Download Closing the Global Water Vapor Budget with AIRS

15 DECEMBER 2011
WONG ET AL.
6307
Closing the Global Water Vapor Budget with AIRS Water Vapor, MERRA Reanalysis,
TRMM and GPCP Precipitation, and GSSTF Surface Evaporation
SUN WONG, ERIC J. FETZER, BRIAN H. KAHN, BAIJUN TIAN, AND BJORN H. LAMBRIGTSEN
Jet Propulsion Laboratory, California Institute of Technology, Pasadena, California
HENGCHUN YE
Department of Geography and Urban Analysis, California State University, Los Angeles, California
(Manuscript received 20 October 2010, in final form 28 May 2011)
ABSTRACT
The authors investigate if atmospheric water vapor from remote sensing retrievals obtained from the Atmospheric Infrared Sounder/Advanced Microwave Sounding Unit (AIRS) and the water vapor budget from
the NASA Goddard Space Flight Center (GSFC) Modern Era Retrospective-analysis for Research and
Applications (MERRA) are physically consistent with independently synthesized precipitation data from the
Tropical Rainfall Measuring Mission (TRMM) or the Global Precipitation Climatology Project (GPCP) and
evaporation data from the Goddard Satellite-based Surface Turbulent Fluxes (GSSTF). The atmospheric total
water vapor sink (S) is estimated from AIRS water vapor retrievals with MERRA winds (AIRS–MERRA S) as
well as directly from the MERRA water vapor budget (MERRA–MERRA S). The global geographical distributions as well as the regional wavelet amplitude spectra of S are then compared with those of TRMM or
GPCP precipitation minus GSSTF surface evaporation (TRMM–GSSTF and GPCP–GSSTF P 2 E, respectively). The AIRS–MERRA and MERRA–MERRA Ss reproduce the main large-scale patterns of global
P 2 E, including the locations and variations of the ITCZ, summertime monsoons, and midlatitude storm tracks
in both hemispheres. The spectra of regional temporal variations in S are generally consistent with those of
observed P 2 E, including the annual and semiannual cycles, and intraseasonal variations. Both AIRS–
MERRA and MERRA–MERRA Ss have smaller amplitudes for the intraseasonal variations over the tropical
oceans. The MERRA P 2 E has spectra similar to that of MERRA–MERRA S in most of the regions except in
tropical Africa. The averaged TRMM–GSSTF and GPCP–GSSTF P 2 E over the ocean are more negative
compared to the AIRS–MERRA, MERRA–MERRA Ss, and MERRA P 2 E.
1. Introduction
Precipitation has a direct impact on society. Changes
of precipitation on global and regional scales have long
been studied and linked to the changing climate (e.g.,
Allen and Ingram 2002; Dai et al. 1997; Sun et al. 2007;
Trenberth et al. 2003). It is important to understand how
the global hydrological cycle is changing with increasing
atmospheric greenhouse gas loading (Schneider et al.
2010; Stephens and Ellis 2008; Trenberth et al. 2003) and
its potential impact on society (Clarke and King 2004;
Corresponding author address: Sun Wong, Jet Propulsion Laboratory, California Institute of Technology, 4800 Oak Grove Dr.,
Pasadena, CA 91109.
E-mail: [email protected]
DOI: 10.1175/2011JCLI4154.1
Ó 2011 American Meteorological Society
Watkins et al. 2007). Such tasks require a long temporal
record of data with global coverage. Satellite-based
observations of various components of the global hydrological cycle (e.g., Stephens et al. 2002) and assimilation of observations for dynamically consistent model
data with global coverage are thus indispensible (e.g.,
Bosilovich et al. 2008; Saha et al. 2010).
Remote sensing instruments on board present-day satellite platforms have measured various components of the
earth’s hydrological cycle. Retrieval and analysis products
are now available for studies of the global hydrological
cycle. Among these products are the atmospheric specific
humidity (q) measured by the Atmospheric Infrared
Sounder/Advanced Microwave Sounding Unit (AIRS;
Divakarla et al. 2006; Fetzer et al. 2004, 2006; Susskind
et al. 2006) on board the Aqua satellite (launched on
6308
JOURNAL OF CLIMATE
VOLUME 24
FIG. 1. Geographical distributions of the (a),(d) TRMM 3B42; (b),(e) GPCP; and (c),(f) MERRA precipitation rates
(mm day21). Averaged precipitation rates for (left) DJF of 2004–08 and (right) JJA of 2004–08.
4 May 2002), the high-frequency precipitation (P) syntheses such as the 3B42 algorithm of the Tropical Rainfall Measuring Mission (TRMM; Huffman et al. 2007)
and the Global Precipitation Climatology Project (GPCP;
Huffman et al. 2009), estimates of surface evaporation
fluxes (E) from the Goddard Satellite-based Surface
Turbulent Fluxes (GSSTF; Chou et al. 2003; Shie et al.
2009), and the reanalysis q, P, and E products such as
those from the National Aeronautics and Space Administration (NASA) Goddard Space Flight Center
(GSFC) Modern Era Retrospective-analysis for Research and Applications (MERRA; Bosilovich et al.
2008). No investigation to date has quantified whether
these datasets are physically consistent with each other
with regard to describing the variability of global hydrological processes.
The net surface water exchange is quantified as precipitation minus evaporation (hereafter referred to as
P 2 E), and is linked to q by the following budget equations (e.g., Peixoto and Oort 1992, chapter 12; Trenberth
and Guillemot 1998):
›q
›q
›q
›q
1u 1y
1w
,
(1)
S(x, y, p, t) 5 2
›t
›x
›y
›p
15 DECEMBER 2011
WONG ET AL.
6309
FIG. 2. Geographical distributions of (a),(c) GSSTF2b and (b),(d) MERRA surface evaporation rates (mm day21).
Averages for (left) DJF of 2004–08 and (right) JJA of 2004–08. Missing data in GSSTF2b are indicated as black areas.
(x, y, t) 5
ðp
s
pt
S(x, y, p, t)
dp
5 P(x, y, t) 2 E(x, y, t).
g
(2)
Here (u, y, v) are the horizontal and vertical wind velocities, respectively; and S is referred to as the apparent
water vapor sink in the literature (Schumacher et al. 2008;
Shige et al. 2008; Yanai et al. 1973) and differs from the
conventional defined Q2 in the literature by a factor equal
to the water latent heat of evaporation (L). The integrated
S from the top of the atmosphere ( pt) to the surface pressure ( ps) (denoted as S) is approximately equal to P 2 E at
the surface. Remote sensing retrievals or reanalysis fields
of q along with wind products from reanalyses should
provide a robust measure of S in the atmosphere such that
independent measured P 2 E can be reproduced. As long
as the satellite observations and reanalysis products are
consistent, the atmospheric branch of the hydrological
cycle can be thought of as ‘‘closed.’’
In this study, we will investigate two datasets of S: 1) the
remotely sounded q from AIRS will be used together with
the MERRA wind fields to estimate global patterns of
S, and 2) the MERRA S directly calculated from the
MERRA water vapor budget. This will help determine if
their column integrals (S) are consistent with the net
surface water exchange (P 2 E) estimated from the
TRMM 3B42 or GPCP Ps and GSSTF E as well as from
the MERRA output of surface fluxes. Variability of the
estimated and reanalysis Ss as well as the observed and
reanalysis P 2 E will be assessed individually by maps of
seasonal averages and by wavelet analyses.
2. Data
The S can be estimated from a combination of remotely
sounded q and reanalysis winds. To do this, the q in Eq.
(1) is obtained from AIRS level 3 (L3) version 5 product
at 18 3 18 horizontal resolution (Olsen et al. 2007), and
the wind fields are obtained from the MERRA product
(version 5.2). Equation (1) is applied to daily data. Since
AIRS makes tropical measurements around 0130 and 1330
local time (LT), the daily averaged q within each 18 3 18
grid is actually the average of the two measurements at
the two time periods. There are two issues that need to
be addressed. The first issue regarding the lack of sampling in cloudy areas hinders the calculation of moisture
gradients, so AIRS daily q are averaged onto 108 3 58
longitude–latitude grid boxes that provide acceptable
6310
JOURNAL OF CLIMATE
VOLUME 24
FIG. 3. Geographical distributions of precipitation minus evaporation (P 2 E, mm day21) of (a),(d) TRMM 3B42
P 2 GSSTF2b E; (b),(e) GPCP P 2 GSSTF2b E; and (c),(f) MERRA P 2 E. Averages for (left) DJF of 2004–08 and
(right) JJA of 2004–08. Missing data in GSSTF2b E are in black area.
spatial coverage of the q gradients needed in the calculation of S over the globe. The second issue regarding
the lack of sampling in cloudy areas is related to cloudinduced sampling biases. Fetzer et al. (2006) showed that
the lack of sampling in opaque and precipitating clouds
cause the total column water vapor obtained from AIRS
retrievals in the tropics to be low by 2%–5% in comparison
with the Advanced Microwave Scanning Radiometer
for Earth Observing System (AMSR-E). These biases
can be higher in tropical deep convective regions and
the midlatitudes since high filaments of water vapor are
disproportionately found in cloudy atmospheric rivers
(Ralph et al. 2004) and frontal systems, which are significantly undersampled in AIRS data. Significant biases in the
vertical coordinate (Soden and Lanzante, 1996; Tian et al.
2006) can also occur from poor sampling in opaque clouds,
even in the tropics, but these effects are not as important
for quantifying column-integrated estimates of P 2 E.
MERRA uses the Goddard Earth Observing System,
version 5 (GEOS-5; Rienecker et al. 2008) to assimilate
observations, including AIRS radiance data, for the
analysis. The 3-hourly instantaneous MERRA winds are
15 DECEMBER 2011
WONG ET AL.
6311
FIG. 4. Geographical distributions of (a),(c) AIRS and (b),(d) MERRA precipitable water vapor (kg m22).
Averages for (left) DJF of 2004–08 and (right) JJA of 2004–08.
averaged daily onto the same 108 3 58 grid boxes as used
for AIRS. The MERRA winds are interpolated or averaged from the 42 pressure levels to the 12 AIRS standard pressure levels for q. After these fields are
consistently processed on a common grid, the right hand
side of Eq. (1) can be calculated on 108 3 58 longitude–
latitude grids, where the gradients are computed by finite
differencing in spherical coordinates. We chose pt in Eq.
(2) to be 200 hPa, the topmost level where AIRS measurements are sensitive to atmospheric water vapor in the
tropics (e.g., Fetzer et al. 2008; Gettelman et al. 2004).
The contribution to the total water vapor burden from
above this altitude is several orders of magnitude smaller
and is neglected in this work. Hereafter, the column integral of S in Eq. (2) that uses AIRS q and MERRA
winds is referred to as AIRS–MERRA S.
While MERRA assimilates AIRS radiance data, it provides fields of q even in cloudy areas. Thus, we also calculate S directly from the MERRA’s water vapor budget
(i.e., from MERRA’s output of q tendencies from dynamics), referred to as MERRA–MERRA S. To compare
with the AIRS–MERRA S, the MERRA–MERRA S is
averaged onto 108 3 58 grid boxes from its original resolution of 1.258 3 1.258.
The P term in Eq. (2) is obtained from the precipitation algorithm 3B-42 (version 6) of TRMM (Huffman
et al. 2007) as well as the GPCP 1DD data (version 1.1;
Huffman et al. 2001, 2009). The TRMM 3B42 algorithm
merges high quality microwave precipitation retrievals
from instruments that include the TRMM Combined
Instrument (TCI), TRMM Microwave Imager (TMI),
Special Sensor Microwave Image (SSM/I), AMSR-E, and
Advanced Microwave Sounding Unit-B (AMSU-B) with
adjusted precipitation estimates from geostationary observations of infrared brightness temperature (Huffman
and Bolvin 2009). The gridded P data are reported at
3-hourly intervals and 0.258 3 0.258 resolution, and extend globally from 508S to 508N. The GPCP 1DD data
product is a 18 3 18 resolution daily precipitation dataset.
It is based on the GPCP version 2.1 satellite-gauge
product, which merges global precipitation-gauge analyses with precipitation retrievals from satellites including
SSM/I, Television and Infrared Observation Satellite
(TIROS) Operational Vertical Sounder (TOVS), AIRS,
and others (Huffman et al. 2009 and references therein).
In this study, the P fields are averaged daily on 108 3 58
grid boxes to facilitate comparison to AIRS–MERRA
and MERRA–MERRA Ss. The P is also obtained from
6312
JOURNAL OF CLIMATE
VOLUME 24
FIG. 5. Geographical distributions of (a),(c) AIRS–MERRA and (b),(d) MERRA–MERRA total water vapor sink
(mm day21). Averages for (left) DJF of 2004–08 and (right) JJA of 2004–08.
the MERRA reanalysis and averaged daily on 108 3 58
grid boxes.
The E term in Eq. (2) is obtained from the GSSTF
version 2b dataset 2, which are daily latent heat flux estimates based on a bulk flux model with inputs of SSM/I
retrievals of 10-m wind speeds, total precipitable water,
and bottom-layer (500 m) precipitable water as well as
sea surface temperature, 2-m air temperature, and sea
level pressure from National Centers for Environmental
Prediction/Depart of Energy (NCEP/DOE) reanalysis-2
(Chou et al. 2003; Shie et al. 2009). The daily latent heat
flux data are reported at 18 3 18 resolution. We converted the latent heat flux to surface water evaporation
flux by dividing the data with the latent heat of evaporation of water. The evaporation fluxes are then averaged on 108 3 58 grid boxes for comparison with the
AIRS–MERRA data.
We compare the estimates of S from AIRS–MERRA
and MERRA–MERRA with three different estimates
of P 2 E: the TRMM 3B42 P minus the GSSTF2b E
(referred to as TRMM–GSSTF P 2 E), the GPCP 1DD
P minus the GSSTF2b E (referred to as GPCP–GSSTF
P 2 E), and the MERRA P minus the MERRA E
(referred to as the MERRA P 2 E). In this way, we not
only investigate the hydrological consistency between
different datasets, but also evaluate the performance of
the MERRA P and E with respect to the observationally
based estimates.
3. Results
a. Precipitation and evaporation
Climatological averages of the TRMM, GPCP, and
MERRA P in 2004–08 are shown in Fig. 1 for boreal
winter [December–January–February (DJF) in Figs. 1a–c]
and summer [June–July–August (JJA) in Figs. 1d–f]. The
geographical distributions of the MERRA P are consistent
with those of the TRMM and GPCP P in both seasons.
However, the MERRA P is larger over the west Pacific,
the Indian, and North American monsoon area.
Similar to the case of P, the MERRA reproduces
geographical distributions of E observed in the GSSTF
for both boreal winter and summer (Fig. 2). However,
the MERRA has smaller E maxima over the subtropical oceans as well as the storm tracks in both
hemispheres.
Seasonal averages of P 2 E are plotted in Fig. 3 for the
TRMM–GSSTF, GPCP–GSSTF, and MERRA. The
15 DECEMBER 2011
WONG ET AL.
6313
area with P 2 E . 0 is where the atmosphere loses water
vapor through precipitation. This area includes the intertropical convergence zone (ITCZ), the west Pacific,
mid- to high-latitude storm tracks in both hemispheres, and
the Indian, North American, and West African monsoon
regions. Areas with P 2 E , 0 are where the atmosphere
gains water vapor from the surface and are primarily located over the subtropical oceans. The MERRA P 2 E
patterns are consistent with the observationally based estimates. However, the MERRA has larger P 2 E over the
west Pacific, the Indian, and North American monsoon
area, because of the larger P. Since MERRA has smaller E
in the subtropical oceans, the MERRA P 2 E in the subtropical oceans is less negative than those of the TRMM–
GSSTF and GPCP–GSSTF. The smaller E in MERRA
over the storm tracks in both hemispheres also yields
larger P 2 E compared to TRMM–GSSTF and GPCP–
GSSTF over mid- to high latitudes in the winter hemisphere (DJF for the Northern Hemisphere and JJA for
the Southern Hemisphere).
b. Apparent water vapor sink
AIRS q is applied to compute the AIRS–MERRA S.
Fetzer et al. (2006) compared the AIRS precipitable
water (column-integrated specific humidity) with AMSRE’s and found that the biases in AIRS precipitable water
depend on the types of cloud present. Compared to the
MERRA precipitable water, the AIRS precipitable water
has similar geographical distributions in both summer
and winter (Fig. 4). The relative difference is about 5%–
10%, consistent with the biases reported by Fetzer et al.
(2006). The AIRS has less precipitable water over the
west Pacific, the Indian, and North American monsoon
area, and the Amazon region. This may be caused by the
lack of sounding over opaque and precipitating clouds in
AIRS.
Geographical distributions of boreal wintertime and
summertime averaged S (for AIRS–MERRA and
MERRA–MERRA) are shown in Fig. 5. The patterns of
S generally reproduce the seasonal cycle of the global
precipitation and evaporation components, including the
variations of the location and strength of the ITCZ, the
variations of Indian, African, and North American summer monsoons, and the evaporation over the subtropical
oceans. The AIRS–MERRA S agrees well with the
MERRA–MERRA S. Detailed discrepancies include the
larger S in AIRS–MERRA over the southern Amazon of
South America during DJF, tropical East Africa, the
Southwest, and northern Mexico during the monsoon
season in JJA. The MERRA–MERRA S is larger than
the AIRS–MERRA S over the midlatitudes during the
respective hemispheric winter seasons. Further tests using
MERRA water vapor and winds averaged for different
FIG. 6. Zonal averages over the oceans for AIRS–MERRA (thin
solid) and MERRA–MERRA (thin dash) total water vapor sink,
MERRA P minus E (thin dash–dot), TRMM 3B42 P minus
GSSTF2b E (thick solid), and GPCP P minus GSSTF2b E (thick
dash). Data are averaged for (a) DJF and (b) JJA of 2004–08.
grid box sizes for calculation of advection show that an
overestimation of AIRS–MERRA in these regions may
be related to the coarse resolution used to compute the
water vapor advection. In general, the similarity between
the AIRS–MERRA and the MERRA–MERRA Ss
suggests that sampling effects caused by the missing data
around cloudy areas in the AIRS L3 dataset (Fig. 4) are
minimal. However, cloud-induced systematic biases may
still exist and cannot be assessed by merely comparing the
two datasets. Microwave remote sensing techniques for
water vapor in cloudy areas may help reduce such systematic biases and are necessary for more accurately
quantifying the atmospheric hydrological cycle.
Zonal averages of AIRS–MERRA and MERRA–
MERRA Ss over the oceans are compared to the
MERRA, TRMM–GSSTF, and GPCP–GSSTF P 2 E in
Fig. 6. The AIRS–MERRA and MERRA–MERRA Ss
are in general agreement with the MERRA P 2 E.
GPCP has higher precipitation rates over the mid- and
high latitudes (Figs. 1a–d); therefore, GPCP–GSSTF
6314
JOURNAL OF CLIMATE
VOLUME 24
FIG. 7. Geographical distributions of differences of (left) AIRS–MERRA, (middle) MERRA–MERRA total water vapor sinks, and
(right) the MERRA P 2 E from the observed P 2 E. The observed P 2 E includes TRMM 3B42 P minus GSSTF2b E (the first and third
rows) and GPCP P minus GSSTF2b E (the second and fourth rows). The data are averaged for DJF (the first two rows) and JJA (the
bottom two rows). Missing data in GSSTF2b are shown in black area.
P 2 E is larger than TRMM–GSSTF P 2 E at these
latitudes. The AIRS–MERRA and MERRA–MERRA
Ss, and the MERRA P 2 E have reasonable latitudinal
variations compared to the TRMM–GSSTF and GPCP–
GSSTF (P 2 E)s. The AIRS–MERRA and MERRA–
MERRA Ss, and the MERRA P 2 E are larger than the
observationally based P 2 E in the tropics during boreal
winter, the subtropics in both seasons, and the wintertime
hemisphere’s mid- to high latitudes.
Geographical distributions of the differences between
AIRS–MERRA and MERRA–MERRA Ss as well
as the MERRA P 2 E from the observationally based
P 2 E are shown in Fig. 7 for both boreal winter and
summer. Large differences are located over the tropics,
15 DECEMBER 2011
TABLE 1. Annually averaged total water vapor sinks and P 2 E
(mm day21) over land and ocean for 508S–508N for 2004–08. The
rightmost column shows the annually averaged data for the whole
globe (908S–908N).
Ocean
Land
Globe
Globe
508S–508N 508S–508N 508S–508N 908S–908N
AIRS–MERRA
MERRA–MERRA
MERRA P 2 E
TRMM P 2
GSSTF E
GPCP P 2
GSSTF E
6315
WONG ET AL.
20.34
20.36
20.17
21.38
0.80
0.79
0.38
—
20.15
20.15
20.07
—
0.03
0.12
0.17
—
21.01
—
—
—
subtropics, and the wintertime hemispheres. In the boreal winter, the AIRS–MERRA has larger S over the
Pacific ITCZ, subtropical Pacific, and Atlantic midlatitude storm track, while the MERRA has larger S and
P 2 E over the tropical west Pacific and the high-latitude
North Pacific storm track. In the boreal summer, the
AIRS–MERRA and MERRA–MERRA Ss, and
MERRA P 2 E all have larger values over the Asian
and North American monsoon area, the subtropical east
Pacific, and the Southern Hemispheric storm track.
In Table 1, we list the averages of AIRS/MERRA and
MERRA–MERRA Ss, MERRA P 2 E, the TRMM–
GSSTF and GPCP–GSSTF P 2 E for 508S–508N over the
ocean. The averaged TRMM–GSSTF and GPCP–GSSTF
P 2 E over the ocean are more negative than the Ss and
MERRA P 2 E. This is consistent with the larger evaporation in GSSTF2b over the subtropical ocean compared
to the MERRA evaporation (Fig. 2). Also shown in Table
1 are the averages of AIRS–MERRA and MERRA–
MERRA Ss, and MERRA P 2 E over the land and the
globe. The averaged MERRA–MERRA Ss are similar to
the averaged AIRS–MERRA Ss and have more land–sea
contrast than the MERRA P 2 E. When averaged over
508S–508N, AIRS–MERRA and MERRA–MERRA Ss
have a value of 20.15 mm day21, while MERRA – is
20.07 mm day21. However, when averaged over the
whole globe (908S–908N), the AIRS–MERRA S gives a
value of 0.03 mm day21, the MERRA–MERRA S yields
0.12 mm day21, and the MERRA P 2 E has a value of
0.17 mm day21. These numbers are comparable to those
obtained by NCEP Climate Forecast System Reanalysis
(CFSR), in which the globally averaged P 2 E is about
a few tenths of a millimeter per day (Saha et al. 2010).
c. Closure of the MERRA water vapor budget
A general circulation model balances its water vapor
budget such that its S should equal to its predicted P 2 E.
Since data reanalysis assimilates observations, the
model will achieve balance among S, P 2 E, and the q
tendency from the analysis. This can be observed by
the discrepancy between MERRA–MERRA S and
MERRA P 2 E by comparing Figs. 3c,f with Figs. 5b,d,
respectively. The difference between MERRA P 2 E
and MERRA–MERRA S is the MERRA q tendency
from the analysis, hdq/dtiana, where the bracket represents column integration (M. Bosilovich 2010, personal
communication):
dq
5 P 2 E 2 S.
dt ana
(3)
The climatological averages of the MERRA hdq/dtiana
for the boreal winter and summer are shown in Figs. 8a,e,
respectively. Areas with positive (negative) hdq/dtiana
indicates regions where the GEOS-5 assimilation system
increases (decreases) q to match the observations.
Model biases in the simulation of hydrological processes are revealed by the discrepancy of model’s P and E
from the corresponding observed values, assuming the
observations represent the truth and the assimilation
processes have no bias. Therefore, one can rewrite Eq. (3)
in the following manner:
dq
5 (P 2 Pobs ) 2 (E 2 Eobs )
dt ana
2 [S 2 (Pobs 2 Eobs )] [ DP 2 DE 2 DS,
(4)
where DP represents the contribution of discrepancy in
the model precipitation process from observations to the
hdq/dtiana, DE represents the contribution of the discrepancy in the model surface evaporation from observations,
and DS represents the discrepancy of MERRA–MERRA
S from the observed P 2 E.
Since the errors of the TRMM and GPCP precipitations and of the GSSTF2b surface evaporation are unknown, implication of the influence of the calculated DP,
DE, and DS on hdq/dtiana becomes less straightforward.
Nevertheless, one can still investigate if spatial correlation exists between hdq/dtiana and the DP, DE, and DS
calculated using the TRMM, GPCP, and GSSTF2b datasets. In Fig. 8, we plot the MERRA hdq/dtiana and
compare its spatial patterns with DP using either TRMM
or GPCP, and with 2DE using GSSTF E. The DS has
been shown in the middle column of Fig. 7. Notice that
hdq/dtiana is the residual of DP, DE, and DS, and its scale is
about half of the others in Fig. 8.
In the boreal winter, positive MERRA hdq/dtiana over
the tropical west to central Pacific and equatorial West
Africa (Fig. 8a) are correlated with the larger model
precipitation over these regions, as compared to both
6316
JOURNAL OF CLIMATE
FIG. 8. Geographical distributions of (a),(e) column-integrated water vapor analysis increment in MERRA;
(b),(f) differences of MERRA P from TRMM 3B42 P; (c),(g) differences of MERRA P from GPCP P; and
(d),(h) negative differences of MERRA E from GSSTF2b E. Averages for (left) DJF of 2004–08 and (right) JJA
of 2004–08. Missing data in GSSTF2b are shown as black areas. Notice that the color scale used for the analysis
increment in (a),(e) is about half of those used for the precipitation and evaporation.
VOLUME 24
15 DECEMBER 2011
WONG ET AL.
FIG. 9. Regions for averaging the total water vapor sinks or P 2 E
for wavelet analyses shown in Fig. 10. The regions are denoted as
(a) tropical Indian Ocean (108S–108N, 508–1108E), (b) west Pacific
(08–208N, 1108–1708E), (c) South Pacific (208S–08, 1308E–1608W),
(d) east Pacific (08–158N, 1608–808W), (e) tropical Atlantic (08–
108N, 458W–08), (f) the India–Bengal area (108–208N, 708–1108E),
(g) tropical South America (108S–58N, 808–458W), and (h) tropical
Africa (108S–108N, 108–408E). Color contours indicate the annual
mean P 2 E from TRMM 3B42 and MERRA E for 2004–08.
TRMM (Fig. 8b) and GPCP P (Fig. 8c). Larger
MERRA–MERRA S over the equatorial east Pacific
compared to Pobs 2 Eobs (Figs. 7e,f) is associated with
the local negative hdq/dtiana (Fig. 8a).
In the boreal summer, positive hdq/dtiana over the
Asian and North American monsoon regions as well as
the tropical west to central Pacific (Fig. 8e) are primarily
associated with larger model precipitation over these
regions (Figs. 8f,g). Smaller model evaporation is found
in the subtropical east Pacific and Atlantic in the Southern Hemisphere (Fig. 8h) compared to observations, and
is associated with positive hdq/dtiana over these regions.
Because of the lack of information on the errors of the
observationally based estimates, the correlation between
hdq/dtiana and DP or DE shown in Fig. 8 may be accidental
or may represent actual realization of the model biases.
We present the correlation results in this study for the
reference of future research.
d. Wavelet spectra of regional hydrological budgets
To further quantify the sources of variability in S and
the different P 2 E, we perform wavelet analyses on regionally averaged results. The S and P 2 E fields are averaged over chosen regions (Fig. 9), and the resultant time
series are transformed by the Morlet wavelet (Farge
1992). In Fig. 9, regions a–e are along the ITCZ, region f
is over the India and Bay of Bengal area (referred to as
the India–Bengal area), region g is over tropical South
America, and region h is over tropical Africa. The wavelet
spectra are shown in Fig. 10. Since surface evaporation
6317
data are not available over land, we only show the wavelet spectra of P for TRMM and GPCP over the Indian–
Bengal area (Fig. 10f), tropical South America (Fig. 10g),
and South Africa (Fig. 10h). We have tested that the
wavelet spectra of MERRA P 2 E over these coastal
or continental regions are almost identical to those of
MERRA P (precipitation only).
Both AIRS–MERRA and MERRA–MERRA Ss have
similar wavelet spectra over all regions. In most regions,
the frequency of variations of S and P 2 E mainly falls
into three regimes: the annual cycle with a period around
365 days, semiannual variations with periods of 100–200
days, and intraseasonal variations with periods of 30–90
days. Except over tropical Africa, AIRS–MERRA,
MERRA–MERRA, and MERRA P 2 E capture the timing of most of the variations when compared to the
TRMM–GSSTF and GPCP–GSSTF P 2 E. However, the
amplitudes of these variations may differ among different
datasets, depending on the regions. The AIRS–MERRA,
MERRA–MERRA Ss, and the MERRA P 2 E have
weaker amplitudes of intraseasonal variations over the
tropical Indian Ocean (Fig. 10a), where the amplitudes
are smaller by up to 50% compared to that seen in the
TRMM–GSSTF and GPCP–GSSTF P 2 E. This is consistent with the analysis of water and energy budgets associated with the intraseasonal oscillation of the Indian
summer monsoon (Wong et al. 2011).
Over the India–Bengal area (Fig. 10f), the variability in
AIRS–MERRA and MERRA–MERRA Ss as well as the
MERRA P 2 E is generally similar to that in the TRMM
and GPCP P, although their intraseasonal variations have
smaller amplitudes compared to observations. The amplitudes of the annual cycle of the MERRA P 2 E over
this region (as well as over the west Pacific) are larger than
any of the others. This discrepancy is also reflected in the
larger summertime precipitation rates in the MERRA
P 2 E (Figs. 1f, 3f, 7k,l, and 8f,g) over these two regions.
Over tropical Africa (Fig. 5h), the AIRS–MERRA
and MERRA–MERRA Ss have very different spectra
from that of the TRMM and GPCP P. The semiannual
cycle in the AIRS–MERRA and MERRA–MERRA S
has weaker variability than the annual cycle. However, the
observed P (TRMM and GPCP) have stronger variability
in the semiannual cycle than the annual cycle. The
MERRA P 2 E is not consistent with either its own
MERRA–MERRA S or the TRMM and GPCP P. This
may suggest that the moisture modules in the GEOS-5
model may continue to require refinement over tropical
Africa and/or the quality of TRMM 3B42 and GPCP
precipitation rates over continents may have limitations.
Future investigations of precipitation data over land
should be carried out (e.g., Dinku et al. 2010, Huffman
and Bolvin 2009).
6318
JOURNAL OF CLIMATE
VOLUME 24
FIG. 10. Wavelet spectra of the (left to right) AIRS–MERRA, MERRA–MERRA total water vapor sinks, MERRA P 2 E, TRMM–
GSSTF P 2 E, and GPCP–GSSTF P 2 E. The unit of the amplitudes is mm day21. The data are averaged over (a) tropical Indian Ocean,
(b) west Pacific, (c) South Pacific, (d) east Pacific, (e) tropical Atlantic, (f) the India–Bengal area, (g) tropical South America, and (h)
tropical Africa. The boundary of each region is listed and shown in Fig. 9. The dashed lines indicate the regimes influenced by the edge
effect, and the solid lines circle the regime with a 95% confidence level.
4. Discussion and conclusions
The total water vapor sinks (S) in the atmosphere are
estimated using water vapor retrievals from the Atmospheric Infrared Sounder/Advanced Microwave Sounding Unit (AIRS) with the wind fields from the NASA
GSFC Modern Era Retrospective-analysis for Research
and Applications (MERRA) reanalysis, as well as from
the water vapor budget in the MERRA (AIRS–MERRA
S and MERRA–MERRA S, respectively). Both estimates of atmospheric water vapor sinks are checked for
physical consistency with the surface water exchange
calculated from the Tropical Rainfall Measuring Mission
(TRMM) 3B42 or the Global Precipitation Climatology
Project (GPCP) precipitation rate, and the Goddard
Satellite-based Surface Turbulent Fluxes (GSSTF2b)
surface evaporation rate (TRMM–GSSTF and GPCP–
GSSTF P 2 E, respectively). We also compare the
MERRA P 2 E with the TRMM–GSSTF and GPCP–
GSSTF (P 2 E)s.
The AIRS–MERRA S is generally consistent with
that from the MERRA water vapor budget, indicating
that cloud-induced sampling effects in AIRS retrieved q
is minimized when averaging the data with large enough
grid boxes (108 3 58 used in this study). However, a systematic retrieval bias within persistently cloudy areas
may still be a potential problem in the estimation of S.
Moreover, averaging q and wind data over large grid
boxes before the calculation of finite difference may also
generate biases over areas with large variation in topography such as the South American Andes. Future research that focuses on comparison of estimates of water
vapor budget using AIRS data with observations of surface water vapor exchange over cloud-cleared area requires collocation of AIRS level-2 product with satellite
soundings of cloud and precipitation. Estimations obtained from datasets with higher spatial resolution, accuracy, and a more complete sampling in all cloudy states
may lead to further improvements in the global estimates
of S using remote sensing data.
15 DECEMBER 2011
WONG ET AL.
6319
FIG. 10. (Continued)
Compared to the TRMM–GSSTF and GPCP–GSSTF
(P 2 E)s, both AIRS–MERRA and MERRA–MERRA
Ss can reproduce distributions and variations of the
main components of the atmospheric branch of the
global hydrological cycle. These components include net
positive water vapor sinks in the intertropical convergence zone (ITCZ), summer monsoons, and midlatitude
storm tracks in both hemispheres. However, values of
S are larger than the observationally based estimates of
P 2 E along the ITCZ, the summer monsoon regions,
the subtropical east Pacific in the Southern Hemisphere,
and the mid- and high-latitude storm track.
MERRA P 2 E is in general similar to TRMM–GSSTF
and GPCP–GSSTF (P 2 E)s. However, MERRA has
larger P 2 E over the Asian and North American monsoon area and tropical west Pacific because of the larger
MERRA P (compared to TRMM and GPCP), and over
the subtropical east Pacific in the Southern Hemisphere
and the wintertime storm tracks because of the smaller
MERRA E (compared to GSSTF). The discrepancies
of MERRA P and E from the observations are correlated with the GEOS-5 analysis water vapor tendency
hdq/dtiana over the Asian and North American monsoon
area, tropical west Pacific, and the subtropical east Pacific
in the Southern Hemisphere. It is difficult to interpret
such correlation as the realization of the model bias, because of the lack of information on the uncertainty of the
observationally based estimates of P 2 E.
The AIRS–MERRA, MERRA–MERRA Ss, and
the MERRA P 2 E have globally averaged values of 0.03,
0.12, and 0.17 mm day21, respectively. Over the ocean, the
TRMM–GSSTF and GPCP–GSSTF averaged (P 2 E)s
are more negative than the averaged AIRS–MERRA,
MERRA–MERRA Ss, and the averaged MERRA
P 2 E, because of the larger GSSTF2b surface evaporation over the subtropical ocean.
Wavelet analysis is used to quantify the temporal
variability and amplitude of P 2 E and S. The AIRS–
MERRA and MERRA–MERRA Ss have reasonable
components of temporal variations along the oceanic
ITCZ and the Maritime Continent. These components
include the annual and semiannual cycles as well as intraseasonal variations. Over the tropical Indian Ocean,
the Indian continent, and the Bay of Bengal area, the
AIRS–MERRA and MERRA–MERRA Ss have smaller
amplitudes of intraseasonal variations, and this is also
6320
JOURNAL OF CLIMATE
reflected in the smaller intraseasonal amplitudes of
MERRA P 2 E over these regions. However, over India,
the Bay of Bengal, and the tropical west Pacific, the
seasonal-cycle amplitudes of MERRA P 2 E (or P alone)
are larger than those of S, and TRMM and GPCP P.
Over tropical Africa, both Ss, MERRA P 2 E (or just
P), TRMM and GSSTF P have inconsistent wavelet
spectra. This suggests future validation of water vapor and
wind profiles, satellite retrieved precipitation rates, and
further investigation of the behavior of moisture modules
in the assimilation model are necessary over this region.
This work is highly relevant to future Joint Polar Satellite System (JPSS; Lee et al. 2010) Cross-Track Infrared
Sounder and Advanced Technology Microwave Sounder
(CrIMSS) observations. The ‘‘closure’’ study presented
here could be used as a test of the satellite retrieval data
with regard to its ability to faithfully represent important
climate processes. Furthermore, continuity could be established with AIRS and JPSS to quantify long-term behavior in trends and variability of S with climate change.
Correct closure of moisture budgets is essential in climate models and reanalysis products. Therefore, this
work is also highly relevant to evaluation of the moisture,
wind, precipitation, and evaporation simulations from
global climate models as well as climate reanalyses (e.g.,
Saha et al. 2010) with the satellite-based data products.
Acknowledgments. We thank Michael Bosilovich at
NASA Goddard Space Flight Center for the discussion
about the MERRA water vapor product. We are grateful
to Chung-Lin Shie at NASA Goddard Space Flight
Center for the discussion of the usage of GSSTF2b data.
We thank NASA Making Earth Science Data Records
for Use in Research Environments (MEaSUREs) for
supporting the data, and Goddard Earth Sciences Data
and Information Services Center (GES DISC) for distributing the data used in this work. The AIRS, MERRA,
TRMM 3B42, and GSSTF2b datasets used in this work
can be downloaded from GES DISC. (The GPCP data
can be downloaded from the Web site http://precip.
gsfc.nasa.gov.) We also thank three anonymous reviewers
for comments that improved the manuscript. The research described in this paper was carried out at the Jet
Propulsion Laboratory, California Institute of Technology, under a contract with the National Aeronautics and
Space Administration. This work is supported by the JPL
AIRS project, NASA Energy and Water Cycle Study
(NEWS), and NASA MEaSUREs.
REFERENCES
Allen, M. A., and W. J. Ingram, 2002: Constraints on future changes
in climate and the hydrological cycle. Nature, 419, 224–232,
doi:10.1038/nature01092.
VOLUME 24
Bosilovich, M., J. Chen, F. R. Robertson, and R. F. Adler, 2008:
Evaluation of global precipitation in reanalyses. J. Appl.
Meteor. Climatol., 47, 2279–2299.
Chou, S.-H., E. Nelkin, J. Ardizzone, R. M. Atlas, and C.-L. Shie,
2003: Surface turbulent heat and momentum fluxes over
global oceans based on the Goddard satellite retrieval, version
2 (GSSTF2). J. Climate, 16, 3256–3273.
Clarke, R., and J. King, 2004: The Water Atlas. The New Press, 122 pp.
Dai, A., I. Y. Fung, and A. D. Del Genio, 1997: Surface observed
global land precipitation variations during 1900–88. J. Climate,
10, 2943–2962.
Dinku, T., F. Ruiz, S. J. Connor, and P. Ceccato, 2010: Validation
and intercomparison of satellite rainfall estimates over Colombia. J. Appl. Meteor. Climatol., 49, 1004–1014.
Divakarla, M., C. D. Barnet, M. D. Goldberg, L. M. McMillin,
E. Maddy, W. Wolf, and L. Zhow, 2006: Validation of AIRS
temperature and water vapor retrievals with matched radiosonde measurements and forecasts. J. Geophys. Res., 111,
D09S15, doi:10.1029/2005JD006116.
Farge, M., 1992: Wavelet transforms and their application to turbulence. Annu. Rev. Fluid Mech., 24, 395–457.
Fetzer, E. J., J. Teixeira, E. T. Olsen, and E. F. Fishbein, 2004:
Satellite remote sounding of atmospheric boundary layer
temperature inversions over the subtropical eastern Pacific.
Geophys. Res. Lett., 31, L17102, doi:10.1029/2004GL020174.
——, B. H. Lambrigtsen, A. Eldering, H. H. Aumann, and M. T.
Chahine, 2006: Biases in total precipitable water vapor climatologies from Atmospheric Infrared Sounder and Advanced Microwave Scanning Radiometer. J. Geophys. Res.,
111, D09S16, doi:10.1029/2005JD006598.
——, and Coauthors, 2008: Comparison of upper tropospheric
water vapor observations from the Microwave Limb Sounder
and Atmospheric Infrared Sounder. J. Geophys. Res., 113,
D22110, doi:10.1029/2008JD010000.
Gettelman, A., and Coauthors, 2004: Validation of Aqua satellite
data in the upper troposphere and lower stratosphere with in
situ aircraft instruments. Geophys. Res. Lett., 31, L22107,
doi:10.1029/2004GL020730.
Huffman, G. J., and D. T. Bolvin, 2009: TRMM and other data
precipitation data set documentation. Laboratory for Atmospheres, NASA Goddard Space Flight Center and Science
Systems and Applications, Inc., 29 pp. [Available online at ftp://
precip.gsfc.nasa.gov/pub/trmmdocs/3B42_3B43_doc.pdf.]
——, R. F. Adler, M. Morrissev, D. T. Bolvin, S. Curtis, R. Joyce,
B. McGavock, and J. Susskind, 2001: Global precipitation at
one-degree daily resolution from multi-satellite observations.
J. Hydrometeor., 2, 36–50.
——, and Coauthors, 2007: The TRMM Multisatellite Precipitation
Analysis (TMPA): Quasi-global, multi-year, combined-sensor
precipitation estimates at fine scales. J. Hydrometeor., 8, 38–55.
——, R. F. Adler, D. T. Bolvin, and G. Gu, 2009: Improving the
global precipitation record: GPCP version 2.1. Geophys. Res.
Lett., 36, L17808, doi:10.1029/2009GL040000.
Lee, T. F., and Coauthors, 2010: NPOESS: Next-generation operational global earth observations. Bull. Amer. Meteor. Soc., 91,
727–740.
Olsen, E. T., S. Granger, E. Manning, and J. Blaisdell, 2007: AIRS/
AMSU/HSB version 5 level 3 quick start, version 1.1. Jet Propulsion Laboratory/California Institute of Technology, 25 pp.
Peixoto, J. P., and A. H. Oort, 1992: Physics of Climate. American
Institute of Physics, 520 pp.
Ralph, F. M., P. J. Neiman, and G. A. Wick, 2004: Satellite and
CALJET aircraft observations of atmospheric rivers over the
15 DECEMBER 2011
WONG ET AL.
eastern North Pacific Ocean during the winter of 1997/98.
Mon. Wea. Rev., 132, 1721–1745.
Rienecker, M. M., and Coauthors, 2008: The GEOS-5 data assimilation system—Document of versions 5.0.1, 5.1.0, and 5.2.0.
Tech. Rep. Series on Global Modeling and Data Assimilation
104606, Vol. 27, NASA, 118 pp.
Saha, S., and Coauthors, 2010: The NCEP climate forecast system
reanalysis. Bull. Amer. Meteor. Soc., 91, 1015–1057.
Schneider, T., P. A. O’Gorman, and X. J. Levine, 2010: Water
vapor and the dynamics of climate changes. Rev. Geophys., 48,
RG3001, doi:10.1029/2009GR000302.
Schumacher, C., P. E. Ciesielski, and M. H. Zhang, 2008: Tropical
cloud heating profiles: Analysis from KWAJEX. Mon. Wea.
Rev., 136, 4289–4300.
Shie, C.-L., and Coauthors, 2009: A note on reviving the Goddard
Satellite-based Surface Turbulent Fluxes (GSSTF) dataset.
Adv. Atmos. Sci., 26 (6), 1071–1080.
Shige, S., Y. N. Takayabu, and W.-K. Tao, 2008: Spectral retrieval
of latent heating profiles from TRMM PR data. Part III: Estimating apparent moisture sink profiles over tropical oceans.
J. Appl. Meteor. Climatol., 47, 620–640.
Soden, B. J., and J. R. Lanzante, 1996: An assessment of satellite
and radiosonde climatologies of upper-tropospheric water
vapor. J. Climate, 9, 1235–1250.
Stephens, G. L., and T. D. Ellis, 2008: Controls of global-mean
precipitation increases in global warming GCM experiments.
J. Climate, 21, 6141–6155.
——, and Coauthors, 2002: The CloudSat mission and the A-train.
Bull. Amer. Meteor. Soc., 83, 1771–1790.
6321
Sun, Y., S. Solomon, A. Dai, and R. W. Portmann, 2007: How often
will it rain? J. Climate, 20, 4801–4818.
Susskind, J., C. Barnet, J. Blisdell, L. Iredell, F. Keita, L. Kouvaris,
G. Molnar, and M. Chahine, 2006: Accuracy of geophysical
parameters derived from Atmospheric Infrared Sounder/
Advanced Microwave Sound Unit as a function of fractional
cloud cover. J. Geophys. Res., 111, D09S17, doi:10.1029/
2005JD006272.
Tian, B., D. E. Waliser, E. J. Fetzer, B. H. Lambrigtsen, Y. L. Yung,
and B. Wang, 2006: Vertical moist thermodynamic structure
and spatial–temporal evolution of the MJO in AIRS observations. J. Atmos. Sci., 63, 2462–2485.
Trenberth, K. E., and C. J. Guillemot, 1998: Evaluation of the atmospheric moisture and hydrological cycle in the NCEP/
NCAR reanalyses. Climate Dyn., 14, 213–231.
——, A. Dai, R. M. Rasmussen, and D. B. Parsons, 2003: The
changing character of precipitation. Bull. Amer. Meteor. Soc.,
84, 1205–1217.
Watkins, K., and Coauthors, 2007: Fighting climate change: Human
solidarity in a divided world. Human Development Rep. 2007/
2008, Palgrave Macmillan for United Nations Development
Program, 399 pp.
Wong, S., E. J. Fetzer, B. Tian, B. H. Lambrigtsen, and H. Ye, 2011:
The apparent water vapor sinks and heat sources associated
with the intraseasonal oscillation of the Indian summer monsoon. J. Climate, 24, 4465–4478.
Yanai, M., S. Esbensen, and J.-H. Chu, 1973: Determination of
bulk properties of tropical cloud clusters from large-scale heat
and moisture budgets. J. Atmos. Sci., 30, 611–627.