2005 Drought event in the Amazon River basin as measured by GRACE and estimated by climate models

Abstract

1] Satellite gravity measurements from the Gravity Recovery and Climate Experiment (GRACE) provide new quantitative measures of the 2005 extreme drought event in the Amazon river basin, regarded as the worst in over a century. GRACE measures a significant decrease in terrestrial water storage (TWS) in the central Amazon basin in the summer of 2005, relative to the average of the 5 other summer periods in the GRACE era. In contrast, data-assimilating climate and land surface models significantly underestimate the drought intensity. GRACE measurements are consistent with accumulated precipitation data from satellite remote sensing and are also supported by in situ water-level data from river gauge stations. This study demonstrates the unique potential of satellite gravity measurements in monitoring large-scale severe drought and flooding events and in evaluating advanced climate and land surface models., 2005 drought event in the Amazon River basin as measured by GRACE and estimated by climate models, J. Geophys. Res., 114, B05404, doi:10.1029/2008JB006056.

Full text

2005 drought event in the Amazon River basin as measured
by GRACE and estimated by climate models
J. L. Chen,
1
C. R. Wilson,
1,2
B. D. Tapley,
1
Z. L. Yang,
2
and G. Y. Niu
2
Received 1 September 2008; revised 8 January 2009; accepted 2 March 2009; published 8 May 2009.
[1]Satellite gravity measurements from the Gravity Recovery and Climate Experiment
(GRACE) provide new quantitative measures of the 2005 extreme drought event in the
Amazon river basin, regarded as the worst in over a century. GRACE measures a
significant decrease in terrestrial water storage (TWS) in the central Amazon basin in the
summer of 2005, relative to the average of the 5 other summer periods in the GRACE era.
In contrast, data-assimilating climate and land surface models significantly underestimate
the drought intensity. GRACE measurements are consistent with accumulated
precipitation data from satellite remote sensing and are also supported by in situ
water-level data from river gauge stations. This study demonstrates the unique potential of
satellite gravity measurements in monitoring large-scale severe drought and flooding
events and in evaluating advanced climate and land surface models.
Citation: Chen, J. L., C. R. Wilson, B. D. Tapley, Z. L. Yang, and G. Y. Niu (2009), 2005 drought event in the Amazon River basin
as measured by GRACE and estimated by climate models, J. Geophys. Res.,114 , B05404, doi:10.1029/2008JB006056.
1. Introduction
[2] In the summer of 2005, the Amazon basin experi-
enced an extreme drought. Many areas, especially in the
west and south, suffered the worst drought in over a century,
leading to official declarations of ‘‘public calamity’’, forest
fires, crop losses, and economic havoc [Rohter, 2005]. The
event appears connected both to the 2002–03 El Nin˜o and
to abnormal warming of the northern tropical Atlantic,
which was up to two degrees warmer than average [Zeng
et al., 2008a]. This paper compares measures of this event
taken from satellite gravity observations and from data-
assimilating hydrologic models.
[3] Understanding and quantification of drought occur-
rence, extent, and intensity is limited by conventional data
resources. Numerical climate models are valuable in ana-
lyzing and diagnosing climate variability, but quantifying
and simulating abnormal events such as droughts remains a
major modeling challenge. Prediction is an even greater
challenge. Conventional observations, especially in situ
meteorological and hydrological samples, are limited in
both space and time. Furthermore, model representations
of dynamical connections between boundary conditions and
extreme climate events tend to be poor.
[4] Terrestrial water storage (TWS) change, a major
component of the global water cycle, includes changes in
water stored in soil, as snow over land, and in ground water
reservoirs. TWS change reflects accumulated precipitation,
evapotranspiration, and surface and subsurface runoff with-
in a given area or basin. TWS change provides a good
measure of abnormal climate conditions such as drought,
and is valuable for agriculture and other water uses. How-
ever, TWS change is difficult to quantify because of limited
fundamental observations (ground water, soil moisture,
precipitation, evapotranspiration, snow water equivalent,
and others) at basin or smaller scales. Numerical model
estimates are useful but exhibit limited accuracy [e.g.,
Matsuyama et al., 1995]. Remote sensing data (such as
TRMM satellite precipitation data) and in situ measure-
ments (such as river level and discharge from gauge
stations) are valuable assets in estimating TWS changes
[e.g., Crowley et al., 2007]. Unfortunately, in situ measure-
ments alone are not sufficient, both because they tend to be
point measurements, and because other hydrological param-
eters (e.g., evapotranspiration) must be estimated separately
to determine TWS change.
[5] The Gravity Recovery and Climate Experiment
(GRACE) is the first dedicated satellite gravity mission,
jointly sponsored by NASA and the German Aerospace
Center (DLR). Launched in March 2002, GRACE has been
measuring Earth gravity change with unprecedented accu-
racy [Tapley et al., 2004] for over 6 years. Early GRACE
time-variable gravity observations showed an accuracy of
1.5 cm of equivalent water thickness change at about
1000-km spatial scale [Wahr et al., 2004]. Various studies
applied early GRACE results to a variety of problems
including TWS change [e.g., Wahr et al., 2004], polar ice
sheets mass balance [e.g., Velicogna and Wahr, 2006; Chen
et al., 2006], and oceanic mass change [e.g., Chambers et
al., 2004; Lombard et al., 2007].
[6] In early 2007, reprocessed GRACE release-04 (RL04)
time-variable gravity fields with improved background
geophysical models and data processing techniques were
released [Bettadpur, 2007a]. RL04 shows significantly
JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 114, B05404, doi:10.1029/2008JB006056, 2009
Click
Here
for
Full
A
rticl
e
1
Center for Space Research, University of Texas at Austin, Austin,
Texas, USA.
2
Department of Geological Sciences, University of Texas at Austin,
Austin, Texas, USA.
Copyright 2009 by the American Geophysical Union.
0148-0227/09/2008JB006056$09.00
B05404 1of9

improved data quality and spatial resolution, now better
than 500 km [e.g., Chen et al., 2007], enabling the study of
a much wider class of problems than before. Improved
quality and spatial resolution of RL04 and optimized data
processing and filtering techniques [e.g., Swenson and
Wahr, 2006] provide new opportunities for quantification
of TWS changes, better monitoring of the global water
cycle, and related understanding of droughts and floods.
[7] In this study, we examine TWS change in the Amazon
basin using RL04 time-variable gravity fields and predic-
tions from major climate and land surface models, including
the National Oceanic and Atmospheric Administration
(NOAA) National Centers for Environmental Prediction
(NCEP) reanalysis II climate model and the global land
data assimilation system (GLDAS) [Rodell et al., 2004].
The goal is to demonstrate the capability of GRACE to
Figure 1. GRACE-averaged August and September water storage changes (in centimeters of water) in
South America in (a) 2002, (b) 2003, (c) 2004, (d) 2005, (e) 2006, and (f) 2007. A 2-step filtering scheme
(P4M6 and 500-km Gaussian smoothing) is applied, as described in the text.
B05404 CHEN ET AL.: 2005 AMAZON DROUGHT FROM GRACE AND MODEL
2of9
B05404

observe and quantify the severe 2005 Amazon drought
event, and to compare GRACE results with climate model
descriptions of the same event.
2. Data Processing
2.1. TWS Changes From GRACE RL04
[8] Our GRACE time series includes 65 approximately
monthly gravity solutions, for the period April 2002 to
December 2007. Spherical harmonic (SH) coefficients (to
degree and order 60) are used to compute monthly mass
change fields on a 1°1°grid. Swenson and Wahr [2006]
showed that GRACE longitudinal stripe noise is associated
with correlations among even or odd degree SH coefficients
at a given order. Here we apply a two-step filter as in
previous studies [Chen et al., 2007]. The first step (called
P4M6) removes this correlated noise by fitting and sub-
tracting a fourth-order polynomial to even and odd coeffi-
cient pairs at SH orders 6 and above. The second step
involves smoothing with a 500-km Gaussian filter [Jekeli,
1981]. After filtering, the mean of the 65-point time series
at each grid point is removed to obtain a time series of
gravity field variations, expressed as equivalent surface mass
variations in cm of water.
[9] Atmospheric and oceanic mass changes have been
largely removed from RL04 using numerical model predic-
tions [Bettadpur, 2007b], so that variations over time scales
of months to years should reflect primarily unmodeled
effects such as TWS change, snow/ice mass changes
(including polar ice sheets and mountain glaciers), plus
other geophysical signals such as postglacial rebound
(PGR) and coseismic and postseismic deformation. Over
the Amazon basin, surface mass variations are expected to
be dominantly due to near-surface water storage changes.
Given this, the major errors in GRACE time series over the
Amazon arise from spatial leakage associated with a finite
range of SH terms, attenuation due to filtering, residual
atmospheric signals, and GRACE measurement errors.
2.2. TWS Changes From Climate and Land Surface
Models
2.2.1. NCEP Reanalysis II Model Estimates
[10] (NCEP/DOE AMIP) II Reanalysis-II was developed
at NCEP and the Department of Energy (DOE) from the
widely used NCEP/NCAR Reanalysis [Kalnay et al., 1996].
NCEP reanalysis II (simply called NCEP in the subsequent
discussion) improves upon earlier results by correcting
errors and refining parameterizations of physical processes.
Soil moisture (volume percentage) and snow fields (cm of
water equivalent) are monthly averages from January 1948
to present, on a Gaussian grid (1.904°latitude by
1.875°longitude). Soil moisture is modeled for an upper
layer of 10 cm and a lower layer of 190 cm thickness.
Because NCEP does not model deeper ground water stor-
age, TWS change at each grid point is the sum of two soil
water layers plus snow water.
2.2.2. GLDAS Model Estimates
[11] GLDAS was developed jointly at the National
Aeronautics and Space Administration Goddard Space
Flight Center and NOAA NCEP [Rodell et al., 2004].
GLDAS parameterizes, forces, and constrains sophisticated
land surface models with ground and satellite products with
the goal of estimating land surface states (e.g., soil mois-
ture and temperature) and fluxes (e.g., evapotranspiration).
In this particular simulation, GLDAS drove the Noah land
surface model [Ek et al., 2003], with observed precipitation
Figure 2. Differences of mean August/September TWS changes (in centimeters of water) in South
America in 2005 relative to mean August/September TWS changes of other years in the period 2002 –
2007: (a) from GRACE, (b) from NCEP, and (c) from GLDAS. GRACE displays a strong 2005 TWS
deficit, while NCEP and GLDAS do not.
B05404 CHEN ET AL.: 2005 AMAZON DROUGHT FROM GRACE AND MODEL
3of9
B05404

and solar radiation included as inputs. Monthly averaged
soil moisture (2 m column depth) and snow water equiv-
alent are available from 1979 to present, and TWS varia-
tion at each grid point is the sum of soil and snow water.
Greenland and Antarctica are excluded because the model
omits ice sheet physics. Ground water is also not modeled
by GLDAS.
2.2.3. TWS Changes From NCEP and GLDAS
[12] A fair comparison with GRACE observations
requires that NCEP and GLDAS fields be spatially filtered
in a similar way. To accomplish this, GLDAS and NCEP
TWS gridded fields were represented in a SH expansion to
degree and order 100, and the same two step filter described
earlier (to remove correlated noise and smooth RL04) was
applied. SH coefficients were truncated above degree and
order 60, and additionally SH coefficients for degree 0
(imposing global mass conservation) and degree 1 (remov-
ing geocenter motion) were set to zero. Finally, GLDAS and
NCEP SH representations were evaluated on a global 1°
1°grid.
3. Results
3.1. GRACE and Climate Models Estimates
[13] Average August and September [(August + Septem-
ber)/2] TWS changes are used to measure the 2005 Amazon
drought. The drought is at its maximum during these two
months, and GRACE observations are available for August
and September starting in 2002. Figure 1 shows GRACE
(August/September) TWS change relative to the mean in
South America for 2002 through 2007. In 2005 (Figure 1d),
a minimum in TWS is evident (in the region circled by a
gray curve). The Orinoco basin (to the north of the Amazon)
shows a significantly wetter 2007 (Figure 1f).
[14] Figure 2a displays the difference between the 2005
(August/September) average and the mean (August/September)
Figure 3. (a) Comparison of TWS changes in the central Amazon basin (average within the red box
[2°S–7°S,294°E –299°E] marked in Figure 2a) from GRACE (blue curve), NCEP (red curve), and
GLDAS (green curve). (b) Comparison of nonseasonal TWS changes in central Amazon basin from
GRACE (blue curve), NCEP (red curve), and GLDAS (green curve).
B05404 CHEN ET AL.: 2005 AMAZON DROUGHT FROM GRACE AND MODEL
4of9
B05404

determined from the other 5 years. This indicates a
GRACE-observed deficiency around 8 –9 cm basin wide
for 2005, or approximately 515 km
3
of water (integrated
over the entire Amazon basin). Figures 2b and 2c show
2005 (August/September) anomalies relative to the other 5
years from NCEP and GLDAS, respectively. NCEP anoma-
lies for 2005 are near zero, while those from GLDAS are
larger, though smaller than the tens of cm seen in Figure 2a.
[15] We examine in more detail a 5°5°region (Figure 2a,
red box) near the center of the 2005 drought region evident in
Figure 2a. Average TWStime series (average equivalent water
height) for this entire region are calculated from GRACE,
NCEP and GLDAS, and compared in Figure 3a. GRACE
shows a strong seasonal signal with peak-to-peak amplitudes
up to 70 cm, with additional nonseasonal variability. Near
the end of 2005, GRACE TWS diminishes (relative to other
years), indicating drought. Both NCEP and GLDAS show
strong seasonal variation, but with magnitudes about one
half of GRACE. Both show significantly less nonseasonal
variability.
[16] Annual and semiannual sinusoids were fit by un-
weighted least squares and removed from each of the three
time series, with results shown in Figure 3b. Additionally,
we removed from the GRACE series a 161-day S2 tide alias
[Knudsen, 2003] determined by least squares. The S2 alias
amplitude (in this GRACE time series) is 1 cm, as com-
pared to 25 cm for the annual component. In Figure 3b,
GRACE shows a steady TWS decrease of about 14 cm,
starting in March/April 2005 through August/September,
marking a clear anomaly. NCEP and GLDAS series each
show some TWS decrease in the same period, but the
change does not appear unusual in the context of variations
over the full 6 year time series.
[17] Estimates of GRACE noise (error bars in Figure 3b) are
determined from root-mean-square (RMS) variability over
tropical oceans (20°S–20°N), a region in which true mass
variability (GRACE signal) is probably near zero, as bara-
tropic ocean mass changes have been removed in GRACE
dealiasing process [Bettadpur, 2007b]. Therefore residuals
over the ocean could approximately represent residual errors
of GRACE data (plus some unmodeled baroclinic ocean
signals) [Wa hr e t a l., 2004]. Another estimate of GRACE
error [Wahr e t a l., 2006] is based on RMS residual of GRACE
variability over land after subtracting seasonal sinusoids and
smoothing, though this overestimates GRACE errors when
there is significant nonseasonal (nonsinusoidal) variability, as
evident in Figure 3b. However, after removing longer-period
interannual variability from GRACE data, the two error bar
estimates are similar (2.4 cm from ocean residuals versus
2.9 cm from land residuals).
3.2. Other Observational Evidences
[18] To validate GRACE-observed significant TWS de-
crease in during the 2005 Amazon drought, we analyze
precipitation data from the Global Precipitation Climatology
Project (GPCP) [Adler et al., 2003], for the GRACE period.
Figure 4. (a) Accumulated precipitation in the central Amazon region (marked by the red box in Figure 2a)
during June to September based on GPCP data. (b) Accumulated precipitation in the central Amazon
region during December to March based on GPCP data.
B05404 CHEN ET AL.: 2005 AMAZON DROUGHT FROM GRACE AND MODEL
5of9
B05404

We compute accumulated precipitation totals in the central
Amazon region (Figure 2a, red box) in the 4 summer
months (i.e., the sum of the 4 months’ totals in June through
September) for each of the 6 years from 2002 to 2007, and
show the results in Figure 4a. The selection of summing
June through September is based on the consideration that
GRACE data show that the 2005 Amazon drought reached
the peak (with minimum TWS) in August/September, and
TWS change could reflect accumulated precipitation
changes of a few months prior to the drought. The use of
4 months summation is arbitrary. We test the calculation in
different cases (including using 3 and 6 months) and all
show the least amount of accumulated precipitation in 2005,
while the 4 months summation appears to show the most
distinctive drop of accumulated summer precipitation in
2005. Similar 4-monthly totals for the same region, but for
December to March are presented in Figure 4b. Clearly, in
the summer 2005, the central Amazon region recorded
significantly less precipitation (up to 14 cm less) than
any other year (during the period 2002 to 2008), consistent
with GRACE observations, as well as reports of the 2005
drought severity cited earlier.
[19] GRACE time series (Figures 3a and 3b) indicate the
2005 drought had ended by the end of 2005 (or beginning
of 2006), and the central Amazon region actually experi-
enced a wetter (as compared to other years) winter season in
early 2006. This is also consistent with GPCP precipitation
data (Figure 4b), as the December–March precipitation total
(in the central Amazon region) in 2005 is significantly less
(up to 30 cm) than other years.
[20] To further verify GRACE measurements, we exam-
ine daily water-level data of 4 selected river gauges in the
Amazon basin (see Figure 5). Among the 4 river gauge
stations, Itapeua and Jatuarana are located close to the
center of the TWS decrease observed by GRACE, and
Parintins and Obidos are spread on the down stream side.
To better focus on nonperiodic variations, we remove
annual and semiannual variations from the 4 water-level
time series (at the 4 gauges marked in Figure 5) using
unweighted least squares fit, and show the nonseasonal
water-level time series in Figure 6.
[21] Each of the 4 gauges show a clear drop of water level
in the summer 2005, bottomed in around August and
September, while Itapeua shows the largest water level
decrease of up to 4.8 m (with respect to the 6-year
temporal mean). In situ water-level data at these 4 river
gauge stations provide additional verification of the signif-
icant central Amazon TWS decrease in 2005 as observed by
Figure 5. Location map of 4 river gauge stations (Itapeua,
Jatuarana, Parintins, and Obidos) in the Amazon basin,
superimposed by GRACE-observed TWS decrease in
August/September 2005 (same as in Figure 2a).
Figure 6. Nonseasonal daily water-level change at 4 selected river gauge stations marked in Figure 5.
Annual and semiannual variations have been removed from these time series using unweighted least
squares fit.
B05404 CHEN ET AL.: 2005 AMAZON DROUGHT FROM GRACE AND MODEL
6of9
B05404

GRACE, although quantitative comparisons (between
GRACE and in situ river gauge data) are difficult because
of different representations of the two different quantities.
We see a gradual decease in magnitudes of the water-level
drops among the 4 gauges, from 4.8 m at Itapeua, 4.4 m
at Jatuarana, 3 m at Parintins, to 2 m at Obidos (Figure 6).
This is consistent with the spatial TWS variation feature
observed by GRACE (Figure 5). In situ water-level data
also indicate that the 2005 Amazon drought had ended by
the end of 2005 and by February or March 2006, the central
Amazon region was actually wetter than normal (see Figure 6),
also consistent with GRACE observations and GPCP pre-
cipitation data.
4. Discussion and Conclusions
[22] GRACE RL04 time series clearly indicate a signif-
icant TWS deficit accompanying the 2005 Amazon
drought, on the order of 14 cm of water equivalent in
the central Amazon. GRACE time series indicate the
drought peaked in the period August to September 2005,
and was relieved by the beginning of 2006, consistent with
independent precipitation observations and in situ water-
level measurements.
[23] NCEP and GLDAS land surface models significantly
underestimate TWS change in the central Amazon relative
to GRACE. These two models show only half the seasonal
variability of GRACE observations, and both lack signifi-
cantly diminished TWS associated with the 2005 drought.
Unfortunately, there are no in situ TWS measurements to
directly validate GRACE estimates. GPCP precipitation
data are helpful for understanding TWS changes, but are
not directly comparable to GRACE observations because
they do not account for other elements affecting TWS.
GPCP data do show diminished precipitation during the
summer of 2005, consistent with GRACE time series.
Figure 7. (a) Comparison of GLDAS and LadWorld estimated TWS changes in the Amazon basin
(averaged over the entire basin). A 500-km Gaussian smoothing is applied to both data sets. (b) Comparison
of GLDAS and LadWorld estimated TWS changes in the central Amazon area (average within the red box
marked in Figure 2a).
B05404 CHEN ET AL.: 2005 AMAZON DROUGHT FROM GRACE AND MODEL
7of9
B05404

Considering these limitations, this study demonstrates the
unique strength of GRACE observations in monitoring
droughts and floods via associated large spatial scale
TWS changes. Additionally, GRACE observations provide
independent measures of TWS for calibrating, evaluating,
and improving climate and land surface models.
[24] Both the GRACE estimate of the 2005 TWS deficit
(Figure 3b) and the discrepancy between GRACE and
model estimates of seasonal variability (Figure 3a) greatly
exceed estimated errors. Spatial leakage errors (due to
filtering and limited SH range) are not likely causes of
differences, given that similar filtering has been applied to
all data sets. Lack of a ground water component in NCEP
and GLDAS may partially account for differences between
GRACE and model estimates, but is probably not a major
cause [Niu et al., 2007].
[25] To further prove this speculation, we show in Figure 7
the comparison of GLDAS TWS estimates and similar
estimates from the LadWorld land surface model, which
includes the groundwater component [Milly and Shmakin,
2002]. LadWorld TWS data (from the Fraser version,
see information at http://www.gfdl.noaa.gov/pcm/project/
ladworld.htm for details) represent the sum of soil (of the
top 6 m), water equivalent snow, and groundwater, and
follow the same data processing (e.g., smoothing and
truncation) as used in GLDAS data. For Amazon basin-
wide average, GLDAS and LadWorld show very similar
TWS estimates, even though groundwater is modeled in
LadWorld. However, in the central Amazon area (marked in
Figure 2a), LadWorld indeed show notably larger TWS
variability than GLDAS (and the differences are consider-
ably larger when no smoothing is applied). The increased
TWS variability of LadWorld estimates is still significantly
smaller than GRACE observations (at both seasonal and
nonseasonal time scales), suggesting that groundwater is
important but may not be the major contributor to models’
underestimation of TWS changes in Amazon.
[26] A recent study [Zeng et al., 2008b] has compared
GRACE estimated TWS change in the Amazon basin with a
few models’ estimates and some other estimates based on
water conservation equation using modeled moisture con-
vergence minus observed runoff (MCR) or using observed
precipitation minus observed runoff and modeled evapo-
transpiration (PER), and indicates that the PER estimate
shows significantly larger seasonal variability (in the
Amazon basin) than both models and MCR estimates and
agrees well with GRACE data. This appears to indicate that,
given the parameterizations of current land surface models,
the more ‘traditional’ PER method can better depict TWS
changes, at least in the Amazon basin.
[27] Model and GRACE estimates are more similar in
other major basins than in the Amazon (e.g., the La Plata
to the south), although model TWS changes are consis-
tently smaller than GRACE estimates in the La Plata as
well. Because the Amazon is the largest river basin in the
world, estimated TWS is likely to be least affected by
GRACE’s spatial resolution limitations. On the other hand,
neither model appears to do a good job of describing TWS
variability in this basin. One reason may be that most
macrohydrological models (including NCEP, GLDAS, and
LadWorld) do not consider horizontal transport of water,
implying instantaneous runoff. In very large basins, such
as the Amazon, precipitation may actually be retained for
some time (in lakes, wet lands, and shallow reservoirs),
leading to underestimation of TWS change. This suggests
that proper modeling of the terrestrial water cycle within
the Amazon will likely require additional consideration of
complexity of its river systems, vegetation, soils, and
floodplains.
[28]Acknowledgments. The authors are grateful to Anny Cazenave
and an anonymous reviewer for their insightful comments, which led to
improved presentation of the results. This research was supported by
NASA’s Solid Earth and Natural Hazards and GRACE Science Program
(NNG04GF10G, NNG04G060G, NNG04GF22G, and NNG04GP70). The
authors would like to thank Flavio Vaz for providing the river gauges water-
level data.
References
Adler, R. F., et al. (2003), The version 2 Global Precipitation Climatology
Project (GPCP) monthly precipitation analysis (1979 – present), J. Hydro-
meteorol.,4, 1147 – 1167.
Bettadpur, S. (2007a), CSR Level-2 Processing Standards Document for
Product Release 04, GRACE 327-742, The GRACE Project, Cent. for
Space Res., Univ. of Tex. at Austin, Austin, Tex.
Bettadpur, S. (2007b), Level-2 Gravity Field Product User Handbook,
GRACE 327-734, The GRACE Project, Cent. for Space Res., Univ. of
Tex. at Austin, Austin, Tex.
Chambers, D. P., J. Wahr, and R. S. Nerem (2004), Preliminary observa-
tions of global ocean mass variations with GRACE, Geophys. Res. Lett.,
31, L13310, doi:10.1029/2004GL020461.
Chen, J. L., C. R. Wilson, and B. D. Tapley (2006), Satellite gravity
measurements confirm accelerated melting of Greenland ice sheet,
Science,313(5795), 1958 – 1960, doi:10.1126/science.1129007.
Chen, J. L., C. R. Wilson, B. D. Tapley, and S. Grand (2007), GRACE
detects coseismic and postseismic deformation from the Sumatra-
Andaman earthquake, Geophys. Res. Lett.,34(13), L13302, doi:10.1029/
2007GL030356.
Crowley, J. W., J. X. Mitrovica, R. C. Bailey, M. E. Tamisiea, and J. L.
Davis (2007), Annual variations in water storage and precipitation in
the Amazon Basin, J. Geodesy,82(1), 9 – 13, doi:10.1007/s00190-007-
0153-1.
Ek, M. B., K. E. Mitchell, Y. Lin, E. Rogers, P. Grunmann, V. Koren,
G. Gayno, and J. D. Tarpley (2003), Implementation of the upgraded
Noah land-surface model in the NCEP operational mesoscale Eta model,
J. Geophys. Res.,108(D22), 8851, doi:10.1029/2002JD003296.
Jekeli, C. (1981), Alternative Methods to Smooth the Earth’s Gravity Field,
Dep. of Geod. Sci. and Surv., Ohio State Univ., Columbus, Ohio.
Kalnay, E., et al. (1996), The NCEP/NCAR 40-year reanalysis project, Bull.
Am. Meteorol. Soc.,77, 437 – 471.
Knudsen, P. (2003), Ocean tides in GRACE monthly averaged gravity
fields, Space Sci. Rev.,108, 261 – 270.
Lombard, A., D. Garcia, G. Ramillien, A. Cazenave, R. Biancale, J. M.
Lemoine, F. Flechtner, R. Schmidt, and M. Ishii (2007), Estimation of
steric sea level variations from combined GRACE and Jason-1 data,
Earth Planet. Sci. Lett.,254(1 – 2), 194 – 202.
Matsuyama, H., T. Oki, and K. Masuda (1995), Applicability of ECMWF’s
4DDA data to interannual variability of the water budget of the Missis-
sippi River Basin, J. Meteorol. Soc. Jpn.,73, 1167 – 1174.
Milly, P. C. D., and A. B. Shmakin (2002), Global modeling of land water
and energy balances. part I: The Land Dynamics (LaD) model, J. Hydro-
meteorol.,3, 283 – 299.
Niu, G.-Y., Z.-L. Yang, R. E. Dickinson, L. E. Gulden, and H. Su (2007),
Development of a simple groundwater model for use in climate models
and evaluation with Gravity Recovery and Climate Experiment data,
J. Geophys. Res.,112 , D07103, doi:10.1029/2006JD007522.
Rodell, M., et al. (2004), The global land data assimilation system, Bull.
Am. Meteorol. Soc.,85(3), 381 – 394.
Rohter, L. (2005), Record drought cripples life along the Amazon, The New
York Times, December 11. (Available at http://www.nytimes.com/2005/
12/11/international/americas/11amazon.html)
Swenson, S., and J. Wahr (2006), Post-processing removal of correlated
errors in GRACE data, Geophys. Res. Lett.,33, L08402, doi:10.1029/
2005GL025285.
Tapley, B. D., S. Bettadpur, M. M. Watkins, and C. Reigber (2004), The
Gravity Recovery and Climate Experiment; mission overview and
early results, Geophys. Res. Lett.,31(9), L09607, doi:10.1029/
2004GL019920.
B05404 CHEN ET AL.: 2005 AMAZON DROUGHT FROM GRACE AND MODEL
8of9
B05404

Velicogna, I., and J. Wahr (2006), Measurements of time-variable gravity
show mass loss in Antarctica, Science,311, 1754 – 1756, doi:10.1126/
science.1123785.
Wahr, J., S. Swenson, V. Zlotnicki, and I. Velicogna (2004), Time-variable
gravity from GRACE: First results, Geophys. Res. Lett.,31, L11501,
doi:10.1029/2004GL019779.
Wahr, J., S. Swenson, and I. Velicogna (2006), Accuracy of GRACE mass
estimates, Geophys. Res. Lett.,33, L06401, doi:10.1029/2005GL025305.
Zeng, N., J. H. Yoon, J. A. Marengo, A. Subramaniam, C. A. Nobre,
A. Mariotti, and J. D. Neelin (2008a), Causes and impacts of the 2005
Amazon drought, Environ. Res. Lett.,3, 014002, 9 pp., doi:10.1088/1748-
9326/3/1/014002.
Zeng, N., J. H. Yoon, and A. Mariotti (2008b), Variability of basin-scale
terrestrial water storage from a P-E-R water budget method: The Amazon
and the Mississippi, J. Clim.,15, 248 – 265.

J. L. Chen, B. D. Tapley, and C. R. Wilson, Center for Space Research,
University of Texas at Austin, 3925 W. Braker Lane, Suite 200, Austin, TX
78759-5321, USA. (chen@csr.utexas.edu)
G. Y. Niu and Z. L. Yang, Department of Geological Sciences, University
of Texas at Austin, 1 University Station C1100, Austin, TX 78712-0254,
USA.
B05404 CHEN ET AL.: 2005 AMAZON DROUGHT FROM GRACE AND MODEL
9of9
B05404