Some of
the assumptions made in the climate-impacts interface are discussed in section 8.1.2.
Section 8.1.3 discusses how some of the conclusions from this work (regarding
adaptation planning) are limited by the current regulatory environment.

8.1.1 Linear simplicity versus non-linear complexity

The modelling framework applied in this thesis can be characterised as mostly linear.
This emanates from using a simple energy balance model, the pattern-scaling technique
and linear transfer functions for climate impacts. There are various advantages and
disadvantages to such a modelling framework. Perhaps the most obvious and pertinent
advantage of this approach, compared to more complex approaches, is the ability to
perform sensitivity analysis. It is impossible as yet to perform global sensitivity analysis
in complex non-linear GCMs in order to determine which parameters dominate the
models response to external forcing. While the scientific community may have an idea
of where the largest uncertainties reside (in cloud modelling according to the survey of
Morgan and Keith 1995), this has never been shown analytically through sensitivity

204
analysis. With the advancement of computational power this might become a reality one
day, but it is currently unavailable. The possibility of conducting sensitivity analysis is
clearly an advantage of simple climate models (as applied here) and intermediate
complexity climate models (e.g., the MIT 2D statistical-dynamical climate model of
Webster et al. 2003) over more complicated GCMs. For instance, no climate modelling
centre in the world has ever run their GCMs with SRES scenarios other than the marker
scenarios. Section 7.1.2 showed that the inclusion of scenarios from the same storyline
(in this case the A2 storyline) other than the marker scenario introduced an uncertainty
of 3.29 Ml/d for water managers (for the ES&E WRZ). While this might be small
compared to the adaptation measures being planned it is difficult to argue that this is
negligible. Currently there is not enough computational power to allow a GCM to
perform such a sensitivity test or maybe it is rather just too expensive of computer time.

Once the major controlling factors of the modelling framework have been determined
(section 7.1.7), it is relatively straightforward to conduct an uncertainty analysis (section
7.2) using a simple linear framework. Limited uncertainty analyses are now being
carried out using more complex GCM (Murphy et al. 2004, Stainforth et al. 2005).
However, these two experiments are not based on what could be called objective
sensitivity analysis, which allows the identification of the most influential parameters of
a certain output variable. Due to the impossibility of performing sensitivity analysis on
these types of models, expert judgement was used to decide which parameters or
schemes to sample as uncertain and how wide the uncertainty range should be. The
advantage of the framework applied here is that it is possible to determine the most
influential elements of the modelling framework from a decision-making perspective.

The fact that the modelling framework is linear and simple allows the performance of
sensitivity and uncertainty analysis, but that is also the source of its major limitations.
For example, the energy balance model MAGICC can emulate the behaviour of more
complex GCMs, but it cannot mimic the variability and other non-linearities that GCMs
display. Pattern-scaling assumes there is a linear relationship between global mean
temperature and local climate variables. This relationship has been shown to hold with
one climate model (HadCM2, see Mitchell et al. 1999, Mitchell 2003) but it is unknown
whether it holds throughout the uncertainty spectrum explored here. Nonetheless, the
pattern-scaling technique is in widespread use in the construction of climate change

205
scenarios. The entire UKCIP02 scenarios were pattern-scaled temporally (to obtain
results for the 2020s and 2050s) and between scenarios (A1FI, B1 and B2; or high,
medium-low and low) from the average of three runs of HadRM3 for the 2080s under
SRES A2. In fact, all previous UK national climate change scenarios have used pattern-
scaling in some way or form (section 5.2). The same has been done in other countries
(see CSIRO 1992, 1996, 2001 for Australia) and also at the global level (Schlesinger et
al. 2000, Ruosteenoja et al. 2003). This has led various impacts studies (e.g., Jones and
Page 2001, Arnell 2004b) to base their climate information about the future on pattern-
scaled results.

Probabilistic pattern-scaling, as applied in this thesis, has also been applied before
(Jones 2000a, New and Hulme 2000, Ekstrm et al. 2005, Hingray et al. 2005), but its
limitations have never been fully discussed. This type of pattern-scaling essentially
treats the various GCM/RCM simulations as a super-ensemble. It assumes that all
GCM/RCM simulations are the realisation of one super model, of which each
simulation displays a certain plausible behaviour (say an x global warming with a y
precipitation change). In doing so, it concatenates the cultural and structural
uncertainties involved in building GCMs. In the sensitivity tests carried out in this
thesis, much wider uncertainty ranges and many more factors were included than in any
previous efforts (cf. Jones 2000a, New and Hulme 2000, Ekstrm et al. 2005, Hingray
et al. 2005). This allows the combination of low aerosol forcing with high climate
sensitivity with large precipitation decreases per degree of global warming, which lead
to considerable drying over the East of England in the next 25 years. While the
uncertainty analysis performed here gives this sort of combination a low probability, it
is difficult to speculate whether a GCM could indeed exhibit such behaviour. The linear
assumption made in probabilistic pattern-scaling will have to be verified when
thousands of GCM simulations become available in the coming years. There is some
preliminary evidence from Murphy et al. (2004) that pattern-scaling of seasonal surface
air temperature is fairly robust around the world, whereas pattern-scaling of seasonal
precipitation is better for the extra-tropics than for the tropics (this evidence arises from
comparing the standard deviation of pattern-scaled results75 with result using a
perturbed physics ensemble approach).

75
Their pattern-scaling methodology is different from the one applied here so results are not entirely
compatible.

206
There is also some preliminary evidence that the relationship between the scaler and the
response pattern might be exponential for areas that show drying according to six
GCMs (Osborn and Wallace 2005). The modelling framework applied here is
sufficiently flexible to take this new information into account. With respect to RCM
simulations, temporal pattern-scaling was a necessity since outputs were only available
for the 2080s and the decision-makers engaged here only plan for the next 25 years.

Even with the advancement of the perturbed physics ensemble approach (Murphy et al.
2004, Stainforth et al. 2005) and further computational power availability, pattern
scaling will remain a necessity for years to come in order to sample the many plausible
emissions scenarios. Furthermore, there are some who argue that GCMs play too central
role in climate change research when there should be a plurality of models of different
complexity (Shackley et al. 1998, Petersen 2000). The framework explored here tries to
bridge the gap between the simple climate models and the complex GCM/RCMs using
the pattern-scaling technique.

In summary, due to computational power restrictions a trade off between being able to
perform sensitivity/uncertainty analysis and model complexity is required. The simpler
the modelling framework is, the easier it is to conduct sensitivity/uncertainty analysis.
The more complex the modelling framework is, the harder it is to conduct
sensitivity/uncertainty analysis.76 The modelling framework applied in this thesis is
somewhere in the middle where sensitivity/uncertainty analysis of intermediate
complexity is still possible to conduct (with the computational power of a contemporary
PC) and relevant to decision-making.

8.1.2 Climate impacts interface

Traditionally, observed climate series have been perturbed with climate change
scenarios to serve as input to hydrological models to determine, for example,
streamflow changes or reservoir yield changes in the future (see, e.g., Arnell and
Reynard 1996, Boorman and Sefton 1997). AWSs hydrological model was not made

76
It is also worth pointing out that the simpler the framework the easier it is to communicate uncertainties
to decision-makers and stakeholders interested in this information.

207
available so in order to establish a relationship between changes in the climate and
reservoir yield a combination of expert judgment and modelling results was used (see
section 6.6). Based on this information a linear transfer function was established (Jones
and Page 2001 apply a similar methodology).77 This is, of course, a limitation as the
model might not behave linearly (cf. Arnell 2000 on the impact of thresholds).
However, assuming the linear relationship was uncertain (see section 7.1.6) makes this
less of a limitation.

Another limitation of this study is that only changes in average seasonal summer
precipitation for a thirty-year period centred on the 2030s was examined. This was a
limitation brought about by using the pattern-scaling technique and the desire to manage
uncertainties consistently. This means this thesis is silent about changes in inter-annual
variability and extremes. It essentially assumes that the transfer function for climate
impacts provides this information. If the variance of extremes changed significantly, for
example with a substantial increase in dry spells, then the results in this thesis would
represent an underestimation of risk for AWS.

Finally, the inclusion of natural multi-decadal climate variability in most of the results
might also be considered a limitation because the GCM/RCM simulations could contain
some natural signal (i.e., they do not represent solely the anthropogenic forcing). This
means the results presented here could be overestimating the variance of additional
water required. Furthermore, anthropogenic forcing may alter the character of multi-
decadal natural climate variability (Mearns et al. 2001). Investigating this issue, which
is likely to be small, would require delving into the detection and attribution literature
(see, e.g., Lambert et al. 2004). There might also be a problem of double counting with
the inclusion of natural variability (as done in this thesis), because the WRP is based on
figures (of deployable output, demand, etc.) from a dry year. It will be important to
find out where in the natural variability PDF this dry year lies and perhaps only
include the estimates beyond this point in the assessment.

77
Too late in the preparation of this thesis, it became apparent that changes in potential evaporation had
also been included in AWSs hydrological modelling. The analysis was re-run with reservoir yields being
dependent on summer precipitation change and summer potential evaporation change simultaneously.
The result was slightly less additional water being required (compared to only using precipitation change)
because potential evaporation estimates (calculated using the Thornthwaite method) were much lower
than the UKWIR (2003) estimates, which is consistent with other findings (see Diaz-Nieto and Wilby
2005).

208
8.1.3 Adaptation planning

A few companies have identified serious impacts from climate change. We welcome
preliminary analysis that these companies have carried out. We have agreed with the
water industry that these analyses are not on their own sufficient to justify significant
changes to investment programmes. We expect all companies that have identified
significant impacts from climate change to carry out more detailed analysis over the
next few years. (EA 2004a, p. 20)

One of the major challenges of this thesis was to link climate change uncertainties to
real world decision-making. This was achieved in previous chapters. However, it is
important to acknowledge that the decisions being contemplated (in particular those
beyond 2015) are not actually being implemented or perused yet (e.g., Alton WTW
extension or a new East Anglian reservoir) as the quote above explains. As discussed
later in more detail, there is currently not enough certainty or evidence to justify
investment in water resources if climate change is the main driver.78 This is a limitation
in terms of the realness of these adaptation measures, but in the opinion of the author
this is as real as it gets in terms of adaptation planning. What this means, and without
wanting to introduce yet another layer of uncertainty, is that some adaptation decisions
(e.g, the ones planned for 2005-2010) are more real (certain) than others (e.g, those
planned beyond 2015). From an analysis perspective this is irrelevant since no
distinction has been made between different types of adaptation measures in this thesis,
but if a reality check is performed this means AWSs WRP might not be robust to
climate change because the regulator has not approved the measures included in the
WRP. In fact the Environment Agency is already concerned that target headroom is too
high in certain WRZs (e.g., ES&E and Lincolnshire Fens WRZ) (EA 2004a).

Another limitation that was observed was the short list of adaptation options almost
exclusively focused on supply increases. This was also noted by the Environment
Agency:

78
Mainly because it is perceived that modelling work based on UKCIP02 only represents a small fraction
of the potential uncertainty (because it only includes one model). Regulators are aware that HadCM3 is
amongst one of the driest models reviewed in UKCIP02 for the UK, so there are probably concerns about
overinvestment.

209
Most companies that have identified a significant impact from climate change seem to
intend to deal with this by developing new water resources and, in particular,
reservoirs. We understand that these are preliminary assessments of need in the late
2020s but we do expect companies to follow Government guidelines and consider a full
range of possible responses to climate change rather than only the more obvious
engineering options. (EA 2004a, p. 20)

This was certainly the case for AWS who focused predominantly on engineering
options such as water treatment works extensions and a potential new reservoir for
2020. Planned demand management strategies are virtually insignificant when
compared to the planned increases in supply. In the Ruthamford WRZ demand
management is expected to make savings of 1.55 Ml/d by 2030, whereas supply is
expected to increase by 138.94 Ml/d (i.e., demand savings are just over 1% of supply
increases). For the ES&E WRZ demand management is estimated to make savings of
0.44 Ml/d and supply is expected to increase by 23 Ml/d (i.e., demand savings are
almost 2% of supply increases). The Government has emphasised on numerous
occasions its preference for a twin track approach of demand management and
development of resources to achieve sustainable management of water resources (see,
e.g., DEFRA 2002). It is unclear whether engineering options are preferred by the water
industry as they have become part of its culture. From the elite interviews conducted,
water managers gave the impression that it was difficult for water companies to pursue
demand management. It is impossible for an analyst outside the field, such as the author
of this thesis, to judge the appropriateness of the adaptation options considered by
AWS, but it is his opinion that the current list of options is restrictive from a decision-
making viewpoint. There are a few reasons why this might be the case.

Anglian Water Services (AWS) Limited is a regulated water, wastewater and

environmental service provider for domestic and industrial customers in eastern
England.79 AWS is also a business company that aims to make profit. Therefore, any
decisions about adapting to climate change have a business interest, and thus are subject
to some confidentiality. This was evident in the elite interviews conducted where not
many of the plans were disclosed to the interviewer. This is perhaps not surprising as

79
And Hartlepool

210
there is commercial interest in the plan and it was the author who contacted the water
company (not the other way around). According to Conde and Lonsdale (2004), the
level of stakeholder engagement reported in this thesis is of the lowest sort, somewhere
between participation in giving information (people are involved in interviews or
questionnaire based extractive research) and participation by consultation (asking for
views on proposals and amending them to take these views into account). If higher
levels of engagement were pursued (joint analysis and joint action planning; or the
decision-maker taking the initiative) perhaps more adaptation options could have been
identified. It is also possible that all their options were not reported in the WRP, but
were considered in their cost/benefit exercise of options (FORWARD; see section 6.5),
thus only reporting on the least cost options. Finally, the WRP and the elite interviews
gave the impression that much more thought has been given to near term options
(especially 2005-10, but also 2010-15) than to the longer term (beyond 2015).80 The
short list of adaptation options is particularly problematic for the robustness approach
(section 7.3) as this method appears most useful in situations where the menu of
alternative options is sufficiently diverse for decision-makers to design strategies
(Lempert et al. 2005). If indeed the adaptation options available to water managers are
as limited as AWS describes in their current WRP, then robustness to uncertainty is
synonymous with bigger water resources development (according to section 7.3).

Eliciting adaptation options has been a convoluted exercise of elite interviews with
experts in the field, in-depth examination of water resource plan documents and
assessment of grey and peer-reviewed literature. The short list gathered in this thesis
represents a limitation on the work carried out, in particular for the robustness analysis.
Future work should aim to gather a larger set of adaptation options without losing sight
of the context (spatial, temporal, cultural, etc.) where they would be implemented. This
would have to be done in collaboration with water companies and other experts in the
field.

80
This is rather problematic since it might be more cost-effective to go for a larger adaptation measure
soon rather than for several smaller adaptations later. But this depends on the discount rate used and the
adaptation options being considered so it is not that simple.

211
8.2 Policy and practical implications

This section discusses the policy and practical implications of the results obtained in
this thesis for the regulatory system, the water industry, climate scenario construction
and adaptation planning. There is also a reflection on the work accomplished in section
8.2.5.

8.2.1 Regulatory system: prediction, investment and

organisational culture

It is neither possible nor sensible to set out detailed requirements to cater for these
eventualities [impacts of climate change on water services] now. Gaining greater
certainty in these areas is a priority for the next few years We believe that the action
proposed for the period 2005-10 will prepare the industry to cope with changes within
the range of possibilities forecast without undertaking inappropriate and unnecessary
investment. (Ofwat 2004, p. 43)

The quote above by the economic regulator (Ofwat) matches closely the previous quote
by the environmental regulator (Environment Agency) in section 8.1.3. This shows that
the regulators of water resources are in synchrony, but is that really the case when it
comes to climate change? This section examines the implications of this thesis for the
water regulatory system. The above excerpt from Ofwat, in its water price review,
justifies single-handedly the being of this thesis, in particular of the uncertainty analysis
conducted in section 7.2. It also emphasises the preference of this particular regulator
for prediction approaches (notice the words greater certainty and forecast in the quote).

One of the major imperatives of the periodic review of water price limits (and by
association water resources planning since that is the source of its funding) is that water
services remains affordable. In the words of the Government: The whole programme
must be carried out at the minimum cost to customers (DEFRA 2004, p. 5). In order to
achieve this, regulators appraise water companies measures (contained in business plans
and WRPs) in order to determine whether individual projects are worthwhile and
represent value for money; and whether the quality programme, taken as a whole,

212
represents a cost-effective package of measures for achieving the desired
improvements (DEFRA 2004, p. 16). Therefore, two core criteria in water resources
decision-making are cost-effectiveness and value for money, which in academic terms
are known as maximising expected utility. This is a culture that values efficiency and
effectiveness over robustness (to uncertainty). A logical interpretation of the quote
above is that in order to invest efficiently, forecasts (or predictions) are required. In the
case of climate change enough certainty is not yet available so investment for this
eventuality is not deemed sensible. This puts much emphasis on the prediction approach
and in quantifying climate change uncertainties in order to gain greater certainty.

The approach taken in this thesis has advanced this agenda, but it is unclear, however,
what exactly greater certainty means in climate change terms. Does it simply mean
moving from scenarios (like UKCIP02) to probabilities (as in section 7.2) or does it
mean using many models rather than just one model (HadRM3)? Or is it about moving
from a 95% confidence range of additional water required between -10 and 50 Ml/d to a
narrower range of 10-20 Ml/d? The work carried out here has tackled the first question
by managing uncertainty explicitly and by taking a multi-model ensemble approach.
This thesis made little advancement on the second question partly because of the small
number of model simulations available, which makes constraining the results with
observations almost negligible (see section 7.1.5). However, this should not detract
from the fact that 95% confidence ranges of climate change risk were estimated here
with a consistent treatment of uncertainty. The uncertainty analysis (in section 7.2)
used best available information to estimate the probability of additional water required
for two WRZs in the Anglian region by the 2030s. At the 95% confidence level the
bounds of additional water required ranged from -6.67 to 26.34 Ml/d for the ES&E
WRZ and from -24.30 to 111.81 Ml/d for the Ruthamford WRZ under SRES A2 and
using the most skilled models (with equal weights). That represents an uncertainty range
of 33.01 Ml/d (33% of demand) for the ES&E WRZ and 136.11 Ml/d (39% of demand)
for the Ruthamford WRZ. These uncertainty ranges are extremely large from an
adaptation/investment perspective.

Is this too uncertain? How much certainty is necessary for investment? It is likely that
the regulators themselves do not have a clear idea of how much certainty is required to
justify investment as adaptation to climate change has never been done before (at least

213
not explicitly under that banner). This puts prediction approaches at the centre of
decision-making as the quotes above illustrate; no investment on adaptation to climate
change will go ahead until x amount of certainty is achieved in the eyes of the regulator.
A vicious circle could easily be established where the regulator asks climate modellers
for more certainty in their projections; climate modellers spend several years
researching and the result is even more uncertainty; the regulator asks for further
certainty and so on. In the meantime potentially important adaptation decisions are
being delayed because of the lack of certainty in future climate information. This
epitomises the problem with the predict-then-act approach (cf. Lempert et al. 2004) and
the importance of being able to make decisions under severe uncertainty (this is where a
robustness approach becomes more useful).

The culture of efficiency and optimisation that dominates the economic regulator filters
down into water resources decision-making and creates a need for predictions or
forecasts of climate change. When predictions are not available, users (Atkins and
AWS) have already made up their own PDFs of the impact of climate change, proving
Schneiders (2002) point that if experts do not provide this information, users will create
their own frequency distribution of future climate impacts or as Pittock et al (2001) put
it, planners will take a gamble. However, coming up with PDFs of climate change and
its impacts is not straightforward as shown by chapters 4-7. Furthermore, the sensitivity
analysis showed that the amount of additional water required is particularly sensitive to
uncertainties in regional climate response and downscaling.

Ofwats planning horizon is essentially five years (2005-2010, with some consideration
till 2015), which makes any consideration of investment for a long term problem, like
climate change, particularly difficult, especially if adaptation strategies take time to be
implemented. Ofwat (2004) argues that it has a long term perspective, but the price
review forces it to concentrate on the next five years. Unless more emphasis is given to
the long term by Ofwat, it is difficult to envisage how any investment on adaptation to
climate change could take place today or in the near future.

The discourse of the environmental regulator, the Environment Agency, is rather

distinct from Ofwats. It focuses on a range of scenarios (instead of forecasting) and in
testing the sensitivity, robustness and flexibility of strategies (see Box 8.1). This

214
approach is consistent with the analyses performed in sections 7.1 and 7.3, in particular
with the robustness analysis. What is not consistent is where it states that a companys
final plans should be based on a single planning scenario. This violates the robustness
approach in its entirety as it assumes there is a scenario that is more likely than other
scenarios and instead of seeking robust strategies across the various scenarios, the
strategy is optimised for this single scenario. In the case of Anglian Water, and probably
many other water companies, there was little evidence of multiple scenarios being used.
The final planning scenario was essentially a disguised forecast for which strategies
were optimised. So while robustness might be an aspiration, optimisation is what comes
out in practice.

Could the robustness analysis demonstrated in section 7.3 be applied to water resources
planning in practice? In principal yes, but it would certainly increase the laboriousness,
complexity and computational effort of an already lengthy process. Furthermore, one of
the major difficulties with robustness analysis is to define how much of the uncertainty
space to sample (e.g., +25 to -50% of summer precipitation). In this thesis the ranges
were conveniently informed by the sensitivity analysis, but that may not always be
available to the analyst. The robustness analysis showed that under AWS demand
projections, the WRP remained vulnerable to certain combinations of climate change
uncertainties. In order to become immune to uncertainties the current WRP would have
to be complemented by a few additional measures. Under different demand scenarios
the robustness of the WRP ranged from being totally inadequate at maintaining levels of
service (under the Alpha and Beta scenarios) to being almost immune to all climate
change uncertainties (under the Gamma and Delta scenarios). The WRP is clearly
vulnerable to uncertainties in demand, at least as portrayed by the Environment Agency
scenarios. But how uncertain is demand? Could it really range from -25% to +86% of
todays values over the next 25 years? And climate change, how uncertain is that?

215
 Box 8.1 Sensitivity testing and developing the final planning
scenario
In developing the final plan, companies should consider a range of planning scenarios to
test the sensitivity, robustness and flexibility of their preferred strategy in maintaining
security of supply in view of the risks and uncertainties within, and outside the final
planning scenario.

At its simplest, upper and lower scenarios about a chosen final or best estimate dry year
planning scenario should be presented as part of the final plan. These upper and lower
limits could be viewed as demonstrating a risk envelope against which companies
should derive a balanced portfolio of (supply and demand management) options. The
UKWIR / Environment Agency report Economics of balancing supply and demand
provides a number of alternative, more detailed approaches for evaluating more complex
supplydemand situations; most companies should be able to use the intermediate
methodologies referred to within the report. The choice of approach is for each company
to determine. However, we expect the approach adopted to relate to the size of the
perceived supplydemand imbalance, the complexities involved in determining a solution
(for example the range of options, zonal connectivity) and the size of the company.

Companies should explore different scenarios of future growth, consumption and demand
management before arriving at a single planning forecast for the final plan and Strategic
Business Plan. For example, companies will need to assess different assumptions about
the pace and type (that is, optional versus selective) of metering or alternative options to
further reduce leakage. In developing different growth scenarios companies could
consider the risks and opportunities of possible changes in political, social, economic and
technological development. Assumptions should be fully described in supporting text.

In evaluating sensitivity, companies will need to consider the phasing of options within
their preferred strategy, the requirement for further options, and any consequent impact
on costs of the preferred strategy, in addition to the assessment of risk and uncertainty of
the individual options, to assess supply-side sensitivity and risks.

Although sensitivity testing is essentially concerned with capturing risk to security of

supply in view of uncertainties in available information, companies will be expected to
provide robust evidence in support of their assessment of risk. Companies will need to
demonstrate that they have taken steps to quantify risk where appropriate, and in so doing
reduce the level of uncertainties within proposed strategies. For example, on supply-side
options companies assessments should draw on feasibility studies, experiences
elsewhere, pilot schemes etc. On the demand-side, companies assessments should be
supported by customer surveys, consumption monitors, extensive metering of demand
(from customer and DMA through to zonal metering) and measures to reduce the
reconciliation factors in the water balance.

Whilst companies final plans should be based on a single planning scenario, the plan
should demonstrate how the chosen strategy is robust and flexible to the risks and
uncertainties identified.

Source: Environment Agency (2003b, p. 20)

216
The different organisational cultures of the two main regulators, who have different
mandates and objectives, appears to lead them to handle uncertainty differently. Ofwat
is mainly concerned with the short term (next five to ten years) and aims to optimise
(least cost schemes) which leads to a culture of forecasting (i.e. quantifying
uncertainties regarding the future) within a risk framework. On the other hand, the
Environment Agency is concerned with the longer term (next 25 years) as well as the
short term, and acknowledges inherent uncertainties and thus prefers a scenario
approach that seeks robustness and flexibility in strategies. However, it seems to cave
into the pressure of the water price review by suggesting the use of only one final
planning scenario. Ofwat seems to work in the realm of risk assessment (where
probabilities can be quantified) whereas the Environment Agency appears to
acknowledge that uncertainties are less tractable. This leads to different approaches to
the inclusion of climate change in water resources planning.

This thesis has demonstrated both approaches, the prediction approach (section 7.2)
which works within risk assessment, and the robustness approach (section 7.3) which
works under deep uncertainty. Whether these differences are a matter of timescale
(short versus long term view), organisational culture or different beliefs and worldviews
needs to be investigated further. Believers that the future is unpredictable might prefer
the robustness approach, but if one believes that the future is knowable then the
prediction approach might be favoured. It seems fair to ask whether current water
resources planning is done under the illusion of risk assessment, when in fact things
could be much more uncertain. This goes deep into the nature of risk (cf. MacGill and
Siu 2004, Macgill and Siu 2005) and in particular how different actors in water
resources management perceive climate change risk. Figure 8.1 shows how different
approaches are taken to deal with uncertainty as a function of knowledge about
likelihoods and outcomes. Whether one is in the deep uncertainty sphere (somewhere
between the uncertainty and ignorance circles) or in the risk sphere depends on their
perception and beliefs in models and the confidence one gives them.

217
Figure 8.1 Approaches to uncertainty as a function of the degree of uncertainty
about probabilities and outcomes. Modified from Dessai and Hulme (2003).

In summary, both approaches taken in this thesis (the predictive and robustness
approaches) fit well with the regulatory environment in the water sector. It is unclear
whether the predictive results provide enough certainty to justify investment from the
regulators viewpoint. If the regulators require an x amount of certainty to make
investment decisions then little adaptation is likely to occur since climate change
uncertainty is only likely to grow, relative to the results analysed in this thesis, as more
of the uncertainty space is explored. The robustness approach provides an alternative to
the prediction paradigm, but is probably only useful if uncertainties are indeed severe
and if there is a large repertoire of options available to craft robust strategies.

8.2.2 Water industry: searching for guidance

It is perhaps not surprising that water companies focus exclusively on the final planning
scenario, instead of considering a range of planning scenarios (as suggested by the
Environment Agency), since putting together a Water Resource Plan for a company as
large as Anglian Water is a considerable endeavour. In this section the implications of
the results obtained are discussed in the context of how companies will incorporate
climate change in their WRP in the future. Some recommendations are put forward.

218
The previous chapters of this thesis have demonstrated how climate change
uncertainties can be included into water resources planning. The method is not
dissimilar to what Anglian Water, and probably several other companies, carried out for
the AMP4 process: a PDF of additional water required (or reduction in deployable
output) due to climate change was included in the calculation of target headroom.81 The
major difference between this thesis and what was done previously relates to how this
PDF of climate change impacts was estimated. For AMP4, Anglia Water constructed a
frequency distribution of resource loss based on expert judgement and hydrological
modelling using the UKWIR scenarios (medium, cool and wet, and warm and dry
scenarios), which are based on UKCIP02. Here, a more consistent treatment of
uncertainties has been provided and a much larger set of uncertainties has been sampled
(including multiple GCMs and RCMs, natural variability, climate sensitivities, etc.). As
Figures 7.22-24 show the results are very different. Current water industry practice, as
demonstrated by Anglian Waters approach, leads to a narrow PDF (or CDF) of water
losses with no possibility of water increases over the next 25 years. The uncertainty
analysis performed here estimates a much larger and flatter uncertainty range of water
losses (or additional water required).

There are ongoing efforts in trying to devise a consistent approach for including climate
change (and its uncertainties) in water resource planning, which would be accepted by
water companies and the regulators and which would provide justification for
investment.82 It might be feasible to get agreement on certain issues (e.g., should
climate change be applied to headroom or deployable output?) while others will be
more difficult (e.g., should an empirical rainfall-runoff model be used or a more
physical based model?). As the sensitivity and uncertainty analysis have shown, of the
components tested it is the uncertainties prior to the impact stage (above it in the
cascade of uncertainty) that have the most influence on adaptation decisions, in
particular GCM and downscaling uncertainties, but also climate sensitivity and to a
certain extent aerosol forcing uncertainties as well. These are uncertainties that the
water industry will find difficult to handle consistently because it is outside their field of
expertise and because there are a plethora of methods and models available. However,

81
Some water companies might have included this reduction directly on deployable output estimates.
82
http://www.futuredrought.org.uk

219
UKCIP might do this for them (as they did for UKCIP98 and UKCIP02) as new
projections are developed for the UK in the next few years (e.g., UKCIP08), but there is
no guarantee that all the uncertainties mentioned will be tackled.

Estimating probabilities of climate change impacts on water resources is complex and

laborious. One could not expect a water company to undergo such a thorough analysis
as in this thesis.83 The Environment Agency (2004a) acknowledges that for many
companies climate change is a new area of work and they may need to seek advice and
expertise from climate change specialists. Some advice is compiled below on how the
water industry can sample a similar amount of uncertainty as in this thesis, but in a
simpler way.

Here are a few recommendations that come out from the sensitivity and uncertainty
analysis that should be useful for future guidelines on how the water industry can
include climate change in water resources planning:
1. It is crucial that natural multi-decadal climate variability be included in any
assessment. Natural variability is a major contributor to the variance of results
and there is evidence that water companies have neglected this important factor,
giving too much confidence (over precision) to results from anthropogenic
climate change alone. Too much emphasis has been put on drought (and loss of
resources) when there is a probability (albeit small; between 13 and 19% in
Figure 7.22) of wetter summers (i.e., more resources available).84
2. Being the largest source of uncertainty and still poorly quantified, it is
imperative to have a range of GCMs be used. As a minimum two models from
the lower and upper range of summer precipitation change should be used. For
the East of England that would mean using HadCM3 (upper range) and CSIRO
Mk2 (lower range).
3. As the second largest source of uncertainty, downscaling ought to be considered
as well. If the dynamical downscaling route is pursued then at least two models
representing the lower and upper range of summer precipitation change should
be used. For the East of England that would mean using HadRM3 (upper range)

83
However, this might be the sort of analysis that could justify investment to regulators.
84
This is almost entirely due to natural multi-decadal climate variability, but also due to a few GCM
simulations that show slight increases of summer precipitation.

220
 and ARPEGE (lower range). Other downscaling techniques (e.g. statistical and
stochastic) ought to be considered too (but these have not been considered
here).
4. As a medium contributor to uncertainty, climate sensitivity also ought to be
considered. If an energy balance model is not used to sample this uncertainty, as
applied in this thesis, then GCMs spanning a large range of climate sensitivities
should be used, thus increasing the numbers of GCMs used to four.
Unfortunately there are no GCM simulations with very high climate
sensitivities (i.e., above 6C) available at present. For the East of England this
would imply using DOE PCM (1.7C effective climate sensitivity) and the
CCSR/NIES2 (5.1C equilibrium climate sensitivity)
5. Climate impacts uncertainty ought to be investigated in more detail by water
companies as usually they are in-house hydrological models
6. The impact of aerosol forcing uncertainty is largely scenario dependent so while
it should not be forgotten it need not be included.
7. The uncertainty introduced by greenhouse gas emissions is relatively small in
the 2030s, but it is advised that a range of scenarios be used, for example, A2
(upper range) and B1 (lower range).85
8. Other uncertainties (such as ocean mixing, carbon cycle, mitigation policies,
conversion from storylines to GHG emissions and sampling strategy) need not
be included as their impact on planning is negligible. However, with new
knowledge being gained in the future, this could change, e.g., if it was
discovered that a specific process in the oceans controlled much of our climate.
9. The inclusion of climate change risk and uncertainty in target headroom is
recommended.

These suggested guidelines already involve a considerable number of runs (two

emissions scenarios x four GCMs x two RCMs = 16 runs) that water companies would
have to examine if they wanted to tackle climate change uncertainties that impact their
planning. Currently these 16 runs are not available because RCMs have not been run for

85
It is important to have a range of scenarios (preferably all four storylines) because this allows the
results to be combined with future local/regional changes in socio-economic conditions (e.g., using the
Foresight environmental future or the Environment Agency water scenarios).

221
that many driving A/GCMs so water companies would have to use the currently
available RCM simulations thus reducing the number of runs.

It seems like the water industry is moving towards a tiered approach (cf. Willows and
Connell 2003) to assessing climate change impacts on water resources.86 The first tier
conducts a rapid assessment (similar to the current UKWIR methodology), while
subsequent tiers go into further detail. This thesis is an example of a high tier
assessment where uncertainties are examined and managed in detail in order to justify
investment decisions. A medium tier assessment could follow the recommendations
above, which are more relaxed and less consistent, but simpler, about uncertainty
management. It is very important that the lowest tier of assessment, which every water
company would go through, includes at least two model simulations that span much of
the uncertainty range explored here (perhaps 95% or 80%) as using models in the
middle of the uncertainty range (the most likely) could provide an illusion of
insignificant climate change impacts to certain companies.

8.2.3 Climate prediction and the future of climate scenario

construction

The work carried out in this thesis has important implications for climate modelling
because it is one of the first attempts at linking real-world decision-making about
adaptation to climate change with climate prediction results that characterise uncertainty
explicitly. Two almost opposing conclusions relevant to climate prediction can be taken
from this thesis. One is that climate change impacts (and therefore adaptation decisions)
are very sensitive to which GCM/RCM is used or which climate sensitivity is chosen
(demonstrated in section 7.1). This emphasises the importance of further climate
modelling and in particular constraining and further reducing uncertainty. The other
conclusion is that even when climate change uncertainties are included in planning (as
in section 7.2), AWS WRP is sufficient to cope with these uncertainties and maintain
levels of service because of the shear size of the adaptation responses being planned
(new reservoirs or extensions for example). This conclusion leads one to believe that
climate change uncertainties do not matter for adaptation. But as section 8.1.2

86
Personal communication from water manager at Anglian Water, 4 November 2005.

222
explained, AWSs WRP has not been approved for investment beyond 2010 (possibly
2015), so the strategies contained in the WRP can be perceived as a wish list by Anglian
Water. The general conclusion, however, is somewhere in the middle of these opposing
conclusions where information on uncertainty ranges is needed because decisions are
sensitive to some uncertainties, but also that adaptation decisions are likely to be driven
equally or more by factors other than climate change (e.g., demand).

There is a need for more rigorous treatment of climate change uncertainties in climate
prediction and climate scenario construction (Hulme and Carter 1999, Katz 2002). This
thesis has demonstrated one approach in section 7.2. It also showed that using different
beliefs about PDFs or parameters can lead to very different results as shown by the
sensitivity analysis in section 7.1. Getting agreement between experts and users on
which models to use or what shape the PDF of a certain parameter should be will be
difficult. This will represent a tremendous challenge to the next set of climate change
scenarios in the UK (UKCIPnext). How much of the uncertainty space the next climate
change scenarios will include is also crucial. If decisions are made based on climate
information such as UKCIP02, then decision-makers run the risk of making bad
decisions, because those decisions are vulnerable to uncertainties not sampled by those
scenarios. Fortuitously, for Anglian Water, and probably other water companies in
England and Wales, their planning for AMP4 was based on one of the driest models,
which makes their decisions fairly robust to climate change uncertainties (because their
main concern is drought). On the other hand, if summer flooding was the major concern
of a particular user, the UKCIP02 scenarios would not lead to robust decisions. This
shows the need for consistent treatment of uncertainties in climate change projections.
However, as mentioned earlier, getting agreement on how to manage climate change
uncertainties will be difficult. Furthermore, consistent treatment of uncertainties leads to
very large and flat PDFs of climate change impacts as shown in section 7.2.

Is there any hope in reducing this large uncertainty? There is an increasing emphasis in
the literature on constraining climate model response using observed data (e.g., surface
temperature, ocean temperature, etc.) via Bayes Theorem (Forest et al. 2002, Murphy
et al. 2004). While some success has been achieved for variables like climate sensitivity,
the uncertainty range is still fairly large: the 90% confidence range in Figure 3.6 is
between 1 and 9.5C. It is unclear whether regional climate response can be constrained

223
any better. Evidence from section 7.1.5 showed that there is some reduction in
uncertainty when the best performing models (compared to observations) are used or
weighted compared to using all the models. However, this is severely limited by the
number of simulations available which makes it difficult to draw out any conclusions.
When using GCMs, the skill attribute made some difference from an adaptation
perspective, whereas with RCMs it was negligible. The difference between a skilled
distribution and a uniform distribution in the uncertainty analysis was also rather small
in terms of adaptation decisions. The problem with pursuing these constraining
exercises within a prediction paradigm is that it will be difficult to get agreement
between experts on what methods to use, which data with which to constrain it, at what
significance level, etc. It will be important to test whether adaptation decisions are
sensitive to these factors.

The sensitivity analysis showed where the largest uncertainties exist from an adaptation
perspective. It makes sense to redefine the cascade of uncertainty pyramid (left in
Figure 8.2) with a rhombus (also called a diamond; right in Figure 8.2). This shows
where uncertainties are largest (regional climate response) and of similar magnitude
(climate sensitivity and impacts) from a decisions points of view. If climate prediction
wants to be more useful to decision-making then it needs to decrease uncertainty where
the diamond is widest. Unfortunately, it is likely that as more of the uncertainty space is
sampled (as is being done in experiments like Murphy et al. 2004, Stainforth et al.
2005), the diamond will only get wider. It is still unknown87 how much of a constraint
observational data can be, but from the literature and the experiments conducted here it
seems decision-makers will have to take action (or decide not to take action) on
adaptation under considerable uncertainty for years, if not decades, to come.

87
Perhaps impossible to known since the climate system is an open system.

224
 World
development

World Greenhouse
development gas emissions

Greenhouse Climate sensitivity

gas emissions
Global climate models
Climate sensitivity
Downscaling
Global climate models
Impacts
Downscaling

Impacts

Figure 8.2 On the left is the traditional depiction of the cascade of uncertainty in
climate change assessments for adaptation. On the right is how a more accurate
depiction of uncertainty could look like from an adaptation decision perspective.

The advice that can be given for future climate scenario construction is not much
different from the previous section: include natural variability (from observations),
include several models (GCMs and RCMs) and sample at least the most important
uncertainties (climate sensitivity and aerosol forcing). This is, of course, based on a
water resources perspective. Since climate change scenarios, like the ones produced by
UKCIP, have a multitude of users from different sectors other uncertainties might have
to be considered as well. All things considered, however, climate change scenarios are
going to be constructed with the models, tools and methods that are available at the time
(that seems to be what history shows; see section 5.2). The important thing is that
uncertainty modelling in climate prediction has contact with users who are going to use
this information in their decision-making. This has been remarkably absent in the past
and has been voiced by other analysts as well (see, e.g., Jones et al. 1999).

Finally, this thesis and the discussion in section 8.2.1 have shown that there will be calls
for robust or sound science in climate prediction in support of decision-making and
policy (see, e.g., Patrinos and Bamzai 2005 for a recent example). This is an issue that
this thesis only touches upon, by testing model skill (in a limited fashion), but which
will represent a major challenge for the climate prediction community. It is probably
clear by now that the author of this thesis favours robustness approaches over
prediction. However, this should not detract from the fact many people will crave
predictions. This opens up numerous questions such as: When is a model good enough

225
to be used in risk management decisions? Should one base decisions on observations or
modelled results? And the list goes one. There is little work that has tackled these
questions in climate change adaptation (for two exceptions that stop at impacts see
Risbey and Stone 1996, Georgakakos 2003), especially from a decision perspective.
Crawford-Brown (2005) examined the concept of sound science in risk management
decisions and proposed the followed characteristics, which are relevant here:
1) a process that satisfies criteria of rational dialogue and reflection
2) an uncertainty distribution that is sufficiently narrow to allow upper-bound risk
estimates to provide apt means to identify benefits of proposed management actions
3) an assessment of overall epistemic status, including the status of causal claims for
any associations, which makes the manager reasonably confident that adjustment of a
purported cause will in fact reduce the risk (p. 7).

It is unclear whether this thesis and wider scientific research in this area possess these
characteristics, but this is certainly something that needs to be investigated further.
There are also some important links here with the model quality assurance literature (cf.
Funtowicz and Ravetz 1990, van der Sluijs et al. 2005)

8.2.4 Adaptation planning

This section discusses what has been learned about adaptation planning from this thesis.
Something that became evident while doing this work, and already mentioned in section
8.1.2, is that adaptation planning in the water sector (beyond 2015) is not being
considered seriously (or it is being delayed or ignored) because of the uncertainty in
climate change projections.88 By quantifying and managing uncertainties in a consistent
manner this work might make the regulators more at ease with planning for the long
term, but this really depends on the degree of certainty they require to make decisions
(which is closely linked to how risk averse they are).

88
It is ironic, however, that so much more planning is being put on climate change mitigation (e.g., Kyoto
Protocol, emissions trading, carbon levy, etc.), which might suggest double standards to the application of
the precautionary principle. However, there is more certainty about the planet getting warmer than how
precipitation will change in the East of England (as shown by chapters four and five). In addition, the
more mitigation that occurs the less adaptation will be required.

226
The analysis showed that water resources were sensitive to several climate change
uncertainties, but that adaptation decisions currently being planned were in fact robust
to those uncertainties. Assuming the WRP is realistic, that does imply that adaptation
planning in this particular context need not worry about climate change uncertainties.
This is partially due to the fact that the WRP was put together using one of the driest
models in the range (HadCM3) and because the adaptation options being considered
(new reservoir and water treatment works extension for example) create plenty of
headroom to deal with climate change uncertainties. However, if one moves
philosophically (on the issue on climate change uncertainty) from the risk paradigm to a
deep uncertainty paradigm, then the robustness analysis (section 7.3) showed that there
are certain combination of climate change uncertainties that make the current WRP
vulnerable. This was particularly evident when demand uncertainties were taken into
account. Under this approach, robustness to uncertainty required more adaptation
actions to be included in the current WRP. This is an important implication of this
thesis: as robustness to uncertainty increases, costs also increase (or aspirations
decrease). This is a general property of decision-making found elsewhere (see, e.g.,
Regan et al. 2005), but which has not been demonstrated for adaptation to climate
change before (at least not in so much detail and using real world decisions). It is clear
that actual adaptation will involve trade-offs between robustness and other criteria that
are used to judge successful adaptation (e.g., efficiency and equity, see Adger et al.
2005).

Ultimately, adaptation decisions that do get implemented by water companies will have
to be negotiated with, and approved by, regulators and civil society. It will be
increasingly important to bring civil society into the debate because it will be necessary
to manage expectations about levels of service and because it is civil society who will
pay for the adapting through their water bills. The Water Act of 2003 does this by
requiring water companies to have a public consultation on their WRP. In truth, this
thesis only focuses on a small, but arguably important, part of the adaptation process,
that with respect to climate information.

Are the conclusions from this thesis valid in other adaptation settings, for example in
other sectors? It was the initial aim of this PhD to conduct two adaptation case-studies.
Unfortunately, quantifying climate change uncertainties and getting immersed into one

227
decision-making context (that of water resources) took much longer than expected so
the second case study had to be dropped. If there had been time to pursue a second case
study, it would have probably focused on the built environment. The life span of
buildings is in the order of decades, typically between 50 and 100 years. This makes this
sector particularly relevant to climate change timescales. In contrast to the water sector,
the built environment is not regulated regarding climate (except schools)89 and is mostly
driven by temperature changes (in particular maximum temperature). Figure 8.3
illustrates how summer mean temperature could change by the 2030s and 2050s in the
East of England. The uncertainty range at the 95% confidence level is 1.42C by the
2030s and 2.54C by the 2050s. The uncertainty range for summer maximum
temperature could be different, but is seems, however, that modelled results have
greater certainty (less uncertainty) ranges for temperature change than for precipitation
change. Without going into the details of adaptation options available to designers and
engineers it is difficult to say whether the conclusions from this thesis would hold in
this sector.

0.06

0.05

0.04

Probability 0.03 2030s

2050s
0.02

0.01

0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0
Summer mean temperature change (C)

Figure 8.3 Probability of summer mean temperature change for the 2030s and
2050s relative to 1961-90 in the East of England, using GCM results (models
equally plausible) and the default scheme used in the sensitivity analysis.

89
Although there is now a Building Regulations Advisory Committee at DEFRA which looks at climate
change.

228
From the experience of conducting this research it is possible to speculate that: 1)
robustness to climate change uncertainties usually means higher costs (and therefore
lower aspirations); 2) different impact sectors will be sensitive to different uncertainties
in climate change assessments; 3) the robustness of adaptation strategies to climate
change uncertainties will likely depend on the pressure exerted on the decision process
by drivers other than climate. This implies that context is crucial and that it is difficult
to generalise any lessons from this case study to wider adaptation planning. Further case
studies need to be investigated to illuminate these speculations.

8.2.5 Post-normal science and interdisciplinarity

This section reflects on the undertaking of this thesis. In particular it addresses the
issues of post-normal science and interdisciplinarity. It could be argued that this is not
an ordinary thesis, with a clear methodology, neat results and straightforward
conclusions. These were the impositions of doing an interdisciplinary thesis in a post-
normal scientific environment. Funtowicz and Ravetz (1993) coined the term post-
normal science when systems uncertainty and decision stakes90 are high, to the point of
dominating the strategies for problem solving (see Figure 8.4). When these variables are
low, Kuhns (1962) normal science focused on puzzle-solving is sufficient (applied
science). As these variables rise, new actors and skills must be brought into play to
forge solutions to policy problems (professional consultancy). When the variables are
high, in the post-normal science region, scientific experts share the platform of
knowledge production with many other actors. Ravetz (2004) argues that under these
conditions, a narrowly trained expertise can be irrelevant, or even counterproductive.
What is required is an extended peer community with extended facts far beyond the
peer-reviewed published literature (e.g., local knowledge, etc).

90
By decision stakes Ravetz (2004) means the investments and commitments, personal as well as
commercial and institutional, that are at stake in the inquiry.

229
Figure 8.4 Post-normal science. Source: Ravetz (2004)

Adaptation to climate change as considered in this thesis is clearly within the post-
normal science realm. Uncertainties have been shown to be high (see Chapter 7) while
decision stakes are also high (see Chapter 6; e.g., building a reservoir or not). The
existence of an extended peer community is also evident: during the preparation of this
thesis the author had to engage with water managers from the business sector (Anglian
Water and Essex and Suffolk Water), regulators (Environment Agency), professional
consultants (Atkins; who prepared much of AWS WRP), other academics/researchers
etc. Clearly, the production of knowledge is not only with climate scientists anymore.
As mentioned in section 8.1.2, the level of stakeholder involvement was low, but it
could still be argued that there was some co-production of knowledge in this thesis as
some analyses would not have been possible without the collaboration of some of the
stakeholders. In this thesis only a limited amount of stakeholders were involved due to
time constraints and because water companies are likely to undertake the bulk of
adaptation measures in the water resources sector. Future work needs to involve a larger
set of stakeholders, in particular those with whom adaptation strategies will have to be
negotiated with (regulators) and those who will be affected by those decisions and who
will pay the costs of adapting (civil society at large).

230
Ravetzs (2004) comment that super specialisation could be a hindrance to problem
solving in post-normal science is a pertinent one for this thesis. A traditional PhD
implies a specialisation and a narrowing of expertise. In this thesis specialisation has
occurred, but there was also a deliberate and explicit trade-off between specialisation
and interdisciplinarity. The thesis crosses the disciplines of climate sciences,91 water
sciences, institutions and decision sciences. In order to deliver the PhD on time, a trade-
off between simplicity and complexity had to be achieved in the several chapters of this
PhD (see discussion in section 8.1.1). Conducting interdisciplinary research was
perhaps one of the greatest challenges of this PhD. The author spent the first few years
quantifying uncertainties in the climate system. Then he had to immerse himself into the
water world so that he could converse with decision-makers. Finally, tools of policy
analysis (such as sensitivity, uncertainty and robustness analysis) had to be learned at
the same time to come up with the final results. Of course any of these subjects could be
researched further (as the next section discusses), but the point of this thesis (and
interdisciplinarity) was to bring together these tools and approaches to produce new
knowledge (and findings) that would not emerge from the single disciplines. Did it
happen? It is the authors conviction that it did, but it is difficult to prove. An
assessment of the knowledge created would have to be carried out to answer this
question more fully (see, e.g., Cash et al. 2003).

8.3 Future research

The work conducted for this thesis opens a Pandoras box of future research questions.
Some ideas have already been mentioned throughout the discussion, including: wider
stakeholder engagement; possibility of enlarging adaptation options; sampling wider
uncertainty spaces with GCMs and constraining results with observations. Each aspect
of the modelling framework used here can also be researched further, for example,
conducting uncertainty analysis on the hydrological model, using more recent climate
sensitivity PDFs, etc.

Many interesting research questions reside in the social science camp. For example,
examining the perception of climate change risk amongst those involved in water

91
Including simple climate modelling, observed climatology and climate scenario construction.

231
resource planning. This research also opens up a number of economic questions. For
example, if one can determine the cost of breaching levels of service then it is possible
to determine the value of reducing uncertainty in climate information and the value of
perfect information. This would allow more informed investment in scientific research
according to actual decision-makers needs (cf. Rubin et al. 1992 on keeping climate
research relevant).

8.4 Concluding remarks

This thesis has demonstrated the application of a framework that links decisions about
adaptation to climate change with a long chain of climate change uncertainties. This has
never been attempted before except in a hypothetical sense or using only one or two
scenarios of climate change (Hobbs et al. 1997, Risbey 1998). It has been possible to do
this because of the thorough and consistent treatment of climate change uncertainties
conducted in Chapters 4 and 5.

In the adaptation context examined, that of water resources planning for the 2030s in the
Anglian region of the UK, losses of water resources were found to be sensitive to
several of the uncertainties examined, namely (in descending order): regional climate
response (from GCM and dynamical downscaling), climate sensitivity, climate impacts,
aerosol forcing and emissions scenarios. Adaptation decisions were insensitive to
uncertainties from sampling strategies, ocean mixing and the carbon cycle. Regardless
of this sensitivity, Anglian Waters WRP remained robust to climate change
uncertainties because of the adaptation options being considered (upgrading water
treatment works, potential new reservoir) and because a climate model (HadCM3 in
UKCIP02) that predicts drier conditions than other models was used for planning. A
robustness analysis was explored for when climate change uncertainties were deemed
unameable to quantification with probabilities. Under this approach, AWS WRP was
found to be vulnerable to particular combination of uncertainties. If demand was
considered uncertain at the same time, the robustness of the plan changed dramatically
depending on which scenario was applied.

232
This thesis has illustrated a methodology that allows real world decisions about
preparing for climate change to be judged against climate change uncertainties. It also
tried to hint that a different decision-making process (based on robustness) might be
required for climate change (and other) uncertainties over the current predict and
optimise approach.

Much improvement is still possible on many aspects of this framework including:

enlargement of adaptation; including further climate change uncertainties by using more
recent GCM simulations and constraining them with observations; wider stakeholder
involvement, etc.

It is worth remembering why this work is important as it easy to get stuck in the
numerous technicalities. Adaptation to climate change, as approached here, is about
planning ahead. Planning is about reducing the risk of being vulnerable to uncertainties
and surprises. Climate change makes planning for adaptation troublesome because of
the many uncertainties involved, but there are (partial) solutions as this thesis has
shown.

233
 APPENDIX 1. OPTIONS IN THE WATER RESOURCES
PLAN

This appendix lists all the options available to AWS, as reported in their Water
Resources Plan and Figures of where the main schemes are located.

Table A1.1. Options to maintain target headroom in planning zones of the ES&E
WRZ

Planning Zone Option

Semer Pressure reduction at Groton / Kersey and Nayland
Woodbridge Upgrade transfer feed from Ipswich
Reallocate Ipswich licences
Pressure reduction at Bradfield
Ipswich Resource development through Alton Extension and
increased effluent re-use at Cliff Quay
Increase Alton output
Pressure reduction schemes at Somersham, Ravenswood,
Felixstowe Tower and Stratford St Mary / Shotley
Halstead Rural Transfer through re-allocation of Deployable Output
from Halstead Urban
Pressure reduction scheme Foxearth / Lamarsh
Halstead Urban Transfer Coggeshall to Parkfield
Braintree Increase transfer from Colchester
Pressure reduction scheme at Kelvedon
Colchester Increase resource take from Ardleigh
Bergholt to Lexden for blending fluoride
Alton Horkesley transfer link
Pressure reduction at Bures

234
Table A1.2. Description of water management options for the ES&E WRZ and
their respective expected increase in Water Available for Use (WAFU) or demand
saving, and total Net Present Cost (NPC). In bold is the sum of each management
style.

Option description Planned Total

gains in NPC
WAFU or (M)
savings in
demand by
2029-30
(Ml/d)
Customer Side Management 0.45 0.23
E58 - Halstead Rural; Cistern Displacement scheme 1 0.01 0.00
E63 - Colchester; Cistern Displacement Scheme 1 0.04 0.01
E63 - Colchester; Domestic Water Audit 1 0.10 0.06
E58 - Halstead Rural; Domestic Water Audit 1 0.02 0.01
E56 - Woodbridge; Domestic Water Audit 1 0.02 0.01
E57 - Ipswich; Domestic water Audit 1 0.21 0.12
E60 - Braintree; Domestic Water Audit 1 0.06 0.03
Distribution Side Management 4.95 0.60
E63 - Colchester; 20 Wks additional ALC 2.12 0.15
E58 - Halstead Rural; E58-P-01 Foxworth/Lamarsh Pressure Red 0.02 0.01
E57 - Ipswich; 12 Wks additional ALC 1.61 0.15
E57 - Ipswich; E57-P-03 Felixstow Tower Pressure Red 0.05 0.01
E58 - Halstead Rural; 3 Wks additional ALC 0.31 0.07
E56 - Woodbridge - 4 Wks additional ALC 0.39 0.07
E60 - Braintree; 4 Wks additional ALC 0.45 0.15
Production Side Management 22.50 9.99
E58 - Halstead Rural; E58-T-01 Reallocated DO from H.Urban Transfer 2.80 0.00
E56 - Woodbridge; E56-T-01 Reallocated Ipswich Licences Transfer 1.80 0.07
E59 - Halstead Urban; E59-T-01 Coggleshall to Parkfield Tower 3.90 1.88
Transfer6
E60 - Braintree; S60-T-01 Increase from Colchester Transfer 3.00 0.00
E63 - Colchester; E63-T-01 Alton-Hawksley Link Transfer 11.00 8.03
Resource management 23.00 50.88
E57 - Ipswich; E57-R-01 Alton Extns (ER Cliff Quay) 20.00 42.85
a
E63 - Colchester; E63-R-01 Ardleigh Increased Take 0.00 4.71
E63 - Colchester; E63-R-02 Bergholt to Lexden for blend 3.00 3.32
a: Temporary increased take of 5 Ml/d between 2006-07 and 2013-14

235
 Key

Existing WTW Existing mains

CAMBRIDGESHIRE & WEST SUFFOLK WRZ New WTW (AMP 4) New scheme
(AMP 4)

ESSEX & SUFFOLK New WTW (AMP 5) New scheme

WATER CO. (AMP 5)
Bury St
A 14
Edmunds

A 14

Ipswich

Felixstowe

5
Petches Bridge 1 Alton
WTW Water

Halstead
Great
Horkesley
WTW

3 2 Ardleigh
WTW
PRINCIPAL SCHEMES
Braintree Lexden
AMP 4
WTW
4 1. E63 - T - 01 Alton - Horkesley link
2. E63 - R- 02 Horkesley to Lexden WTW for blend
3. E59 - T - 01 Coggeshall to Parkfield Tower
TENDRING
Colchester HUNDRED 4. E63 - R - 01 Ardleigh Reservoir increased take
WATER CO.
AMP 5
5. E57 - R - 01 Alton WTW extension
A 12

ESSEX & SUFFOLK Figure A1.1

WATER CO.
EAST SUFFOLK & ESSEX WRZ
March 2004

236
Table A1.3 Options to maintain target headroom in planning zones of the
Ruthamford WRZ.

Planning Zone Option

Peterborough Proportion of Wing Extensions
Mains replacement schemes at Thorney, Retail Park, and
Rholm Bank
Proportion of Wing Extensions
Northampton Future Wing Extensions Phase 2 or Bourne transfer
Mains replacement schemes at Northampton Brackmills
Phase 1 and 2, Boughton Reservoir, Pitsford to A508 and
South Oval
Proportion of Wing Extensions
Oundle Future Wing Extensions Phase 2 or Bourne transfer
Proportion of Wing Extensions
Corby Future Wing Extensions Phase 2 or Bourne transfer
Mains replacement at Gretton Brook, Sexton Way, and
White Post Court,.
Proportion of Wing Extensions
Kettering Future Wing Extensions Phase 2 or Bourne transfer
Mains replacement at The Avenue and London Road to
Wicksteed Park
Proportion of Wing Extensions
Desborough Future Wing Extensions Phase 2 or Bourne transfer
Transfer from Clapham
Proportion of Wing Extensions
Ravensthorpe Future Wing Extensions Phase 2 or Bourne transfer
Mains replacement at Long Buckby Lodge Lane
Proportion of Wing Extensions
Wellingborough Future Wing Extensions Phase 2 or Bourne transfer
Mains replacement at Farndish 3 main, Little Addington
and Brafield On The Green
Proportion of Wing Extensions
Rushden Future Wing Extensions Phase 2 or Bourne transfer
Proportion of Wing Extensions
Thrapston Future Wing Extensions Phase 2 or Bourne transfer
Mains replacement at Stanwick
Proportion of Wing Extensions
Clapham Future Wing Extensions Phase 2 or Bourne transfer
Clapham WTW Extension
Mains replacement at Stagsden West End
Proportion of Wing Extensions
Bedford Future Wing Extensions Phase 2 or Bourne transfer
Transfer from Clapham
Mains replacement at Kempston
Proportion of Wing Extensions
Meppershall Future Wing Extensions Phase 2 or Bourne transfer
Proportion of Wing Extensions

237
Flitwick Future Wing Extensions Phase 2 or Bourne transfer
Mains replacement at Willow walk
Proportion of Wing Extensions
Wobourn Future Wing Extensions Phase 2 or Bourne transfer
Proportion of Wing Extensions
Daventry Future Wing Extensions Phase 2 or Bourne transfer
Mains replacement at Borough Hill
Proportion of Wing Extensions
Buckingham Future Wing Extensions Phase 2 or Bourne transfer
Proportion of Wing Extensions
Mursley Future Wing Extensions Phase 2 or Bourne transfer
Mains replacement at Quainton to Botolph
Proportion of Wing Extensions
Newport Future Wing Extensions Phase 2 or Bourne transfer
Proportion of Wing Extensions
Milton Keynes Future Wing Extensions Phase 2 or Bourne transfer
Sandhouse WTW
Transfer from Pulloxhill
Pressure reduction for the city of Milton Keynes
Proportion of Wing Extensions
Leighton Future Wing Extensions Phase 2 or Bourne transfer
Linslade Transfer from Pulloxhill
Mains replacement at Linslade SR supply
Proportion of Wing Extensions
Biggleswade Future Wing Extensions Phase 2 or Bourne transfer
Proportion of Wing Extensions
Towcester Future Wing Extensions Phase 2 or Bourne transfer
Proportion of Wing Extensions
Huntingdon Future Wing Extensions Phase 2 or Bourne transfer
Transfer from Clapham
Proportion of Wing Extensions
March Future Wing Extensions Phase 2 or Bourne transfer
Transfer from Peterborough to March

238
Table A1.4. Description of water management options for the Ruthamford WRZ
and their respective expected increase in Water Available for Use (WAFU) or
demand saving, and total Net Present Cost (NPC). In bold is the sum of each
management style.

Option description Extra Total NPC (M)

WAFU/se
mand
savings by
2029-30
Customer Side Management 1.53 0.69
E98 - March; Cistern Displacement Scheme 1 0.01 0.00
E98 - March; Domestic Water Audit 1 0.02 0.02
E98 - March; Domestic Water Audit 2 0.03 0.01
R74 - Peterborough; Cistern Displacement Scheme 1 0.06 0.02
R89 - Daventry; Domestic Water Audit 0.03 0.01
R80 - Ravensthorpe; Domestic Water Audit 0.01 0.01
R87 - Flitwick; Domestic water Audfit 0.02 0.08
R74 - Peterborough; Domestic Water Audit 0.15 0.01
R96 - Towcester; Domestic Water Audit 0.02 0.03
R77 - Corby; Domestic Water Audit 0.05 0.01
R86 - Meppershall; Domestic Water Audit 0.02 0.10
R75 - Northampton; Domestic Water Audit 0.18 0.02
R82 - Rushden; Domestic Water Audit 0.03 0.03
R95 - Biggleswade; Domestic Water Audit 0.05 0.01
R84 - Clapham; Domestic Water Audit 0.02 0.01
R92 - Newport; Domestic Water Audit 0.02 0.01
R91 - Mursley; Domestic Water Audit 0.02 0.00
R92 - Newport; Cistern Displacement Scheme 1 0.01 0.00
R82 - Rushden; Cistern Displacement Scheme 1 0.01 0.00
R87 - Flitwick; Cistern Displacement Scheme 1 0.01 0.04
R85 - Bedford; Domestic Water Audit 0.09 0.02
R75 - Northampton; Cistern Displacement Scheme 0.07 0.00
R77 - Corby; Cistern Displacement Scheme 0.02 0.01
R78 - Kettering; Cistern Displacement Scheme 0.02 0.03
R78 - Kettering; Domestic Water Audit 0.07 0.00
R79 - Desborough; Cistern Displacement Scheme 0.01 0.01
R79 - Desborough; Domestic Water Audit 0.02 0.00
R80 - Ravensthorpe; Cistern Displacement Scheme 0.01 0.00
R86 - Meppershall; Cistern Displacement Scheme 0.01 0.01
R88 - Woburn; Domestic Water Audit 0.01 0.00
R88 - Woburn; Cistern Displacement Scheme 0.01 0.00
R89 - Daventry; Cistern Displacement Scheme 0.01 0.02
R90 - Buckingham; Domestic Water Audit 0.04 0.00
R90 - Buckingham; Cistern Displacement Scheme 0.01 0.00
R91 - Mursley; Cistern Displacement Scheme 0.01 0.01
R93 - Milton Keynes; Cistern Displacement Scheme 0.05 0.08
R93 - Milton Keynes; Domestic Water Audit 0.15 0.00
R95 - Biggleswade; Cistern Displacement Scheme 0.02 0.00

239
R97 - Huntingdon; Domestic Water Audit 0.09 0.04
R97 - Huntingdon; Cistern Displacement Scheme 0.03 0.01
Distribution Side Management 10.55 2.17
E98 - March; 6 Wks ALC 0.38 0.07
R74 - Peterborough; 4 Months Additional ALC 1.33 0.15
R95 - Biggleswade; 6 Wks additional ALC 0.50 0.07
R96 - Towcester; 1 Wks additional ALC 0.11 0.07
R75 - Northampton; 4 Months additional ALC 1.77 0.22
R87 - Flitwick; 1 Wk additional ALC 0.11 0.15
R77 - Corby; 6 Wks additional ALC 0.47 0.07
R92 - Newport; 1 Wk additional ALC 0.11 0.07
R80 - Ravensthorpe; 2 Wks additional ALC 0.22 0.07
R89 - Daventry; 1 Wk additional ALC 0.16 0.07
R91 - Mursley; 3 Wks additional ALC 0.26 0.07
R82 - Rushden; 3 Wks additional ALC 0.25 0.07
R86 - Meppershall; 3 Wks additional ALC 0.33 0.07
R93 - Milton Keynes; 4 Months additional ALC 1.35 0.07
R85 - Bedford; 2 Wks additional ALC 0.23 0.15
R78 - Kettering ; 6 Wks additional ALC 0.49 0.15
R79 - Desborough; 2 Wks additional ALC 0.16 0.15
R88 - Woburn; 1 Wk additional ALC 0.12 0.07
R90 - Buckingham; 2 Wks additional ALC 0.20 0.07
R93 - Milton Keynes; R93-P-01 Pressure Reduction 1.59 0.07
R96 - Towcester; Cistern Displacement Scheme 0.01 0.06
R97 - Huntingdon; 4 Wks additional ALC 0.40 0.14
Production Side Management 21.23 9.28
E98 - March; E98-T-01 Peterborough to March Transfer 9.70 9.28
R85 - Bedford; Water from Clapham Transfer 3.00 0.00
R93 - Milton Keynes; Water from Pulloxhill 4.00 0.00
R76 - Oundle; Freed up by Clapham Transfer 1.10 0.00
R79 - Desborough; Water Freed up by Clapham Transfer 1.23 0.00
R83 - Thrapston; Water freed up from Clapham Transfer 1.70 0.00
R97 - Huntingdon; Water from Clapham 0.50 0.00
Resource management 182.22 345.90
R90 - Buckingham; Future Wing Ph2 or Bourne 2.00 7.11
R84 - Clapham; R84-R-01 Clapham WTW Extns 10.00 8.60
R97 - Huntingdon; Proportion of Wing WTW Extns & 2.50 0.00
TM's
R88 - Woburn; Proportion of Wing WTW Extns & TM's 1.40 0.00
R81 - Wellingborough; Proportion of Wing WTW Extns 5.20 0.00
& TM's
R74 - Peterborough; Proportion of Wing WTW Extns & 13.20 38.76
TM's
R78 - Kettering; Proportion of Wing WTW Extns & TM's 4.20 0.00
R95 - Biggleswade; Proportion of Wing WTW Extns & 5.30 3.88
TM's
R96 - Towcester; Proprtion of Wing WTW Extns & TM'S 2.20 1.29
R82 - Rushden; Proportion of Wing WTW Extns & TM's 2.20 1.62
R91 - Mursley; Proportion of Wing WTW Extns & TM's 2.11 2.75

240
R89 - Daventry; Proprtion of Wing WTW Extns & TM's 2.50 0.32
R87 - Fltiwick; Proprtion of Wing WTW Extns & TM's 2.20 2.42
R86 - Meppershall; Proprtion of Wing WTW Extns & 3.40 2.58
TM's
R92 - Newport; Proportion of Wing WTW Extns & TM's 3.10 2.91
R77 - Corby; Proprtion of Wing WTW Extns & TM's 7.20 6.62
R75 - Northampton; Proprtion of Wing WTW Extns & 11.30 11.31
TM's
R80 - Ravensthorpe; Proprtion of Wing WTW Extns & 1.50 2.26
TM's
R74 - Peterborough; Future Wing Ph2 or Bourne 21.00 94.79
R87 - Flitwick; Future Wing Ph2 or Bourne 1.70 5.21
R76 - Oundle; Future Wing Ph2 or Bourne 0.80 0.00
R96 - Towcester; Future Wing Ph2 or Bourne 1.80 3.13
R86 - Meppershall; Future Wing Ph2 or Bourne 3.00 7.29
R97 - Huntingdon; Future Wing Ph2 or Bourne 4.00 18.23
R88 - Woburn; Future Wing Ph2 or Bourne 1.00 2.60
R82 - Rushden; Future Wing Ph2 or Bourne 2.00 2.60
R95 - Biggelswade; Future Wing Ph2 or Bourne 4.10 10.42
R89 - Daventry; Future Wing Ph2 or Bounre 1.60 3.13
R93 - Milton Keynes; Proprtion of Wing WTW 12.00 4.04
R78 - Kettering; Future Wing Ph2 or Bourne 2.80 5.21
R93 - Milton Keynes; Future Wing Ph2 or Bounre 12.00 34.38
R80 - Raventhorpe; Future Wing Ph2 or Bourne 1.20 3.13
R75 - Northampton; Future Wing Ph2 or Bourne 5.00 28.65
R91 - Mursley - Future Wing Ph2 or Bourne 1.10 7.81
R79 - Desborough; Future Wing Ph2 or Bourne 0.80 2.08
R77 - Corby; Future Wing Ph2 or Bourne 2.10 6.25
R81 - Wellingborough; Future Wing Ph2 or Bourne 3.00 0.00
R83 - Thrapston; Future Wing Ph2 or Bourne 1.50 0.00
R85 - Bedford; Future Wing Ph2 or Bourne 3.00 0.00
R90 - Buckingham; Proportion of Wing WTW 3.10 4.74
R92 - Newport; Future Wing Ph2 or Bourne 2.21 7.81
R93 - Milton Keynes; R93-R-01 Sandhouse WTW 9.90 1.97
Refurbishment

241
PRINCIPAL SCHEMES
Lincolnshire Fens WRZ
1. RUT-R-01 Wing WTW extension. Rutland Water
2. RUT-T-01 Wing strategic main. Wing Etton Booster Fenland WRZ
3. Ecton Booster.
4. E98-In-01 March Chatteris reinforcement. Wing
5. R93-IN-01 UpgradeOld Stratford PS. WTW 1
Peterborough
6. RUT-IN-03 Deanshanger to Greatmirth reinforcement.
7. E98-T-01 Peterborough to March link.
Export to
AMP 5 Severn Trent March
8. Clapham WTW Extension Water
7
4
Severn Trent
Water Services Corby
A1 Chatteris
2

Kettering
A 14 Cambridgeshire &
West Suffolk WRZ

Pitsford A 14
WTW Huntingdon

Grafham Water

Wellingborough
Northampton Grafham
WTW

Daventry
3
Clapham
8 WTW

M1 Cambridge Water
Co

Bedford

Brackley
6 5
Key
Deanshanger
Reservoir
Milton Keynes Existing WTW Existing mains

New WTW (AMP 4) New scheme

(AMP 4)
Buckingham
New WTW (AMP 5) New scheme
(AMP 5)
Thames Water
Major Export To 3
Valleys Water Co
Leighton Buzzard

Three Valleys Water

Co
Figure A1.2
RUTHAMFORD WRZ
March 2004

242
 APPENDIX 2. QUESTIONNAIRE

This appendix shows the questionnaire sent to Anglian Water, including replies in red.

Dear Gerry,

If you are reading this document then you have decided to collaborate in this research
which is investigating how serious of a risk climate change is for your business. While
your company has taken into account climate change in its recent Water Resources Plan
(WRP), the Environment Agency (2004) encourages all companies to carry out more
detailed analyses to help them understand the scale and scope of the impact, as well as
to help them to choose appropriate adaptation measures. Good long-term planning is
essential because many of the best options to manage future water supplies take many
years to implement. The purpose of this research is to investigate the sensitivity of
adaptation options you have envisaged in the WRP and other options you can imagine
to the risks associated with climate change.

In your WRP the scenarios you used suggest that by 2030 there could be a significant
impact on surface water sources with a loss of some 40 Ml/d. These results are based
on the UKCIP02 data which uses several socio-economic scenarios in a single Regional
Climate Model driven by a single Global Climate Model. In this research, we attempt to
quantify the uncertainties associated with climate change projections explicitly using a
multi-model ensemble approach. The idea is to link state-of-the-art climate science to
actual decision-making, an idea much talked about but rarely tried out in practice. I
could easily chose some options from the WRP you have sent me and made some
assumptions about the link to climate, but then it would remain a purely academic
exercise. By eliciting options from you and your colleagues this work will be more
grounded in real world decision-making and hopefully more useful for your businesss
long term planning. As the Environment Agency (2004) notes, it is, of course, difficult
to assess the true risk to supply and resilience to drought when faced with the range of
uncertainties in the supply-demand balance. This is particularly the case for the
uncertainties introduced by climate change in the calculation of risk to the levels of
service. Your company has defined the Level of Service as:

243
 ? Restriction of the use of hosepipes not more than 1 in 10 years
? Use of Drought Orders to enforce restriction on non-essential uses and
secure raw water resources not more than 1 in 40 years
? Imposition of the use of standpipes not more than 1 in 100 years

I have devised this options elicitation exercise through a questionnaire that follows. I
also provide a worked example of question 6 in the end.

Do not hesitate to contact me should you have any questions. Looking forward to
hearing back from you.

Best regards,
Suraje

244
QUESTIONNAIRE

0. How long have you been working in water management? And in Anglian Water?
27 years for both

1. Do you think climate change is already happening? If yes, how?

Yes. Temperature, evaporation, rainfall, drought frequency.

2. Do you think it is important to plan ahead (i.e., in anticipation; in a non-reactive way)

in the water resources sector in order to cope with an uncertain changing climate? Yes.
If yes, do you think there is enough knowledge to do this? No. Do you think
probabilistic climate change projections would be more useful than climate change
scenarios? Yes.

3. Do you currently use seasonal climate forecasting (next 2 months to a year) data for
your decision-making? 2-4 months, but only loosely.

4. When do you initiate your drought plan? What are the triggers? See Drought Plan
(copy attached). Trigger curves for reservoirs explicitly linked to Levels of Service.

5. What is the current demand and supply (in Ml/d) for the whole region you supply and
for the Norwich and the Broads WRZ? Whole region: 1172 Ml/d. Norwich & Broads
WRZ8: 73 Ml/d. Suffolk & Essex WRZ10: 131 Ml/d (2004/05 figures).

6. As mentioned earlier, the idea is to link your companies adaptation

options/strategies to potential changes in the climate of the Anglian region. The unit of
choice to make this link is fairly important to the assessment. The closer it is to climate
variables, e.g., temperature, precipitation or effective rainfall change, the less
uncertainty there is e.g., compared to recharge or groundwater levels or even further
down the chain such as deployable output or WAFU. When we met in Norwich, you
told me you could work with precipitation change as you could translate that into
groundwater level and then assess options. Is that correct? At what temporal scale
would you require this; annual or seasonal (summer, winter)? Annual presents us with
some problems because there will actually be more precipitation in the future due to

245
larger increases in the winter (compared to decreases in the summer). I would prefer to
work with seasonal means, but I can easily aggregate things into annual means if that is
better for you. These are the climate variables that I can work with:

? temperature change
? precipitation change (preferred)
? potential evapotranspiration (PET) change (calculated using the
Thornthwaite method)
? effective rainfall change (precipitation minus PET)
? recharge change (calculated using the soil water budgeting approach)

Once we have identified which climate variable to use you can proceed with answering
the questionnaire. Based on the strategies identified in the Water Resources Plan and
other strategies that you can imagine please list your adaptation options to cope with a
drier climate that ranges from 0% to 50% change of the present value of the climate
variable youve identified. Please think hard about strategies other than the ones in the
WRP; consult with your colleagues if you feel necessary and have the time.

Please note that a certain level of aggregation of options might be necessary since your
WRP identifies hundreds of options. This is where your expert knowledge is invaluable.
Also, please try to abstract yourself from all the other pressures and drivers that affect
your sector (e.g., time-limited abstractions, environmental improvements, water quality,
etc.) when thinking about options to cope with a drier climate. Try to think of your
options only in terms of change in the climate variable in question and how you would
cope with that; assume other pressures, drivers and uncertainties remain at present
levels in the future.

Wed like you to do this for two spatial and temporal scales. In terms of spatial scale,
wed like you to do this for the whole region (where there will be more uncertainty
because of the aggregation of the options) and for a specific water resource zone such as
Norwich and The Broads (Ive listed the adaptation strategies I picked up from the WRP
in the end of the document). In terms of temporal scale, please consider your first
planning horizon the 2030s (what you planned for in your WRP) for the decrease in the
climate variable from 0% to 50%. Then imagine your planning horizon is the 2050s for

246
the same decrease in the climate variable from 0% to 50%. Assume demand remains
relatively stable as described in your WRP for the 2030s (although demand management
strategies will of course mitigate changes in the climate if the future is drier than the
present). Make the same assumption (that demand will remain stable) for the 2050s. In
principle, changing the planning horizon to 2050s should only change the
implementation date of the strategies you have identified so you dont need to list the
strategies all over again, just write down the implementation dates (however, I might be
wrong in assuming this).

Suffolk & Essex (WRZ10) selected instead of Norwich and the Broads
For each strategy (or bulk of strategies) please describe these characteristics:
a) How much percentage change of the climate variable (compared to the present) can
the strategy cope with? (e.g., 2%)
2030s: -10% JJA precipitation based upon Arnell medium scenario, relative to 1961-90
average (UKCIP02).
2050s: -22% JJA precipitation based upon UKWIR CL04C project output (2005), using
UKCIP02 medium-high scenario, relative to 1961-90 average, and regionalising to
CAMS catchments. Figure is for North Essex.

b) How much extra water (deployable output or WAFU) are you getting through this
strategy? (e.g., 10-25 Ml/d)
2030s: As I pointed out previously it is very difficult to unravel the degree to which
change in the climate variable (principally rainfall) contributes to the strategy because it
went into the 'melting pot' (HOURUS, the Monte Carlo simulation method - see WR
Plan Vol 3) with a number of other variables to determine target headroom. It is
possible to get a rough idea of the climate change component by comparing the
Headroom in Yr28 for Suffolk & Essex (25%) with WR zones where climate change is
not included (7 zones average ~12% [exceptions being Linc Fens, Ruthamford and S
Humberside]) i.e. ~ 50% HR due to CC. Given that WAFU would rise by 23 Ml/d in
WRZ10 (WRP Table: DRY YEAR_FP_EAST SUFFOLK & ESSEX_FINAL_ V.2
Sheet WRP 1), 0.5 x 23 =11.5 Ml/d through strategy due to CC.
2050s: No strategy developed in detail beyond 2030 yet! But likely to involve transfer
from the R Trent, development of pumped-storage reservoir in Lower Witham

247
catchment and adjacent to the Cut-Off Channel with a new transfer link. Possible need
for implementation of this strategy just prior to 2030. See WRP Vol 1, 4.10.20.

c) When does the strategy need to be implemented? (e.g., now, 2007, 2015)
2030: + 8 water resource yield Ml/d in 2006/07; net +15 Ml/d in 2016/17 (WRP Table:
DRY YEAR_FP_EAST SUFFOLK & ESSEX_FINAL_ V.2 Sheet WRP 1)
2050: possibly just prior to 2030, then phased 2030-2050.

d) How confident are you of the link between the strategy and the climate variable?
(e.g., low, medium, high confidence)
2030: medium
2050: low

e) Can you pursue this strategy on your own? If no, who do you have to negotiate
with?
2030: OFWAT, EA, E&SWCo, Tendring Hundred WS.
2050: OFWAT, EA, E&SWCo, Tendring Hundred WS.

f) How much does this strategy cost? (e.g., 100-250k )

2030: 33,870k (Alton extensions and effluent re-use resource scheme at Cliff Quay/or
possible ASR WRP Vol 1, 4.10.22)
2050: Has not been costed but capital costs for 2 new reservoirs and associated pumping
installations and trunk mains very substantial.

g) How energy intensive is this strategy? (e.g., low, medium, high)

2030: medium (but high if ASR replaces effluent re-use)
2050: high

h) Does this strategy have harmful environmental impacts? (e.g. yes, no, dont know,
explain)
2030: possibly yes, due to loss of effluent discharge to the Colne estuary.
2050: intended pump storage reservoirs would be filled from Trent transfers and
therefore not impact on local catchments, so environmental impacts likely to be low.

248
Id like you to collate options/strategies until the sum of all strategies for question 6(a)
is at least 50%. Im expecting options/strategies that are in the WRP to be better defined
than options beyond the WRP although aggregation of options (for the whole region)
could create a problem. When certain characteristics are not well defined please provide
a range (e.g., 10-25 Ml/d) or an estimated guess (e.g, the cost could be something
between 100-250k ). I acknowledge that sometimes you might be unable to describe
certain characteristics, in which case write unknown.

7. Where strategies are heavily influenced by climate change, who do you think should
pay for its implementation? Should the costs be externalised to customers?
No long-term strategic view due to effect of 5-year regulatory cycle. Currently costs borne by government
and new customers.

8. You are currently a bulk supply exporter of water to other companies (of the order of
108 Ml/d); do you think youll have to suspend this export if climate change has a
significant impact?
The export is re-negotiable in either direction.

9. Did you do a sensitivity analysis of what drives your target headroom? If yes, which
are the main drivers and how does climate change rank?
Yes, see WR Plan: e.g. Figure 2 in Vol 2 (ref. Climate, nitrate, pollution etc). CC ranks high in
Ruthamford and Suffolk & Essex, medium-high in Lincolnshire Fens and S Humberside, low in other
WRZs.

END

249
Here is a worked example of question 6 for the whole Anglia region based on my
interpretation of the available strategies:

A. Demand management through:

1. Household metering
2. Leakage control
3. Promotion of water efficiency
B. Development of water resources (supply management):
1. Transfer water from areas of surplus to areas of deficit
2. Development of licensed resources (either groundwater or surface) at
Clapham, Ixworth, Norwich and Colchester
3. Reinforcement of trunk mains
4. Import of water from the west of the country via the Midlands and the
River Trent
5. Re-use of treated effluent currently discharged to tidal waters

Characteristics of strategy B5:

a) How much effective rainfall change can the strategy cope with? 2%
b) How much extra water are you getting through this strategy? 10-25 Ml/d
c) When does the strategy need to be implemented? Between 2007 and 2010
d) How confident are you of the link between the strategy and the climate variable?
medium confidence
e) Can you pursue this strategy on your own? If no, who do you have to negotiate
with? With the Environment Agency and Ofwat
f) How much does this strategy cost? 250k
g) How energy intensive is this strategy? High
h) This strategy has a positive environmental impact by avoiding discharge of
effluent into the sea; e.g., through improvement of bathing water quality.

Other adaptation strategies elicited from our interview in 12 February 2004:

New reservoir
Reservoir extensions
De-salination plant

250
 Winter storage (rainwater harvesting)
Wetland creation

For WRZ8, Norwich and The Broads

Base year demand: 42.3 Ml/d (total domestic) + 15.7 Ml/d (commercial)
2030 demand: 41.8 Ml/d (total domestic) + 15.4 Ml/d (commercial)
5 out of 6 planning zones are projected to have headroom deficits against dry year
average and critical period forecasts.

Options to maintain target headroom:

? Generic demand management options (additional customer metering and
targeted domestic water audits and cistern displacement devices as well as the
use of additional leakage control to reduce leakage to below the current ELL)
? Resource development:
o a number of internal trunk main and storage schemes needed
o development of 2 existing, but unused, boreholes in the Yare Valley at
Marlingford and Barford and uprating of the Thorpe St Andrews/
Mousehold sourceworks
o import water to Norwich through the strategic development of the Stoke
Ferry WTW in the longer term

References

Anglian Water Services Ltd. (2004) Water Resources Plan, April 2004. Anglian Water.

Environment Agency (2004) Maintaining water supply, July 2004. Environment

Agency.

251
 LIST OF REFERENCES

Adger, W. N. 1999. Social vulnerability to climate change and extremes in coastal

Vietnam. World Development 27:249-269.
Adger, W. N., N. W. Arnell, and E. L. Tompkins. 2005. Successful adaptation to
climate change across scales. Global Environmental Change 15:77-86.
Ahmand, Q. K., R. A. Warrick, T. E. Downing, S. Nishioka, K. S. Parikh, C. Parmesan,
S. H. Schneider, F. Toth, and G. Yohe. 2001. Methods and tools. Pages 105-143
in IPCC, editor. Climate change 2001: impacts, adaptation, and vulnerability.
Cambridge University Press, Cambridge.
Alcamo, J., editor. 1994. IMAGE 2.0: Integrated modeling of global climate change.
Kluwer Academic Press, Dordrecht.
Alcamo, J., G. J. J. Kreileman, and R. Leemans, editors. 1998. Global change scenarios
for the 21st century. Pergamon/Elsevier Science, Oxford.
Alexander, L. V., and P. D. Jones. 2001. Updated precipitation series for the U.K. and
discussion of recent extremes. Atmospheric Science Letters 1.
Allen, M. R., and W. J. Ingram. 2002. Constraints on future changes in climate and the
hydrologic cycle. Nature 419:224-232.
Allen, M. R., P. A. Stott, J. F. B. Mitchell, R. Schnur, and T. L. Delworth. 2000.
Quantifying the uncertainty in forecasts of anthropogenic climate change. Nature
407:617-620.
Andronova, N. G., and M. E. Schlesinger. 2001. Objective estimation of the probability
density function for climate sensitivity. Journal of Geophysical Research-
Atmospheres 106:22605-22611.
Andronova, N. G., M. E. Schlesinger, S. Dessai, M. Hulme, and B. Li. 2005. The
Concept of Climate Sensitivity: History and Development. in M. E. Schlesinger,
editor. Human-Induced Climate Change: An Interdisciplinary Assessment.
Cambridge University Press, Cambridge.
Arnell, N. 2000. Thresholds and response to climate change forcing: the water sector.
Climatic Change 46:305-316.
Arnell, N. W. 1998. Climate change and water resources in Britain. Climatic Change
39:83-110.

252
Arnell, N. W. 2003a. Effects of climate change on river flows and groundwater
recharge: UKCIP02 scenarios. Report No. 03/CL/04/2, UK Water Industry
Research.
Arnell, N. W. 2003b. Relative effects of multi-decadal climatic variability and changes
in the mean and variability of climate due to global warming: future streamflows
in Britain. Journal of Hydrology 270:195-213.
Arnell, N. W. 2004a. Climate change and global water resources: SRES emissions and
socio-economic scenarios. Global Environmental Change-Human and Policy
Dimensions 14:31-52.
Arnell, N. W. 2004b. Climate-change impacts on river flows in Britain: The UKCIP02
scenarios. Water and Environment Journal 18:112-117.
Arnell, N. W., M. J. L. Livermore, S. Kovats, P. E. Levy, R. Nicholls, M. L. Parry, and
S. R. Gaffin. 2004. Climate and socio-economic scenarios for global-scale
climate change impacts assessments: characterising the SRES storylines. Global
Environmental Change-Human and Policy Dimensions 14:3-20.
Arnell, N. W., N. Reynard, R. King, C. Prudhomme, and J. Branson. 1997. Effects of
climate change on river flows and groundwater recharge: guidelines for resource
assessment. Report No. 97/CL/04/1, UKWIR/Environment Agency.
Arnell, N. W., and N. S. Reynard. 1996. The effects of climate change due to global
warming on river flows in Great Britain. Journal of Hydrology 183:397-424.
Arribas, A., C. Gallardo, M. A. Gaertner, and M. Castro. 2003. Sensitivity of the Iberian
Peninsula climate to a land degradation. Climate Dynamics 20:477-489.
AWS. 2003. Impact of climate change on Anglian Water surface water sources. Anglian
Water Services Ltd, Huntingdon & Cambridge.
AWS. 2004a. FORWARD - Description of the scheme costing scheduling and
calculation of average incremental social cost. RPS Water Services, Edinburgh.
AWS. 2004b. Water Resources Plan. Anglian Water Services Ltd.
Bakker, K. 1999. Deconstructing discourses of drought. Transactions of the Institute of
British Geographers 24:367-372.
Bakker, K. J. 2000. Privatizing water, producing scarcity: The Yorkshire drought of
1995. Economic Geography 76:4-27.
Bakker, K. J. 2003. From public to private to ... mutual? - Restructuring water supply
governance in England and Wales. Geoforum 34:359-374.

253
Bankes, S. 1993. Exploratory Modeling for Policy Analysis. Operations Research
41:435-449.
Bankes, S. C. 2002. Tools and techniques for developing policies for complex and
uncertain systems. Proceedings of the National Academy of Sciences of the
United States of America 99:72637266.
Barnett, J. 2001. Adapting to climate change in Pacific Island Countries: The problem
of uncertainty. World Development 29:977-993.
Barrauqu, B. 2003. Past and future sustainability of water policies in Europe. Natural
Resources Forum 27:200-211.
Barrow, E., and M. Hulme. 1997. Describing the surface climate of the British Isles.
Pages 33-62 in M. Hulme and E. Barrow, editors. Climate of the British Isles:
present, past and future. Routledge, London.
Bass, B., G. H. Huang, and J. Russo. 1997. Incorporating climate change into risk
assessment using grey mathematical programming. Journal of Environmental
Management 49:107-123.
Benestad, R. E. 2004. Tentative probabilistic temperature scenarios for northen Europe.
Tellus 56A:89-101.
Ben-Haim, Y. 2001. Information-gap decision theory: decisions under severe
uncertainty. Academic Press.
Berkhout, F., and J. Hertin. 2000. Socio-economic scenarios for climate impact
assessment. Global Environmental Change 10:165-168.
Berkhout, F., J. Hertin, and A. Jordan. 2002. Socio-economic futures in climate change
impact assessment: using scenarios as 'learning machines'. Global
Environmental Change 12:83-95.
Blaikie, P., T. Cannon, I. Davis, and B. Wisner. 1994. At risk: natural hazards, people's
vulnerability, and disasters. Routledge, London.
Boorman, D. B., and C. E. M. Sefton. 1997. Recognising the uncertainty in the
quantification of the effects of climate change on hydrological response.
Climatic Change 35:415-434.
Boville, B. A., J. T. Kiehl, P. J. Rasch, and F. O. Bryan. 2001. Improvements to the
NCAR CSM-1 for transient climate simulations. Journal of Climate 14:164-179.
Burton, I. 1997. Vulnerability and adaptive response in the context of climate and
climate change. Climatic Change 36:185-196.

254
Burton, I., S. Huq, B. Lim, O. Pilifosova, and E. L. Schipper. 2002. From impacts
assessment to adaptation priorities: the shaping of adaptation policy. Climate
Policy 2:145-159.
Caldeira, K., A. K. Jain, and M. I. Hoffert. 2003. Climate sensitivity uncertainty and the
need for energy without CO2 emission. Science 299:2052-2054.
Campolongo, F., J. Kleijnen, and T. Andres. 2000. Screening methods. Pages 65-80 in
A. Saltelli, K. Chan, and E. M. Scott, editors. Sensitivity Analysis. Wiley,
Chichester.
Carnell, J., J. D. Lawson, P. H. von Lany, and R. M. J. Scarrott. 1999. Water supply and
demand balances: Converting uncertainty into headroom. Journal of the
Chartered Institution of Water and Environmental Management 13:413-419.
Carter, T. R., E. L. La Rovere, R. N. Jones, R. Leemas, L. O. Mearns, N. Nakicenovic,
A. B. Pittock, S. M. Semenov, and J. Skea. 2001. Developing and applying
scenarios. Pages 145-190 in IPCC, editor. Climate change 2001: impacts,
adaptation, and vulnerability. Cambridge University Press, Cambridge.
Carter, T. R., M. Parry, H. Harasawa, and S. Nishioka. 1994. IPCC technical guidelines
for assessing climate change impacts and adaptation. Department of Geography,
University College London, UK and the Centre for Global environmental
Research, National Institute for Environmental Studies, Japan.
Cash, D. W., W. C. Clark, F. Alcock, N. M. Dickson, N. Eckley, D. H. Guston, J. Jager,
and R. B. Mitchell. 2003. Knowledge systems for sustainable development.
Proceedings of the National Academy of Sciences of the United States of
America 100:8086-8091.
Casman, E. A., M. G. Morgan, and H. Dowlatabadi. 1999. Mixed levels of uncertainty
in complex policy models. Risk Analysis 19:33-42.
CCIRG. 1991. The potential effects of climate change in the United Kingdom.
Department of the Environment, London.
CCIRG. 1996. Review of the potential effects of climate change in the United Kingdom.
Department of the Environment, London.
Chao, H. P. 1995. Managing the risk of global climate catastrophe - an uncertainty
analysis. Risk Analysis 15:69-78.
Christensen, J. H., J. Raisanen, T. Iversen, D. Bjorge, O. B. Christensen, and M.
Rummukainen. 2001. A synthesis of regional climate change simulations - A
Scandinavian perspective. Geophysical Research Letters 28:1003-1006.

255
CIRIA. 2005. Climate change risks in building - an introduction. Construction Industry
Research and Information Association, London.
Clark, M. P., and R. S. Pulwarty. 2003. Devising resilient responses to potential climate
change impacts. Ogmius - Newsletter of the Center for Science and Technology
Policy Research:No. 5, 2-3.
Clemen, R. T., and R. L. Winkler. 1999. Combining probability distributions from
experts in risk analysis. Risk Analysis 19:187-203.
Comission. 2005. Winning the battle against global climate change: background paper.
Comission of the European Communities, Brussels.
Conde, C., and K. Lonsdale. 2004. Stakeholder engagement in the adaptation process. in
B. Lim, editor. Adaptation Policy Framework. Cambridge University Press.
Conway, D. 1998. Recent climate variability and future climate change scenarios for
Great Britain. Progress in Physical Geography 22:350-374.
Craig, S. G., and K. J. Holmen. 1995. Uncertainties in Future CO2 Projections. Global
Biogeochemical Cycles 9:139-152.
Crawford-Brown, D. 2005. The concept of 'sounds science' in risk management
decisions. Risk Management: An International Journal 7:7-20.
CSIRO. 1992. Climate Change Scenarios for Australia. Climate Impact Group, CSIRO,
Division of Atmospheric Research, Melbourne.
CSIRO. 1996. Climate Change Scenarios for Australia. Climate Impact Group, CSIRO,
Division of Atmospheric Research, Melbourne.
CSIRO. 2001. Climate Change Scenarios for Australia. Climate Impact Group, CSIRO,
Division of Atmospheric Research, Melbourne.
Cubasch, U., G. A. Meehl, G. J. Boer, R. J. Stouffer, M. Dix, A. Noda, C. A. Senior, S.
Raper, and K. S. Yap. 2001. Projections of Future Climate Change. Pages 525-
582 in J. T. Houghton, Y. Ding, D. J. Griggs, M. Noguer, P. J. van der Linden,
X. Dai, K. Maskell, and C. A. Johnson, editors. Climate Change 2001: The
Scientific Basis. Cambridge University Press, Cambridge.
Davies, T., P. M. Kelly, and T. J. Osborn. 1997. Explaining the climate of the British
Isles. Pages 11-32 in M. Hulme and E. Barrow, editors. Climate of the British
Isles: present, past and future. Routledge, London.
de Lande Long, P. G., and C. W. Scott. 2000. Engineering feasibility studies for a large
reservoir in the Fens. Proceedings of the Institution of Civil Engineers-Water
and Maritime Engineering 148:25-37.

256
de Lo, R., R. Kreutzwiser, and L. Moraru. 2001. Adaptation options for the near term:
climate change and the Canadian water sector. Global Environmental Change
11:231-245.
de Vries, B., M. Janssen, and A. Beusen. 1999. Perspectives on Global Energy Futures -
Simulations with the TIME model. Energy Policy 27:477-494.
DEFRA. 2002. Directing the flow: priorities for future water policy. Department for
Environment, Food and Rural Affairs.
DEFRA. 2004. Principal guidance from the Secretary of State to the Director-General of
Water Services: 2004 periodic review of water price limits. Department for
Environment, Food and Rural Affairs, London.
Dessai, S. 2003. Heat stress and mortality in Lisbon Part II. An assessment of the
potential impacts of climate change. International Journal of Biometeorology
48:37-44.
Dessai, S., and M. Hulme. 2001. Climatic implications of revised IPCC emission
scenarios, the Kyoto Protocol and quantification of uncertainties. Integrated
Assessment 2:159-170.
Dessai, S., and M. Hulme. 2003. Does climate policy need probabilities? Tyndall
Working Paper No. 34, Tyndall Centre for Climate Change Research, Norwich.
Dessai, S., and M. Hulme. 2004. Does climate adaptation policy need probabilities?
Climate Policy 4:107-128.
Dessai, S., X. Lu, and M. Hulme. 2005a. Limited sensitivity analysis of regional climate
change probabilities for the 21st century. Journal of Geophysical Research-
Atmospheres 110:D19108.
Dessai, S., X. F. Lu, and J. S. Risbey. 2005b. On the role of climate scenarios for
adaptation planning. Global Environmental Change-Human and Policy
Dimensions 15:87-97.
Diaz-Nieto, J., and R. L. Wilby. 2005. A comparison of statistical downscaling and
climate change factor methods: Impacts on low flows in the River Thames,
United Kingdom. Climatic Change 69:245-268.
Dixon, K. W., and J. R. Lanzante. 1999. Global mean surface air temperature and North
Atlantic overturning in a suite of coupled GCM climate change experiments.
Geophysical Research Letters 26:1885-1888.
Dowlatabadi, H., and M. G. Morgan. 1993. A Model Framework for Integrated Studies
of the Climate Problem. Energy Policy 21:209-221.

257
Downing, T. E., R. E. Butterfield, B. Edmonds, J. W. Knox, S. Moss, B. S. Piper, E. K.
Weatherhead, and a. t. C. D. p. team). 2003. Climate Change and the Demand
for Water. Stockholm Environment Institute Oxford Office, Oxford.
DTI. 1999. Environmental Futures. Foresight Energy and Natural Resoureces Panel,
Office of Science and Technology, Department of Trade and Industry, London.
Dubrovsky, M., I. Nemesova, and J. Kalvona. 2005. Uncertainties in climate change
scenarios for the Czech Republic. Climate Research 29:139-156.
EA. 2001a. A scenario approach to water demand forecasting. Environment Agency,
Worthing.
EA. 2001b. Water resources for the future: a strategy for Anglian region. Environment
Agency, Peterborough.
EA. 2001c. Water resources for the future: a strategy for England and Wales.
Environment Agency.
EA. 2003a. Securing water supply: The Environment Agency's advice to Ministers on
the draft water resources plans submitted by water companies as part of the 2004
periodic review. Environment Agency.
EA. 2003b. Water Resources Planning Guideline. Environment Agency, Bristol.
EA. 2003c. Water resources planning guidelines version 3.3 (+Supplementary guidance
notes). Environment Agency.
EA. 2004a. Maintaining water supply:The Environment Agency's advice to Ministers
on the final water resources plans submitted by water companies as part of the
2004 periodic review. Environment Agency.
EA. 2004b. Review of water company water resources plans. Environment Agency.
Edmonds, J., M. Wise, H. Pithcer, R. Richels, T. Wigley, and C. N. MacCracken. 1996.
An integrated assessment of climate change and the accelerated introduction of
advanced energy technology. Mitigation and Adaptation Strategies for Global
Change 1:311-339.
EERA, and SDRT. 2004. Living with climate chage in the East of England: summary
report. East of England Regional Assembly and Sustainable Development
Round Table.
Ekstrm, M., B. Hingray, A. Mezghani, and P. D. Jones. 2005. Regional climate model
data used within the SWURVE project 2: addressing uncertainty in regional
climate model data for five European case study areas. Hydrology & Earth
System Sciences:(in press).

258
Fankhauser, S. 1995. Valuing climate change: the economics of the greenhouse.
Earthscan, London.
Fankhauser, S., J. B. Smith, and R. S. J. Tol. 1999. Weathering climate change: some
simple rules to guide adaptation decisions. Ecological Economics 30:67-78.
Feenstra, J. F., I. Burton, J. B. Smith, and R. S. J. Tol, editors. 1998. Handbook on
methods for climate change, impact assessment and adaptation strategies. UNEP
and IVM.
Flato, G. M., and G. J. Boer. 2001. Warming asymmetry in climate change simulations.
Geophysical Research Letters 28:195-198.
Flato, G. M., G. J. Boer, W. G. Lee, N. A. McFarlane, D. Ramsden, M. C. Reader, and
A. J. Weaver. 2000. The Canadian Centre for Climate Modelling and Analysis
global coupled model and its climate. Climate Dynamics 16:451-467.
Forest, C. E., M. R. Allen, A. P. Sokolov, and P. H. Stone. 2001. Constraining climate
model properties using optimal fingerprint detection methods. Climate
Dynamics 18:277-295.
Forest, C. E., M. R. Allen, P. H. Stone, and A. P. Sokolov. 2000. Constraining
uncertainties in climate models using climate change detection techniques.
Geophysical Research Letters 27:569-572.
Forest, C. E., P. H. Stone, A. P. Sokolov, M. R. Allen, and M. D. Webster. 2002.
Quantifying uncertainties in climate system properties with the use of recent
climate observations. Science 295:113-117.
Frederick, K. D. 1997. Adapting to climate impacts on the supply and demand for
water. Climatic Change 37:141-156.
French, S. 1995. Uncertainty and imprecision: modelling and analysis. Journal of the
Operational Research Society 46:70-79.
Funtowicz, S. O., and J. R. Ravetz. 1990. Uncertainty and quality in science for policy.
Kluwer Academic Publishers, Dordrecht ; Boston.
Funtowicz, S. O., and J. R. Ravetz. 1993. Science for the Post-Normal Age. Futures
25:739-755.
Georgakakos, K. P. 2003. Probabilistic climate-model diagnostics for hydrologic and
water resources impact studies. Journal of Hydrometeorology 4:92-105.
Gibelin, A. L., and M. Deque. 2003. Anthropogenic climate change over the
Mediterranean region simulated by a global variable resolution model. Climate
Dynamics 20:327-339.

259
Giorgi, F., and R. Francisco. 2000. Evaluating uncertainties in the prediction of regional
climate change. Geophysical Research Letters 27:1295-1298.
Giorgi, F., B. Hewitson, J. H. Christensen, M. Hulme, H. Von Storch, P. H. Whetton, R.
Jones, L. O. Mearns, and C. Fu. 2001a. Regional climate information -
evaluation and projections. Pages 583-638 in IPCC, editor. Climate change
2001: the scientific basis. Cambridge University Press, Cambridge.
Giorgi, F., Y. Huang, K. Nishizawa, and C. B. Fu. 1999. A seasonal cycle simulation
over eastern Asia and its sensitivity to radiative transfer and surface processes.
Journal of Geophysical Research-Atmospheres 104:6403-6423.
Giorgi, F., M. R. Marinucci, and G. T. Bates. 1993a. Development of a 2nd-Generation
Regional Climate Model (Regcm2) .1. Boundary-Layer and Radiative-Transfer
Processes. Monthly Weather Review 121:2794-2813.
Giorgi, F., M. R. Marinucci, G. T. Bates, and G. Decanio. 1993b. Development of a
2nd-Generation Regional Climate Model (Regcm2) .2. Convective Processes
and Assimilation of Lateral Boundary-Conditions. Monthly Weather Review
121:2814-2832.
Giorgi, F., and L. O. Mearns. 2002. Calculation of average, uncertainty range, and
reliability of regional climate changes from AOGCM simulations via the
"reliability ensemble averaging'' (REA) method. Journal of Climate 15:1141-
1158.
Giorgi, F., and L. O. Mearns. 2003. Probability of regional climate change based on the
Reliability Ensemble Averaging (REA) method. Geophysical Research Letters
30:art. no.-1629.
Giorgi, F., L. O. Mearns, C. Shields, and L. McDaniel. 1998. Regional nested model
simulations of present day and 2 x CO2 climate over the central plains of the
U.S. Climatic Change 40:457-493.
Giorgi, F., P. H. Whetton, R. G. Jones, J. H. Christensen, L. O. Mearns, B. Hewitson, H.
von Storch, R. Francisco, and C. Jack. 2001b. Emerging patterns of simulated
regional climatic changes for the 21st century due to anthropogenic forcings.
Geophysical Research Letters 28:3317-3320.
Goodess, C. M., T. J. Osborn, and M. Hulme. 2003. The identification and evaluation of
suitable scenario development methods for the estimation of future probabilities
of extreme weather events. Tyndall Centre Technical Report 4, Tyndall Centre
for Climate Change Research, Norwich.

260
Goodman, D. 2002. Extrapolation in risk assessment: Improving the quantification of
uncertainty, and improving information to reduce the uncertainty. Human and
Ecological Risk Assessment 8:177-192.
Gordon, C., C. Cooper, C. A. Senior, H. Banks, J. M. Gregory, T. C. Johns, J. F. B.
Mitchell, and R. A. Wood. 2000. The simulation of SST, sea ice extents and
ocean heat transports in a version of the Hadley Centre coupled model without
flux adjustments. Climate Dynamics 16:147-168.
Gordon, H. B., and S. P. O'Farrell. 1997. Transient climate change in the CSIRO
coupled model with dynamic sea ice. Monthly Weather Review 125:875-907.
Gregory, J. M., P. D. Jones, and T. M. L. Wigley. 1991. Precipitation in Britain - an
Analysis of Area-Average Data Updated to 1989. International Journal of
Climatology 11:331-345.
Gregory, J. M., R. J. Stouffer, S. C. B. Raper, P. A. Stott, and N. A. Rayner. 2002. An
observationally based estimate of the climate sensitivity. Journal of Climate
15:3117-3121.
Gritsevskyi, A., and N. Nakicenovic. 2000. Modeling uncertainty of induced
technological change. Energy Policy 28:907-921.
Groves, D. G., S. Matyac, and T. Hawkins. 2005. Quantified scenarios of 2030
California water demand. California Department of Water Resources.
Grubler, A., and N. Nakicenovic. 2001. Identifying dangers in an uncertain climate.
Nature 412:15.
Ha Duong, M. 2003. Imprecise probability bridges scenario-forecast gap. (unpublished
manuscript).
Hams, T., and D. Taylor. 1999. Water resources project report. East of England
Sustainable Development Round Table and the Water Resources Group.
Handmer, J. W., S. Dovers, and T. E. Downing. 1999. Societal vulnerability to climate
change and variability. Mitigation and Adaptation Strategies for Global Change
4:267-281.
Hashimoto, T., D. P. Loucks, and J. R. Stedinger. 1982. Robustness of water resources
systems. Water Resources Research 18:21-26.
Hassan, J. 1998. A history of water in modern England and Wales. Manchester
Unievrsity Press, Manchester.

261
Haughton, G. 1998. Private profits - public drought: the creation of a crisis in water
management for West Yorkshire. Transactions of the Institute of British
Geographers 23:419-433.
Helton, J. C., and D. E. Burmaster. 1996. Guest editorial: Treatment of aleatory and
epistemic uncertainty in performance assessments for complex systems.
Reliability Engineering & System Safety 54:91-94.
Helton, J. C., and F. J. Davis. 2002. Illustration of sampling-based methods for
uncertainty and sensitivity analysis. Risk Analysis 22:591-622.
Helton, J. C., and F. J. Davis. 2003. Latin hypercube sampling and the propagation of
uncertainty in analyses of complex systems. Reliability Engineering & System
Safety 81:23-69.
Hingray, B., A. Mezghani, and T. A. Buishand. 2005. Development of probability
distributions for regional climate change from uncertain global mean warming
and an uncertain scaling relationship. Hydrology & Earth System Sciences:(in
press).
Hobbs, B. F. 1997. Bayesian methods for analysing climate change and water resource
uncertainties. Journal of Environmental Management 49:53-72.
Hobbs, B. J., P. T. Chao, and B. N. Venkatesh. 1997. Using decision analysis to include
climate change in water resources decision making. Climatic Change 37:177-
202.
Hoffert, M. I., K. Caldeira, G. Benford, D. R. Criswell, C. Green, H. Herzog, A. K. Jain,
H. S. Kheshgi, K. S. Lackner, J. S. Lewis, H. D. Lightfoot, W. Manheimer, J. C.
Mankins, M. E. Mauel, L. J. Perkins, M. E. Schlesinger, T. Volk, and T. M. L.
Wigley. 2002. Advanced technology paths to global climate stability: Energy for
a greenhouse planet. Science 298:981-987.
Holman, I. P., R. J. Nicholls, P. M. Berry, P. A. Harrison, E. Audsley, S. Shackley, and
M. D. A. Rounsevell. 2005a. A regional, multi-sectoral and integrated
assessment of the impacts of climate and socio-economic change in the UK. Part
2. Results. Climatic Change 71:43-73.
Holman, I. P., M. D. A. Rounsevell, S. Shackley, P. A. Harrison, R. J. Nicholls, P. M.
Berry, and E. Audsley. 2005b. A regional, multi-sectoral and integrated
assessment of the impacts of climate and socio-economic change in the UK. Part
1. Methodology. Climatic Change 71:9-41.

262
Hope, C., J. Anderson, and P. Wenman. 1993. Policy Analysis of the Greenhouse-Effect
- an Application of the Page Model. Energy Policy 21:327-337.
Hulme, M., E. M. Barrow, N. W. Arnell, P. A. Harrison, T. C. Johns, and T. E.
Downing. 1999. Relative impacts of human-induced climate change and natural
climate variability. Nature 397:688-691.
Hulme, M., and T. R. Carter. 1999. Representing uncertainty in climate change
scenarios and impact studies. Pages 11-37 in T. R. Carter, M. Hulme, and D.
Viner, editors. Representing uncertainty in climate change scenarios and impact
studies - ECLAT-2 Workshop Report. Climatic Research Unit, Norwich, UK.
Hulme, M., and G. J. Jenkins. 1998. Climate change scenarios for the UK: scientific
report. Climatic Research Unit, Norwich.
Hulme, M., G. J. Jenkins, X. Lu, J. R. Turnpenny, T. D. Mitchell, R. G. Jones, J. Lowe,
J. M. Murphy, D. Hassell, P. Boorman, R. McDonald, and S. Hill. 2002. Climate
Change Scenarios for the United Kingdom: The UKCIP02 Scientific Report.
Tyndall Centre for Climate Change Research, Norwich.
Hulme, M., and P. D. Jones. 1989. Climatic change scenarios for the UK. A report
prepared for the Institute of Hydrology under contract No. F3CR05-C1-31-01,
Climatic Research Unit, Norwich.
Hulme, M., and J. Turnpenny. 2004. Understanding and managing climate change: The
UK experience. Geographical Journal 170:105-115.
IPCC. 2001a. Climate change 2001: impacts, adaptation, and vulnerability. Cambridge
University Press, Cambridge.
IPCC. 2001b. Climate change 2001: the scientific basis. Cambridge University Press,
Cambridge.
IPCC-TGCIA. 1999. Guidelines on the Use of Scenario Data for Climate Impact and
Adaptation Assessment. Version 1. Intergovernmental Panel on Climate Change,
Task Group on Scenarios for Climate Impact Assessment.
Jacob, D. 2001. A note to the simulation of the annual and inter-annual variability of the
water budget over the Baltic Sea drainage basin. Meteorology and Atmospheric
Physics 77:61-73.
Johns, T. C., R. E. Carnell, J. F. Crossley, J. M. Gregory, J. F. B. Mitchell, C. A. Senior,
S. F. B. Tett, and R. A. Wood. 1997. The second Hadley Centre coupled ocean-
atmosphere GCM: Model description, spinup and validation. Climate Dynamics
13:103-134.

263
Johnson, C., and J. Handmer. 2002. Water supply in England and Wales: whose
responsibility is it when things go wrong? Water Policy 4:345-366.
Jones, C. D., P. Cox, and C. Huntingford. 2003. Uncertainty in climate-carbon-cycle
projections associated with the sensitivity of soil respiration to temperature.
Tellus Series B-Chemical and Physical Meteorology 55:642-648.
Jones, P. D. 1984. Riverflow Reconstruction from Precipitation Data. Journal of
Climatology 4:171-186.
Jones, P. D., and D. Conway. 1997. Precipitation in the British Isles: An analysis of
area-average data updated to 1995. International Journal of Climatology 17:427-
438.
Jones, P. D., and D. H. Lister. 1998. Riverflow reconstructions for 15 catchments over
England and Wales and an assessment of hydrologic drought since 1865.
International Journal of Climatology 18:999-1013.
Jones, P. D., D. H. Lister, and E. Kostopoulou. 2004. Reconstructed river flow series
from 1860s to present: updating previously reconstructed series to 2002. Science
Report SC040052/SR, Environment Agency, Bristol.
Jones, R. G., J. M. Murphy, and M. Noguer. 1995. Simulation of climate change over
Europe using a nested regional-climate model. I: assessment of control climate,
including sensitivity to location of lateral boundaries. Quarterly Journal of the
Royal Meteorological Society 121.
Jones, R. N. 2000a. Analysing the risk of climate change using an irrigation demand
model. Climate Research 14:89-100.
Jones, R. N. 2000b. Managing uncertainty in climate change projections - Issues for
impact assessment. Climatic Change 45:403-419.
Jones, R. N. 2001. An environmental risk assessment/management framework for
climate change impact assessments. Natural Hazards 23:197-230.
Jones, R. N., and C. M. Page. 2001. Assessing the risk of climate change on the water
resources of the Macquire river catchment. Pages 673-678 in F. Ghassemi, P. H.
Whetton, R. Little, and M. Littleboy, editors. Integrating models of natural
resources management across disciplines, issues and scales. Modelling and
Simulation Society of Australia and New Zealand, Canberra.
Jones, S. A., B. Fischhoff, and D. Lach. 1999. Evaluating the science-policy interface
for climate change research. Climatic Change 43:581-599.

264
Kahneman, D., P. Slovic, and A. Tversky. 1982. Judgement under uncertainty :
heuristics and biases. Cambridge University Press, Cambridge.
Katz, R. W. 2002. Techniques for estimating uncertainty in climate change scenarios
and impact studies. Climate Research 20:167-185.
Kelly, P. M., and W. N. Adger. 2000. Theory and practice in assessing vulnerability to
climate change and facilitating adaptation. Climatic Change 47:325-352.
Kittel, T. G. F., F. Giorgi, and G. A. Meehl. 1998. Intercomparison of regional biases
and doubled CO2-sensitivity of coupled atmosphere-ocean general circulation
model experiments. Climate Dynamics 14:1-15.
Klein, R. J. T., E. L. F. Schipper, and S. Dessai. 2005. Integrating mitigation and
adaptation into climate and development policy: three research questions.
Environmental Science & Policy 8:579-588.
Knight, F. 1922. Risk, uncertainty and profit. Houghton Mifflin, London.
Knutti, R., T. F. Stocker, F. Joos, and G. K. Plattner. 2002. Constraints on radiative
forcing and future climate change from observations and climate model
ensembles. Nature 416:719-723.
Knutti, R., T. F. Stocker, F. Joos, and G. K. Plattner. 2003. Probabilistic climate change
projections using neural networks. Climate Dynamics 21:257-272.
Kriegler, E., and H. Held. 2005. Utilizing belief functions for the estimation of future
climate change. International Journal of Approximate Reasoning 39:185-209.
Kuhn, T. 1962. The structure of scientific revolutions. University of Chicago Press,
Chicago.
Lambert, F. H., P. A. Stott, M. R. Allen, and M. A. Palmer. 2004. Detection and
attribution of changes in 20th century land precipitation. Geophysical Research
Letters 31:art. no.-L10203.
Lambert, S. J., and G. J. Boer. 2001. CMIP1 evaluation and intercomparison of coupled
climate models. Climate Dynamics 17:83-106.
Leggett, J., W. Pepper, A. Sankovski, J. B. Smith, R. S. J. Tol, and T. M. L. Wigley.
2003. Climate change risk analysis framework (CCRAF) - a probabilistic tool
for analyzing climate change uncertainties. in Poster presented at the EGS-
AGU-EUG Joint Assembly, Nice, France.
Lempert, R., N. Nakicenovic, D. Sarewitz, and M. Schlesinger. 2004. Characterizing
climate-change uncertainties for decision-makers - An editorial essay. Climatic
Change 65:1-9.

265
Lempert, R. J. 2002. A new decision sciences for complex systems. PNAS 99:7309-
7313.
Lempert, R. J., D. G. Groves, S. W. Popper, and S. C. Bankes. 2005. A General,
Analytic Method for Generating Robust Strategies and Narrative Scenarios.
Managment Science:(in press).
Lempert, R. J., S. W. Popper, and S. C. Bankes. 2003. Shaping the next one hundred
years: new methods for quantitative, long-term policy analysis. RAND, Santa
Monica, CA.
Lempert, R. J., and M. E. Schlesinger. 2000. Robust strategies for abating climate
change. Climatic Change 45:387-401.
Lempert, R. J., M. E. Schlesinger, and S. C. Bankes. 1996. When we don't know the
costs or the benefits: Adaptive strategies for abating climate change. Climatic
Change 33:235-274.
Lempert, R. J., M. E. Schlesinger, S. C. Bankes, and N. G. Andronova. 2000. The
impacts of climate variability on near-term policy choices and the value of
information. Climatic Change 45:129-161.
Lenderink, G., B. van den Hurk, E. van Meijgaard, A. van Ulden, and H. Cuijpers.
2003. Simulation of present-day climate in RACMO2: first results and model
development. Technical Report TR-252, KMNI.
Lim, B., E. Spanger-Siegfried, I. Burton, E. L. Malone, and S. Huq. 2005. Adaptation
policy frameworks for climate change: developing strategies, policies, and
measures. Cambridge University Press, Cambridge.
Linacre, E. 1992. Climate data and resources : a reference and guide. Routledge,
London.
Lorenz, E. N. 1993. The essence of chaos. University of Washington Press, Seattle.
Luo, W. B., and B. Caselton. 1997. Using Dempster-Shafer theory to represent climate
change uncertainties. Journal of Environmental Management 49:73-93.
Lutz, W., W. Sanderson, and S. Scherbov. 1997. Doubling of world population unlikely.
Nature 387:803-805.
Lutz, W., W. Sanderson, and S. Scherbov. 2001. The end of world population growth.
Nature 412:543-545.
MacGill, S. M., and Y. L. Siu. 2004. The nature of risk. Journal of Risk Research 7:315-
352.

266
Macgill, S. M., and Y. L. Siu. 2005. A new paradigm for risk analysis. Futures 37:1105-
1131.
Major, D. C., and K. D. Frederick. 1997. Water resources planning and climate change
assessment methods. Climatic Change 37:25-40.
Manley, G. 1974. Central England temperatures - monthly means 1659 to 1973.
Quarterly Journal of the Royal Meteorological Society 100:389-405.
Marsh, T. J., R. A. Monkhouse, N. W. Arnell, M. L. Lees, and N. S. Reynard. 1994.
The 1988-92 drought. Institute of Hydrology, Wallingford, UK.
Mastrandrea, M. D., and S. H. Schneider. 2001. Integrated assessment of abrupt climatic
changes. Climate Policy 1:433-449.
McAvaney, B. J., C. Covey, S. Joussaume, V. Kattsov, A. Kitoh, W. Ogana, A. J.
Pitman, A. J. Weaver, R. A. Wood, and Z.-C. Zhao. 2001. Model Evaluation.
Pages 471-523 in J. T. Houghton, Y. Ding, D. J. Griggs, M. Noguer, P. J. van
der Linden, X. Dai, K. Maskell, and C. A. Johnson, editors. Climate Change
2001: The Scientific Basis. Cambridge University Press, Cambridge.
McIntyre, N., M. Lees, H. Wheater, C. Onof, and B. Connorton. 2003. Evaluation and
visualisation of risk to water resources. Proceedings of the Institution of Civil
Engineers-Water and Maritime Engineering 156:1-11.
McKay, M. D., R. J. Beckman, and W. J. Conover. 1979. A comparison of three
methods of selecting values of input variables in the analysis of output from a
computer code. Technometrics 21:239-245.
Mearns, L. O., I. Bogardi, F. Giorgi, I. Matyasovszky, and M. Palecki. 1999.
Comparison of climate change scenarios generated from regional climate model
experiments and statistical downscaling. Journal of Geophysical Research-
Atmospheres 104:6603-6621.
Mearns, L. O., F. Giorgi, P. Whetton, D. Pabon, M. Hulme, and M. Lal. 2003.
Guidelines for Use of Climate Scenarios Developed from Regional Climate
Model Experiments. IPCC TGCIA.
Mearns, L. O., M. Hulme, T. R. Carter, R. Leemans, M. Lal, and P. H. Whetton. 2001.
Climate Scenario Development. Pages 739-768 in J. T. Houghton, Y. Ding, D. J.
Griggs, M. Noguer, P. J. van der Linden, X. Dai, K. Maskell, and C. A. Johnson,
editors. Climate Change 2001: The Scientific Basi. Cambridge University Press,
Cambridge.

267
Meyer, W. B., K. W. Butzer, T. E. Downing, B. L. Turner II, G. W. Wenzel, and J. L.
Wescoat. 1998. Reasoning by analogy. Pages 217-289 in S. Rayner and E. L.
Malone, editors. Human choice and climate change: v. 3. The tools for policy
analysis. Battelle Press, Columbus, Ohio.
Mitchell, J. F. B., T. C. Johns, M. Eagles, W. J. Ingram, and R. A. Davis. 1999.
Towards the construction of climate change scenarios. Climatic Change 41:547-
581.
Mitchell, T. D. 2003. Pattern scaling - An examination of the accuracy of the technique
for describing future climates. Climatic Change 60:217-242.
Mitchell, T. D., and M. Hulme. 1999. Predicting regional climate change: living with
uncertainty. Progress in Physical Geography 23:57-78.
Morgan, M. G., P. J. Adams, and D. W. Keith. 2005. Elicitation of expert judgements of
aerosol forcing. Climatic Change:(in press).
Morgan, M. G., and M. Henrion. 1990. Uncertainty: a guide to dealing with uncertainty
in quantitative risk and policy analysis. Cambridge University Press, Cambridge.
Morgan, M. G., and D. W. Keith. 1995. Climate-Change - Subjective Judgments by
Climate Experts. Environmental Science & Technology 29:A468-A476.
Morgan, M. G., L. F. Pitelka, and E. Shevliakova. 2001. Elicitation of expert judgments
of climate change impacts on forest ecosystems. Climatic Change 49:279-307.
Mori, S., and M. Takahashi. 1998. An Integrated Assessment Model for New Energy
Technologies and Food Production - An Extension of the MARIA Model.
International Journal of Global Energy Issues 11:1-17.
Morita, T., Y. Matsuoka, M. Kainuma, and H. Harasawa. 1994. AIM - Asian Pacific
integrated model for evaluating policy options to reduce GHG emissions and
global warming impacts. Pages 254-273 in S. Bhattacharya and others, editors.
Global warming issues in Asia. AIT, Bangkok.
Morita, T., and J. Robinson. 2001. Greenhouse gas emission mitigation scenarios and
implications. Pages 115-166 in B. Metz, O. Davidson, R. Swart, and J. Pan,
editors. Climate change 2001: mitigation. Cambridge University Press,
Cambridge.
Moss, R. H. 2000. Ready for IPCC-2001: innovation and change in plans for the IPCC
Third Assessment Report. Climatic Change 45:459-468.

268
Murphy, J. M. 1995. Transient-Response of the Hadley-Center Coupled Ocean-
Atmosphere Model to Increasing Carbon-Dioxide .1. Control Climate and Flux
Adjustment. Journal of Climate 8:36-56.
Murphy, J. M., and J. F. B. Mitchell. 1995. Transient-Response of the Hadley-Center
Coupled Ocean-Atmosphere Model to Increasing Carbon-Dioxide .2. Spatial and
Temporal Structure of Response. Journal of Climate 8:57-80.
Murphy, J. M., D. M. H. Sexton, D. N. Barnett, G. S. Jones, M. J. Webb, M. Collins,
and D. A. Stainforth. 2004. Quantifying uncertainties in climate change from a
large ensemble of general circulation model predictions. Nature 430:768-772.
Nakicenovic, N., J. Alcamo, G. Davis, B. de Vries, J. Fenhann, S. Gaffin, K. Gregory,
A. Grbler, T. Y. Jung, T. Kram, E. L. La Rovere, L. Michaelis, S. Mori, T.
Morita, W. Pepper, H. Pitcher, L. Price, K. Raihi, A. Roehrl, H.-H. Rogner, A.
Sankovski, M. Schlesinger, P. Shukla, S. Smith, R. Swart, S. van Rooijen, N.
Victor, and Z. Dadi. 2000. Emissions Scenarios. A Special Report of Working
Group III of the Intergovernmental Panel on Climate Change. Cambridge
University Press, Cambridge.
New, M., and M. Hulme. 2000. Representing uncertainty in climate change scenarios: a
Monte-Carlo approach. Integrated Assessment 1:203-213.
Nordhaus, W. D. 1994. Expert Opinion on Climatic-Change. American Scientist 82:45-
51.
Nordhaus, W. D., and D. Popp. 1997. What is the value of scientific knowledge? An
application to global warming using the PRICE model. Energy Journal 18:1-45.
Nozawa, T., S. Emori, T. Takemura, T. Nakajima, A. Numaguti, A. Abe-Ouchi, and M.
Kimoto. 2000. Coupled ocean-atmosphere model experiments of future climate
change based on IPCC SRES scenarios. Pages 353-355 in Preprints of the 11th
Symposium on Global Change Studies, Long Beach, USA.
Ofwat. 2004. Future water and sewerage charges 2005-10: Final determinations. Office
of Water Services, Birmingham.
O'Hare, G. 2001. Hurricane 07B in the Godavari delta, Andhra Pradesh, India:
vulnerability, mitigation and the spatial impact. Geographical Journal 167:23-38.
O'Neill, B. C. 2004. Conditional probabilistic population projections: An application to
climate change. International Statistical Review 72:167-184.
O'Neill, B. C., and M. Oppenheimer. 2002. Climate change - Dangerous climate
impacts and the Kyoto protocol. Science 296:1971-1972.

269
Oreskes, N., K. Shraderfrechette, and K. Belitz. 1994. Verification, Validation, and
Confirmation of Numerical-Models in the Earth-Sciences. Science 263:641-646.
Osborn, T. J., and M. Hulme. 2002. Evidence for trends in heavy rainfall events over the
UK. Philosophical Transactions of the Royal Society of London A 360:1313-
1325.
Osborn, T. J., M. Hulme, P. D. Jones, and T. A. Basnett. 2000. Observed trends in the
daily intensity of United Kingdom precipitation. International Journal of
Climatology 20:347-364.
Osborn, T. J., and C. J. Wallace. 2005. Linear and exponential relationships between
global temperature change and patterns of precipitation change. Journal of
Geophysical Research-Atmospheres:(submitted).
OST. 2002. Foresight Futures 2002: Revised Scenarios and Guidance. Office of Science
and Technology, London.
Pal, J. S., E. E. Small, and E. A. B. Eltahir. 2000. Simulation of regional-scale water
and energy budgets: Representation of subgrid cloud and precipitation processes
within RegCM. Journal of Geophysical Research-Atmospheres 105:29579-
29594.
Palmer, T. N. 2000. Predicting uncertainty in forecasts of weather and climate. Reports
on Progress in Physics 63:71-116.
Palmer, T. N. 2001. A nonlinear dynamical perspective on model error: A proposal for
non-local stochastic-dynamic parametrization in weather and climate prediction
models. Quarterly Journal of the Royal Meteorological Society 127:279-304.
Parker, D., and B. Horton. 2005. Uncertainties in central England temperature 1878-
2003 and some improvements to the maximum and minimum series.
International Journal of Climatology 25:1173-1188.
Parker, D. E., T. P. Legg, and C. K. Folland. 1992. A New Daily Central England
Temperature Series, 1772-1991. International Journal of Climatology 12:317-
342.
Parmesan, C., and G. Yohe. 2003. A globally coherent fingerprint of climate change
impacts across natural systems. Nature 421:37-42.
Parry, G. W. 1996. The characterization of uncertainty in probabilistic risk assessments
of complex systems. Reliability Engineering & System Safety 54:119-126.
Parry, M., N. Arnell, M. Hulme, R. Nicholls, and M. Livermore. 1998. Adapting to the
inevitable. Nature 395:741-741.

270
Parry, M. L., and T. R. Carter. 1998. Climate impact and adaptation assessment: a guide
to the IPCC approach. Earthscan, London.
Pat-Cornell, M. E. 1996. Uncertainties in risk analysis: Six levels of treatment.
Reliability Engineering & System Safety 54:95-111.
Patrinos, A., and A. Bamzai. 2005. Policy needs robust climate science. Nature 438:285.
Patwardhan, A., and M. J. Small. 1992. Bayesian Methods for Model Uncertainty
Analysis with Application to Future Sea-Level Rise. Risk Analysis 12:513-523.
Peck, S. C., and T. J. Teisberg. 1996. Uncertainty and the value of information with
stochastic losses from global warming. Risk Analysis 16:227-235.
Pepin, N. C. 1997. Scenarios of future climate change: Effects on frost occurrence and
severity in the maritime uplands of northern England. Geografiska Annaler
Series a-Physical Geography 79A:121-137.
Perry, M., and D. Hollis. 2005a. The development of a new set of long-term climate
averages for the UK. International Journal of Climatology 25:1023-1039.
Perry, M., and D. Hollis. 2005b. The generation of monthly gridded datasets for a range
of climatic variables over the UK. International Journal of Climatology 25:1041-
1054.
Petersen, A. C. 2000. Philosophy of climate science. Bulletin of the American
Meteorological Society 81:265-271.
Pielke Jr., R. A. 1998. Rethinking the role of adaptation in climate policy. Global
Environmental Change-Human and Policy Dimensions 8:159-170.
Pielke Sr., R. A. 2002. Overlooked issues in the US National climate and IPCC
assessments. Climatic Change 52:1-11.
Pilling, C. G., and J. A. A. Jones. 2002. The impact of future climate change on
seasonal discharge, hydrological processes and extreme flows in the Upper Wye
experimental catchment, mid-Wales. Hydrological Processes 16:1201-1213.
Pittock, A. B., and R. N. Jones. 2000. Adaptation to what and why? Environmental
Monitoring and Assessment 61:9-35.
Pittock, A. B., R. N. Jones, and C. D. Mitchell. 2001. Probabilities will help us plan for
climate change. Nature 413:249.
Pope, V. D., M. L. Gallani, P. R. Rowntree, and R. A. Stratton. 2000. The impact of
new physical parametrizations in the Hadley Centre climate model: HadAM3.
Climate Dynamics 16:123-146.

271
Price, M. 1998. Water storage and climate change in Great Britain - the role of
groundwater. Proceedings of the Institution of Civil Engineers-Water Maritime
and Energy 130:42-50.
PRUDENCE. 2002. Prediction of Regional scenarios and Uncertainties for Defining
EuropeaN Climate change risks and effects. in.
Prudhomme, C., D. Jakob, and C. Svensson. 2003. Uncertainty and climate change
impact on the flood regime of small UK catchments. Journal of Hydrology
277:1-23.
Prudhomme, C., N. Reynard, and S. Crooks. 2002. Downscaling of global climate
models for flood frequency analysis: where are we now? Hydrological Processes
16:1137-1150.
Pulwarty, R. S., and T. S. Melis. 2001. Climate extremes and adaptive management on
the Colorado River: Lessons from the 1997-1998 ENSO event. Journal of
Environmental Management 63:307-324.
Risnen, J., U. Hansson, A. Ullerstig, R. Doscher, L. P. Graham, C. Jones, H. E. M.
Meier, P. Samuelsson, and U. Willen. 2004. European climate in the late twenty-
first century: regional simulations with two driving global models and two
forcing scenarios. Climate Dynamics 22:13-31.
Risnen, J., and T. N. Palmer. 2001. A probability and decision-model analysis of a
multimodel ensemble of climate change simulations. Journal of Climate
14:3212-3226.
Raper, S. C. B., and U. Cubasch. 1996. Emulation of the results from a coupled general
circulation model using a simple climate model. Geophysical Research Letters
23:1107-1110.
Raskin, P., G. Gallopin, P. Gutman, A. Hammond, and R. Swart. 1998. Bending the
Curve: Toward Global Sustainability. Stockholm Environment Institute,
Stockholm.
Ravetz, J. 2004. The post-normal science of precaution. Futures 36:347-357.
Regan, H. M., Y. Ben-Haim, B. Langford, W. G. Wilson, P. Lundberg, S. J. Andelman,
and M. A. Burgman. 2005. Robust decision-making under severe uncertainty for
conservation management. Ecological Applications 15:1471-1477.
Reilly, J., R. Prinn, J. Harnisch, J. Fitzmaurice, H. Jacoby, D. Kicklighter, J. Melillo, P.
Stone, A. Sokolov, and C. Wang. 1999. Multi-gas assessment of the Kyoto
Protocol. Nature 401:549-555.

272
Riahi, K., and R. A. Roehrl. 2000. Robust Energy Technology Strategies for the 21st
Century - Carbon dioxide mitigation and sustainable development.
Environmental Economics and Policy Studies 3.
Risbey, J., and M. Kandlikar. 2002. Expert assessment of uncertainties in detection and
attribution of climate change. Bulletin of the American Meteorological Society
83:1317-1326.
Risbey, J., M. Kandlikar, H. Dowlatabadi, and D. Graetz. 1999. Scale, context, and
decision making in agricultural adaptation to climate variability and change.
Mitigation and Adaptation Strategies for Global Change 4:137-165.
Risbey, J. S. 1998. Sensitivities of water supply planning decisions to streamflow and
climate scenario uncertainties. Water Policy 1:321-340.
Risbey, J. S., M. Kandlikar, and D. J. Karoly. 2000. A protocol to articulate and
quantify uncertainties in climate change detection and attribution. Climate
Research 16:61-78.
Risbey, J. S., and P. H. Stone. 1996. A case study of the adequacy of GCM simulations
for input to regional climate change assessments. Journal of Climate 9:1441-
1467.
Roeckner, E., L. Bengtsson, J. Feichter, J. Lelieveld, and H. Rodhe. 1999. Transient
climate change simulations with a coupled atmosphere-ocean GCM including
the tropospheric sulfur cycle. Journal of Climate 12:3004-3032.
Roeckner, E., J. M. Oberhuber, A. Bacher, M. Christoph, and I. Kirchner. 1996. ENSO
variability and atmospheric response in a global coupled atmosphere-ocean
GCM. Climate Dynamics 12:737-754.
Root, T. L., J. T. Price, K. R. Hall, S. H. Schneider, C. Rosenzweig, and J. A. Pounds.
2003. Fingerprints of global warming on wild animals and plants. Nature
421:57-60.
Roughgarden, T., and S. H. Schneider. 1999. Climate change policy: quantifying
uncertainties for damages and optimal carbon taxes. Energy Policy 27:415-429.
Royer, J.-F., D. Cariolle, F. Chauvin, M. Deque, H. Douville, R.-M. Hu, S. Planton, A.
Rascol, J.-L. Ricard, and D. Salas Y Melia. 2002. Simulation des changements
climatiques au cours du XXIe siecle incluant l'ozone stratospherique: Simulation
of climate changes during the 21st century including stratospheric ozone.
Comptes Rendus Geosciences 334:147-154.

273
Rubin, E. S., L. B. Lave, and M. G. Morgan. 1992. Keeping Climate Research Relevant.
Issues in Science and Technology 8:47-55.
Ruosteenoja, K., T. R. Carter, K. Jylh, and H. Tuomenvirta. 2003. Future climate in
world regions: an intercomparison of model-based projections for the new IPCC
emissions scenarios. Finnish Environment Institute, Helsinki.
Saltelli, A. 2000. What is sensitivity analysis? Pages 3-13 in A. Saltelli, K. Chan, and E.
M. Scott, editors. Sensitivity Analysis. Wiley, Chichester.
Sankovski, A., W. Barbour, and W. Pepper. 2000. Quantification of the IS99 emission
scenario storylines using the atmospheric stabilization framework.
Technological Forecasting and Social Change 63:263-287.
Santer, B. D., T. M. L. Wigley, M. E. Schlesinger, and J. F. B. Mitchell. 1990.
Developing climate scenarios from equilibrium GCM results. Report No. 47,
Max-Planck-Institut fr Meteorologie, Hamburg.
Sarewitz, D., R. A. Pielke Jr., and M. Keykhah. 2003. Vulnerability and risk: some
thoughts from a political and policy perspective. Risk Analysis 23:805-810.
Scheraga, J. D., and A. E. Grambsch. 1998. Risks, opportunities, and adaptation to
climate change. Climate Research 11:85-95.
Scherm, H. 2000. Simulating uncertainty in climate-pest models with fuzzy numbers.
Environmental Pollution 108:373-379.
Schimmelpfennig, D. 1996. Uncertainty in economic models of climate-change impacts.
Climatic Change 33:213-234.
Schlesinger, M. E., S. Malyshev, E. V. Rozanov, F. L. Yang, N. G. Andronova, B. De
Vries, A. Grubler, K. J. Jiang, T. Masui, T. Morita, J. Penner, W. Pepper, A.
Sankovski, and Y. Zhang. 2000. Geographical distributions of temperature
change for scenarios of greenhouse gas and sulfur dioxide emissions.
Technological Forecasting and Social Change 65:167-193.
Schneider, S. H. 2001. What is 'dangerous' climate change? Nature 411:17-19.
Schneider, S. H. 2002. Can we estimate the likelihood of climatic changes at 2100?
Climatic Change 52:441-451.
Schrter, D., and others. 2005. Ecosystem Service Supply and Vulnerability to Global
Change in Europe. Science:(in press).
Schumann, A. H., and M. Antl. 2001. Probabilistic characterization of uncertainties of
climate change impact assessments. Pages 247-254 in Sixth IAHS Scientific
Assembly. IAHS, Maastricht, The Netherlands.

274
Schwartz, P. 1991. The art of the long view. Doubleday, New York.
Scott, M. J., R. D. Sands, J. Edmonds, A. M. Liebetrau, and D. W. Engel. 1999.
Uncertainty in integrated assessment models: modeling with MiniCAM 1.0.
Energy Policy 27:855-879.
Sen, A. 1981. Poverty and famines: an essay on entitlement and deprivation. Oxford
University Press, Oxford.
Shackley, S., P. Young, S. Parkinson, and B. Wynne. 1998. Uncertainty, complexity
and concepts of good science in climate change modelling: Are GCMs the best
tools? Climatic Change 38:159-205.
Sheriff, J. D., J. D. Lawson, and T. E. A. Askew. 1996. Strategic resource development
options in England and Wales. Journal of the Chartered Institution of Water and
Environmental Management 10:160-169.
Slaughter, R. A. 1994. Why we should care for future generations now. Futures
26:1080.
Smith, B., I. Burton, R. J. T. Klein, and R. Street. 1999. The science of adaptation: a
framework for assessment. Mitigation and Adaptation Strategies for Global
Change 4:199-213.
Smith, B., I. Burton, R. J. T. Klein, and J. Wandel. 2000. An anatomy of adaptation to
climate change and variability. Climatic Change 45:223-251.
Smith, E. J. 1997a. The balance between public water supply and environmental needs.
Journal of the Chartered Institution of Water and Environmental Management
11:8-13.
Smith, J. B. 1997b. Setting priorities for adapting to climate change. Global
Environmental Change-Human and Policy Dimensions 7:251-264.
Smith, J. B., and S. Hitz. 2003. Background Paper: Estimating Global Impacts from
Climate Change. Pages 101 in OECD Workshop on the Benefits of Climate
Policy: Improving Information for Policy Makers, Paris.
Smith, L. A. 2002. What might we learn from climate forecasts? Proceedings of the
National Academy of Sciences of the United States of America 99:2487-2492.
Stainforth, D. A., T. Aina, C. Christensen, M. Collins, N. Faull, D. J. Frame, J. A.
Kettleborough, S. Knight, A. Martin, J. M. Murphy, C. Piani, D. Sexton, L. A.
Smith, R. A. Spicer, A. J. Thorpe, and M. R. Allen. 2005. Uncertainty in
predictions of the climate response to rising levels of greenhouse gases. Nature
433:403-406.

275
STARDEX. 2002. Statistical and Regional dynamic Downscaling of Extremes for
European regions. in.
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. Meteorology and Atmospheric Physics 82:75-96.
Stewart, T. R. 2000. Uncertainty, judgement, and error in prediction. Pages 41-57 in D.
Sarewitz, R. A. Pielke Jr., and R. Byerly Jr., editors. Prediction: science,
decision making and the future of nature. Island Press, Washington, D.C.
Stott, P. A., and J. A. Kettleborough. 2002. Origins and estimates of uncertainty in
predictions of twenty- first century temperature rise. Nature 416:723-726.
Subak, S. 2000. Climate change adaptation in the UK water industry: Managers'
perceptions of past variability and future scenarios. Water Resources
Management 14:137-156.
Swart, R., J. Mitchell, T. Morita, and S. Raper. 2002. Stabilisation scenarios for climate
impact assessment. Global Environmental Change 12:155-165.
Taylor, D., and T. Hams. 2001. 'Water and you': a stakeholder action plan for the East
of England. East of England Sustainable Development Round Table and the
Water Resources Group.
Tebaldi, C., L. O. Mearns, D. Nychka, and R. L. Smith. 2004. Regional probabilities of
precipitation change: A Bayesian analysis of multimodel simulations.
Geophysical Research Letters 31:L24213.
Tebaldi, C., R. L. Smith, D. Nychka, and L. O. Mearns. 2005. Quantifying uncertainty
in projections of regional climate change: A Bayesian approach to the analysis
of multi-model ensembles. Journal of Climate:1524-1540.
Titus, J. G., and V. Narayanan. 1996. The risk of sea level rise. Climatic Change
33:151-212.
Tol, R. S. J., and A. F. de Vos. 1998. A Bayesian statistical analysis of the enhanced
greenhouse effect. Climatic Change 38:87-112.
Tol, R. S. J., B. C. O'Neill, and D. P. van Vuuren. 2005. A critical assessment of the
IPCC SRES scenarios. Working paper FNU, Centre for Marine and Climate
Research, Hamburg University, Hamburg.
Tompkins, E. L., and W. N. Adger. 2003. Building resilience to climate change through
adaptive management of natural resources. Tyndall Centre Working Paper No.
27, Tyndall Centre Working Papers, Norwich.

276
Toth, F. 2000. Decision analysis frameworks. Pages 53-68 in R. K. Pachauri, T.
Taniguchi, and K. Tanaka, editors. Guidance papers on the cross cutting issues
of the third assessment report of the IPCC. IPCC, Geneva.
Toth, F. L., T. Bruckner, H. M. Fussel, M. Leimbach, and G. Petschel- Held. 2003.
Integrated assessment of long-term climate policies: Part 2 - model results and
uncertainty analysis. Climatic Change 56:57-72.
Trigo, R. M., and J. P. Palutikof. 1999. Simulation of daily temperatures for climate
change scenarios over Portugal: a neural network model approach. Climate
Research 13:45-59.
Tsang, F. T., and H. Dowlatabadi. 1995. A Bayesian Technique for Refining the
Uncertainty in Global Energy-Model Forecasts. International Journal of
Forecasting 11:43-61.
UKCIP. 2000. Socio-economic scenarios for climate change impact assessment: a guide
to their use in the UK. UK Climate Impact Programme, Oxford.
UKWIR. 2003. Effects of climate change on river flows and groundwater recharge:
UKCIP02 scenarios. 03/CL/04/02, UK Water Industry Research Limited,
London.
van Asselt, M. B. A. 2000. Perspectives on uncertainty and risk: the PRIMA approach
to decision support. Kluwer Academic Publishers, Boston.
van der Heijden, K. 1996. Scenarios: the art of strategic conversation. John Wiley &
Sons.
van der Sluijs, J. 1996. Integrated Assessment Models and the Management of
Uncertainties. International Institute for Applied Systems Analysis WP-96-119,
International Institute for Applied Systems Analysis, Laxenburg.
van der Sluijs, J., J. van Eijndhoven, S. Shackley, and B. Wynne. 1998. Anchoring
devices in science for policy: The case of consensus around climate sensitivity.
Social Studies of Science 28:291-323.
van der Sluijs, J. P., M. Craye, S. Funtowicz, P. Kloprogge, J. Ravetz, and J. Risbey.
2005. Combining quantitative and qualitative measures of uncertainty in model-
based environmental assessment: The NUSAP system. Risk Analysis 25:481-
492.
van der Sluijs, J. P., P. H. M. Janssen, A. C. Peterson, P. Kloprogge, J. S. Risbey, W.
Tuinstra, and J. R. Ravetz. 2004. RIVM/MNP Guidance for Uncertainty

277
 Assessment and Communication: Tool Catalogue for Uncertainty Assessment.
NWS-E-2004-37, Copernicus Institute & RIVM, Utrecht/Bilthoven.
van der Sluijs, J. P., J. S. Risbey, P. Kloprogge, J. R. Ravetz, S. O. Funtowicz, S. Corral
Quintana, . Guimares Pereira, B. De Marchi, A. C. Peterson, P. H. M.
Janssen, R. Hoppe, and S. W. F. Huijs. 2003. RIVM/MNP Guidance for
Uncertainty Assessment and Communication: Detailed Guidance. NWS-E-2003-
163, Copernicus Institute & RIVM, Utrecht/Bilthoven.
van Lenthe, J., L. Hendrickx, W. Biesiot, and C. Vlek. 1997. A decision-analytic
approach to the integrated assessment of climate change. Risk Decision and
Policy 2:213-234.
Vaughan, D. G., and J. R. Spouge. 2002. Risk estimation of collapse of the West
Antarctic Ice Sheet. Climatic Change 52:65-91.
Vellinga, M., and R. A. Wood. 2002. Global climatic impacts of a collapse of the
Atlantic thermohaline circulation. Climatic Change 54:251-267.
Venkatesh, B. N., and B. F. Hobbs. 1999. Analyzing investments for managing Lake
Erie levels under climate change uncertainty. Water Resources Research
35:1671-1683.
Vidale, P. L., D. Luthi, C. Frei, S. I. Seneviratne, and C. Schar. 2003. Predictability and
uncertainty in a regional climate model. Journal of Geophysical Research-
Atmospheres 108:-.
Visser, H., R. J. M. Folkert, J. Hoekstra, and J. J. De Wolff. 2000. Identifying key
sources of uncertainty in climate change projections. Climatic Change 45:421-
457.
Walker, W. E., P. Harremos, J. Rotmans, J. P. van der Sluijs, M. B. A. van Asselt, P.
H. M. Janssen, and M. P. K. von Krauss. 2003. Defining uncertainty: a
conceptual basis for uncertainty management in model-based decision support.
Integrated Assessment 4:5-17.
Warrick, R. A., and E. M. Barrow. 1991. Climate change scenarios for the UK.
Transactions - Institute of British Geographers 16:387-399.
Washington, W. M., J. W. Weatherly, G. A. Meehl, A. J. Semtner, T. W. Bettge, A. P.
Craig, W. G. Strand, J. Arblaster, V. B. Wayland, R. James, and Y. Zhang.
2000. Parallel climate model (PCM) control and transient simulations. Climate
Dynamics 16:755-774.

278
Webster, M., C. Forest, J. Reilly, M. Babiker, D. Kicklighter, M. Mayer, R. Prinn, M.
Sarofim, A. Sokolov, P. Stone, and C. Wang. 2003. Uncertainty analysis of
climate change and policy response. Climatic Change 61:295-320.
Webster, M. D., M. Babiker, M. Mayer, J. M. Reilly, J. Harnisch, R. Hyman, M. C.
Sarofim, and C. Wang. 2002. Uncertainty in emissions projections for climate
models. Atmospheric Environment 36:3659-3670.
Webster, M. D., and A. P. Sokolov. 2000. A methodology for quantifying uncertainty in
climate projections. Climatic Change 46:417-446.
Weyant, J., O. Davidson, H. Dowlatabadi, J. A. Edmonds, M. Grubb, E. A. Parson, R.
Richels, J. Rotmans, P. R. Shukla, and R. S. J. Tol. 1996. Integrated assessment
of climate change: an overview and comparision of approaches and results.
Pages 367-396 in IPCC, editor. Climate change 1995: economic and social
dimensions of climate change. Cambridge University Press, Cambridge.
WFD. 2000. Directive 2000/60/EC of the European Parliament and of the Council of 23
October 2000 establishing a framework for Community action in the field of
water policy. Official Journal L 327:0001 - 0073.
Whitehead, P. G., R. L. Wilby, D. Butterfield, and A. J. Wade. 2005. Impacts of climate
change on nitrogen in a lowland chalk stream: an appraisal of adaptation
strategies. Science of the Total Environment:(submitted).
Wigley, T. M. L., and P. D. Jones. 1987. England and Wales Precipitation - a
Discussion of Recent Changes in Variability and an Update to 1985. Journal of
Climatology 7:231-246.
Wigley, T. M. L., and S. C. B. Raper. 1987. Thermal-Expansion of Sea-Water
Associated with Global Warming. Nature 330:127-131.
Wigley, T. M. L., and S. C. B. Raper. 1992. Implications for Climate and Sea-Level of
Revised IPCC Emissions Scenarios. Nature 357:293-300.
Wigley, T. M. L., and S. C. B. Raper. 1995. An Heuristic Model for Sea-Level Rise Due
to the Melting of Small Glaciers. Geophysical Research Letters 22:2749-2752.
Wigley, T. M. L., and S. C. B. Raper. 2001. Interpretation of high projections for
global-mean warming. Science 293:451-454.
Wilby, R. L. 1994. Stochastic Weather Type Simulation for Regional Climate-Change
Impact Assessment. Water Resources Research 30:3395-3403.
Wilby, R. L. 2005. Uncertainty in water resource model parameters used for climate
change impact assessment. Hydrological Processes 19:3201-3219.

279
Wilby, R. L., S. P. Charles, E. Zorita, B. Timball, P. Whetton, and L. O. Mearns. 2004.
Guidelines for use of climate scenarios developed from statistical downscaling
methods. Intergovernmental Panel on Climate Change.
Wilby, R. L., and I. Harris. 2006. A framework for assessing uncertainties in climate
change impacts: low flow scenarios for the River Thames, UK. Water Resources
Research:(in press).
Wilby, R. L., L. E. Hay, W. J. Gutowski, R. W. Arritt, E. S. Takle, Z. T. Pan, G. H.
Leavesley, and M. P. Clark. 2000. Hydrological responses to dynamically and
statistically downscaled climate model output. Geophysical Research Letters
27:1199-1202.
Wilby, R. L., L. E. Hay, and G. H. Leavesley. 1999. A comparison of downscaled and
raw GCM output: implications for climate change scenarios in the San Juan
River basin, Colorado. Journal of Hydrology 225:67-91.
Wilby, R. L., and T. M. L. Wigley. 1997. Downscaling general circulation model
output: a review of methods and limitations. Progress in Physical Geography
21:530-548.
Wilby, R. L., T. M. L. Wigley, D. Conway, P. D. Jones, B. C. Hewitson, J. Main, and
D. S. Wilks. 1998. Statistical downscaling of general circulation model output:
A comparison of methods. Water Resources Research 34:2995-3008.
Wilks, D. S., and R. L. Wilby. 1999. The weather generation game: a review of
stochastic weather models. Progress in Physical Geography 23:329-357.
Williams, L. J., D. Shaw, and R. Mendelsohn. 1998. Evaluating GCM output with
impact models. Climatic Change 39:111-133.
Willows, R. I., and R. Connell, editors. 2003. Climate adaptation: Risk, Uncertainty and
Decision-making. UKCIP, Oxford.
Winkler, R. L. 1996. Uncertainty in probabilistic risk assessment. Reliability
Engineering & System Safety 54:127-132.
Wood, A. W., D. P. Lettenmaier, and R. N. Palmer. 1997. Assessing climate change
implications for water resources planning. Climatic Change 37:203-228.
Woodbury, P. B., J. E. Smith, D. A. Weinstein, and J. A. Laurence. 1998. Assessing
potential climate change effects on loblolly pine growth: A probabilistic regional
modeling approach. Forest Ecology and Management 107:99-116.
Yohe, G. 1991. Uncertainty, Climate Change and the Economic Value of Information -
an Economic Methodology for Evaluating the Timing and Relative Efficacy of

280
 Alternative Response to Climate Change with Application to Protecting
Developed Property from Greenhouse Induced Sea-Level Rise. Policy Sciences
24:245-269.
Yohe, G. 2000. Assessing the role of adaptation in evaluating vulnerability to climate
change. Climatic Change 46:371-390.
Yohe, G., N. Andronova, and M. Schlesinger. 2004. Climate - To hedge or not against
an uncertain climate. Science 306:416-417.
Yohe, G., and H. Dowlatabadi. 1999. Risk and uncertainties, analysis and evaluation:
lessons for adaptation and integration. Mitigation and Adaptation Strategies for
Global Change 4:319-329.
Yohe, G., and J. Neumann. 1997. Planning for sea level rise and shore protection under
climate uncertainty. Climatic Change 37:243-270.
Yukimoto, S., M. Endoh, Y. Kitamura, A. Kitoh, T. Motoi, and A. Noda. 2000. ENSO-
like interdecadal variability in the Pacific Ocean as simulated in a coupled
general circulation model. Journal of Geophysical Research-Oceans 105:13945-
13963.
Zapert, R., P. S. Gaertner, and J. A. Filar. 1998. Uncertainty propagation within an
integrated model of climate change. Energy Economics 20:571-598.

281