Professional Documents
Culture Documents
Research Team
1
Jet Propulsion Laboratory, MS 300-227
2
Earth System Research Laboratory, Physical Science Division
3
National Drought Mitigation Center
University of Nebraska-Lincoln
4
Climate Prediction Center, Room 604
Abstract
This chapter focuses on soil moisture measurements using wide-swath microwave sensors
including satellite passive radiometer and active scatterometer. First, a review is presented on
the background research and state-of-the-art methods for remote sensing of soil moisture.
Emission and scattering physics is then presented to explain the underlying science of the
different approaches of passive and active estimates of soil moisture. Results are obtained at
local, regional, and continental scales to show patterns of soil moisture variations in daily,
weekly, and seasonal time scales. Practical applications to assessments of drought conditions are
illustrated in specific examples in the United States and other countries. For these applications,
radar meteorology and to identify issues related to dry rain or virga, which is relevant to
1. Introduction
Soil moisture is a fundamental link between global water and carbon cycles, and has major
applications in predicting natural hazards such as floods and droughts [National Research
Council, 2007]. Precipitation data can be used to estimate soil wetness. In the United States
(US), preliminary precipitation data are based on measurements gathered from many active
stations nationwide each month, and it takes three to four months to assemble final, quality-
controlled data. In western US, some climate divisions may have no stations reporting in a
particular month or may lack a first- or second-order station altogether. A first-order station
4
reports all climate variables three times per day, and a second-order station reports maximum
and minimum temperatures and rainfall once per day and other variables twice per day. For
direct soil moisture measurements, a limited number of observations are provided from the
Oklahoma Mesonet System [Illston et al., 2004] and the Soil Climate Analysis Network [USDA,
2009a], but the data are generally too sparse spatially with different data quality and accuracy for
their assimilation to be of much use. Also, there are soil-moisture observations from enhanced
additions in the SNOTEL network [USDA, 2009b], but fully calibrated data are not yet available
routinely. Given the very limited number of stations collecting point-based, in-situ data, the use
of such information may not be representative of regional water or soil moisture amounts and
variations.
Soil moisture measurements over large spatial extent (areal data rather than point data) with
little or no missing gaps are crucial to represent land surface water distribution from regional to
continental scales. Recognizing the importance of soil moisture, especially as a key variable for
drought monitoring, the void in adequate soil moisture information in the U.S. and elsewhere
demands immediate and extensive measurements from satellite microwave remote sensing of
soil moisture with both passive and active sensors. A radiometer is used to measure the natural
emission from land surface in passive remote sensing, while a radar, including synthetic aperture
radar (SAR) and scatterometer, is used to transmit signals to a targeted surface area and measure
the scattering return in active remote sensing. From passive or active microwave data, soil
Passive methods use data from microwave radiometers such as the Scanning Multichannel
Microwave Radiometer (SMMR), Special Sensor Microwave Imager (SSM/I), the Tropical
Rainfall Measuring Mission (TRMM) Microwave Imager (TMI), the Advanced Microwave
5
Scanning Radiometer on the Earth Observing System (AMSR-E), and the Soil Moisture and
Ocean Salinity sensor (SMOS) [Wang, 1985; Owe et al.,1988; Kerr and Njoku, 1990; Teng et al.,
1993; van de Griend and Owe, 1994; Engman, 1995; Jackson, 1997; Kerr et al., 2001; Njoku et
al., 2003].
For active remote sensing, many approaches have been used for various datasets from SARs
including Seasat, Spaceborne Imaging Radar-C (SIR-C), European Remote Sensing (ERS),
(ALOS) [Blanchard and Chang, 1983; Dubois et al., 1995; Cognard et al., 1995; Shrivastava et
al., 2009; Loew et al., 2006; Takada et al., 2009], and from scatterometers such as ERS and
QuikSCAT [Wagner et al., 1999; Wagner and Scipal, 2000; Nghiem et al., 2000]. In this
chapter, we review the science principle of active and passive remote sensing of large-scale soil
moisture and illustrate the results from AMSR-E and QuikSCAT for drought applications.
The principle of microwave remote sensing of soil moisture is based on the sensitivity of soil
permittivity to the amount of liquid water. The permittivity of a medium, like moist soil,
characterizes electromagnetic wave propagation and attenuation in the medium. Both brightness
the soil permittivity. Empirical models of dielectric constant, which is the permittivity relative to
that of the free space, for different soil types as a function of volumetric moisture content (mv) at
microwave frequencies between 1.4 and 18 GHz was developed [Hallikainen et al., 1985;
While soil dielectric constant can be measured in-situ with a probe [Jackson, 1990], satellite
remote sensors do not directly provide soil dielectric measurements. Instead, these sensors
acquire brightness temperature or backscatter signatures, which are dependent on soil dielectric
and thus on soil moisture. Such a relationship enables the inversion of soil moisture from
brightness temperature or backscatter data, which can be complicated by natural effects such as
vegetation cover, surface roughness, rainfall, and anthropogenic effects such as radio-frequency
interference (RFI), which have different impacts on the accuracy of soil moisture retrieval at
The retrieval of soil moisture from brightness temperature, measured by a satellite passive
radiometer, was long formulated by many researchers and has been summarized by Njoku et al.
[2003]. Here is a brief review for an isothermal vegetated soil surface, whose brightness
where the soil emissivity is esp = 1 rsp for soil reflectivity rsp, which is influenced by soil
moisture through the effect of moisture on the soil dielectric constant. The emitting depth is
controlled by the near-surface moisture profile, and is smaller for higher microwave frequencies
and for wetter soils. Although microwaves can only sense soil moisture in the top surface layer,
there is a correlation to soil moisture in deeper soil at night when the soil moisture and
temperature profiles are more uniform than in the early afternoon. In (1), c and p are the
vegetation opacity and the vegetation single scattering albedo, respectively. Multiple scattering
7
in the vegetation layer is neglected, and a quasi-specular soil surface and no reflection at the air-
vegetation boundary are assumed in (1). Vegetation opacity and multiple scattering have lesser
For surface roughness, based on a fixed viewing angle, an empirical formulation has been
found practically useful for relating the reflectivity of a rough soil surface, rsp, to that of the
equivalent smooth surface, rop [Wang and Choudhury, 1981; Wang, 1983], which is expressed
as:
where p and q represent either of the orthogonal polarization states (V or H), and Q and h are
roughness parameters. Q may be approximated as zero at low frequencies (e.g., L and C bands).
The simple separation of soil moisture and roughness effects through (2) is not precise, and the
parameter h has a residual moisture dependence [Li et al., 2000; Wigneron et al., 2001].
To normalize the surface temperature (Ts) dependence in (1), the polarization ratio (PR) is
obtained by:
which is suitable for multi-channel data taken at the same incidence angle [Kerr and Njoku,
1990]. At large incidence angles (e.g., >50o), the difference between the vertically and
horizontally polarized brightness temperatures for bare soils is large, giving rise to a large PR
signal. Nevertheless, the observation path length through the vegetation becomes longer at large
incidence angles, increasing the vegetation attenuation and thus decreasing the sensitivity to the
soil moisture.
8
While (1-3) form a general theoretical basis for soil moisture retrieval from passive
microwave data, different approaches have been used for different satellite datasets with
different correction methods for the effects of soil types, roughness, vegetation, and surface
temperature [Njoku et al, 2003] although further advances should be considered for various non-
isothermal conditions and multiple interactions between soil surface and vegetation cover at
different growth stages. In practice for data from the Advanced Microwave Scanning
Radiometer on the EOS Aqua satellite (AMSR-E), the soil moisture retrieval utilizes primarily
the frequency channels of 10.7 and 18.7 GHz to consider the effects of atmospheric and
vegetative attenuation and to minimize the requirement for ancillary data inputs. The Tropical
Rainfall Measuring Mission (TRMM) Microwave Imager (TMI) has 10.7 and 19.3-GHz
channels, which can be used to obtain PR for soil moisture applications with a better consistency
at the lower frequency [Njoku et al, 2003]. Further details of the retrieval were published [Njoku
In active remote sensing, soil moisture can be derived from backscatter measured by a radar such
as a SAR at a high spatial resolution with a small and infrequent coverage, and by a
scatterometer at a low spatial resolution with a large and frequent coverage. Many theoretical
models (references given in the introduction) have been developed to characterize backscatter
signatures of vegetated soil. Here, a scattering model, based on the analytic vector wave theory
Backscatter 0 from moist soil with a vegetation cover is determined from an ensemble
(4)
where subscript 0 represents the air space above the vegetation, and subscript 1 indicates the
vegetation cover occupying volume V1 over the soil surface. The dyadic Green's function G and
the mean field F are obtained as described by Nghiem et al. [1990]. The correlation function C
characterizes the vegetation scatterers having different size, shape, and orientation angle f in
elevation and f in azimuth. For the vegetation canopy, the effective permittivity is calculated
under the strong permittivity fluctuation theory, which accounts for wave attenuation including
scattering and absorption loss [Nghiem et al., 1993a]. The analytic vector wave theory accounts
for fully polarimetric scattering, preserves the phase information, and includes multiple
reflection and transmission interactions of up-going and down-going electromagnetic waves with
the soil surface, thereby conveying the soil moisture information. This is because the soil
transmissivity and the soil reflectivity are controlled by the soil dielectric constant as a function
Rough surface scattering can be included in the contribution to the total backscatter. The
small-scale roughness of the soil surface is described with a standard deviation height and a
slope. When a large-scale roughness also exists, the overall roughness is accounted for by a joint
probability density function for both roughness scales [Nghiem et al., 1995]. The vegetation
volume scattering and the soil surface scattering are assumed to be uncorrelated due to
independent statistic representations of vegetation scatterers (e.g., leaves, twigs, branches) and
soil surface roughness. Then, the total backscatter is a sum of the vegetation volume backscatter
10
and soil surface backscatter. In the layer scattering configuration, such as a vegetation layer over
a rough soil surface, contributions from the rough surface scattering are considered with wave
interactions, differential propagation delay, and wave attenuation in the vegetation layer [Nghiem
et al., 1995], which can be effectively anisotropic when vegetation scatters have a preferential
distribution) [Nghiem et al., 1993b]. The backscatter from rough soil surface depends strongly
on the soil dielectric constant and also on transmissivity and reflectivity due to wave interaction
with the soil boundary, and thus the surface scattering also contains soil moisture signature in
addition to the soil moisture information in the volume scattering component. Nevertheless, a
dense vegetation canopy can have a large imaginary part in its effective permittivity, which
attenuates both the soil surface scattering and soil interactions in the volume scattering, and
consequently masks the soil moisture signature. Specific mathematical details of the volume and
surface scattering in layered media can be found in earlier publications [Nghiem et al., 1990,
While the above formulation provides physical insights and serves as a theoretical basis for
active remote sensing of soil moisture, in practice, due to the complexities of natural
environments in different climate regimes, it is not possible to set up an inversion method of soil
0 = amv + b (5)
where coefficients a and b are dependent on incidence angle, polarization, vegetation conditions,
soil type, surface variation, and climate regime [Mo et al., 1984, Prevot et al., 1993, Shrivastava
11
et al., 1997; Shoshany et al., 2000; Hutchingson, 2003]. Particularly for Ku-band backscatter
data from the SeaWinds scatterometer aboard the QuikSCAT satellite (QSCAT), the bias term b
in (5) contains a signature of seasonal vegetation change while changes in volumetric soil
moisture mv from rainwater are detectable in backscatter variations in a time scale consistent
with the initial impulse increase in wetness from the precipitation input throughout the
subsequent drying process [Nghiem et al., 2005]. Thus, soil moisture change (SMC) can be
directly inferred from (5) using the temporal backscatter-change method, which removes most of
As presented in sections 2.1 and 2.2, passive and active sensors measure different parameters:
passive brightness temperature and active radar backscatter, which represent different responses
to vegetated soil with different sensitivities to soil moisture and vegetation cover.
In a theoretical ideal case of smooth bare soil (c = 0), (1) dictates that the brightness
temperature is directly proportional to the emissivity esp, which is determined by soil dielectric
constant and is thereby most sensitive to soil moisture. As opposed to the passive signature in
this idealized case, there is no backscatter since there is no vegetation (V1 = 0 in (4) without
vegetation) and no rough surface; hence, the soil moisture is not measurable by a radar for bare
soil without any roughness. In reality, there is some surface roughness or some vegetation cover
on moist soil, which affects the sensitivity to soil moisture differently in passive data [Njoku et
al., 2003] and in active data [Nghiem et al., 1993a] until the vegetation cover becomes
sufficiently dense to start masking the soil effects. Therefore, information conveyed in passive
12
and active signatures together may yield more comprehensive results compared to the capability
As an illustration of the passive and active blending, correlation analysis results of observed,
satellite-based remote sensing signatures versus in-situ soil moisture and vegetation
Service (NRCS) Soil Climate Analysis Network (SCAN) site in Lonoke, Arkansas (91.867oW
and 34.833oN) is presented. The vegetation cover in the Lonoke region is primarily agricultural
crops including soybeans (41%), rice (21%), and wheat (11%) [Njoku et al., 2003]. TMI passive
microwave data at 10.7 and 19.3 GHz were collected for all points with their centers located
within 25 km of the Lonoke SCAN site location. More than a year of time series data in 1999-
2000 were analyzed. Results show a wide range of sensitivity in the response of instantaneous
PR (obtained at each local overpass time) and in that of transient SMC after rain events. The
variance between measurements and linear fit values of daily PR (10.7 GHz) versus the
contemporaneous daily mv becomes large at larger values of soil moisture so much that PR can
34% for the same PR around 0.017. This is consistent with the conclusion that a number of
transient soil moisture events recorded in the SCAN data are not evident in the TMI data [Njoku
et al., 2003], from which the retrieved soil moisture can be inconsistent among various transient
precipitation events.
For seasonal trends, seasonal TMI PR (90-day running average) is well correlated with
seasonal soil moisture (90-day running average) measured at a depth of 5 cm at the Lonoke
SCAN site, as indicated by high values of correlation coefficient (Table 1). Plots of seasonal data
(90-day running average) for contemporaneous SCAN mv and TMI PR at both frequencies reveal
13
a hysteresis behavior (Figure 1). Theoretically, (1) suggests that the hysteresis is caused by
attenuation effects on the passive microwave signatures under different vegetation conditions in
different seasons. In Figure 1, the linear fit for all data in the whole year is used as a reference
for each frequency. The polarization ratio PR is mostly above the annual linear fit in fall and
summer seasons from fall equinox in 1999 to spring equinox in 2000, when vegetation starts to
decrease in early fall down to a minimum in winter and then increases toward the spring
equinox. In contrast, PR is mostly below the annual linear fit during spring and summer from
spring equinox in 2000 to fall equinox in 2000, when vegetation increases in spring to a peak in
summer then slightly decrease toward the fall equinox. In different seasons, PR can be larger for
less vegetation in winter and smaller for more vegetation in summer, thereby causing the
hysteresis in seasonal PR versus mv at both frequencies. Also, the vegetation attenuation effects
are less severe at the lower frequency. This is evident by two distinctive characteristics in the
plots at 10.7 and 19.3 GHz (Figure 1): steeper slope at 10.7 GHz, and larger hysteresis at 19.3
GHz. Note that the annual correlation coefficient declines significantly from 0.936 at 10.7
GHz to 0.792 at 19.3 GHz because the hysteresis is much larger, corresponding to strong
vegetation attenuation at the higher frequency. Thus far, this discussion has been theoretical
since the passive microwave analysis has not included any independent data characterizing the
For the active microwave analysis, time-series QSCAT data are extracted within 25 km
around the same SCAN site in the same manner as done in the case for TMI data. In contrast to
the passive microwave case, daily QSCAT backscatter change correlates well with
contemporaneous SMC from rainwater. In fact, daily QSCAT data capture 91% of the rain
events recorded at the Lonoke SCAN site in 1999-2000. To illustrate the high correlation of
14
QSCAT backscatter to transient soil moisture, a regression analysis is carried out for the period
of 4 October to 19 November 1999 when two major rain events occurred as observed in daily in-
situ soil moisture measured at the SCAN site. With the linear formulation in the inverted form of
(5) such that mv = a'0 + b' for backscatter 0 in dB and mv in percent, a' = 8.921%/dB and b' =
111.11% with a high correlation coefficient of 0.907 and a small standard deviation of 3.7% for
the case of backscatter at the horizontal polarization (0HH), which well represents both the initial
impulse of soil moisture increase from rain and the subsequent soil moisture decrease in the
entailed drying process. For backscatter at the vertical polarization (0VV), the result is similar
with SMC of 8.369% for a dB change in 0VV, and thus the backscatter at the vertical polarization
is slightly less sensitive than that at the horizontal polarization to transient soil moisture. This is
consistent with (4) in which the dyadic Greens function includes soil reflection which is
stronger at the horizontal polarization compared to that at the vertical polarization due to the
Brewster effects. Also, the incidence angle at 54o for 0VV is larger than 46o for 0HH, which
means that 0VV suffers from higher attenuation effects due to the longer path length in the
vegetation cover. Nevertheless, QSCAT data can identify rain events even in peak vegetation
conditions when the rain is sufficiently heavy for the backscatter to increase above the seasonal
level of the background backscatter, explaining the high percentage of success in the QSCAT
capability for identification of transient SMC. This capability to capture transient rain events
illustrates the complimentary information that active microwave remote sensing provides in
combination with the seasonal SMC that can be estimated from passive microwave remote
sensing to provide a better representation of soil moisture conditions when both types of satellite-
Regarding seasonal trends, active backscatter data primarily contain vegetation information.
To verify this, normalized difference vegetation index (NDVI) data representing seasonal
vegetation change [Justice et al. 1985; Verdin et al., 1999; Zhang et al., 2010] are used to
compare with seasonal QSCAT backscatter data (90 running average). NDVI is defined by
NDVI = (NIRVIS) / (NIR+VIS), where NIR is the near infrared band and VIS is the visible band
from a spectral sensor such as the Advanced Very High Resolution Radiometer (AVHRR)
aboard the National Oceanic and Atmospheric Administration (NOAA) operational polar-
orbiting satellites. AVHRR NDVI data are collected and averaged within 25 km around the
SCAN site so that the spatial scale of AVHRR NDVI data become compatible with that of
QSCAT data. Results from the linear regression analysis of contemporaneous QSCAT
backscatter and AVHRR NDVI around Lonoke show a very high correlation coefficient of 0.946
for linear 0VV and a lower correlation coefficient of 0.864 for linear 0HH. Therefore, seasonal
QSCAT backscatter can be used to characterize seasonal vegetation change regardless of cloud
cover, which is transparent to QSCAT at the Ku-band frequency of 13.4 GHz. This is consistent
with earlier results on the relation of Ku-band backscatter with NDVI [Moran et al., 1997], with
green leaf area index [Moran et al., 1998], and with above-ground biomass [Nghiem, 2001].
Here, QSCAT data, conveying vegetation change information independent of the passive
data, are used to substantiate the theoretical discussion of vegetation cover effects on passive
microwave presented earlier. For this purpose, seasonal running averaged QSCAT 0VV (more
sensitive to seasonal vegetation change compared to 0HH) and TMI PR at 10.7 GHz (more
sensitive to seasonal SMC compared to 19.3 GHz data) are collocated in time and partitioned in
fall-winter seasons and spring-summer seasons. The hysteresis behavior is clearly observed in
the curve of 0VV versus PR (Figure 2). In fall and winter, PR is below the annual linear fit,
16
corresponding to less vegetation cover as compared to spring-summer PR above the linear fit
with more vegetation cover. The less vegetation cover indicated by lower backscatter in fall and
winter supports the fact that PR is above the annual linear fit in the PR-mv hysteresis (Figure 1)
for less vegetation attenuation effects on PR. Similarly, the more vegetation cover characterized
by higher backscatter in spring and summer confirms the stronger vegetation attenuation forcing
PR to the lower section of the PR-mv hysteresis below the annual linear fit (Figure 1). In terms
of timing, the vegetation peak observed in 0VV occurs in summer after the seasonal soil moisture
SMC in passive polarization ratio data. The active and passive combination can thereby provide
different information that will be missed if only one type of data is used separately.
3. Drought Applications
In U.S., two major factors have been identified as important impediments to early detection and
monitoring of drought and its impacts at the county level where key drought-related decisions are
made: the lack of accurate and objective data source and the coarse level of spatial detail in
drought analyses and results [Nghiem et al., 2010]. These issues are even more severe in many
other countries across the world where there is a significant lack of both data and drought
monitoring systems.
The Lincoln Declaration on Drought Indices in 2009 declared a consensus that the
around the world [World Meteorological Organization, 2009]. The U.S. Drought Monitor
(USDM) [Svoboda et al., 2001, 2002] uses SPI, which is based on preliminary precipitation data
obtained from surface observations from the NOAA Climate Prediction Center (CPC) and
National Climatic Data Center (NCDC). The use of rain gauge data from stations (point
amounts, and rain gauge data can also be in considerable error [Story, 2009]. Rain radars may
provide better coverage for rain rate estimations; however, surface radars suffer from anomalous
rainwater. Satellite precipitation estimates can provide nearly global coverage [Kummerow et
al., 1996; Sorooshian et al., 2000] without AP problems in surface radar meteorology; however,
A more fundamental issue is that, while precipitation data are useful for assessing
meteorological drought, such data may not represent rainwater that actually reaches to land
surface and accumulates in soil, especially in drought-prone regions where virga (dry rains, dry
thunderstorms, or dry lightning) are prevalent. The United Nations World Meteorological
Organization stressed the need for undertaking a comprehensive review to develop common
indices for better early drought warnings in the agricultural and water sectors [World
Meteorological Organization, 2009]. For hydrological and agricultural drought assessment and
monitoring, water on land surface and in soil is the most relevant, and thus soil moisture data
must have a key role. Nevertheless, the current in-situ station network is inadequate, and soil
moisture measurements are too sparse for effective use or even non-existent in many areas
[NIDIS, 2007].
18
Information System (NIDIS) [Western Governors Association, 2004; NIDIS 2006, 2007], the
National Weather Service (NWS) has determined that an effective Cooperative Observer
Network would require a minimum spatial density of one observing site per 1000 km2 across the
country, or a separation of about 24 to 32 km [NIDIS 2007]. The location of each in-situ sensor
must be carefully selected such that the measured soil moisture is representative of the
surrounding area. Furthermore, consistency and persistency in data collections are important in
terms of data quality and data availability across different agencies and across different nations.
In view of the above issues in drought monitoring, recent efforts have enabled certain use of soil
moisture measurements derived from satellite remote sensing data for enhancing drought
monitoring systems [Nghiem et al., 2010]. Several specific results are presented in this section to
illustrate various uses of satellite data with different temporal and spatial scales.
Regarding the issue of the lack of accurate and objective data from limited in-situ station
measurements, we use temporal QSCAT observations together with in-situ station measurements
in the local station area to illustrate how satellite data can help to enhance the drought monitoring
capabilities. Figure 3 presents results at the NCDC Global Summary Of the Day (GSOD)
Station 727760 in Great Falls, Montana (47.467oN, 111.383oW). The NCDC GSOD station
precipitation, temperature, humidity, dew point, pressure, and other meteorological data
19
measured by a global network of weather stations [NCDC, 2010a]. Time series of QSCAT data
together with in-situ measurements around this station are constructed with the Special Satellite-
Station Processor (SSSP) [Nghiem et al., 2003]. SSSP uses an innovative algorithm based on
pointers in computer coding, which enables rapid satellite data extraction and collocation with
in-situ data over numerous global stations represented by a global station mask allowing satellite
data selection to any radius within 60 km around each station. In Figure 3, daily QSCAT data
whose centroids within 25 km around Station 727760 are selected for the horizontal polarization
(more sensitive to soil moisture compare to the vertical polarization) from ascending orbits (~6
am local overpass, and more correlated with soil moisture compared to data from descending
orbits).
QSCAT 0HH data around Great Falls (top panel in Figure 3) clearly identify rain events
before and after the dry period between 9 July and 5 September 2000 when there was very little
rain. In August 2000, the long-term Palmer Drought Index for the region around Great Falls was
Agriculture (USDA) issued Natural Disaster Determinations for drought for the entire state of
Montana in 2000, when severe and persistent drought caused significant losses to agriculture and
other sectors [Resource Management Service, 2004]. The summer drought period observed by
QSCAT is validated by the lack of rain in in-situ precipitation data (bottom panel in Figure 3)
from Station 727760 in the same period, when there were several heat waves indicated by spells
of high temperatures (bottom panel in Figure 3). Both before and after the drought period in the
summer of 2000, QSCAT detected a number of significant rain events which increased
backscatter by about 3 dB, which is equivalent to 26.8% in volumetric soil moisture increase per
the Lonoke rating value of a' = 8.921%/dB. Thus, rainwater from these rain events reached the
20
land surface and significantly increased the moisture in soil. However, precipitation data from
the station rain gauge corresponding to these significant rain events disparately ranged across
one order of magnitude from low values (< 0.2 cm) to high values (> 2.0 cm).
In contrast to SMC, which has a consistent relationship with backscatter change (section 2.3),
the inconsistency in in-situ precipitation data measured by a rain gauge at this station
manner, especially for hydrological droughts. The inconsistency of precipitation data compared
to SMC exists not only in satellite measurements of SMC but also in in-situ data of SMC from
SCAN, which also has rain gauges to measure local precipitation. This problem can be
compounded by the measurement location where data may be affected by localized effects (e.g.,
wind effects, shading effects, equipment maintenance) and by the point-based nature of rain
gauge measurement itself where a local rainfall event may not be recorded if the rain does not
fall over the exact station location. These issues can be addressed with the satellite data to
enhance the drought monitoring in an objective and consistent manner. Moreover, similar time
series of QSCAT signature can be obtained at all grid cells for a full and continuous spatial
coverage while station data are only available at local station points, which may or may not be
Satellite microwave remote sensing data, such as AMSR-E or QSCAT, can be used to monitor
drought and water resources at regional to global scales. Both have swath widths of 1400 km or
larger [Njoku et al., 2003; Tsai et al., 2000], which allows a nearly daily coverage across the
world and as many as two times per day at high latitudes. Several attributes related to water can
21
be obtained from microwave satellite data for drought monitoring from local areas to large
scales. Results are presented here to illustrate the use of satellite data for Texas (TX), an
important agricultural state. Pertaining to the regional climatic regime, drought is inevitable in
TX, where water demand will significantly increase and drought will become more critical
A relevant attribute for water resource and drought assessments is precipitation frequency,
which quantifies the recurrence of rain events in a given period [Gonzlez and Valds, 2004].
Instead of an apparent precipitation frequency (APF) derived from in-situ rain gauge data or
surface rain radar data, a different measure of precipitation frequency is derived from satellite
scatterometer data. This measure is defined as effective precipitation frequency (EPF) because it
accounts for rainwater that effectively reaches land surface and increases soil moisture as
opposed to APF that may have problems with AP, virga, or inconsistent point data. For
applications to QSCAT data, EPF = 100 (NW / NC) is defined as the percentage of the number of
wet days (NW), when soil moisture increase in the top soil layer (5 cm) is 5 % such that the
corresponding backscatter increase is 0.56 dB for 0HH or 0.60 dB for 0VV above the
background level, over the total number of satellite coverage days (NC) without counting missing
EPF is retrieved from QSCAT data across the state of TX over the period of 1 June to 31
August 2009 (left panel in Figure 4). In summer 2009, TX suffered exceptional drought over
most of the southern region of the state as observed in the USDM maps from June to August
2009 (right panels in Figure 4). By August 2009, extreme and exceptional drought conditions
(D3 and D4, respectively) remained persistent across south central TX, where the topsoil
conditions were very dry and river levels were near historic lows [NCDC, 2009]. Consistent
22
with these drought conditions, QSCAT EPF showed little to no rain event across most of
southern TX (black to magenta areas, left panel of Figure 4). In contrast, the soil in a part of the
TX Panhandle was shown to be wetted by rainwater the most often during this time period (light
blue to green and yellow areas, left panel of Figure 4), where abnormally dry (D0) depicted in
the June USDM map had improved to no drought by late August 2009.
While EPF carries information on wet precipitation frequency or how often land surface
becomes wet due to rainwater, daily SMC from QSCAT data represents the quantitative change
in soil moisture or the amount (intensity) of rainwater that accumulates on land surface each day.
to water on land rather than rain drops in the atmosphere (a meteorological parameter). Figure 5
presents maps of daily SMC compared to the semi-monthly average over the state of TX from
early September to early October 2009. Intense SMC (yellow areas in maps), which reflect large
increases in soil moisture conditions occurred across large areas of central TX on 10, 11, 13, and
14 September, as well as 4 October. The SMC results on these days are consistent with torrential
rainfall events reported across central and south TX (rainfall totals up to 20 inches were
recorded in some locations)in September 2009, causing flash flooding [NWS, 2009]. With this
new water input, drought conditions in central and south TX had significantly improved by early
October 2009 (as shown in the USDM map for October 6 in Figure 6).
Complementary to the transient change observed in the QSCAT daily SMC, AMSR-E
passive microwave data provide good measurements of seasonal soil moisture (as discussed in
section 2.3). Figure 6 (left panel) shows the difference in seasonal soil moisture between the
September-October and June-July periods in 2009. AMSR-E seasonal soil moisture results
reveal a large region of increased soil moisture in south and south central TX (blue areas). This
23
July 2009 as seen on the USDM drought maps (right panels in Figure 6). In contrast, an area in
western TX has a substantial reduction in soil moisture by the September-October period (red-
brown areas in left panel of Figure 6) compared to more moist conditions in the June-July period.
This area had a larger EPF observed by QSCAT in the earlier months as seen in the left panel of
Figure 4 for June-August 2009. Therefore, the independent attributes derived from different
remote sensing datasets (QSCAT and AMSR-E) are consistent and thereby cross-verifying the
results. Although the USDM map (lower right panel in Figure 6) suggests a slight improvement
by 6 October 2009, this improved condition was a result of recent transient wetting events
observed in daily SMC from QSCAT (e.g., SMC map for 4 October 2009 in Figure 5). The
above results demonstrate the capability and consistency of different parameters to depict the
state of soil moisture and its transient as well as seasonal changes, which are relevant to drought
A major advantage of satellite data is its large spatial coverage across the continental scale to the
global scale compared to surface in-situ measurements from station networks. Here, the pattern
of soil moisture change, observed by QSCAT and AMSR-E satellites, is examined across the
contiguous United States (CONUS) and compared to rainfall patterns from the regional multi-
sensor precipitation analysis assembled into a national product (stage 4). Maps of stage-4 daily
precipitation (SDP) are available from the National Mosaic and Multi-Sensor QPE [NMQ, 2009].
With large swaths of measurements, QSCAT and AMSR-E can provide coverage over
CONUS on a near daily basis. Nonetheless, data gaps exist and a full coverage of the entire
24
CONUS is not possible within a day, especially when ascending and descending orbit data are
used separately. Figure 7 includes daily SMC maps in May 2009 from QSCAT ascending-orbit
data (upper panel) at about 6 am local overpass time, and from AMSR-E descending-orbit data
(lower panel) at about 1:30 am local overpass time. Both maps have data gaps, which are larger
in AMSR-E data due to its narrower swath compared to that of QSCAT at the vertical
polarization. There are also missing AMSR-E data in most of West Virginia due to invalid
Overall, the patterns of daily SMC from QSCAT and AMSR-E are similar. Both reveal
precipitation water on land surface in the Midwest and the Great Lakes states extending toward
the northeastern U.S., whereas most of the western U.S. was dry. An extensive wet region is
observed across Kansas (KS) and Nebraska (NE) in both SMC maps (marked by the circles in
Figure 7). Interestingly, a well-defined dry area is detected by both QSCAT and AMSR-E just
south of Lake Michigan at the Illinois-Indiana border. Nevertheless, there are discrepancies.
First, the amount of SMC observed by QSCAT can be more than volumetric 10 % (yellow areas
in the upper panel of Figure 7) compared to AMSR-E SMC that barely exceeds 5 % (blue areas,
lower panel, Figure 7). In New York, the region east of Lake Ontario had a large positive SMC
(wet) in the QSCAT map while the AMSR-E SMC showed a slightly negative value (dry).
These differences, given the better sensitivity of QSCAT data to transient SMC, are not
In the case of the discrepancy between QSCAT and AMSR-E SMC in New York, it could be
hypothesized that the difference was due to the different observation times of the two
instruments (6 am for QSCAT and 1:30 am for AMSR-E). However, SDP maps indicate
significant rainfall on 27 May continuing to 28 May 2009 in New York (Figure 8). Thus, the
25
lower sensitivity in AMSR-E data to transient SMC is likely the cause of the differences in SM
results. The SDP map on 28 May 2009 (Figure 8a) also shows a large-scale overall pattern
similar to the SMC observed by both satellites with band of heavier rainfall across the upper
Midwest and Great Lakes region extending into the northeastern states. However, the 28-May
SDP map indicates no precipitation in KS and NE where both QSCAT and AMSR-E detected
rainwater on land surface resulting from the intense rainfall on the prior day (marked by circles
in Figures 7 and 8). This case illustrates that SMC can represent the rainwater accumulated from
preceding strong precipitation events so that the corresponding large amount of water can stay in
the top soil for a period of time after the rain events. As such, SMC is also an indicator of the
There are discrepancies between SMC and SDP. For example, in New Mexico (NM), SDP
observed extensive precipitation across the state while SMC from both QSCAT and AMSR-E
found wetness only in some areas of NM (such as in northeastern NM). This difference suggests
either the rainwater did not fully reach land surface (virga problem) or SDP has uncertainties in
surface radar data (AP problem). Similarly, the SDP pattern was much more widespread
compared to the SMC pattern (Figures 7 and 8) in TX, where AP problems can cause significant
difficulties in precipitation mapping [Story, 2009]. These observations suggest that SMC is more
monitoring.
In an operational environment, science results need to be transitioned into data and image
products with appropriate formats and protocols that can be used by drought experts, such as the
26
USDM authors, in the process of making drought monitoring maps in conjunction with a suite of
attributes from multiple data sources (e.g., vegetation conditions, stream flow, and precipitation).
Here, examples of various SMC products are presented and compared with other traditional
Data from the QuikSCAT satellite are processed at the Jet Propulsion Laboratory in
Pasadena, California (CA), which are then automatically and routinely uploaded to the NOAA
Physical Science Division (PSD) in Boulder, Colorado (CO), where multiple SMC products are
generated. For the data link, a file-transfer-protocol (ftp) is established between JPL using an
automated code and PSD using a semi-automatic approach. In the data transfer process, human
interventions are needed to account for anomalies in satellite data collections and operations;
therefore, a fully automated system requires a full integration with the satellite operation and
The development of various SMC products for use in USDM is an iterative process where
many interim products are made for different attributes of SMC in different formats with various
color protocols. In this process, three SMC attributes, including daily SMC, weekly maximum
SMC, and weekly mean SMC, in five different formats (gif, jpg, png, NetCDF, and kmz) using
five different color protocols, result in 75 different products when all combinations are produced.
Several raster sizes were also considered so that images could be displayed on a drought authors
computer terminal without a missing section in each of the images. The timing of the routine
SMC product delivery is important in the operational environment so that updated results can be
used in time. This timing dictates which period of SMC data should be used to compile weekly
results. For USDM, the weekly SMC products are made ready by Monday afternoon to meet the
27
operational timing requirement for USDM authors to have the inputs by late Monday or early
Tuesday for the preparation of the drought monitor for the current week.
The various SMC products have been evaluated and specific products useful for USDM have
been identified. We present an example in Figure 9, which compares 8-day mean and 8-day
maximum SMC products with the rain-gauge precipitation (RGP) product used in USDM for the
period of 14 October 2008 and the ensuing seven days. The RGP product is made by CPC with
an Oracle database that houses observations from a multitude of sources such as surface weather
measurements from the Automated Surface Observing Systems (ASOS) and from co-operative
observers. These data are run through several quality-control steps and through Geographic
Information System (GIS) processing to obtain RGP maps. About 7000 daily reports of in-situ
rain gauge data are included in the making of the RGP product [Higgins et al., 2000]. RGP maps
are made with different time periods from 5 to 8 days for operational use in the USDM drought
assessment. In this particular example, we compare the full 8-day RGP product with the
corresponding 8-day SMC map. The mean and maximum SMC maps (Figures 9a and 9b), based
on the color protocol with yellow-to-brown for drier conditions and green-to-blue for wetter
conditions, are shown together with the corresponding USDM D-level contours from the USDM
for 14 October 2008. This color protocol is selected to enhance the visibility of the different
The mean SMC map (Figure 9a) reveals a significant value of soil moisture increase (light
blue area) extending from western KS toward its border with CO. In Figure 9a, noticeable
increase in soil moisture (green areas) is observed around the TX Panhandle, in southeastern TX,
central Oklahoma (OK), eastern New Mexico (NM), and in some areas of Montana (MT). The
mean SMC map also shows significant drying across several states in the upper midwest US.
28
Compared to the mean SMC, the maximum SMC map (Figure 9b) indicates much more intensive
soil moisture increase over extensive regions (blue to magenta areas) since it is the peak soil
moisture increase detected at any time in the 8-day period, representing the largest value of any
rainwater detectable on land surface on any given day including the remnant of rainwater from
previous days. In the maximum SMC map, a caveat is that low SMC values (gray and light
While the maximum SMC corresponds to the peak water accumulation on land surface, the
mean SMC provides an assessment of the persistency of the rainwater in soil because the more
the number of days when a significant amount of soil moisture increase occurs the larger the
mean SMC becomes in the given time period. Therefore, it possible to have a large peak SMC
due to an intensive rain event in a day in an area (e.g., blue area between Indiana and Ohio in
Figure 9b) where the mean SMC is low because there is no rain water accumulation in other days
during the 8-day period. Since the persistency of SMC (how long rainwater accumulates and
stays in soil) depends on soil type, infiltration, and runoff processes among other factors, the
mean SMC carries hydrological information and is thus relevant for hydrological drought
monitoring.
For benchmarking, the traditional RGP product used in USDM is included in Figure 9c to
compare with the mean and maximum SMC products. The comparison of the RGP and SMC
products in Figure 9 clearly points to the different characteristics of these measurements: RGP
consists of point data localized at each separated rain gauge station, while SMC are composed of
25-km pixel with a continuous spatial coverage within each satellite swath. Here, an important
note is that a goal of NIDIS is to resolve county-level drought conditions because many
important drought-related decisions are made at that spatial scale. The average size of a county
29
area). Thus, to resolve the county scale of 50 km, the Nyquist sampling theorem requires a
spatial scale of 25 km, which is satisfied by the SMC data. However, RGP can have hourly data
while SMC is available only two times per day at most. Although SMC temporal scale is suitable
for the weekly time scale of USDM, better temporal coverage can improve the overall result,
especially in the tropics where current satellite data gaps are the largest.
While the spatial patterns in both RGP and maximum SMC maps (Figures 9b and 9c) agree
in general over the areas of extensive precipitation in the states of NM, TX,OK, KS, NE, and
Iowa (IA), there is however a large area of discrepancy between maximum SMC and RGP in
MT, Wyoming (WY), and a part of North Dakota (ND). A reason for this discrepancy is SMC
has a memory of any precipitation water as long as it remains on land surface at the time of the
rain gauge data at the in-situ station location. In particular for the case in Figure 9, the wetness
in the maximum SMC was observed the area of discrepancy only on the first day (14 October
2008). Another difficulty is the sparse number of stations in certain regions (e.g., south TX, MT,
Regarding the mean SMC (Figure 9a), a high value requires sufficient rain water to
accumulate on land surface over a significant duration during the period under consideration. As
such, the mean SMC can truly represent the significance in both quantity and persistence of new
precipitation water in soil that are most relevant to hydrological drought monitoring, compared to
both maximum SMC and RGP. For example, maximum SMC and RGP show some precipitation
pattern in NE and IA; however, the same region appears dry in the mean SMC, indicating that the
30
transient rainwater may not be sufficient to sustain the presence of soil moisture over a
significant fraction of the 8-day period to have any significant impact on the overall condition.
For the 2009 growing season (i.e., June to early October), Figures 10 and 11 present a
comparison of QSCAT SMC and USDM results across CONUS. There is an overall consistency
between the two sets of results on a regional scale. For example, there was not much water from
precipitation detected on land surface (Figure 10) in the West and the Southwest, where USDM
maps (Figure 11) show either no improvement (e.g., CA) or more severe drought (e.g., Arizona).
For south TX, no significant wet events occurred in the first part of the growing season, which is
reflected by the severe to extreme drought conditions in the USDM, while the rainfall events in
August and September that lessened the drought conditions are represented by the positive SMC
In the Midwest, Figure 10 reveals extensive SMC in South Dakota (SD), southern Minnesota
(MN), eastern NE, and western IA in June and July 2009. During this same time period, USDM
results consistently indicate some improvement in SD, NE, and IA (primarily change from D0 to
no drought classification), and USDM maps suggest drought levels in MN remained unimproved
or became slightly worse. In the first half of June 2009, SDP results showed extensive rain
pattern in the Midwest, which supports the existence of rain water on land seen in SMC (top left
two panels in Figure 10). Since the mean SMC represents a persistent amount of rainwater on
land surface, SMC inherently reflects information about temperature, wind, insolation, and other
parameters that affect the soil wetness. Therefore, SMC may supplement information in synergy
with other parameters currently used in USDM to enhance the results, especially regarding
In addition, with the participation from the NIDIS team, further advances have been made in
order to demonstrate SMC products in the NIDIS operational environment. For this NIDIS
demonstration, a set of nearly three years of SMC products (2007-2009) are made in the Network
Common Data Form (NetCDF) and delivered for transition into the environment of the NIDIS
portal. The first working prototype is available for the demonstration of SMC maps together
Currently working primary features include zoom, pan (via the grab tool), layer selection (both
SMC and USDM and other layers are selectable separately), transparency control for each layer
(via a dialog box), locator by geographic names, etc. Figure 12 illustrates the SMC
demonstration at the national level with USDM and AHP River Gauge layers, and at the county
level in Nebraska as an example. Further work is on going to include more operational features.
Soil moisture change as measured by satellite can benefit not only drought monitoring, but
also benefits drought forecasting. Skillful forecasts of drought or soil moisture would have
significant uses for agriculture and hydrology (water planning). Recognizing the importance of
seasonal forecasts of drought, NOAA CPC has been issuing such forecasts since March 2000.
These forecasts are designed to indicate whether existing droughts will persist or improve, and
whether a new drought will form. An important first step in creating an improved forecast would
improved knowledge of short-term moisture trends can contribute positively to drought forecasts.
Although there is no certainty that short-term trends will persist, forecasters need to know if
moisture conditions are deteriorating, and how fast they are deteriorating. Such trends serve to
4. Concluding Remarks
As presented in this chapter, both passive and active microwave remote sensing have advantages
and limitations for soil moisture monitoring, which necessitates using them in combination with
each other, as well as in situ observations to more fully characterize soil moisture conditions.
First, in-situ measurements can be obtained many times in a day (e.g., hourly measurements)
whereas a satellite sensor typically collects data one or two times per day depending on latitudes.
Most in-situ stations have a longer time-series compared to satellite data. In particular, AMSR-E
data have been collected since 2002 to the present and QSCAT data were obtained over a decade
(1999-2009), while many rain gauge stations were established several decades ago. Regarding
spatial coverage, satellite data, such as seasonal soil moisture from AMSR-E or SMC from
QSCAT, have two important advantages: (1) areal data represent the condition over the pixel
size, and (2) continuous coverage from the local area to regional and continental scales, whereas
in-situ RGP data or soil moisture network such as SCAN consists of sparse point data separated
in space with different numbers of stations and different data quality in different areas.
Furthermore, there are differences in the characteristics of different attributes measured by in-
situ gauges and by satellite sensors as presented in the benchmark study in the previous section.
Given the advantages and limitations in both surface and remote sensing measurements, the
data should be combined in an optimal synergistic manner to improve the quality of both kinds
of data (in-situ versus remote sensing). Here, an approach is to use satellite data to assess the
representativeness of in-situ point measurement in the surrounding area. For example, soil
moisture data from SCAN can be compared or correlated in time (across months, seasons, or
years) to satellite soil moisture signatures collected over areas with different radii away from the
33
in-situ station, to determine whether and how far the different measurements are correlated. This
is valuable in the selection of station location for long-term maintenance so that the surface data
are valid over the largest area as possible (not just in a close proximity of the station) in view of
Because of the differences in the different data types in time and in space, the role of data
assimilation system [Mitchell et al., 2004; Kumar et al., 2008] is important to integrate various
ground measurements and satellite observations using multiple community land surface models.
The modeling approach allows the various time scales and spatial coverage to be incorporated in
a systematic manner. Since in-situ networks are changing and improving while new satellite
data and products are being developed, land data assimilation needs to be continuously evolved
to account for changes in land surface data from all data sources and to provide enhanced
products that can be used to advance drought monitoring. Furthermore, new measurements can
allow better cross-verifications and validations among different models used in the land data
assimilation systems in an effort to produce the most accurate and high quality products.
Drought is a common climatic phenomenon throughout the world and is not stopped or
limited at geo-political boundaries between different nations, and therefore it is a global problem
requiring international efforts for drought assessment, forecast, and mitigation. In this regard,
satellite data from different nations can contribute to the overall goal. The QSCAT antenna
ceased to spin in November 2009 after its continuous operation collecting global data over a
decade since July 1999. Meanwhile, the Indian Space Agency successfully launched another
scatterometer similar to QSCAT aboard the Oceansat-2 Satellite [Jayaraman et al., 1999] in
September 2009. These satellite events, together with scatterometer data agreement signed
among the different nations, highlight the importance of international collaborations in the use of
34
satellite data to extend the observational record for various applications including drought
monitoring.
Currently, QSCAT is still measuring good backscatter data along narrow tracks at a fixed
azimuth, which are valuable to assist in the calibration and validation of OSCAT. Once
consistently calibrated with QSCAT, OSCAT can continue the QSCAT time-series in producing
SMC results, which are important for drought monitoring based on both present and long-term
data for a consistent assessment of drought conditions within its climatic perspective. In the near
future, China will launch the Haiyan-2 (HY-2) satellite carrying another scatterometer [Dong et
al., 2005]. Moreover, the development of another advanced satellite scatterometer is being
studied in U.S., stemming from the recommendation of the Decadal Survey [National Research
Council, 2007]. Regarding soil moisture measurements, the current European Soil Moisture and
Ocean Salinity (SMOS) mission [Kerr et al., 2010] and the proposed U.S. Soil Moisture Active
and Passive (SMAP) mission [Entekhabi et al., 2010] to be launched in this decade could
provide global measurements critical for drought monitoring. Collectively, these successive
satellite missions would be important to provide multi-decadal data to address the non-stationary
issue in climate change. Moreover, long-term data are necessary for the probabilistic
standardized index approach with multiple time scales of soil moisture variability to be used for
drought monitoring.
Regarding drought monitoring systems, experiences in using satellite data to enhance USDM
and NIDIS are valuable for the development and improvement of international drought
monitoring systems, such as the North American Drought Monitor (NADM) that is a cooperative
effort between drought experts in Canada, Mexico and the United States to monitor drought
across the continent on an ongoing basis [NCDC, 2010b]. In developing countries, the lack of
35
in-situ or surface measurement networks further emphasizes the value of satellite data for
drought monitoring because satellite products such as AMSR-E soil moisture or QSCAT SMC
can be retrieved from global satellite data across international boundaries. For example, Figure
13 presents SMC patterns over the African continent [Nghiem, 2009], from which it is possible to
produce multiple SMC derivatives similar to those developed from USDM (Figure 9). Such
products can contribute to the global drought monitoring as a common goal, for which the Global
overall integrator.
to forecasting is the use of climatology. Given the limited skill of seasonal forecasts of
current conditions forward based on what has happened in the past. In this regard, careful
attention should be paid to the issue of non-stationary due to significant changes in regional
climatic trends in recent years. Once the SMC products are obtained for a suitable number of
years to capture contemporary changes, forecasters may gain knowledge of the probabilities that
soil moisture conditions will likely improve or deteriorate during the following season. One of
the goals of drought forecasting is to cast the forecasts in terms of probabilities so as to provide a
more accurate portrayal of confidence levels, and the statistics of historical soil moisture
conditions can contribute to this effort. In short, improved knowledge of initial moisture
conditions, short-term trends, and climatology has the potential to further the skill of current and
future drought forecasts for not only the United States, but globally as well.
36
Acknowledgments
The research carried out at the Jet Propulsion Laboratory, California Institute of Technology,
was supported by the National Aeronautics and Space Administration (NASA) Water Resources
Area of the NASA Applied Sciences Program. The contribution of the NIDIS team in the SMC
product demonstration in the NIDIS portal is acknowledged. Thanks to Robert Rabin from the
NOAA National Severe Storms Laboratory for his direction to Stage-4 daily precipitation
References
and A. Vidal, Evaluation of the ERS-1 Synthetic Aperture Radar capacity to estimate surface
soil-moisture 2-year results over the Nazin watershed, Water Resources Res., 31(4), 975-
982, 1995.
behavior of wet soil, Part II: Dielectric mixing models, IEEE Trans. Geosci. Remote Sens.,
Dong, X., K. Xu, H. Liu, and J. Jiang, The radar altimeter and scatterometer of Chinas HY-2
doi:10.1109/IGARSS.2004.1370659, 2004.
37
Dubois, P., J. J. van Zyl and T. Engman, Measuring soil moisture with imaging radars, IEEE
Engman, E. T., Recent advances in remote-sensing in hydrology, Rev. Geophys., 33, Part 2,
Spencer, S. W. Thurman, L. Tsang, Leung, and J. J. Van Zyl, The Soil Moisture Active
events for drought analyses, Natural Hazards and Earth Sys. Sci., 4(1), 17-28, 2004.
dielectric behavior of wet soil, Part I: Empirical models and experimental observations, IEEE
Higgins, R. W., W. Shi, E. Yarosh, and R. Joyce, Improved United States precipitation quality
control system and analysis, NCEP/Climate Prediction Center ATLAS No. 7, U.S.
http://www.cpc.noaa.gov/products/outreach/research_papers/ncep_cpc_atlas/7/, 2000.
Hutchinson, J. M. S., Estimating near-surface soil moisture using active microwave satellite
imagery and optical sensor inputs, Trans. ASAE, 46(2), 225-236, 2003.
Jackson, T. J., Soil moisture estimation using SSM/I satellite data over a grassland region, Water
Jayaraman, V., V. S. Hedge, M. Rao, and H. H. Gowda, Future Earth observation missions for
1999.
global vegetation using meteorological satellite data, Int. J. Remote Sens., 6(8), 1271-1318,
1985.
Kerr, Y. H. and E. G. Njoku, A semiempirical model for interpreting microwave emission from
semiarid land surfaces as seen from space, IEEE Trans. Geosci. Rem. Sens., 28, 384-393,
1990.
Kerr, Y. H., P. Waldteufel, J.-P. Wigneron, J.-M. Martinuzzi, J. Font, and M. Berger, Soil
moisture retrieval from space: The Soil Moisture and Ocean Salinity (SMOS) mission, IEEE
S. Mecklenburg, The SMOS mission: New tool for monitoring key elements of the global
Eylander, and P. R. Houser, A land surface data assimilation framework using the land
39
1419-1432, 2008.
Kummerow, C., W. S. Olson, and L. Giglio, A simplified scheme for obtaining precipitation and
vertical hydrometeor profiles from passive microwave sensors, IEEE Trans. Geosci. Remote
Lautenbacher, C. C., The Global Earth Observation System of Systems: Science serving society,
Li, Q., L. Tsang, J. Shi, and C. H. Chan, Application of physics-based two-grid method and
sparse matrix canonical grid method for numerical simulations of emissivities of soils with
rough surfaces at microwave frequencies, IEEE Trans. Geosci. Rem. Sens., 38, 1635-1643,
2000.
Loew, A., R. Ludwig, and W. Mauser, Derivation of surface soil moisture from ENVISAT
ASAR wide swath and image mode data in agricultural areas, IEEE Trans. Geosci. Remote
Ramsay, and A. A. Bailey, The multi-institution North American Land Data Assimilation
System (NLDAS): Utilizing multiple GCIP products and partners in a continental distributed
doi:10.1029/2003JD003823, 2004.
40
Inoue, and C. M. U. Neale, , Combining multifrequency microwave and optical data for crop
Moran, M. S., A. Vidal, D. Troufleau, Y. Inoue, and T. A. Mitchell, Ku- and C-band SAR for
discriminating agricultural crop and soil conditions, IEEE Trans. Geosci. Remote Sens., 36,
265-272, 1998.
Myneni, R. B., C. D. Keeling, C. J. Tucker, G. Asrar, and R. R. Nemani, Increased plant growth
in the northern high latitude from 1981 to 1991, Nature, 386(6626), 698-702, 1997.
National Research Council, Earth Science and Applications from Space: National Imperatives
for the Next Decade and Beyond. The National Academies Press, Washington, D.C., 2007.
NCDC, State of the Climate Drought, August 2009, National Climatic Data Center, NESDIS,
year=2009&month=8, 2009.
NCDC, Global Surface Summary of the Day GSOD, National Climatic Data Center, NESDIS,
2010a.
NCDC, North American Drought Monitor, National Climatic Data Center, NESDIS, NOAA,
Research - Polarimetric Remote Sensing, ed. by J. A. Kong, Elsevier, New York, Vol. 3,
Nghiem, S. V., T. Le Toan, J. A. Kong, H. C. Han, and M. Borgeaud, Layer model with random
spheroidal scatterers for remote sensing of vegetation canopy, J. Electromag. Waves Applic.,
Nghiem, S. V., R. Kwok, S. H. Yueh, and M. R. Drinkwater, Polarimetric signatures of sea ice,
Nghiem, S. V., J. J. Van Zyl, W.-Y. Tsai, and G. Neumann, Potential application of
scatterometry to large-scale soil moisture monitoring, JPL Doc. D-19523, Jet Propulsion
Nghiem, S. V., Advanced scatterometry for geophysical remote sensing, JPL Doc. D-23048, Jet
Nghiem, S. V., E. G. Njoku, J. J. Van Zyl, and Y. Kim, Global energy and water cycle Soil
moisture variability pattern over continental extent observed with active and passive satellite
data, JPL Doc. D-26225, Jet Propulsion Lab., California Inst. Tech., Pasadena, California,
2003.
42
Nghiem, S. V., E. G. Njoku, G. R. Brakenridge, and Y. Kim, Land surface water cycles observed
with satellite sensors, J6.14, Proc. 19th Conference on Hydrology and 16th Conference on
Climate Variability and Change, 85th Amer. Met. Soc. Meeting, San Diego, California, 2005.
Nghiem, S. V., NASA satellite data for applications to early warning of droughts across the
world Examples for Africa, Inter-Regional Workshop on Indices and Early Warning
Engman, and D. Toll, Improved drought monitoring with NASA satellite data, EWRI
Currents, 12(3), 7, Environ. Water Res. Inst., Amer. Soc. Civil Eng., 2010.
NIDIS, U.S. Public Law 109-430, 109th Congress, 2nd sess., National Integrated Drought
NIDIS, The National Integrated Drought Information System Implementation Plan A Pathway
Njoku, E. G., and L. Li, Retrieval of land surface parameters using passive microwave
measurements at 6-18 GHz, IEEE Trans. Geosci. Rem. Sens., 37, 79-93, 1999
Njoku, E. G., T. J. Jackson, V. Lakshmi, T. K. Chan, and S. V. Nghiem, Soil moisture retrieval
from AMSR-E, IEEE Trans. Geosci. Remote Sens., 41(2), 215-229, 2003.
Njoku, E., AMSR-E/Aqua L2B Surface Soil Moisture, Ancillary Parms, & QC EASE-Grids
NMQ, Stage-4 24hr QPE Accumulation, National Mosaic & Multi-Sensor QPE, NOAA National
NWS, September 2009 Weather in Review, NOAA National Weather Service, Southern Region
Owe, M., A. Chang, and R. E. Golus, Estimating surface soil moisture from satellite microwave
measurements and a satellite-derived vegetation index, Rem. Sens. Environ., 24, 131-345,
1988.
the characteristics of vegetation canopies with airborne radar measurements, Int. J. Rem.
Resource Management Services, State of Montana multi-hazard mitigation plan and statewide
hazard assessment, Land and Water Consulting, Big Sky Management, 251 pp., 2004.
between ERS-2 SAR backscatter and soil moisture: Generalization from a humid to semi-arid
PERSIANN system satellite-based estimates of tropical rainfall, Bull. Amer. Meteorol. Soc.,
Story, G., The difficulty of achieving good precipitation estimates for use in real-time drought
monitoring, 6th U.S. Drought Monitor Forum, Lower Colorado River Authority, Austin, TX,
Svoboda, M. D., M. J. Hayes, and D. A. Wilhite, The role of integrated drought monitoring in
Palecki, D. Stooksbury, D. Miskus, and S. Stephens, The Drought Monitor, Bull. Amer.
Takada, M., Y. Mishima, and S. Natsume, Estimation of soil surface properties in peatland using
Texas Water Development Board, Highlights of the 2007 State Water Plan, Water for Texas
Teng, W. L., Wang, J. R., and Doriaswamy, P. C., Relationship between satellite microwave
radiometric data, antecedent precipitation index, and regional soil moisture, Int. J. Rem.
USDA, SCAN - Soil Climate Analysis Network, SCAN Brochure, Natural Resources
Conservation Service, National Water & Climate Center National Soil Survey Center, 2009a.
USDA, SNOTEL And Snow Survey & Water Supply Forecasting, SNOTEL Brochure, Natural
Resources Conservation Service, National Water & Climate Center, NWCC Rev. 3, 2009b.
van de Griend, A. A. and M. Owe, Microwave vegetation optical depth and inverse modelling of
soil emissivity using Nimbus/SMMR satellite observations, Meteorol. Atmos. Phys., 54,
225-239, 1994.
Verdin, J., C. Funk, R. Klaver, and D. Roberts, Exploring the correlation between Southern
Africa NDVI and Pacific sea surface temperatures: results for the 1998 maize growing
Wagner, W., J. Noll, M. Borgeaud, and H. Rott, Monitoring soil moisture over the Canadian
prairies with the ERS scatterometer, IEEE Trans. Geosci. Remote Sens., 37(1), 206-216,
1999.
Wagner, W., and K. Scipal, Large-scale soil moisture mapping in western Africa using the ERS
scatterometer, IEEE Trans. Geosci. Remote Sens., 38(4), Part 2, 1777-1782, 2000.
Wang, J. R., and B. J. Choudhury, Remote sensing of soil moisture content over bare field at 1.4
Wang, J. R., Passive microwave sensing of soil moisture content: the effects of soil bulk density
Wang, J. R., Effect of vegetation on soil moisture sensing observed from orbiting microwave
Western Governors Association, Creating a Drought Early Warning System for the 21st
Century, The National Integrated Drought Information System, 13 pp., Denver, Colorado,
2004.
microwave emission from rough agricultural soils, IEEE Trans. Geosci. Rem. Sens., 39,
1697-1707, 2001.
World Meteorological Organization, Expert agree on a universal drought index to cope with
climate risks, Press Release, WMO No.-872, United Nations, Copenhagen, Geneva, 2009.
Drought-induced vegetation stress in southwestern North America, Environ. Res. Lett., 5(2),
List of Table
Table 1. Correlation results between seasonal TMI polarization ratio PR and seasonal SCAN
volumetric soil moisture mv at 5-cm depth from linear regression analysis in the form of PR =
List of Figures
Figure 1. Seasonal TMI polarization ratio versus seasonal SCAN soil moisture at 5-cm depth in
an agricultural area at Lonoke, Arkansas. All data are contemporaneous (collocated in time), and
are 90-day running averages. The upper plot is for 10.7 GHz, and the lower one is for 19.3 GHz.
Figure 2. Seasonal QSCAT backscatter 0VV versus seasonal TMI polarization ratio PR at 10.7
GHz in an agricultural area within 25 km around Lonoke, Arkansas. All data are
Figure 3. Measurements around the NCDC GSOD Station 727760 (47.467oN 111.383oW) at
Great Falls in Montana. Top panel is QSCAT 0HH within 25 km around the station, middle
panel is in-situ air temperature (magenta for minimum, black for average, and red for maximum),
and bottom panel is precipitation from station rain gauge (there were missing data). Thin vertical
lines align rain events to backscatter impulses. Yellow marks the period between 7/9/2000 and
Figure 4. Effective precipitation frequency (%) measured by QSCAT the period of June-August
2009 (left panel), and drought levels from D0 to D4 from the U.S. Drought Monitor for weeks
ending on the marked dates in 2009 (right panels). The USDM drought levels include D0 for
abnormally dry, D1 for moderate drought, D2 for severe drought, D3 for extreme drought, and
Figure 5. Soil moisture change (SMC) measured by QSCAT with the vertical polarization along
ascending orbits in September to early October 2009. The color scale represents backscatter
10/6/2009) mv(6/29/2009-7/28/2009) showing seasonal SMC (left panel), and drought condition
change between USDM drought maps in July and in September 2009 (right panels).
Figure 7. Soil moisture change (SMC) on 28 May 2009 compared to the two-week average
between 14-28 May 2009 observed by: (a) QSCAT SMC represented by backscatter change in
dB and by volumetric moisture change in % from the Lonoke rating, and (b) AMSR-E by
volumetric moisture change in % with yellow-brown for drier and cyan-blue for wetter
conditions.
Figure 8. Stage-4 24-hour precipitation measurements [NMQ, 2009] at 12:00 UTC in inches for:
(a) 28 May 2009 in the upper panel, and (b) and 27 May 2009 in the lower panel.
49
Figure 9. Comparison of QSCAT SMC with rain-gauge precipitation (RGP) for the period of 14
October 2008 and the ensuing 7 days: (a) mean SMC, (b) max SMC, and (c) RGP used in making
USDM maps.
Figure 10. Weekly QSCAT mean SMC maps for the growing season in June-October 2009.
Figure 11. Weekly USDM maps for the growing season in June-October 2009.
Figure 12. SMC product demonstration in the NIDIS portal at the national level with USDM
and AHP River Gauge layers (upper panel), and at the county level for Nebraska (lower panel).
Figure 13. Soil moisture change (SMC) across Africa [Nghiem, 2009] measured by QSCAT
with the vertical polarization along ascending orbits on 16 October 2009. The color scale
represents backscatter change in dB, and volumetric SMC in % with the Lonoke rating. Data
gaps between orbits are seen in dark bands on land. High SMC is observed in an extensive
pattern (yellow areas) across South Africa curving northward to Botswana, Namibia, Zambia,
Table 1. Correlation results between seasonal TMI polarization ratio PR and seasonal SCAN
volumetric soil moisture mv at 5-cm depth from linear regression analysis in the form of PR =
(a)
(b)
Figure 1. Seasonal TMI polarization ratio versus seasonal SCAN soil moisture at 5-cm depth in
an agricultural area at Lonoke, Arkansas. All data are contemporaneous (collocated in time), and
are 90-day running averages. The upper plot is for 10.7 GHz, and the lower one is for 19.3 GHz.
52
Figure 2. Seasonal QSCAT backscatter 0VV versus seasonal TMI polarization ratio PR at 10.7
GHz in an agricultural area within 25 km around Lonoke, Arkansas. All data are
contemporaneous (collocated in time), and are 90-day running averages.
53
Figure 3. Measurements around the NCDC GSOD Station 727760 (47.467oN 111.383oW) at
Great Falls in Montana. Top panel is QSCAT 0HH within 25 km around the station, middle
panel is in-situ air temperature (magenta for minimum, black for average, and red for maximum),
and bottom panel is precipitation from station rain gauge (there were missing data). Thin vertical
lines align rain events to backscatter impulses. Yellow marks the period between 7/9/2000 and
9/5/2000 when there was very little rain.
54
Figure 4. Effective precipitation frequency (%) measured by QSCAT the period of June-August
2009 (left panel), and drought levels from D0 to D4 from the U.S. Drought Monitor for weeks
ending on the marked dates in 2009 (right panels). The USDM drought levels include D0 for
abnormally dry, D1 for moderate drought, D2 for severe drought, D3 for extreme drought, and
D4 for exceptional drought [Svoboda et al., 2001, 2002].
55
Figure 5. Soil moisture change (SMC) measured by QSCAT with the vertical polarization along
ascending orbits in September to early October 2009. The color scale represents backscatter
change in dB, and volumetric SMC in % with the Lonoke rating.
56
Figure 7. Soil moisture change (SMC) on 28 May 2009 compared to the two-week average
between 14-28 May 2009 observed by: (a) QSCAT SMC represented by backscatter change in
dB and by volumetric moisture change in % from the Lonoke rating, and (b) AMSR-E by
volumetric moisture change in % with yellow-brown for drier and cyan-blue for wetter
conditions.
58
Figure 8. Stage-4 24-hour precipitation measurements [NMQ, 2009] at 12:00 UTC in inches for:
(a) 28 May 2009 in the upper panel, and (b) and 27 May 2009 in the lower panel.
59
Figure 9. Comparison of QSCAT SMC with rain-gauge precipitation for the period of 14
October 2008 and the ensuing 7 days: (a) mean SMC, (b) max SMC, and (c) precipitation used in
making USDM maps.
60
Figure 10. Weekly QSCAT mean SMC maps for the growing season in June-October 2009.
61
Figure 11. Weekly USDM maps for the growing season in June-October 2009.
62
Figure 12. SMC product demonstration in the NIDIS portal at the national level with USDM
and AHP River Gauge layers (upper panel), and at the county level for Nebraska (lower panel).
63
Figure 13. Soil moisture change (SMC) across Africa [Nghiem, 2009] measured by QSCAT
with the vertical polarization along ascending orbits on 16 October 2009. The color scale
represents backscatter change in dB, and volumetric SMC in % with the Lonoke rating. Data
gaps between orbits are seen in dark bands on land. High SMC is observed in an extensive
pattern (yellow areas) across South Africa curving northward to Botswana, Namibia, Zambia,
and Mozambique, with a wet region seen in Madagascar.
64
Chan, and G. Neumann, Microwave Remote Sensing of Soil Moisture Science and
Applications, book chapter, 53 pp., in Drought and Water Crises Book Series - Remote
Sensing and Drought New and Emerging Monitoring Approaches, Taylor and Francis Pub.,
in review, 2011.
Nghiem, et al. (authors from JPL, USGS, NDMC, NOAA PSD, DFO, and others), Pattern and
Frequency of Soil Moisture Variability over the Continental United States, manuscript in
Engman, and D. Toll, Improved Drought Monitoring with NASA Satellite Data, EWRI
doi:10.1016/j.isprsjprs.2009.01.004, 2009.
Nghiem, S. V., and G. Neumann, Remote Sensing of the Global Environment with Satellite
Environment VI, ed A. Valinia, P. H. Hildebrand, and S. Uratsuka, Proc. of SPIE, Vol. 7154,
Press Release
JPL Photo Journal, Rapid Dry-Up of Rainwater on Land Surface Leading to the Santa Barbara
Nghiem, S. V., NASA Satellite Data for Applications to Early Warning of Droughts across the
World Examples for Africa, Inter-Regional Workshop on Indices and Early Warning
Nghiem, S. V., Satellite Observation of Soil Moisture Change and Applications to Drought
Monitoring, 6th U.S. Drought Monitor Forum, Lower Colorado River Authority, Redbud
Nghiem, S. V., Geophysical Information from NASA Satellite Scatterometry Western U.S.
and California, ESRI International User Conference, San Diego Convention Center, San
Nghiem, S. V., Satellite Remote Sensing of Soil Moisture for Drought Applications, invited
Nghiem, S. V., G. R. Brakenridge, and G. Neumann, Drought, wetland, and Flood Monitoring
with Satellite Scatterometer, EOS Trans, AGU, 88(23), Jt. Assem. Suppl., Abst. U53B-05,
May 2007.
Yueh, and T. Zhang, Global Observations of Land Surface Water with Satellite Active and
Passive Microwave Sensors, Satellite Observations of the Global Water Cycle, Irvine,
Nghiem, S. V., Drought Monitoring with NASA Satellite Data, National Integrated Drought
U.S. Geological Survey, EROS Data Center, Sioux Falls, SD 57198, (605) 594-6018
Summary
New soil moisture and vegetation monitoring products were developed for integration into the
weekly production of the U.S. Drought Monitor (USDM) map (http://drought.unl.edu/dm), the
recognized national reference for drought conditions in the United States. A soil moisture
change (SMC) product was developed using NASA scatterometer and radiometer data (led by
JPL, see Section 1). The Vegetation Drought Response Index (VegDRI), originally formulated
to use imagery from NOAA AVHRR instruments, was updated to use imagery from NASA
Production of VegDRI with MODIS imagery was implemented on a very fast turn-around (less
than 12 hours after satellite acquisition) by taking advantage of the new NASA LANCE system
for rapid delivery of swath surface reflectance data, and the implementation of a new expedited
MODIS (eMODIS) processing chain at USGS EROS. Frequency and timing of production were
designed to meet USDM author weekly schedules, and automated procedures were implemented
to ingest SMC and VegDRI products into the GIS environment used by authors to make weekly
adjustments to drought category zones. Based on positive feedback from USDM authors, USGS
68
EROS is now investing in the transition of VegDRI production from science to operations.
Systems engineering staff are implementing a robust, automated operational VegDRI production
chain to support the U.S. drought monitoring community on into the future, beyond the life of
this project. The operational VegDRI system is expected to be actively supporting USDM
Initial development of the Vegetation Drought Response Index (VegDRI) began in 2002, funded
by a seed grant from the U.S. Geological Survey (USGS). The grant funded prototype efforts, in
collaboration with the National Drought Mitigation Center (NDMC), to develop methods to
improve drought monitoring in the north central plains of the U.S. The initial study (outlined by
vegetation based on phenological indicators using data from a series of optical sensorsthe
Advanced Very High Resolution Radiometer (AVHRR). This seed money also supported
writing several research proposals. Subsequent funding awarded in 2006 by the NASA Applied
and geographic expansion of the VegDRI modeling approach to the conterminous U.S.
(CONUS). The project goal was to improve drought early warning response and mitigation
through the use of timely monitoring products designed to meet the needs of the U.S. community
of drought-sensitive decision-makers.
The VegDRI methodology builds upon the monitoring traditions of both the climate and remote
sensing communities. The approach provides improvement in spatial detail, spatial coverage,
69
and the timely delivery of information in a variety of accessible formats (for example, maps,
descriptive text, and statistics), increasing the value, spatial detail, and relevancy of drought
information available to decision makers. A key project goal was to integrate VegDRI into the
weekly operational process of making the U.S. Drought Monitor (USDM) map by providing
Since its launch in 1999, the USDM (Svoboda et al. 2002) has been the state-of-the-practice
drought monitoring tool used in the United States. The USDM provides a general assessment of
weekly drought conditions (both agricultural and hydrologic) across the nation. It is produced
using a hybrid approach that involves a number of variables including short- and long-term
climate-based drought indicators, hydrologic indices, and remote sensing information. Ten years
ago, the USDM map was produced at the effective scale of climate divisions (multi-county
aggregations with similar climate), however developments in recent years, including finer
resolution point source and gridded climatic data, have led to improvement in resolution
approaching individual county scale or better when we incorporate remote sensing products
Climate and meteorological data have long been the primary sources for creating drought indices
for monitoring, including the USDM. Climate-based drought indices characterize the intensity
of dryness as compared to long-term average or normal condition, and are usually calculated
from one or more of the following variables: rainfall, temperature, snow pack, stream flow, soil
moisture, and other water supply indicators. The spatial coverage and detail of climate-based
drought indices is limited as these depend upon meteorological data collected at stations that are
characterize relatively broad-scale drought patterns and the level of accuracy and spatial detail in
70
the information they provide depends on the density and geographic placement of stations across
the landscape.
Satellite-based observations have proven very useful for assessing broad-scale vegetation
condition anomalies (that is, apparent changes in vegetation health), but the specific cause or
causes for the anomalies may not always be determined solely from the remotely sensed data. A
number of natural (for example, drought, flooding, fire, pest infestation, and hail damage) and
anthropogenic (for example, land cover/land use conversion) events can produce these anomalies
(Asner et al. 2000; Breshears 2005; Goetz et al. 2006; Kogan 1990; Parker et al. 2005; Peters et
al. 2000; Peters et al. 2002; Wang et al. 2003). Therefore, effective drought monitoring
approaches must consider both climate station and satellite-based information, as well as other
environmental parameters that may influence the effects of drought and its severity on
vegetation, such as soil properties or land cover type. It follows that the integration of coarser-
resolution climate data and higher-resolution satellite-based vegetation observations will provide
an improvement to monitoring and characterization of the spatial extent, intensity, and local
Because the effects of drought vary from region to region and season to season, there are many
approaches for early warning and monitoring that range from sub-national, to national, regional,
71
and even global in geographic scope. To be successful, monitoring approaches must deal with a
multitude of challenging drought characteristics (e.g., the difficulty in determining drought onset
and termination, the multiple, varying definitions of drought, and the existence of multiple
the United States, largely through the coordination and leadership of the U.S. Drought Monitor
(USDM) authors (Svoboda et al. 2002) and the National Integrated Drought Information System
(WGA 2004), has suggested product/data requirements for successful drought monitoring shown
in Table 1.
The VegDRI system was designed to meet many of these requirements. The 1-km2 resolution
VegDRI maps (Figure 1) depict drought-specific information related to vegetation condition with
higher spatial detail than traditional drought monitoring tools such as the Palmer Drought
Severity Index (PDSI) and USDM. Thanks to its spatial detail and coverage, VegDRI can
72
support drought assessment at local to national levels. Because of its weekly production and
release before noon (Eastern Time) on each Monday, VegDRI supports the weekly production
VegDRI Overview
VegDRI characterizes the level of drought stress on vegetation by integrating two satellite-based
time-series vegetation index-based metrics, two climate-based drought indices, and five
73
biophysical characteristics of the environment. Further details on the basic input variables and
the methodology are found in (Brown et al. 2008). The VegDRI methodology consists of three
main steps:
1. Processing, summarization, and organization of the data for each input variable into a
database
regression tree (CART) analysis technique to the historical information in the database
VegDRI maps portray drought severity in seven categories reflecting varying levels of drought-
induced vegetation stress that are based on the PDSI classification scheme (Wells et al. 2004;
Palmer, 1965). The maps also include three additional classes that depict areas over which
VegDRI values cannot be calculated. These additional classes include water, out of season (i.e.,
time when the vegetation is not photosynthetically active for a location), and no season (i.e.,
locations in the southwestern U.S. where there was no detectable vegetation response in the
historical VI data).
Several significant improvements to the VegDRI system were made in carrying out this NASA
(1) use of expedited MODIS (eMODIS) imagery instead of AVHRR to calculate the satellite
phenological indicators,
(2) an automated production flow for the near-real time geospatial model,
The satellite VI data ingested into VegDRI are collected from two daily polar-orbiting earth
observing systems, AVHRR and MODIS. These instruments provide frequent, rapid, and
synoptic measurements of land surface conditions across large geographic regions. A more than
20-year history of AVHRR time-series data across the United States provides historical context
for monitoring drought conditions (Eidenshink, 2006). However, the newer MODIS instrument
has improved sensor characteristics specifically designed for global land surface monitoring.
The MODIS sensors carried aboard the NASA Terra and Aqua satellite platforms have 36
spectral channels and have been extensively used for land and atmospheric monitoring since
Normalized Difference Vegetation Index (NDVI)] are similar to AVHRR products but provide
higher spatial and spectral resolutions and improved atmospheric and geometric corrections
(Townshend and Justice 2002). AVHRR data have a longer historical record, allowing more
anomalous conditions. Therefore, both AVHRR and MODIS NDVI data are valuable inputs for
operational monitoring.
Since standard MODIS products are often not delivered to users quickly enough to aid
operational decisions (typically 6-10 day lag times between satellite overpass and standard
product data availability have been noted in the past), USGS has designed and is now operating
an expedited MODIS, or eMODIS system that takes advantage of NASAs new Land
75
Atmosphere Near-real time Capability for EOS (LANCE). The eMODIS system at EROS
(Figure 2) provides the near-real time MODIS VI and surface reflectance data (typically less than
10 hours after satellite overpass) needed to supply VegDRI products on a schedule that meets
eMODIS Historical
Terra MODIS
+ 2-4hrs
LAADS USGS Drought
VegDRI
Monitoring
MODAPS
EDOS
eMODIS Expedited
+ 4hrs NIDIS Drought
MODIS L0 Data Portal
+ 3hrs
Each expedited NDVI composite product from eMODIS is derived from data delivered in near
real time by LANCE. eMODIS ingests level-2 surface reflectance swaths from LANCE as soon
as they are available (3-6 hours after acquisition). The swaths are gridded per VegDRI
specifications, and NDVI is calculated. The gridded NDVI pixels are evaluated by an enhanced
maximum value compositing routine to select the best available data for each 7-day composite.
76
Expedited eMODIS NDVI is produced daily using the most recent 7 days of data. The expedited
products are later replaced in the permanent archive by precision eMODIS data generated from
standard reflectance inputs within 3-10 days as they are delivered from the Level 1 and
The VegDRI System consists of a set of software components, primarily based on C code, that
can be initiated independently and manually. For near-real time processing, the components are
run in a processing stream that is automated using scripting. Appendix 1 shows the components
Operational Status
VegDRI production is being transitioned by USGS to operational status. In the current status,
VegDRI components are run in a processing stream that is initiated at 00:01 a.m. each Monday.
Automated processing ordinarily completes by 10:30 a.m. each Monday morning, 24 hours after
Terra MODIS satellite data acquisition. When a problem is encountered with one of the required
inputs or one of the processing components, completion of processing is delayed, but can be
analyzed, completed, and/or re-run within a day. The initiation of each component for each
interim product is automated from scripts but can be done manually. Any component testing or
verification of results is done by staff, nominally during normal business hours (8:00 a.m. 4:00
p.m.).
77
Maintenance
Current VegDRI production streams [7-day (Monday a.m.) CONUS] are maintained without
interruption and fully supported by the necessary hardware and software components.
Software development completed to date is included in the VegDRI prototype operations and
production in the form of logs are emailed to staff. Software engineering staff remains
responsible for production tasks as well as for troubleshooting and catastrophic recovery actions.
system is currently a manual process because components were not developed to support a
recovery methodology. Each component was developed independently, which allows flexibility
of use, but they were not optimized with operational considerations. They are joined together
with automation scripts to create an operational work flow. Failures can occur at various steps in
processing, often without warning or notification when external inputs are unavailable. Previous
failures have involved hardware and network issues, problems with collaborator inputs, and
problems with external interfaces. The VegDRI system relies on software engineering staff to
recognize and mitigate these issues, which requires a growing knowledge base to facilitate
recognition of expected anomalies versus actionable system failures. As each individual issue
has been accommodated, processing has become more stable, but some issues are due to external
dependencies.
78
Model Recalibration
Annual recalibration of VegDRI models is desired to incorporate new time series data and new
regression tree models incorporating data inputs (eMODIS NDVI, gridded climate surfaces, and
biophysical parameters).
Extensibility
and input data. VegDRI models and data can be extended to additional geographic windows.
Agriculture and Agri-Food Canada. Finally, satellite continuity is critical and in the post-
MODIS era, extension to the Visible Infrared Imager Radiometer Suite (VIIRS) will be essential
System Configuration
The VegDRI system is installed on a Dell Server (purchased in 2010) running the linux operating
system with the following components: a dual-processor CPU with 32GB of memory, 2.5TB of
RAID disk storage, redundant power supplies. The server is covered by a five year warranty.
The eMODIS expedited and historical composites ingested into VegDRI are stored for future use
in event of problems or new processing requirements. The weekly products are stored for use in
subsequent years ranking. SPI data are stored for future use when models are recalibrated.
79
Model parameter and input files are stored for comparative analysis. RAID storage is used for all
data and is estimated to provide adequate storage until the end of 2015.
Data continuity from daily polar orbiting satellites is extremely important to successful drought
monitoring. Since drought reveals itself in relation to the normal climate condition at a
specific location, the continuity of repeated observations allows for the objective calculation of
both normal and anomalous conditions. Both AVHRR and MODIS sensors are incorporated
into VegDRI for drought monitoring. In the United States, regular vegetation monitoring started
in 1989 (well before the MODIS era) and continues today, facilitated by EROS direct reception
and processing of AVHRR data into national composites on weekly and 14-day time steps
(Eidenshink 2006). A data processing system produces and delivers AVHRR data to users
within 30 hours of satellite overpass. The 20+ year satellite record created by this system is used
The long-term continuity of the AVHRR and MODIS data records (and extending into the future
with data from planned future instruments) is critical for monitoring land surface conditions. It
is a goal of the USGS to develop a linkage between AVHRR and MODIS to ensure the
and in other cases, on methods of inter-calibrating the pertinent surface reflectance bands. Inter-
sensor compatibility was examined and translation equations were derived to seamlessly extend
the multi-sensor data record. Multi-sensor translations were based on an overlapping period of
80
observations using geometric mean regressions (GMRs) to treat variations in both AVHRR and
Two AVHRR-MODIS data translation equations (Eq. A and Eq. B ) have been developed to
support United States drought monitoring (T. Miura, pers. communication). Phenological regions
(Gu et al., 2010) were used as a geographic framework for extracting data to derive the
equations. Eq. A was developed using a combination of canonical correlation analysis and GMR
NDVIp AVHRR = 9.89 + 0.76 * (0.99 * NDVI MODIS + 0.002 * NDVI MODIS 2)
Where NDVIp AVHRR is the NDVI from AVHRR data, and NDVI MODIS is the NDVI from
Eq. B was developed based on a maximum likelihood regression or estimation method and the
phenological regions framework. Eq. B formulation was accomplished with 1 km2 resolution
AVHRR and eMODIS Terra NDVI data for 2005-2007. There are 19 phenological class-
dependent translation equations and one general (or default) translation equation in Eq. B. Eq. B
was evaluated with 2008 satellite data using mean difference and root mean square error (RMSE)
between NDVIp AVHRR and NDVI MODIS. Overall the RMSE between translated AVHRR
NDVI and MODIS NDVI was <0.03 and the mean differences in NDVIp AVHRR for all but one
Operational Performance
As of 8/10/2010, the VegDRI system was fully functional and producing weekly drought
products delivered to meet the schedule requirements of the USDM authors. USGS EROS has
transitioned the system to operational status under the direction of the EROS Long Term Archive
project as of 12/31/2010.
VegDRI is based on an empirical modeling strategy and, as such, is not a true turn-key system.
Model recalibration is advised on a yearly schedule to update the historical archive that is used to
calculate the averages and anomalies of the time-series satellite VI measures. This is especially
needed because of the fairly short historical period for the satellite data (now 21 years for the
AVHRR and 10 years for Terra MODIS). Model recalibration requires participation and support
from science staff at the NDMC and EROS. Risks include inadequate funding or losing staff
institutional knowledge. However, the current system can continue to operate without model
recalibration.
There are additional risks to input data streams incorporated into VegDRI. The weekly gridded
incomplete raster file or delivered later than the 00:01 a.m. Monday deadline due to source data
or software issues at the High Plains Regional Climate Center (which is co-located with the
National Drought Mitigation Center). The NDMC staff are dedicated to solving these problems
as they occur.
82
The eMODIS system at EROS is going through a transition to operations at the USGS EROS at
the time of this report. Efforts are ongoing to improve and fortify the eMODIS hardware and
software systems, improving redundancies and reducing risk. An important risk reduction
strategy is to have access to multiple satellite VI records for input into VegDRI. Both Terra
(morning) and Aqua (afternoon) composite data are being produced from the eMODIS system.
This project has not directly addressed a potential future transition of the VegDRI system after
the conclusion of the lifespans of the Terra and Aqua MODIS instruments. Given the success of
the index, we recommend a transition of VegDRI to ingest VIIRS observations, given the
similarities in spatial resolution and multi-spectral band widths of this instrument compared to
MODIS.
Accessibility/Distribution
The VegDRI is currently distributed as raster image data (geotiff format), graphical maps, and
Starting in mid-2010, EROS has also been staging weekly graphic maps (see Figure 1) and
sending notices out to the USDM authors each week via the Drought Exploder listserve.
Currently, around 290 local experts across 45 states are members of this listserve.
During the course of the project (2006-2010), various methods have been used to carry out
project outreach and communication. Project team members have led workshops, given multiple
NIDIS Workshop
Over thirty researchers, scientists, and natural resource managers representing a variety of
federal and non-federal agencies convened in at the National Integrated Drought Information
Project Presentations
1. Brown, J.F., VegDRI evaluation: Focus on Owyhee and Upper Colorado Basins,
2. Brown, J.F., Advantages of near-real time satellite data for operational drought
monitoring: the utility of MODIS and AVHRR, Annual Meeting of the Association of
3. Brown, J.F., EROS NASA drought project status and plans, Drought Forum, Austin TX,
4. Brown, J.F., Anderson, M., Wardlow, B.D., and Svoboda, M., Remote sensing
techniques for monitoring drought hazards: an intercomparison, 2009 AGU Fall Meeting,
5. Brown, J.F., Miura, T., Gu, Y., Jenkerson, C., and Wardlow, B.D., Utilizing a multi-
sensor satellite time series in real-time drought monitoring across the United States, 2009
Joint Assembly of the American Geophysical Union, Toronto, Ontario, Canada May 24-
6. Gu, Y., Brown, J.F. Phenological classification of the United States: A framework for
7. Brown, J.F., Wardlow, B.D., Tadesse, T., Callahan, K., and Pervez, M.S., Monitoring
recent drought effects on corn yields across the Corn Belt with the Vegetation Drought
Response Index, 2009 Annual Meeting of the Association of American Geographers, Las
8. Brown, J.F., Wardlow, B.D., Tadesse, T., and Gu, Y., Improving decision support for
drought using new geospatial models and online tools, ACES: A Conference for
9. Brown, J.F., Pervez, S., Wardlow, B., Tadesse, T. and Callahan, K., Assessment of 2006
and 2007 drought patterns in the vegetation drought response index across Nebraska,
10. Brown, J.F., and Hayes, M.J., National drought monitoring progress and future plans,
Geography staff briefing, USGS Headquarters, Reston, VA, Aug 17, 2008 [ORAL
PRESENTATION].
11. Brown, J.F., Wardlow, B., Gu, Y., Tadesse, T., Pervez, S., Hayes, M., and Verdin, J.,
2008, The value of satellite observations in drought and phenology research and
applications, Climate Change Workshop 2008, North Platte, NE, May 19-22, 2008
[POSTER].
12. Brown, J.F., Wylie, B., Homer, C., and Zhu, Z., Integrating remote sensing tools for
National inventory and monitoring, Briefing for M. Myers, USGS Director, Sioux Falls,
13. Brown, J.F., Verdin, J., Wardlow, B., and Tadesse, T., Remote sensing tools for
14. Gu, Y., Brown, J.F., Verdin, J.P., and Wardlow, B., A five-year analysis of MODIS
NDVI and NDWI for rangeland drought assessment, Global Vegetation Workshop,
Publications
1. Brown, J.F. 2010, Drought monitoring with VegDRI, U.S. Geological Survey Fact Sheet
2. Tadesse, T., Wardlow, B.D., Hayes, M.J., Svoboda, M.D., and Brown, J.F., 2010, The
3. Gu, Y., Brown, J.F., Miura, T., van Leeuwen, W., and Reed, B.C., 2010, Phenological
4. Gu, Y., Hunt E., Wardlow, B., Basara, J., Brown, J.F., and Verdin, J.P., 2008, Evaluation
of MODIS NDVI and NDWI for vegetation drought monitoring using Oklahoma
5. Brown, J.F., Wardlow, B.D., Tadesse, T., Hayes, M.J., and Reed, B.C., 2008, The
monitoring drought stress in vegetation: GIScience & Remote Sensing, 45 (1), p. 16-46.
6. Gu, Y., Brown, J.F., Verdin, J.P. and Wardlow, B., 2007, A five-year analysis of MODIS
NDVI and NDWI for grassland drought assessment over the central Great Plains of the
References
Asner, G.P., Townsend, A.R., & Braswell, B.H. (2000). Satellite observation of El Nino effects
on Amazon forest phenology and productivity. Geophysical Research Letters, 27, 981-984
Brown, J.F., Wardlow, B.D., Tadesse, T., Hayes, M.J., & Reed, B.C. (2008). The Vegetation
Drought Response Index (VegDRI): A new integrated approach for monitoring drought stress in
Eidenshink, J. (2006). A 16-year time series of 1 km AVHRR satellite data of the Conterminous
United States and Alaska. Photogrammetric Engineering & Remote Sensing, 72, 1027-1035
Goetz, S., Fiske, G., & Bunn, A. (2006). Using satellite time-series data sets to analyze fire
disturbance and forest recovery across Canada. Remote Sensing of Environment, 101, 352-365
Jenkerson, C.B., Maiersperger, T., Schmidt, G., 2010. eMODIS: A user-friendly data source:
Parker, M.D., Ratcliffe, I.C., & Henebry, G.M. (2005). The July 2003 Dakota hailswaths:
Creation, characteristics, and possible impacts. Monthly Weather Review, 133, 1241-1260
Peters, A.J., Griffin, S.C., Vina, A., & Ji, L. (2000). Use of remotely sensed data for assessing
crop hail damage. Photogrammetric Engineering and Remote Sensing, 66, 1349-1355
Peters, A.J., Walter-Shea, E.A., Ji, L., Via, A., Hayes, M., & Svoboda, M.D. (2002). Drought
Svoboda, M., LeComte, D., Hayes, M., Heim, R., Gleason, K., Angel, J., Rippey, B., Tinker, R.,
Palecki, M., Stooksbury, D., Miskus, D., & Stephens, S. (2002). The Drought Monitor. Bulletin
Wang, J., Rich, P.M., & Price, K.P. (2003). Temporal responses of NDVI to precipitation and
temperature in the central Great Plains, USA. International Journal of Remote Sensing, 24,
2345-2364
Wells, N., Goddard, S., & Hayes, M.J. (2004). A self-calibrating Palmer Drought Severity Index.
Appendix 1 (Section 2)
91
92
Monitoring
University of Nebraska-Lincoln
The National Drought Mitigation Center (NDMC) served as the lead integrator of both soil
moisture and Vegetation Drought Response Index (VegDRI) products generated from this
project within the current operational U.S. Drought Monitor (USDM) framework
(http://drought.unl.edu/dm). The NDMC was specifically tasked with ingesting these results
from our project partners into the USDM (and National Integrated Drought Information System
decision support. The NDMC also helped direct development and benchmarked the results and
value of the products developed as part of this project. Within the USDM framework, case
studies were conducted by soliciting and obtaining feedback from the USDM monitoring
community (~290 members), including the weekly USDM authors themselves. The USDM
author team consists of a total of 10 members from the National Drought Mitigation Center,
93
USDA, NOAA (National Climatic Data Center and Climate Prediction Center) and the Western
Regional Climate Center (also sponsored by NOAA). In addition, the NDMC developed and
currently hosts and maintains the projects website at http://drought.unl.edu/nasa. The following
sections summarize in more detail the specific tasks and deliverables from the NDMC that were
highlighted above.
In the initial stages of the project, the NDMC provided guidance to establish data and processing
requirements for the remote sensing products generated by the project participants for effective
integration into the operational USDM system. This guidance included defining specific product
requirements for the USDM system including product types, data format, update (frequency)
schedule, and a classification color scheme for the maps utilized within a GIS system. As
products developed using this initial set of requirements became available, the NDMC worked
with the USDM authors and USGS EROS, NASA JPL, and NOAA PSD project participants to
ensure that operational needs were met. This was an iterative process where adjustments and
modifications were made to various products and deliverables based on an interactive feedback
For the NASA JPL and NOAA PSD derivative products involving QuikSCAT soil moisture
change (SMC), information on the USDM ranking percentile criteria, cartographic color scheme
and operational USDM GIS methods, and the delta map and temporal production needs of the
USDM were provided to apply to the SMC product. In addition, feedback was provided on the
94
SMC maps on map legend development and the incorporation of USDM vector overlays on the
For the USGS EROS eMODIS VegDRI product, similar feedback and evaluation of VegDRI
maps were provided related to map color scheme, data format, updating schedule, and data
delivery mechanism. Product definition and formatting was prioritized early in the project in
order to ensure more effective integration of project data deliverables into the USDM system and
NIDIS portal. This product is currently ingested automatically by the NIDIS Portal Map Viewer
for the USDM author community on a weekly basis within the core suite of overlay products
and is currently hosted and maintained by the NDMC for project-related information. To date, it
has not been made open to the public at the request of the Principal Investigators while it
contains information related product under development and testing, but it can be made fully
(nasadrought@unl.edu) was also created by the NDMC to facilitate communication among all
project participants throughout the life of the project. The intent of the project web site was to
serve as central access point and repository for the project partners remote sensing-derived
products, general information, project reports and presentations that were developed for the
advanced USDM prototype system that was being supported. Once the data feeds were
95
established and automated from USGS EROS and NOAA PSD to NIDIS, demonstrations of the
advanced USDM system that included these new NASA products was conducted for the USDM
community and all project partners. In addition to the project website, the NDMC worked with
the NIDIS Portal team to post eMODIS-based VegDRI and SMC delta maps within the USDM
Authors Community map viewer to allow USDM authors to visualize the information and
assess the validity and potential use of these new products into the making of the USDM.
The eMODIS VegDRI continues to be fed in near real-time to the NIDIS Portal and the USDM
authors on a weekly basis while the SMC was stopped because of an interruption of the
QuikSCAT data stream due to QuikSCAT instrument problems. At the completion of the
project, the NDMC will continue to update, enhance and maintain the project website (and list
server if needed) for project communication and facilitate the continued integration of project
products, as well as other NASA Earth Science data, into the USDM after this project expires, as
long as the NDMC has the in-house resources to maintain this ongoing activity.
In support of eMODIS VegDRI development, the NDMC prepared and maintained a database of
data inputs for eMODIS VegDRI and generated the models that were operationally implemented
in this project. The NDMC also coordinated with USGS and the High Plains Regional Climate
Center (HPRCC) the automated delivery of current climate data (i.e., Standardized Precipitation
Index, SPI) on a customized time schedule (early Monday morning) to allow the data to be
96
acquired and ingested into the eMODIS system for near real-time, weekly VegDRI map
production. Beta eMODIS VegDRI maps/data were evaluated by USDM authors at the NDMC
and feedback was provided regarding their formatting, cartographic presentation, and ease of
use. This information was used to define the final formatting of the weekly operational eMODIS
VegDRI maps that began to be routinely ingested into the USDM system in 2010.
Product Evaluation
The NDMC engaged USDM authors and other drought experts within the USDM drought
community to assess the accuracy and utility of the remote sensing-derived products as they
became available. Initial assessments were completed for both the SMC and VegDRI products,
but only for an abbreviated period because operational products were only available towards the
Case studies were undertaken with JPL to evaluate how the SMC responded to drought
conditions and helped determine its potential for being utilized in the USDM system with several
conclusions reported earlier in Section 1. However, interruption of the SMC creation because of
sensor issues led us to rely only on a number of case studies instead of a near-real-time
operational growing season approach. A full investigation for the full capability of the SMC in
enhancing the USDM was thus impacted by the discontinuity of the QuikSCAT dataset. Future
data from the Oceansat-2 scatterometer (similar to QuikSCAT) will be useful to continue the
For eMODIS VegDRI, full assessment of the accuracy and utility of these products began in the
summer 2010 and continued through the Fall. The eMODIS maps were presented to USDM
authors for their review within the NIDIS Portal map viewer and USDM authors at the NDMC
provided more directed feedback on the performance and complimentary information that
VegDRI could provide to the USDM process. Given the short evaluation period and the
relatively drought-free conditions across most of the United States in 2010, a complete and
thorough review of VegDRI was not possible. However, the initial investigation showed that the
drought conditions/patterns depicted by VegDRI were relatively consistent with the USDM and
other indicators, which suggests that VegDRI represents a higher spatial resolution product that
can be used to improve the characterization of agricultural drought patterns within the USDM.
This is critical in those vast areas which lack observed climate data, particularly over rangeland.
VegDRI has begun to establish traction in being utilized by the USDM author in their weekly
suite of USDM overlay maps. An extended, post-project evaluation of near real-time VegDRI
maps for the 2011 growing season by NDMC, USGS EROS, and the USDM community is
planned to more fully evaluate the accuracy and contribution of the eMODIS VegDRI tool.
Coordination with USDM Authors to Implement Remote Sensing Data Products into an
Throughout the project, the NDMC continually engaged the USDM authors to communicate and
describe the information content of each project deliverable as they became available and
assisted them with the evaluation/integration of this information into the USDM development
process and within the USDM and NIDIS Portals. Feedback from USDM authors was collected,
98
organized, and communicated to other project partners by the NDMC to facilitate product
development and to promote the increased and continual application of the information. The
initial response, based on a limited amount of operational integration, was sufficiently favorable
to begin to include and test the eMODIS VegDRI product within their weekly suite of drought
monitoring products. Authors use the products with overlays of the USDM to aid them in the
placement of the D0-D4 line work on the map. This process has been automated and will
continue to operate after this project has ended, as long as resources can be leveraged by the
partners.
The NDMC has presented and highlighted project activities and the new tools and products
through numerous presentations and publications. The NDMC (Wardlow and Svoboda) worked
with other project participants to prepare book chapters for the VegDRI (lead author: Brian
Wardlow) and SMC (lead author: Son Nghiem) approaches for submission to a new book
entitled Remote Sensing and Drought: Innovative Monitoring Approaches by CRC Press (co-
editors: Brian Wardlow - NDMC, Jim Verdin USGS & NIDIS, and Martha Anderson USDA
ARS). The NDMC (Svoboda and Wardlow) is currently assisting NASA JPL lead, Son Nghiem,
in the writing of a peer-reviewed journal article on some of the project deliverables. Brian
Wardlow, of the NDMC, has also summarized the VegDRI work in a World Meteorological
Mark Svoboda and Brian Wardlow have presented on various aspects of the NDMCs role in this
project at the past three Drought Monitor Forums (both U.S. and North American venues), which
are held annually and bring together ~100 drought experts and media from Canada, Mexico, and
the United States. Potential continental/global applications of the VegDRI and SMC information
have also been discussed at these same U.S. and North American Drought Monitor (USDM and
NADM) Forums. Project deliverables and products have also been shared with the National
Integrated Drought Information System (NIDIS) Program Office (tasked with, among other
things, developing a national drought early warning system) and presented at each of the NIDIS
Pilot Drought Early Warning System (DEWS) basin meetings for the Upper Colorado River and
Project, scheduled to start in 2011. Internationally, the NDMC has presented project activities at
a WMO-sponsored workshop in West Africa and at various meetings involving the Global Earth
Observations (GEO) office and its potential usage within the Global Earth Observation System
of Systems (GEOSS).
Publications
Wardlow, B.D., T. Tadesse, J. Brown, K. Callahan, S. Swain, and E. Hunt, 2012. Integration of
Satellite, Climate, and Biophysical Data for Drought Monitoring of Vegetation. In B. Wardlow
et al. (Eds), Remote Sensing of Drought: Innovative Monitoring Approaches, CRC Press, Boca
Nghiem, S., B.D. Wardlow, D. Allured, M.D. Svoboda, D. LeComte, M. Rosencrans, S.K. Chan,
and G. Neumann, 2012. Microwave remote sensing of soil moisture Science and Applications.
Presentations
Wardlow, B.D., GIScience Activities at the National Drought Mitigation Center, USDA
National Agricultural Statistics Service (NASS), Fairfax, VA., August 12, 2010.
Wardlow, B.D., An Overview of Remote Sensing Activities at the National Drought Mitigation
Center, Hydrology and Remote Sensing Laboratory, USDA Agricultural Research Service (ARS),
Wardlow, B.D., The Future of Remote Sensing Applications for Drought Monitoring, World
Meteorological Organization (WMO) Drought Workshop, Lincoln, NE, December 10, 2009.
Wardlow, B.D., Vegetation Drought Response Index (VegDRI): 2009 Update and Ongoing
Wardlow, B.D., Remote Sensing and Drought An Overview and Opportunities for Mali, WMO
Wardlow, B.D., Remote Sensing and Drought Monitoring - New Tools and Future Directions,
Wardlow, B.D., VegDRI A New Hybrid Drought Index for Monitoring Vegetation in the U.S.,
U.S.-Canada GEO Bilateral Workshop on Ice and Water, National Science Foundation (NSF),
Wardlow, B.D., Vegetation Drought Response Index (VegDRI): A Hybrid-Based Approach for
Vegetation Drought Monitoring, North American Drought Monitor Workshop, Ottawa, Canada,
G. Robert Brakenridge
http://floodobservatory.colorado.edu/
Email: Robert.Brakenridge@Colorado.edu
The Dartmouth portion of this project was to develop MODIS surface water status products, and
use them as a metric for evaluating the system performance of the Drought Monitor. This
project is a transfer from its former location at Dartmouth College to University of Colorado and
the work plan is extended for completion of the remaining tasks in the current year. This section
contains the research progress to date. As noted in the original proposal, U.S. stream gaging
stations, which also measure surface water, provide a critical input to the present drought
geomorphology and geology, and antecedent conditions. The U.S. has been divided into 5348
103
hydrologic accounting units (watersheds) and the present array of gaging stations provide
information on the status of moisture deficits or surpluses for many of these units, and including
contributing units further upstream. However, many of these stations measure major rivers and
their linked tributary systems, and data from these are not independent measurements of the local
contributing areas until the imported water from upstream watersheds is accounted for. This is
not accomplished in the 7 day streamflow averages prepared by the USGS (used in the USDM)
and presented in the form of percentiles for period of record: instead, as one moves downstream
along all streams and rivers, the additional gaging station discharge percentile information
applies to the entire upstream contributing areas instead of the local incremental contributing
areas. Remote sensing offers, instead, the opportunity for discrete land parcel or catchment-unit
measurements of existing surface water status. As well, there exist hundreds of U.S. ponds and
reservoirs, including many small water bodies < 100 km2, which are ungauged (Figure 1)
(left) and southern Texas (right). Different shades of blue show MODIS hydrographic mapping
104
during normal years (light blue) and during intervals of moisture surpluses (darker blues).
Red crosses show all operating gaging stations: note that many reservoirs are not measured by
the in situ network, but can be by MODIS. Also shown are the hydrologic units, whose
Across much of the U.S, these water bodies are critical to local agriculture, livestock, recreation,
and other uses, and their areal extents are also observable indicators of drought status. Increases
in contributing watershed runoff to many reservoirs (depending on its mode of operation) are
accompanied by changes in reservoir water surface areas. If these areas can be consistently,
frequently, and economically monitored, through time, then such data can provide significant
enhancement and testing of map evaluations of regional drought conditions (such as the U.S.
Drought Monitor, or USDM). We proposed to obtain such measurements, and to facilitate their
During the first project year, we evaluated the capability of MODIS data to numerically estimate
the surface water status of selected U.S. hydrologic accounting units. We began as planned, by
testing change detection/measurement approaches (Figure 2). Band 2 on both operating MODIS
sensors very sensitively discriminates surface water changes at a spatial resolution of 250 m: the
challenge is to design a processing methodology that, working around the abundant noise
(changes in cloud cover and cloud shadows, terrain shadowing, seasonal changes in vegetation
and illumination, atmospheric changes), can produce a weekly MODIS-based product covering
all of the U.S. (excluding Alaska, Hawaii, and territories) and which can be directly compared to
the USDM).
105
Figure 2. MODIS band 2 near-IR geocoded 250 m image of southern Indiana (Ohio River is in
lower portion of scene) during dry conditions (left, September 18, 2004), compared to an
interval of moderate moisture surplus (right, change detection product using the dry MODIS
scene and one from July 16, 2003). Red shows increasing water as inferred by declining band 2
radiances between the two dry and wet scenes. Surface water variability at several small
During further work, we developed several other methods for surface water classification and
mapping. Change detection for water can proceed either by direct comparison of raster-based
comparison of mapped water. Mapping the water, in this case, makes use of remote sensing
image classification procedures to discriminate water, then polygons are fit to the classified
water areas, and then, within the GIS, map displays can show areas of expanded or decreased
water area (Figures 3 and 4). Partly with the support of this projects first year, we have
106
developed an efficient water classification algorithm using a widely available MODIS data
Figure 3. Sample of the planned MODIS-based Reduced Surface Water display, for one of the
twelve 10 deg. x 10 deg. map sheets covering the coterminous U.S. This display is here shown at
much reduced scale. Small areas of yellow are dry land, as imaged in March of 2009, compared
to that imaged by the same sensor using the same classification techniques in March of 2005,
107
when no drought conditions were occurring according to the USDM. Display thus shows at
maximum sensor resolution all areas of shrinkage of ponds, reservoirs, lakes, and rivers. Careful
choice of non-drought comparison dates is essential. These GIS data can be easily ingested into
an all-U.S. mosaic at a scale comparable to the USDM, or overlain onto the USDM map. Where
actual dimensions of water area shrinkage are too small to show at USDM scale, GIS
information. Planned work for year 2 will provide detailed, regional maps depicting which water
bodies are affected, while also providing a unified weekly all-US map showing the locations of
During the first year, the advantage of the GIS-based approach became apparent. It offers the
possibility to retain the relatively fine spatial resolution of the two MODIS 250 m bands, by
108
performing new data retrieval and water classification and mapping over discrete, 10 deg x 10
deg portions of the U.S. (Figure 3). It also allows for easy combination of such detailed, regional
displays into all-U.S. displays that can be directly compared to the USDM (to be accomplished
Note that in the original proposal, we also considered a gauging reach approach, whereby
changes in reach surface water area along rivers could be used as a predictor of river discharge
(when river discharge rises, mean river width and total reach water area increase), and a paired
time series of local water area changes, including small reservoir water surface areas. We
proposed to incorporate this approach within a semi-automated system that could ingest these
same NASA MODIS data over selected U.S. hydrologic accounting units, with the first year of
work planned to develop this system, the second to provide demonstrations of its use for
surface water data. However, NASA funding limitations de-scoped the effort to a two year
run, and with the input being limited to USDM validation rather than work toward actual
MODIS water area data input into the USDM operational product. We have thus focused on the
MODIS mapping approach, as per Figures 3 and 4, and because the remaining work then
requires only expansion of our trial approach to cover 12 US map sheets, and production of
weekly surface water change maps at the much-reduced spatial scale of the USDM (Figure 5).
These remain feasible within the time and budgetary constraints of the second and final project
year.
109
Figure 5. Left: location of the planned map sheets. Right: example of USDM output. Surface
water data from each sheet will be depicted at much-reduced resolution on a U.S. map at same
During the first project year, we also prepared a web site location to host these products (for
above. Joint publications have been listed in the first section of this report.
110
Participants in the 6th U.S. Drought Monitor Forum, Lower Colorado River Authority (Redbud
Center), Austin, TX October 7-8, 2009. Members of the NASA drought project team, and Brad