Hydr JHM D 11 088 - 1

VOLUME 13 JOURNAL OF HYDROMETEOROLOGY OCTOBER 2012
The Hydrological Cycle in Three State-of-the-Art Reanalyses:

Intercomparison and Performance Analysis
CHRISTOF LORENZ
Karlsruhe Institute of Technology, Institute of Meteorology and Climate Research, Department of Atmospheric
Environmental Research (IMK-IFU), Garmisch-Partenkirchen, Germany
HARALD KUNSTMANN
Karlsruhe Institute of Technology, Institute of Meteorology and Climate Research, Department of Atmospheric
Environmental Research (IMK-IFU), Garmisch-Partenkirchen, and Institute of Geography,
Regional Climate and Hydrology, University of Augsburg, Germany
(Manuscript received 21 July 2011, in final form 12 March 2012)
ABSTRACT
The three state-of-the-art global atmospheric reanalysis models—namely, ECMWF Interim Re-Analysis
(ERA-Interim), Modern-Era Retrospective Analysis for Research and Applications (MERRA; NASA), and
Climate Forecast System Reanalysis (CFSR; NCEP)—are analyzed and compared with independent ob-
servations in the period between 1989 and 2006. Comparison of precipitation and temperature estimates from
the three models with gridded observations reveals large differences between the reanalyses and also of the
observation datasets. A major source of uncertainty in the observations is the spatial distribution and change
of the number of gauges over time. In South America, active measuring stations were reduced from 4267 to
390. The quality of precipitation estimates from the reanalyses strongly depends on the geographic location, as
there are significant differences especially in tropical regions. The closure of the water cycle in the three
reanalyses is analyzed by estimating long-term mean values for precipitation, evapotranspiration, surface
runoff, and moisture flux divergence. Major shortcomings in the moisture budgets of the datasets are mainly
due to inconsistencies of the net precipitation minus evaporation and evapotranspiration, respectively, (P 2
E) estimates over the oceans and landmasses. This imbalance largely originates from the assimilation of
radiance sounding data from the NOAA-15 satellite, which results in an unrealistic increase of oceanic P 2 E
in the MERRA and CFSR budgets. Overall, ERA-Interim shows both a comparatively reasonable closure of
the terrestrial and atmospheric water balance and a reasonable agreement with the observation datasets. The
limited performance of the three state-of-the-art reanalyses in reproducing the hydrological cycle, however,
puts the use of these models for climate trend analyses and long-term water budget studies into question.
1. Introduction To bring these various observations into a consistent

scheme, computation of the reanalysis models is performed
Global and regional atmospheric retrospective anal-
via state-of-the-art data assimilation methods like three-
ysis models (reanalyses) play a crucial role in today’s
or four-dimensional variational data assimilation (3DVAR
hydrological and hydrometeorological research. These
or 4DVAR) that constrain the observations with physi-
global atmospheric reanalyses aim at assimilating a large
cally reasonable time evolution and budget equations.
amount of historical observation data to provide a
These reanalyses can be used to analyze the global climate
physically consistent basis for the most important hydro-
system, atmosphere, and land surface processes on large to
logical, hydrometeorological, and atmospheric quantities.
continental scales and to understand exchange processes
between these different regimes. Global atmospheric
reanalyses also are often used as forcing data for regional
Corresponding author address: Christof Lorenz, Karlsruhe In- hydrological or hydrometeorological simulations, such as
stitute of Technology, Institute for Meteorology and Climate Re-
search, Kreuzeckbahnstraße 19, 82467 Garmisch-Partenkirchen, numerical weather predictions and regional climate sim-
Germany. ulations. Three of the most widely used reanalyses are the
E-mail: christof.lorenz@kit.edu European Centre for Medium-Range Weather Forecasts
DOI: 10.1175/JHM-D-11-088.1
Ó 2012 American Meteorological Society 1397

Unauthenticated | Downloaded 09/11/23 08:19 AM UTC
1398 JOURNAL OF HYDROMETEOROLOGY VOLUME 13
(ECMWF) Interim Re-Analysis (ERA-Interim), the as precipitation, temperature, and atmospheric water
Modern-Era Retrospective Analysis for Research and vapor—with observational datasets from the Global
Applications (MERRA) from the National Aeronautics Precipitation Climatology Centre (GPCC), the GPCP,
and Space Administration (NASA), and the Climate the Climate Prediction Center (CPC), the Climate Re-
Forecast System Reanalysis (CFSR) from the National search Unit (CRU), the University of Delaware (DEL),
Centers for Environmental Prediction (NCEP). and the Hamburg Ocean Atmosphere Parameters and
Reanalyses represent an approximation of the real Fluxes from Satellite Data (HOAPS). Differences in the
world. Because of the changing amount of assimilated total amount, spatial variability, and distribution of
observational data, different data assimilation methods, gauges of the gridded rainfall observations are analyzed
and different model equations and assumptions, results in order to estimate the uncertainties incorporated in
of reanalysis models deviate significantly, even if they these datasets. Special emphasis is devoted to the com-
should be similar in principle. Therefore, it is necessary parison of precipitation estimates from the reanalyses
to validate these global atmospheric models with ob- because of their importance in the hydrological cycle.
servational datasets. In addition, the closure of the water budgets in the
Such comparisons were made by, for example, Janowiak three reanalyses are analyzed and it will be estimated
et al. (1997), Poccard et al. (2000), or Higgins et al. (2010), how well the transport processes between the oceans
with rainfall estimates from the CFSR and its predecessor, and the continents as well as moisture exchange between
the NCEP– National Center for Atmospheric Research the land surface and the atmosphere are balanced. For
(NCAR) reanalysis, being validated against precipitation this purpose, long-term mean values of precipitation,
observations. Bosilovich et al. (2008) compared precipi- evapotranspiration, surface runoff, and atmospheric mois-
tation from the 40-yr ECMWF Re-Analysis (ERA-40), ture flux divergences are computed. As evapotranspiration
the two older NCEP reanalyses (which are often referred and surface runoff are the dominating quantities of mois-
to as NR1 and NR2), and the Japanese 25-yr Reanalysis ture transport from the surface back into the atmosphere
(JRA-25) with data from the Global Precipitation Cli- and oceans, respectively, the estimates from the re-
matology Project (GPCP) and the widely used Climate analyses are used to investigate how well the water
Prediction Center Merged Analysis of Precipitation budgets in the models are closed.
(CMAP) on both the continents and the oceans. In
Hagemann et al. (2005), different quantities contribut- 2. Data and methods
ing to the global hydrological cycle of ERA-Interim’s
a. Reanalysis data
predecessor ERA-40 were analyzed in detail, while
Chido and Haimberger (2009) or Mueller et al. (2010) For comparison, three different global atmospheric
investigated the closure of water and energy budgets in retrospective analyses are used—namely, ERA-Interim
the ERA-Interim reanalysis. A more detailed compar- from ECMWF (Simmons et al. 2006; Berrisford et al.
ison is given in, for example, Trenberth et al. (2007), 2009), MERRA from the NASA Goddard Space Flight
where estimates of the most important quantities of the Center (GSFC) [National Oceanic and Atmospheric
global water cycle are presented. On regional scales, Administration (NOAA)] (Rienecker et al. 2011), and
Yeh and Famiglietti (2008) concentrated on the esti- CFSR (Saha et al. 2010) from NOAA/NCEP. The two
mation of evapotranspiration. Considerations relating latter reanalyses cover the satellite period from 1979 to
to the hydrological cycle over the United States were the present, while ERA-Interim was intended to cover
presented by Roads et al. (1994). Seneviratne et al. (2004) the period from 1989 to the present to provide a bridge
analyzed the water budget closure over the Mississippi between ECMWF’s previous reanalysis ERA-40 (Uppala
basin and presented estimates of monthly water storage et al. 2005) and a forthcoming next-generation reanalysis.
variations based on water vapor flux convergence, at- Recently, the ERA-Interim archive was extended to
mospheric water vapor content, and river runoff. Similar cover the years between 1979 and 1989 as well. The CFSR
work was performed by Betts et al. (1999, 2003, 2005, dataset succeeds the widely used NCEP–NCAR rean-
2009), who analyzed energy and mass budgets of ERA-40 alysis (Kalnay et al. 1996). The novelties of this reanalysis
and ERA-Interim over several river basins (especially in are the coupling to the ocean during the generation of the
North America). An assessment of the applicability of the 6-h guess field, an interactive sea ice model, and the as-
ERA-40 model for the detection of climate trends was similation of satellite radiances for the entire period
made by Bengtsson et al. (2004). (Saha et al. 2010). Furthermore, the analysis system used
In this study, the three state-of-the-art reanalyses in CFSR for the atmosphere, the Gridpoint Statistical
ERA-Interim, MERRA, and CFSR are compared. The Interpolation (GSI) scheme, is nearly the same as the
reanalyses are evaluated by comparing quantities—such one used by MERRA at NASA GSFC. The MERRA

OCTOBER 2012 LORENZ AND KUNSTMANN 1399
TABLE 1. Summary of the three reanalyses.
Interim MERRA CFSR

Horizontal resolution T255 (;78 km) 1/ 2
83 8 2/ 3 T382 (;38 km)
Vertical levels 60 72 64
Top level 0.1 hPa 0.01 hPa 0.26 hPa
Period 1979–present 1979–present 1979–2009
Output times 6 h, daily, monthly 1 h, 6 h, daily, monthly 1 h, 6 h, monthly
Assimilation scheme 4DVAR 3DVAR 3DVAR
atmosphere-only reanalysis is being conducted over the estimates of oceanic precipitation, evaporation, and
same years with nearly the same input data (Saha et al. other atmospheric variables.
2010). However, observation processing, model equa- The continental precipitation and temperature data-
tions, and the main scope of the reanalyses differ signifi- sets contain at least daily (CPC) or monthly (GPCC,
cantly. The resulting differences in modeled variables CRU, and DEL) means at a spatial resolution of 0.58 3
thus reveal uncertainty ranges of present-day reanalysis 0.58 for the whole world (see Table 2 for further details
models (see Table 1 for further details of these datasets). of the gridded observation products).
According to Kalnay et al. (1996) or Kistler et al. (2001), In principle, the different datasets should provide simi-
gridded variables from reanalyses can be separated into lar precipitation and temperature values. Differences of
three classes, which vary by the influence of assimilated global fields must be considered as uncertainty ranges,
observations on the variable. The type A variables (e.g., which can be expected when using such datasets for vali-
upper-air temperatures or horizontal winds) are strongly dation purposes. To generate gridded observations from in
influenced by the observations, and are thus assumed to situ measurements, the different data centers apply similar
be the most reliable variables. Type B variables (e.g., interpolation algorithms and may therefore exhibit similar
surface and 2-m temperatures) are influenced by both the biases (particularly in areas with complex terrain).
observations and the model while type C variables (e.g., Two main error sources lead to uncertainties in pre-
precipitation or surface runoff) are derived solely from cipitation observations. The sampling error, which is due
the model. to the irregular distribution of gauges, has a magnitude
of about 67%–40% of the true area-mean precipitation
b. Gridded observation data
(Schneider et al. 2008). Rudolf and Rubel (2005) report
To validate the three different reanalyses, we com- that sampling errors between 15% and 100% can be
pare precipitation and temperature estimates from the expected for sparsely gauged regions (less than 3 gauges
reanalyses with gridded observations from GPCC per 2.58 3 2.58 grid cell). The second error is due to the
(Rudolf and Schneider 2005), GPCP (Adler et al. 2003), undercatch of precipitation gauges, which results from
CRU (Mitchell and Jones 2005), the Unified Gauge- wind-field deformation above the gauge orifice, losses
based Analysis of Global Daily Precipitation from the from wetting on internal walls of the collector and in the
CPC (Chen et al. 2008), and DEL (Matsuura and Will- container, and losses due to evaporation from the con-
mott 2009). For validation of the atmospheric water tainer (Rudolf and Rubel 2005). The gauge undercatch
vapor over the oceans, data from the HOAPS product error might be large especially during winter in the high-
(Andersson et al. 2010) are used, which is based on latitude regions or over mountain ranges, as there will be
satellite observations from the Special Sensor Micro- a high amount of solid precipitation. This leads to an
wave Imager (SSM/I) on satellites of the Defense Me- underestimation of the true precipitation of up to 50%.
teorological Satellites Program and provides reliable Since 2007, GPCC has been providing event-based
TABLE 2. Summary of the observation datasets containing precipitation P, near-surface temperature T2, and the atmospheric
water vapor content W.
Variables Horizontal resolution Period Output times Version number

GPCC P 0.58 3 0.58 1901–2009 Monthly 4.0
GPCP P 2.58 3 2.58 1979–2009 Monthly 2.1
CRU P, T2 0.58 3 0.58 1901–2009 Monthly 3.0
CPC P 0.58 3 0.58 1979–present Daily 1.0
DEL P, T2 0.58 3 0.58 1900–2008 Monthly 2.01
HOAPS P, E, W 0.58 3 0.58 1987–2005 Daily, five daily, monthly 3

FIG. 1. Land–sea mask used for computing the spatial averages over North America, South America, Europe, Africa,
Asia, and Australia. The table shows the areas of the regions considered within each continent.
correction factors (Fuchs et al. 2001; Schneider et al. d. The global water balance
2008) to account for the systematic gauge undercatch er-
The terrestrial large-scale water balance (mm month21)
ror. Before 2007, the corrections consisted of monthly
can be written as
climatologies as proposed in Legates and Willmott (1990),
which are still applied to the GPCP precipitation product. dS
5 P 2 E 2 R, (1)
The original GPCC full data product used for this study dt
does not include such corrections (A. Becker 2011,
personal communication). where dS/dt is the change in the terrestrial water storage,
In the course of this study, the GPCC precipitation P is precipitation, E is evapotranspiration, and R is total
product was updated from version 4.0 to 5.0. Even though runoff—that is, the sum of surface runoff and subsurface
the new dataset is based on a denser station network, the runoff (Willmott et al. 1985; Peixoto and Oort 1992).
differences of area-averaged values or long-term mean According to Yeh and Famiglietti (2008), the change
fields are not significant (not shown here). Therefore, the in the total terrestrial water storage depends on its sur-
GPCC v4.0 product was used for reference observations face component, the soil moisture, and the groundwater
in this study, but the differences in the distribution and components. The total soil water depends largely on the
total number of gauges between version 4.0 and 5.0 are characteristics of the land surface model in the rean-
discussed briefly (see section 3a). alyses. Interim and CFSR provide soil moisture values
divided into multiple levels. MERRA does not consider
c. Area averaging of gridded data a multilevel soil model for computing the interaction
For the validation of the reanalyses’ rainfall estimates between the land surface and the atmosphere. The water
with the observation datasets, all fields were remapped storage term dS/dt can thus be computed as a residual
to the resolution of the GPCC dataset (i.e., 0.58 3 0.58) term like in, for example, Roads et al. (1994). When an-
using a first-order conservative interpolation (Jones alyzing water budgets in atmospheric reanalysis models, it
1999). From these fields, area-weighted averages were is convenient to further consider a term for the analysis
computed over different regions using the continental increment—that is, the increment that is due to the forcing
mask shown in Fig. 1. As GPCC only contains gauge- of the models toward the observations. The terrestrial
based observations, the oceans or the poles were not water balance equation is then modified to
considered for comparison of the precipitation fields. dS
Consequently, the global and hemispheric averages do 5 P 2 E 2 R 1 RES9s , (2)
dt
only represent the rainfall over land. For investigating the
water budget closure, a correct differentiation between where RES9s is accounting for surface residual water
the processes over land and the oceans is crucial. We did forcing. As further proposed in Roads et al. (2002) and
not perform any additional interpolation for this analysis, Szeto et al. (2008), the water storage tendency term
but used the fields in the models’ native resolutions. The is combined with residual forcing—that is, RESs 5
area-averaged values over the continents and oceans RES9s 2 dS/dt. According to Kleidon and Schymanski
were calculated using the land–sea masks from the three (2008) or Seneviratne et al. (2004), it can be assumed
reanalyses. For the evaluation of the oceanic water cycle that for climatic time scales the total soil water content
components, we used a dynamic land–sea mask, as the remains constant and its tendency can be neglected. This
satellite observations from HOAPS are available over assumption results in the simplified terrestrial water
ice-free ocean only. balance equation of

R ’ P 2 E, (3) moisture flux divergences are directly linked to the

vertical exchange terms of the terrestrial water balance:
where overbars indicate averaging over a climatic time
scale. The imbalance of this equation provides infor- $ Q 5 E 2 P. (8)
mation on the magnitude of the disturbances of the water
As in case of the surface water balance, misclosure of the
budgets introduced by surface water forcing RES9s .
equation when using long-term averages is an estimate
Apart from the continental water budget, exchange of
of the magnitude of the analysis increment of the at-
water between the continents and oceans is balanced as
mospheric water forcing RES9w . Equation (1) can be
well. Over multiyear averages, the global water budget
combined with Eq. (5) to obtain another linked balance
should be closed—that is, the convergence (P 2 E) of
equation:
moisture over land should equal the divergence (E 2 P) of
moisture over the ocean (Hagemann et al. 2005); that is, dS dW
52 2 $ Q 2 R. (9)
dt dt
Pland 2 Eland 5 Eocean 2 Pocean . (4)
As the atmospheric and terrestrial tendency terms can
In general, this value must be positive over the conti- be assumed to be negligible over longer time scales, the
nents because of a surplus of precipitation, while there is atmospheric net input of moisture in a certain area must
more evaporation over the oceans. be balanced by a terrestrial net outflow at the surface;
These terrestrial budgets can be linked to the atmo- that is, R ’ 2$ Q (Roads et al. 1994). The imbalance
sphere by the atmospheric water balance equation of this equation is a rough estimate of the total atmo-
spheric and surface water analysis increments.
dW
1 $ Q 5 E 2 P, (5)
dt e. Computation of spatial correlations
where E and P are actual gridpoint evapotranspiration To analyze the agreement of spatial patterns between
(or evaporation over the oceans) and precipitation at the two datasets, spatial correlations are computed for fur-
surface (Roads et al. 1994). Here W denotes the total ther analysis. This yields information about the extent to
column water content in the atmosphere and $ Q is the which certain events (e.g., large-scale rainfall) agree in
net balance of moisture flux (i.e., moisture flux di- terms of location, dimension, and magnitude when using
vergence), which is defined as various datasets. We compute the spatial correlations
ðp between two datasets x and y according to
1 sfc
$Q [ $ vh (p)q(p) dp, (6) nx
g p50 1 T 1 (xi,t,x 2 xt,x )(yi,t,x 2 yt,x )
corrx 5 å å
T t51 nx 2 1 i51 sx,t,x sy,t,x
,
with air pressure p (Pa), the gravitational acceleration
(10)
g (m s22), the horizontal wind vector vh (m s21), and the
specific humidity q (kg kg21). When computed from
where n is the number of grid points of a given area x,
reanalyses, moisture flux divergences are based on type
xi,t,x and yi,t,x are the actual gridpoint values at the time t,
A and type B variables only, precipitation, evapora-
the overbar denotes the spatial mean value of the area,
tion, and evapotranspiration are type C variables.
and sx,t,x and sy,t,x are the standard deviations of the two
Again, it is convenient to add a term accounting for the
datasets x and y of the area x at the time t. Here T is the
analysis increment to the atmospheric water balance
number of time steps contributing to a temporal subset
equation:
like, for example, all Januaries of the considered time
dW series or all months of a specific year.
1 $ Q 5 E 2 P 1 RES9w . (7) Apart from time series of spatial correlations, we use
dt
the Taylor diagrams (Taylor 2001) to analyze the level of
The atmospheric tendency term dW/dt can be combined agreement of rainfall patterns from different data sour-
with the residual forcing; that is, RESw 5 RES9w 2 dW/dt ces. In this case, the standard deviation (the radial
(Roads et al. 2002; Szeto et al. 2008). On annual or longer distance of a data point from the origin) is a measure of
time scales, the variations of the atmospheric water the intensity and variability of the patterns, while the
storage W are often assumed to be negligible (Peixoto correlations (the angle between the x axis and a data
and Oort 1992). For monthly time scales, this assump- point) reflect how well the analyzed datasets repro-
tion does not hold, however. The vertically integrated duce the rainfall patterns from a reference dataset.

The root-mean-square difference (radial distance be-

tween the reference data point and another data point)
is a measure of the average pixelwise differences be-
tween two datasets and computed from the standard
deviations and the correlations.
f. Computation of CFSR evapotranspiration
In contrast to Interim and MERRA, CFSR does not
provide fields of total evapotranspiration. Therefore,
these fields are computed from latent heat flux, which is FIG. 2. Relative difference between CFSR evaporation (over the
given in energy flux form (i.e., in units of W m22). The oceans) and evapotranspiration (over the continents), with and
transformation into units of mm (i.e., mass flux form) without considering changes in the near-surface temperature. The
impact is generally higher over the continents, while over the oceans,
was performed via the amplitude of the intra-annual variations is larger.
lE
E5 , (11) can be neglected on these spatial and temporal scales.
Le
For small-scale studies and especially during summer
where E is evapotranspiration (mm), lE is the latent months, however, the influence might be significant and
heat flux (W m22), and Le is the latent heat of evapo- should be taken into account.
ration (J kg21), which can be approximated through
3. Results
Le ’ 2:501 3 106 2 2370Tc , (12)
a. Distribution of gauges in the observation datasets
with Tc being the near-surface temperature in degrees To validate the global atmospheric reanalyses, rainfall
Celsius (e.g., Jacobson 2007). The latent heat flux fields observations, interpolated to a regular grid as described
from CFSR include both evaporative flux from liquid in, for example, Chen et al. (2008) and Rudolf and
and snow sublimation from snow surface. Equation (11) Schneider (2005), are used. As a matter of fact, the
must consequently be corrected for sublimation: quality of these gridded precipitation fields depends pri-
marily on the number of active gauges and their spatial
lE 2 lS lS distribution. The interpretation of interpolated gridded
E5 1 , (13)
Le Ls observations in regions with a few gauges only or a dis-
advantageous spatial distribution of such observation
where lS is sublimation and Ls is the latent heat of stations remains open.
sublimation, which is the sum of the latent heats of Figure 3 shows the number of gauges per grid cell at
evaporation and melting (e.g., Jacobson 2007): the beginning and the end of the considered time series
for GPCC v4.0, GPCC v5.0, and CPC. In 1989 (Figs.
Ls ’ 2:501 3 106 1 3:35 3 105 2 Tc (340 1 10:46Tc ). 3a,c,e), a dense network of observation stations existed
(14) over North America, central Europe, coastal regions of
Australia, and the eastern part of Brazil. The GPCC
According to CFSR, the influence of the temperature products also exhibit a good spatial coverage of South
can be neglected (R. Yang 2011, personal communica- Africa, while only few gauges are located in the rain-
tion). Our computations support this assumption. On laden regions of tropical Africa, South America, and
monthly time scales, the temperature causes an increase Southeast Asia and large parts of the subtropics, Eurasia,
of the continental evapotranspiration of about 2.1% and high-latitude regions. Depending on the geographic
with a maximum during the summer months (Fig. 2). location of these ungauged regions, an interpolation
Over the oceans, the impact is smaller (about 1.7%), but might introduce large uncertainties. This is particularly
has a strong semiannual signal with its maxima in the true if the complex cycle of tropical precipitation or the
summer and winter months. By considering the magni- high spatial variability of rainfall over mountain ranges is
tudes of other major water cycle components, the considered. Figures 3b,d,f shows the amount of gauges
functional dependencies of Le and Lc on temperatures per grid cell in December 2006. Spatial coverage with
are found to exert little impact on the calculations of observation stations has changed drastically, especially
evaporation and evapotranspiration, respectively, from for the GPCC v4.0 data in North America, South
latent heat flux. Thus, the corrections for temperature America, and Africa. Large parts of the tropics and

FIG. 3. Number of monitoring stations per 0.58 3 0.58 grid cell in (a),(c),(e) January 1989 and (b),(d),(f) December
2006 for the (a),(b) GPCC v4.0; (c),(d) GPCC v5.0; and (e),(f) CPC datasets. A good spatial coverage with obser-
vation stations can be observed over North America (GPCC v5.0 and CPC) and Europe (GPCC v4.0 and v5.0), while
the number of gauges over North America is significantly reduced in GPCC v4.0. Over most of the tropical regions
like the Congo or Amazon basin, high-latitude regions, and large parts of Asia, the three datasets use a maximum of
1–2 gauges per grid cell, whereas some areas are completely ungauged.
deserts remain completely ungauged over hundreds of modeled precipitation fields. On the other hand, the
kilometers in both GPCC and CPC datasets. The update reliability of the observation datasets remains ques-
from version 4.0 to 5.0 of the GPCC product significantly tionable especially over the tropics, deserts, mountain
improved spatial coverage of North America and Aus- ranges, and large parts of the Asian continent because of
tralia, while there is only little improvement over South the decreasing number of active gauges and their sparse
America, central Africa, or large parts of Eurasia. As is spatial distribution.
obvious from Fig. 4, the number of gauges decreased
significantly for all three observation datasets over most b. Precipitation
of the regions. At the end of the period studied, only
1) LONG-TERM MEAN ANNUAL PRECIPITATION
1314 (CPC), 390 (GPCC v4.0), and 555 (GPCC v5.0)
gauges remain, which are used to compute the pre- The long-term mean annual rainfalls (Fig. 5) obtained
cipitation fields over South America. Although the de- from the different reanalyses are in general agreement
crease in active gauges is not that significant over the with the observational references when looking at the
Asian continent, comparison of the numbers of gauges spatial precipitation patterns. The large-scale rain-laden
over Europe and Asia again illustrates the very sparse regions of the tropics in South and Central America,
distribution of gauges in the latter regions. In contrast to central Africa, and Southeast Asia show precipitation
this, the CPC dataset is based on about 10 000 gauges rates of up to 11 mm day21 in all products. These moist
over North America in the beginning and the end of the regions are clearly separated from the large subtropic
time period, while in between, the number of gauges desert regions with very limited precipitation. In addi-
increases up to ;17 000 in 2003. This shows that there tion, good agreement in the precipitation patterns can
are certain regions where the gridded GPCC and CPC be found, for example, in Australia and over the moist
products are based on a dense network of gauges and, regions in the southeastern part of North America
hence, provide a scientifically sound basis to validate and the drier Great Plains. Large mountain ranges,

In general, the highest differences in spatial variability

and the amount of rainfall can be found over tropical
South America, central Africa, Southeast Asia, and the
large mountain ranges of the Andes and the Himalayas.
These differences cause large-scale deviation patterns,
which can reach magnitudes of up to 64 mm day21
(Fig. 6). Even when focusing on the ensemble of the ob-
servation datasets GPCC, GPCP, CRU, DEL, and CPC,
differences of up to 3 mm day21 result in central Africa
(Fig. 7a), which are likely introduced by uncertainties in
the observations due to the sparse distribution of gauges
in these regions. In the three reanalyses, the differences
in spatial variability and the amount of precipitation are
even larger compared to the observations. The mid- to
high-latitude rainfall estimates by CFSR appear to be
significantly biased, as there are deviations of up to
2 mm day21 (Fig. 6e). Higgins et al. (2010) investigated
the reliability of CFSR precipitation over North America
and concluded that parts of this bias can be explained by
an overactive diurnal cycle in the atmospheric com-
ponent of CFSR. The observation datasets are based
on a dense network of gauges and show only small
deviations in these regions. It can thus be assumed that
there are some inaccuracies in the CFSR estimates.
A significant discrepancy exists between the pre-
cipitation patterns from GPCC, Interim, and MERRA
in South America and central Africa (Figs. 6c,d). These
differences were also noted in Trenberth et al. (2011).
FIG. 4. (a)–(f) Mean annual number of gauges used in the pre- Interim overestimates rainfall over the Andes and
cipitation observations from GPCC v4.0, GPCC v5.0, and CPC. central Africa up to 2.5 mm day21, while MERRA
Version 4.0 shows a significant drop in the number of gauges between shows a large-scale underestimation in central South
2000 and 2001 over North America and Australia, while version 5.0 of
America and central Africa and an overestimation over
the GPCC product is based on a nearly constant number of obser-
vation stations during the complete time series. Over South America, coastal regions. It is well known that tropical pre-
Europe, and Africa, the update from v4.0 to v5.0 results in little im- cipitation in the MERRA reanalysis over South
provement only, as both versions show a nearly constant decline over America has its shortcomings. Therefore, a corrected
time. Over Asia, GPCC v5.0 is using about 1000 gauges more than dataset for land hydrology will be released in the near
version 4.0 until 2000. The CPC product is based on about 1000 gauges
future (Reichle et al. 2011). CFSR indicates conditions
over Europe and 500 gauges over the whole of Africa, while more
than 14 000 gauges are used to generate the gridded precipitation that are too moist in the center and underestimates
observations over North America between 1991 and 2003. rainfall east of the Congo basin along the course of the
Nile (Fig. 6e). Poccard et al. (2000) and Sylla et al.
(2010) discuss that rainfall simulations in these re-
especially the Andes, the Alps, and the Himalayas, can gions is a very complex task and might lead to large
be identified because of their wet conditions compared discrepancies. As the largest part of precipitable wa-
to the surrounding regions. Except for MERRA, all ter in central Africa arises from evapotranspirating
datasets show a maximum in annual precipitation at the water in the tropical rain forests (Van der Ent et al.
headwaters of the Amazon River, which extends along 2010), these deviations might be due to shortcomings in
the course of the river down to the Atlantic. In the the models’ land–atmosphere interactions in this com-
MERRA dataset, this maximum is shifted eastward. In plex environment. On the other hand, the number of
the regions between southern Brazil and the southern active gauges (Fig. 3d) shows that their spatial density
foothills of the Andes, GPCC, CRU, CPC, Interim, and decreased significantly during the period considered.
CFSR show a mean precipitation rate of about 4– This means that uncertainty is up to 63 mm day21 in
5 mm day21, contrary to a distinguished dry region with these regions because of the variability of the ensemble
less than 2 mm day21 predicted by MERRA. of observations (Fig. 7a). Only the Interim precipitation

FIG. 5. Long-term mean annual precipitation between 1989 and 2006 (mm day21). The three observation datasets
(a) GPCC, (b) CRU, and (c) CPC are in good agreement over most of the regions, even if CPC assumes less rainfall
over central Africa. The precipitation estimates from the three reanalyses (d) Interim, (e) MERRA, and (f) CFSR
show similar large-scale patterns, while significant differences exist in the spatial distribution and the amount of
rainfall on smaller scales.
exceeds the uncertainty given by the observations over of gauges used for generating the observation datasets
a large area. The other two reanalyses are within the and the amount of observations assimilated in the re-
bounds given by the observations and are therefore as- analyses changed significantly between 1989 and 2006,
sumed to be more realistic. there is only a minor impact on the agreement with
Precipitation over the Andes is generally overestimated GPCC on these scales. Over the Southern Hemisphere
in the reanalyses, while all datasets show less rainfall and the tropics, the spatial correlations (Figs. 8c,d) ex-
over the Himalayas than to GPCC. This might be due to hibit a wider range between the datasets. It is difficult
the impact of orography on convective events caused by to determine, however, whether this range is due to
the differences in resolution and the description of the the reduction of gauges or changes in the assimilated
underlying terrain model. On the other hand, the high observations.
spatial variability of precipitation in mountain ranges The CFSR series show a general wet bias over the
aggravates reliable areawide observations. Because of the Northern Hemisphere and a drop in the global conti-
sparse distribution of gauges and the errors caused by the nental rainfall toward GPCC from 1998 (Figs. 9a,b).
undercatch of solid precipitation, the quality of in- Interestingly, the differences between CFSR and GPCC
terpolated rainfall values from GPCC, CPC, and CRU increase again after 2000 over both the tropical regions
remains questionable in these regions. between 158N and 158S and the Southern Hemisphere
(Figs. 9c,d), while the bias over the Northern Hemi-
2) TIME EVOLUTION OF GLOBAL, HEMISPHERIC, sphere is slightly reduced. The decline between 1998 and
AND TROPIC PRECIPITATION 2000 is also evident from the Interim rainfall over the
tropics and the Southern Hemisphere but not in MERRA.
The global and Northern Hemispheric correlations During these years, all three reanalyses and the CRU
and differences (Figs. 8a,b and 9a,b) of the four obser- observations show a sudden increase in the near-surface
vation datasets and the Interim and MERRA reanalyses temperature (not shown), which might be related to the
are relatively constant over time. Although the number gaps in the precipitation estimates.

FIG. 6. Absolute differences of the mean annual precipitation (mm day21) from 1989 to 2006 between GPCC and
(a) CRU, (b) CPC, (c) Interim, (d) MERRA, and (e) CFSR. CRU shows a good agreement with GPCC. CPC is drier
over the Congo basin, the Himalayas, and the northern part of the Andes. The largest differences between GPCC and
the three reanalyses can be observed over the tropics and the mountain ranges. CFSR also has a wet bias over mid- to
high northern latitudes, while MERRA shows a dry pattern that extends over large parts of South America.
The CPC observations show a dry bias of about Hemispheric Interim precipitation rates are higher
20.3 mm day21 over the Northern and 20.4 mm day21 throughout the year with a distinct peak during the pe-
over the Southern Hemisphere in relation to GPCC riod from September to December where deviations
(Figs. 9b,d). In general, precipitation over the Southern from GPCC of up to 0.75 mm day21 can be observed.
Hemisphere and the tropics has a higher variability in Thus, on the global scale, Interim assumes slightly higher
spatial correlations and deviations relating to GPCC. In precipitation rates than GPCC, with the largest differ-
contrast to the reanalyses, the observation datasets are ences occurring in the periods from March to May and
in good agreement, even if the deviations from GPCC from September to December—that is, in the Northern
are larger because of the higher precipitation rates in and Southern Hemispheric spring months (Figs. 10a–d).
these regions. Spatial variability of rainfall observations MERRA does not exhibit a clear annual cycle over the
over the Southern Hemisphere correlates with a corre- Northern or Southern Hemisphere; deviations in the
lation coefficient of at least 0.8 and, hence, is a reliable tropics are maximal during the period from November to
basis for validating the reanalyses on large spatial scales. April. The intra-annual spatial correlations between
It can also be noticed that the reanalyses’ spatial cor- MERRA and GPCC show a clear annual cycle especially
relations over the Southern Hemisphere are dominated over the Southern Hemisphere, which is mainly domi-
by the variations between 158N and 158S. Especially nated by variations between 158N and 158S (Figs. 11c,d).
after 1995, the MERRA dataset depicts a quasi-periodic CFSR and Interim show a similar annual cycle with
signal in the spatial correlations over the Southern a generally higher correlation except for the period from
Hemisphere and the tropics (Figs. 8c,d). September to November over the Southern Hemisphere,
The global intra-annual differences between the CFSR where Interim agrees better with GPCC than CFSR.
rainfall estimates and the GPCC observations have an Over the Northern Hemisphere, the reanalyses are in
annual cycle with maximal deviations in the period good agreement with an average correlation coefficient
from March to June (Fig. 10a). Interim tends to slightly of about 0.8 (Fig. 11b).
overestimate the GPCC rainfall over the Northern The Taylor diagrams in Fig. 12 show that on a global
Hemisphere with largest deviations of about 0.3 mm day21 scale, all three reanalyses and the observations reproduce
occurring from March to May. The tropical and Southern the spatial rainfall patterns from GPCC with a correlation

FIG. 7. (a),(c) Variability of rainfall P (mm day21) and (b),(d) range of temperature T2 (8C) of (a),(b) the ensemble
of gridded observations and (c),(d) the ensemble of reanalyses. The ensemble of rainfall observations is generated
from GPCC, GPCP, DEL, CRU, and CPC, while the temperature range is based on DEL and CRU. The reanalysis
ensemble consists of Interim, MERRA, and CFSR for both precipitation and temperature. The reanalyses generally
produce a larger variability especially over the tropics and the whole of South America. Over the Congo basin,
however, the precipitation variability of the observation ensemble reaches values of up to 3 mm day21. The tem-
perature range from the three reanalyses shows the largest values over South America, the Congo basin, the Sahara,
and Greenland, where differences of more than 58C can be observed.
coefficient of .0.7. The statistics over the Northern or January. On the other hand, the differences in total pre-
Southern Hemisphere and the tropics indicate that the cipitation between MERRA and GPCC (Fig. 10h) show
level of agreement between GPCC and the other datasets a reduced annual cycle compared to the correlations. This
decreases when the area of interest is reduced. It is also indicates that intra-annual variations in the amount of
evident that CFSR predicts a too-high spatial variability precipitation are in good agreement with the observa-
compared to GPCC during the summer months of the tions, while there are major differences in spatial vari-
Northern and Southern Hemisphere, which is indicated ability. The Taylor diagrams (Fig. 12) confirm the
by higher RMSD values. MERRA agrees best with problems of the MERRA dataset in this respect, which
GPCC during the boreal summer. Over the Southern cannot be explained by outliers exclusively. MERRA’s
Hemisphere and the tropics, MERRA shows the lowest annual mean correlations and deviations (Figs. 8h and
correlation coefficients (,0.6 for some years) of the 9h) converge toward GPCC and the other datasets over
three reanalyses. The performance of the models in re- South America, leading to better precipitation esti-
producing the GPCC rainfall patterns changes signifi- mates at the end of the time series. The time when the
cantly depending on the region and time (month) MERRA precipitation estimates improve coincides
but even from year to year. This is also true for the with the assimilation of observations from the Ad-
observation datasets although the other gridded rainfall vanced Microwave Sounding Unit (AMSU) on the
observations on these scales agree better with GPCC NOAA-15 satellite. This assimilation is performed only
than the reanalyses. over the oceans, but Bosilovich et al. (2011) note that
such satellite epoch changes might indirectly affect the
3) TIME EVOLUTION OF CONTINENTAL-SCALE
MERRA water balances over land through altered
PRECIPITATION
moisture.
Over South America, the correlations of Interim and Over North America, Europe, and Asia, significant
CFSR are in good agreement with an average correlation wet biases in the mean annual and intra-annual CFSR
coefficient of about 0.7, while MERRA predicts com- precipitation are found (Figs. 9e,f,g and 10e,f,g). While
pletely different rainfall patterns, resulting in a low spa- spatial correlations decrease during the period from
tial correlation coefficient of ;0.5 (Fig. 8h). As regards May to August over North America, they are in good
the intra-annual spatial variability (Fig. 11h), the lowest agreement with the other reanalyses (Figs. 8e,f,g and
correlations of MERRA are found between October and 11e,f,g). Over Europe, the Interim reanalysis matches

FIG. 8. (a)–(j) Area-averaged spatial correlations in mean annual precipitation of GPCP, CRU, CPC, Interim, MERRA, and CFSR in
relation to GPCC for the various regions. Values close to 1 indicate that the precipitation patterns from the respective dataset are in good
agreement with the spatial distribution of rainfall from GPCC. In most of the regions, all datasets reproduce the spatial rainfall patterns
from GPCC with a correlation coefficient .0.7 between 1989 and 2006. The largest deviations can be observed over South America, where
especially MERRA shows correlation coefficients ,0.6 (until 1998) and ;0.6 (from 1998). Compared to the reanalyses, the agreement
between GPCC and the observation datasets is generally better. Over most of the regions, Interim shows the highest correlation co-
efficients in relation to GPCC.
well with the GPCC observations with an average with the lowest correlation coefficients occurring in
spatial correlation coefficient .0.8 and a deviation of May (Europe; Fig. 11f) and from July to August (North
less than 0.1 mm day21 on interannual and intra-annual America; Fig. 11e).
time scales. The Taylor diagrams representing July over Over Asia, South America, and Africa, the differences
North America and Europe reveal that some data points between CFSR and GPCC decrease significantly, which
predict a correlation coefficient ,0.7, which is likely due might be explained by the assimilation of AMSU data
to an increase of convective precipitation. This is con- (Figs. 9g,h,i). After 1998, the wet bias of CFSR over Asia is
firmed by the reanalyses’ intra-annual spatial correlations, constantly reduced to 0.6 mm day21 while bias reduction

FIG. 9. Same as Fig. 8 but for area-averaged differences in the mean annual precipitation (mm day21). On a global scale, MERRA
agrees best with GPCC, but shows significant deviations especially over the Southern Hemisphere (likely due to the large differences over
South America). Good agreement between GPCC and MERRA can also be observed over North America, while over Europe, Interim
performs best. Over Australia, MERRA and CFSR are of comparable agreement with the observations, but the differences between
GPCC and Interim increase toward the end of the time series. In general, the observations from CPC show large deviations from GPCC
over South America and Africa due to CPC’s drier conditions in these regions. GPCP has a slight wet bias especially over the northern
hemispheric regions (North America, Europe, and Asia), which might be due to corrections for the gauge undercatch error.
over the other continents is only temporary, as the differ- 11e). Over Europe, the precipitation estimates from
ences between CFSR and GPCC increase toward the end Interim are superior to the other two reanalyses (Figs.
of the time period. 8f, 9f, 10f, and 11f), while Interim has a wet bias over
Over North America, the MERRA precipitation es- Africa because of the overestimation of precipitation
timates show the smallest deviations from the GPCC in the Congo basin (Fig. 6c). A general dry bias of the
observations on both interannual and intra-annual time CPC observations can be noticed over all regions except
scales even though the slightly biased Interim estimates for North America and Australia. It is mentioned in the
tend to display higher spatial correlations (Figs. 8e and dataset description that especially over large parts of

FIG. 10. Same as Fig. 8 but for long-term (17 yr) averaged differences of monthly precipitation (mm day21). Over most of the regions,
the differences between CFSR and GPCC show an intra-annual cycle that is obvious in South America and to a lesser extent over North
America, Europe, and Asia. Both Interim and MERRA show a good agreement with GPCC over North America, Europe, and Asia, while
over South America, Interim and MERRA assume too-moist and too-dry conditions, respectively, during an intra-annual cycle. A sig-
nificant dry bias between GPCC and Interim can be observed over Australia during the period from January to March, whereas the other
datasets are in good agreement with GPCC during the same period.
Africa and South America the observations should be GPCP and CPC, but to a smaller extent. As CRU and CPC
treated carefully, as there is a very sparse distribution of are based solely on gauge observations, the Australian
gauges, even if the spatial correlations are in good rainfall patterns from GPCC have to differ from the other
agreement with the other datasets. datasets. However, this difference had not yet been detected.
Over Australia, it can be seen that the three reanalyses
c. 2-m temperature
and the observations from CRU show similar spatial
correlations, which differ from GPCC especially in April The mean annual differences of the reference temper-
and November (Fig. 11j). These drops are also evident for atures given by the CRU dataset and the three reanalyses

FIG. 11. Same as Fig. 8 but for area-averaged spatial correlations of the long-term (17 yr) mean monthly rainfall. Values close to 1
indicate that the respective dataset is in good agreement with the spatial distribution of rainfall from the GPCC product. Globally, the
three reanalyses and observations reproduce the variations in the intra-annual rainfall patterns with a correlation coefficient .0.7 and
.0.8, respectively. A significant intra-annual cycle can be observed over North America and Europe, which has its lowest values during
the period from July to August (North America). Over South America, MERRA shows correlation coefficients ,0.5 during September–
February. A similar intra-annual cycle, but less pronounced, can be observed for Interim and CFSR with its minimum between October
and November. Over Australia, a significant drop in the spatial correlations can be observed in April and the period from October to
November. As this drop is evident in both the reanalyses and the observations, there might be some shortcomings in the GPCC pre-
cipitation patterns during these months.
are shown in Fig. 13. The patterns of larger temperature MERRA and CRU reaches values of up to 68C. This
differences are closely related to the differences in the warm bias may cause an increased saturation deficit of
precipitation fields (Fig. 6). The MERRA tempera- the air, which might explain the underestimation of
ture estimates (Fig. 13b) seem to have a warm bias South American precipitation in the MERRA dataset.
especially in South America, where the difference between A similar effect can be noticed over central Africa,

FIG. 12. Taylor plots of spatial statistics of the mean monthly precipitation in January and July for GPCP, CRU, CPC, Interim,
MERRA, and CFSR with respect to GPCC; each data point in a plot displays the correlation as the angle between the x axis and the data
point, the standard deviation (normalized) as the y coordinate, and the root-mean-square difference (normalized) as the radial distance of
one month of a specific year with respect to GPCC.
where MERRA predicts too-warm conditions and too- not in CRU. Over the largest parts of the continents,
little rainfall. The Interim field (Fig. 13a) shows a cold a general uncertainty of about 18C can be expected.
bias in central Africa and South America and, thus, When considering this value as an uncertainty bound,
a decreased saturation deficit, which results in larger the large-scale deviations of MERRA and Interim over
rainfall compared to the other datasets. The relation South America or the Congo basin and the general cold
between the temperature and precipitation biases might bias in the CFSR dataset over the whole Sahara indicate
also be explained by the reduced clouds and pre- significant inaccuracies in the reanalyses. On the other
cipitation in these regions, leading to excess solar radi- hand, only slight deviations are encountered over North
ation reaching the surface, which results in an increased America, Europe, and Australia. Overall, Interim shows
temperature. the best agreement with CRU.
In general, it can be concluded that the temperature
range (Fig. 7d) of the three reanalyses is similar to that d. Closure of the water budgets
of the precipitation fields (Fig. 7c). The widest range of
1) SURFACE WATER BUDGET
temperature observations (Fig. 7b) can be detected over
the large mountain ranges. This might be due to an el- Table 3 summarizes the computed long-term mean
evation correction performed in the DEL dataset, but values of the different quantities contributing to the

TABLE 3. Mean global water cycle components over land and

ocean between 1989 and 2006 (1015 kg yr25); the values in the
rightmost columns are the long-term estimates from Trenberth
et al. (2007; TB) and Oki and Shinjiro (2006; OKI) and are printed
here as a reference.
Interim MERRA CFSR TB OKI

Pland 119.7 115.4 127.4 113 111.0
Eland 82.1 86.7 70.4 73 65.5
Pocean 418.0 381.3 459.3 373 391.0
Eocean 455.3 411.5 470.0 413 436.5
R 46.6 31.3 38.3 40 45.5
(P 2 E)land 37.6 28.7 57.3 40 45.5
(P 2 E)ocean 237.4 230.2 210.7 240 245.5
(P 2 E)ocean, before 98 229.1 248.9 226.2 — —
(P 2 E)ocean, after 98 247.7 26.8 8.6 — —
$ Qland 32.3 38.2 34.6 40 45.5
$ Qocean 232.6 238.2 234.6 240 245.5
RESs, 9.0 2.6 19.0 — —
land
RESw, 5.3 29.5 22.7 — —
land
may be due to differences in the used land–sea mask or,

when compared to the estimates from, for example,
Bosilovich et al. (2011), a different time period.
On the other hand, CFSR leaves an imbalance of
about 80% of the continental surface water budget due
to an overestimation and underestimation of continental
and oceanic P 2 E values, respectively. It can be as-
sumed that the too-small oceanic P 2 E value of CFSR
FIG. 13. Differences in the annual mean temperatures (8C) at 2 m
between (a) Interim, (b) MERRA, and (c) CFSR and CRU. The mainly arises from an overestimation of rainfall, as both
largest differences between MERRA and CRU can be observed Interim and MERRA assume an evaporation surplus
over South America, where MERRA shows a warm bias .48C. of about 8% with respect to the water that precipitates
The deviation pattern agrees well with the differences in the long- over the oceans, while CFSR predicts only 2%. This is
term mean precipitation estimates, where MERRA revealed a dry
confirmed by the evaluation with the HOAPS dataset
bias in these regions. This is also true for Interim, which assumes
too-cold and too-wet conditions over the Congo basin. All three (Table 4), as CFSR predicts significantly more rainfall
reanalyses show a warm bias over Siberia, while large parts of than the other datasets. This might be due to the high
Greenland and the mountain ranges are generally too cold. moisture convergence in the oceanic domain of the in-
tertropical convergence zone (ITCZ) (Fig. 14c). There
global and continental-scale water budgets. The estimates are also patterns of large positive P 2 E values south
from Trenberth et al. (2007) and Oki and Shinjiro (2006) east and west of South America, which are absent in
are presented here as well for reference. In the long-term Interim and MERRA, assuming a significantly larger
mean, Interim and MERRA show a reasonable closure of depletion of water over the oceans.
the global surface water balance, as P 2 E over land According to Fig. 15a, there is a shift in the global P 2
equals the divergence of moisture E 2 P over the oceans. E moisture budgets of CFSR and MERRA in 1998. Both
Interim generally predicts more oceanic precipitation and depict a significant increase of oceanic P 2 E (Fig. 15c),
evaporation. Both datasets achieve a closure of the with CFSR reaching values of about 20 3 1015 kg yr21 in
combined continental–oceanic water budget [Eq. (4)] 2001. In both cases, this is caused by an increase in
with a remaining residual of about 1% (Interim) and 5% oceanic rainfall, while Interim predicts a decrease (not
(MERRA) of the continental P 2 E moisture budget. shown). The increase in MERRA and CFSR is likely
Similar values for Interim and MERRA were also re- due to the assimilation of sounding radiances from
ported by Trenberth et al. (2011). Jung et al. (2010) es- AMSU-A on the NOAA-15 satellite from 1998 (Nicolas
timated a mean total land surface evapotranspiration of and Bromwich 2011; Bosilovich et al. 2011). Robertson
65 6 3 3 1015 kg yr21 between 1982 and 2008, which et al. (2011) detected that the assimilation of AMSU-A
agrees with the estimates from Oki and Shinjiro (2006). data has a significant impact on the MERRA water
It should be noted that small deviations of the estimates vapor increments, leading to an increased amount of

TABLE 4. Mean oceanic precipitation, evaporation, and P 2 E

between 1989 and 2006 from Interim, MERRA, CFSR, and satel-
lite observations from the HOAPS dataset (mm day21). The
numbers in the brackets denote the standard deviations.
P E P2E
HOAPS 2.86 (60.16) 3.54 (60).18 20.68 (60.23)
Interim 3.29 (60.10) 3.71 (60.09) 20.43 (60.13)
MERRA 2.99 (60.20) 3.37 (60.08) 20.39 (60.20)
CFSR 3.56 (60.16) 3.81 (60.11) 0.08 (60.15)
moisture in the model. This agrees with the time evo-

lution of the total atmospheric water vapor content
(Figs. 17a,c), which shows a sudden increase of both
MERRA and CFSR in 1998. Robertson et al. (2011)
further concluded that the additional moisture causes
an increase of precipitation especially over the tropic
oceans. If so, there should also be an increase in oceanic
evaporation for compensating the shift in oceanic rain-
fall, which cannot be detected in MERRA and CFSR.
The significant changes of many CFSR variables in 1998
are discussed by Wang et al. (2010) and Xue et al. (2010).
The continental P 2 E estimates (Fig. 15b) are in much
better agreement than the oceanic moisture budgets. In
1998, however, changes are significant, as MERRA pre-
dicts an increase of 6 3 1015 kg yr21. The CFSR budgets
decrease by about 12 3 1015 kg yr21 between 1998 and
2000. The gap in the continental P 2 E budgets is less
distinct and becomes smaller toward the end of the time FIG. 14. Vertically integrated moisture fluxes and moisture flux
convergences (mm day21) from (a) Interim, (b) MERRA, and (c)
series. The differences and spatial correlations of pre-
CFSR. Positive values depict areas with a surplus of precipitation
cipitation estimates from MERRA and GPCC show (i.e., P . E), while evaporation and evapotranspiration, re-
a sudden change in 1998 over South America only (Figs. spectively, are larger than precipitation over regions with negative
8h, 9h). CFSR precipitation deviations from GPCC, values. In general, negative P 2 E values over the continents
however, exhibit significant gaps over Asia, South should only be expected in regions containing large lakes or inland
seas. The direction and the amount of moisture transported are
America, and Africa (Fig. 9g,h,i). The reason has not yet
represented by the vector field. Large differences of the reanalyses
been revealed, but if the assimilation of AMSU data can be observed along the ITCZ over the oceans, where CFSR
causes these changes in the precipitation estimates, the shows larger moisture flux convergences than Interim and
impact of assimilation would differ significantly for both MERRA. There is also a positive pattern east of Brazil, which is
MERRA and CFSR. absent in the other reanalyses. This results in a generally increased
depletion of water in the CFSR over the oceans.
As a result of the oceanic P 2 E increase, the annual
budgets between the oceans and the continents are
highly distorted in MERRA and CFSR. CFSR shows a 16 satellite are introduced. Indeed, there is a distinct
positive oceanic P 2 E average of 8.6 3 1015 kg yr21 increase in the oceanic P 2 E estimates of MERRA
between 1999 and 2006, which obviously is not reason- between 2000 and 2001. In the CFSR dataset, there also
able. The MERRA P 2 E oceanic moisture budgets is an upward shift of oceanic P 2 E estimates between
exhibit a change in sign after 2005. According to 1999 and 2001, but the effect seems to weaken at the end
Rienecker et al. (2011), Interim does not use these ob- of the time series.
servations and, thus, shows no shift in 1998. Interim’s When analyzing the continental surface water balance
oceanic P 2 E moisture budgets reveal a permanent [Eq. (3)], MERRA shows the best performance in
downward trend until 1998, after which the budgets closing the long-term water balance. The surplus of
fluctuate around 245 3 1015 kg yr21, which agrees with evaporation over the continents is balanced by reduced
the reference values in Table 3. runoff compared to the other reanalyses and the refer-
Another significant shift in both MERRA and CFSR ence estimates. This leads to a significantly smaller sur-
is assumed to occur in 2001 when data from the NOAA- face water forcing residual in the terrestrial water storage

storage change dS/dt may be significant during shorter

periods and, hence, a large part of RESs might be due to
natural processes rather than artificial forcing in-
crements. CFSR and Interim have larger residuals of
19.0 3 1015 kg yr21 and 9 3 1015 kg yr21, respectively,
between runoff and continental P 2 E even if the runoff
estimates seem to be more realistic compared to MERRA.
The global intra-annual water budgets (Fig. 16a) show
a clear annual cycle with the minimum P 2 E in June
due to the increased evaporation and reduced pre-
cipitation during the Northern Hemispheric summer
months. Compared to the annual P 2 E moisture bud-
gets, the intra-annual variations of MERRA and Interim
are in much better agreement. Even if CFSR reproduces
a similar annual cycle, there is a significant deviation of
about 30 3 1014 kg month21 from the other reanalyses,
which causes a remaining imbalance during an intra-an-
nual cycle of the global P 2 E moisture budgets.
2) ATMOSPHERIC AND COMBINED ATMOSPHERIC–
TERRESTRIAL WATER BALANCE
The global atmospheric water budgets $ Q (Fig. 15a) are

nearly constant during the analyzed period. Consequently,
the atmospheric moisture exchange between the oceans
and the continents is a fully closed cycle in the three
reanalyses. As the time series of moisture flux diver-
gences obtained by the three reanalyses are in good
agreement, it is concluded that using the atmospheric
budgets for quantifying the exchange of moisture be-
tween the oceans and landmasses is more reliable than
the modeled P 2 E moisture budgets. The closure of the
global combined atmospheric–terrestrial water balance
[Eq. (8)] reveals some shortcomings: Interim predicts
FIG. 15. (a) Global, (b) continental, and (c) oceanic annual water
too-high (too low) P 2 E estimates until 1996 (from
balance (1015 kg yr21). Here P 2 E budgets are plotted as black
lines, while the atmospheric budgets are represented by the dotted 1996). CFSR has a significant moist bias (i.e., too-high
gray lines. On annual time scales, the difference between P 2 E and global P 2 E values) over the whole time series with a
the moisture flux divergences is an estimate of the atmospheric sudden increase likely due to the assimilation of AMSU
water forcing increment in the reanalysis models. The closure of data in 1998, while MERRA shows the same gap with
the combined atmospheric–terrestrial water budget would require
too-dry (too wet) conditions until 1999 (from 1999).
both P 2 E and 2$ Q to be equal. MERRA and CFSR show an
unrealistic increase of oceanic P 2 E from 1998, which is likely due Over the continents (15b), both the atmospheric and
to changes in the assimilated observations. Over the continents, the terrestrial budgets are in good agreement, leading to the
P 2 E budgets from Interim and MERRA are in better agreement closure of the combined atmospheric–terrestrial water
with the moisture fluxes, leading to a nearly closed continental balance. Hence, the largest part of the global imbalance
atmospheric–terrestrial water budget. CFSR overestimates conti-
comes from the large gaps between the P 2 E moisture
nental P 2 E as well, which causes an imbalanced residual of about
30 3 1015 kg yr21. budgets and the moisture flux divergences over the oceans
(Fig. 15c). As the changes in the tendency terms of atmo-
spheric and terrestrial water storage can be neglected on
of 2.6 3 1015 kg yr21. As the time evolution of both the annual time scales, the differences between the moisture
annual P 2 E moisture budgets and the total annual flux divergences and the P 2 E budgets are an estimate of
runoff (not shown) are not constant over time, however, the atmospheric water forcing due to the assimilation of
RESs changes as well. For water budget studies on observations. Hence, the impact on MERRA and Interim
shorter time scales, this changing imbalance should not is less pronounced over the continents than over the
be neglected. As reported by Roads et al. (2002), the oceans.

water cycle is not closed in CFSR). It should be noted,

however, that CFSR does not provide fields of evapo-
transpiration. The imbalance might be affected largely
by the approximation of E from fields of latent heat flux
and sublimation [Eq. (13)]. As the long-term average of
continental evapotranspiration agrees with the model
estimates from Trenberth et al. (2007), however, it is
likely that the too-high P 2 E values arise from the CFSR
precipitation.
3) ATMOSPHERIC WATER VAPOR
Figure 17 shows the monthly mean of total preci-
pitable water over the complete time series. Globally,
Interim predicts more atmospheric vapor before 1998
and less vapor after 1998 compared to MERRA and
CFSR (Fig. 17a). The main differences between the
datasets result from deviations over the oceans (Fig.
17c) that can be divided clearly into three periods. Be-
fore 1992, the Interim water vapor exceeds the estimates
from CFSR and MERRA. Between 1992 and 1998, the
three reanalyses are in good agreement, as the models
use similar observational data in this period. After 1998,
CFSR and MERRA show an increase of the preci-
pitable water over the oceans, which has already been
discussed in section 1. Compared to the differences in
the reanalyses’ water budgets, the time series of pre-
cipitable water are in good agreement. This is empha-
FIG. 16. (a) Global, (b) continental, and (c) oceanic intra-annual sized by Fig. 18 where also satellite observations from
water balance (1014 kg month21). Here P 2 E budgets are plotted
as black lines, while the atmospheric budgets are represented by
the HOAPS dataset representing the total atmospheric
the dotted gray lines. For the monthly budgets, changes in the at- water vapor over the ice-free ocean are shown. Espe-
mospheric water vapor content dW/dt were considered as well. cially during the period between 1992 and 1998, the re-
Interim and MERRA show a good closure of the global atmo- analyses successfully reproduce the annual cycle of
spheric–terrestrial intra-annual budgets. Over the continents, there water vapor over the oceans. After 1998, MERRA and
is a clear cycle with its minimum (maximum) in June (January) in
both P 2 E and the atmospheric moisture budgets, which might be
CFSR overestimate the amount of precipitable water,
explained by the decrease (increase) of evapotranspiration (pre- while Interim still shows a good agreement with HOAPS.
cipitation) during the boreal winter (summer) over the large con- On the other hand, Interim clearly overestimates atmo-
tinental areas of Northern Hemisphere. The CFSR P 2 E estimates spheric water vapor before 1992, while MERRA and
show a bias both over the oceans and the continents, causing CFSR agree well with HOAPS. This changing level of
a significant imbalance of the global intra-annual atmospheric–
terrestrial moisture budget.
agreement is likely due to the assimilation of different
data sources, as all three reanalyses use similar observa-
tions between 1992 and 1998 only.
As regards long-term monthly moisture budgets, the Changes in the atmospheric water vapor content
global, continental, and oceanic Interim and MERRA dW/dt were considered when computing the monthly
P 2 E values and atmospheric moisture fluxes are in water budgets. There is a strong annual cycle especially
very good agreement (Fig. 16). We assume that even if over the continents, leading to maximal values of dW/dt
the annual variations of the water cycle show significant in spring and autumn. This agrees with, for example,
shortcomings, the modeled processes are balanced well Rasmusson (1968) and is also considered in Seneviratne
on a monthly time scale. The CFSR budgets show a sig- et al. (2004). However, changes in the vertically inte-
nificant overestimation of P 2 E over both the continents grated water vapor usually are smaller by several orders
and the oceans, resulting in a significant imbalance that of magnitude compared to other quantities of the hy-
has its maximum between September and October, drological cycle and do not vary on annual time scales, as
where the global water budget leaves a monthly imbal- the intra-annual cycle of dW/dt is closed with sufficient
ance of up to 48 3 1014 kg month21 (i.e., the intra-annual accuracy in all three reanalyses (not shown). It is therefore

FIG. 18. Precipitable water (mm) from the reanalyses and the
HOAPS dataset over the ice-free ocean. Between 1992 and 1998,
the three reanalyses are in good agreement with the HOAPS ob-
servations. It is obvious that the increase in 1998 of both MERRA
and CFSR causes an overestimation of precipitable water when
compared to HOAPS. Interim shows an overestimation before
1992, but reveals a good agreement with HOAPS for the rest of the
time period.
0.3 mm day21 higher compared to HOAPS, the mean

P 2 E moisture budget is positive, showing an over-
estimation of 0.7 mm day21. All three reanalyses over-
estimate the P 2 E moisture budget, but in Interim, this
is due to an overestimation of precipitation and evapo-
ration, while MERRA underestimates evaporation.
Hence, it is impossible to make a general statement
about the origins of the too-large oceanic P 2 E mois-
ture budgets in the analyzed reanalyses.
4. Summary and conclusions
FIG. 17. (a) Global, (b) continental, and (c) oceanic precipitable The present study demonstrated major differences of
water estimates (mm) from the three reanalyses. Globally, Interim differences between the three reanalyses ERA-Interim
shows higher values until 1995. Between 1992 and early 1998, the from ECMWF, MERRA from NASA, and CFSR from
three reanalyses are in very good agreement, which is likely due to NCEP. Precipitation is one of the most important quan-
the assimilation of similar observations during that period. After
tities of the water cycle. Its estimates are highly uncertain
1998, the MERRA and CFSR estimates are higher when compared
to Interim. This also holds for the time evolution of the atmo- in terms of spatial variability and total amount. The
spheric water vapor over the oceans. Again, it is presumed that the largest discrepancies occur in the summer months of the
increase in 1998 is due to changes in the assimilated observations in respective hemisphere, because convective effects still
both MERRA and CFSR but not in Interim. When compared to are a large source of uncertainty. However, a validation
other components of the large-scale water cycle, the water vapor
of, for example, tropical or mountainous rainfall also re-
estimates are in much better agreement, which is likely due to the
forcing of the estimates toward observations in all three reanalyses. mains difficult because of the large differences of the
observation datasets in these areas. This can be attributed
to the irregular distribution of gauges especially in such
proposed to neglect dW/dt for large-scale and long-term complex and highly variable regions. Hence, there are
water budget studies. large parts of South America and Africa that are com-
pletely ungauged. For regions with a dense network of
4) EVALUATION OF HYDROLOGICAL VARIABLES
gauges like North America, Australia, or Europe, it may
OVER THE OCEANS
be concluded that Interim still provides the most reliable
Mean estimates of modeled precipitation, evapora- rainfall estimates. The largest problem in these areas is
tion, and P 2 E over the ice-free oceans are presented the decrease of active gauges during the period considered.
and compared with satellite observations from HOAPS The quality of an interpolated product over a continuously
in Table 4. As another reference, the GPCP dataset changing network of observations remains questionable.
predicts a mean precipitation rate of about 3.0 mm day21. On the other hand, validation of the coarsely resolved
CFSR shows an overestimation of oceanic rainfall of GPCC and reanalysis precipitation estimates with in situ
0.6 mm day21. As evaporation from CFSR is only about rainfall observations is not yet meaningful because of the

precipitation’s high spatial variability and dependence on are much larger than those of their atmospheric coun-
surrounding terrain. terpart. It may therefore be concluded that they are still
In data-sparse regions, the general tendency of the more reliable than the terrestrial P 2 E values.
datasets considered can be regarded examined only. It Because of the limitations presented, the performance
may be concluded that major shortcomings exist in the of all three reanalyses in reproducing the hydrological
spatial patterns of South American rainfall in MERRA cycle still causes doubts in the use of such models for
and in the total amount of mid- to high-latitude pre- climate trend analyses and long-term water budget
cipitation in CFSR where a significant bias was detected. studies.
In the case of the MERRA reanalysis, these shortcom-
ings are well known. Thus, NASA is currently per- Acknowledgments. We thank the European Centre
forming a land-only rerun of the MERRA reanalysis in for Medium-Range Weather Forecast, the Global
which observational data from GPCP and other in- Modeling and Assimilation Office, the Goddard Earth
dependent global data products are used to correct and Sciences Data and Information Services Center, the
evaluate MERRA’s land surface hydrology (Rienecker National Centers for Environmental Prediction, and the
et al. 2011). The corrected precipitation estimates are National Center for Atmospheric Research for creating
already available and will replace the original MERRA and providing the reanalysis data. GPCC data were ob-
products in the near future (Reichle et al. 2011). tained from http://gpcc.dwd.de. GPCP rainfall observa-
The differences in the amount and spatial patterns tions are developed and computed by the NASA God-
of continental rainfall between GPCC and the other dard Space Flight Center’s Laboratory for Atmospheres
datasets remain more or less constant during the whole as a contribution to the GEWEX Global Precipitation
period (except for South American precipitation in the Climatology Project (http://precip.gsfc.nasa.gov/). The
MERRA dataset). This is important, as the assimilation CRU high-resolution precipitation and temperature
of observations, which became available during the pe- data were downloaded from the British Atmospheric
riod considered, does not seem to significantly affect the Data Centre (http://badc.nerc.ac.uk/view/badc.nerc.ac.uk__
precipitation estimates over the landmasses. The situa- ATOM__dataent_1256223773328276). The daily rainfall
tion is clearly different for oceanic precipitation that observations from the Climate Prediction Center are
exhibits a significant shift in late 1998 when sounding provided at ftp.cpc.ncep.noaa.gov/precip/CPC_UNI_PRCP/
radiances from AMSU were assimilated into both MERRA GAUGE_GLB/. Precipitation and temperature observa-
and CFSR. Atmospheric reanalysis models are still sensitive tions from the University of Delaware were created at the
to the introduction of observational data. Similar find- Center for Climatic Research at the Department of Ge-
ings were also presented in Bengtsson et al. (2004) for ography (http://climate.geog.udel.edu/;climate/). We
the ERA-40 reanalysis. would also like to thank Prof. Dr. Tonie van Dam (Uni-
The uncertainties in the precipitation estimates of the versity of Luxemburg), Dr. Patrick Laux (Karlsruhe In-
three reanalyses are highly correlated with the vari- stitute of Technology), and Balaji Devaraju (University
ability in the temperature fields. An obvious connection of Stuttgart), who improved this paper with their com-
was found between the large-scale underestimation of ments and corrections. This work is part of the Direct
South American precipitation from MERRA and a Waterbalance subproject of the DFG priority pro-
general warm bias in these regions. To make statements gramme 1257 Mass Transport and Mass Distribution in
about the validity of an atmospheric reanalysis, it is not the System Earth (http://www.massentransporte.de).
sufficient to consider one quantity like, for example,
precipitation only, but also other variables must be
REFERENCES
taken into account. This is crucial to the closure of the
water budgets of these datasets. When introducing new Adler, R. F., and Coauthors, 2003: The Version-2 Global
observations, CFSR and MERRA became significantly Precipitation Climatology Project (GPCP) Monthly Pre-
imbalanced after 1998. Oceanic precipitation increased cipitation Analysis (1979–present). J. Hydrometeor., 4,
1147–1167.
because of the assimilation of sounding radiances from Andersson, A., K. Fennig, C. Klepp, S. Bakan, H. Graßl, and
AMSU. In both CFSR and MERRA, the increase in J. Schulz, 2010: The Hamburg Ocean Atmosphere Parameters
total water is not balanced by an increase in oceanic and Fluxes from Satellite Data—HOAPS-3. Earth Syst. Sci.
evaporation or a decrease of continental P 2 E esti- Data Discuss., 3, 143–194.
mates, leaving a large gap in the global water balance. Bengtsson, L., S. Hagemann, and K. I. Hodges, 2004: Can climate
trends be calculated from reanalysis data? J. Geophys. Res.,
The atmospheric budgets do not show such a sudden 109, D11111, doi:10.1029/2004JD004536.
shift, but tend to increase during the time series consid- Berrisford, P., D. Dee, K. Fielding, M. Fuentes, P. Kallberg,
ered. Furthermore, the differences of the P 2 E estimates S. Kobayashi, and S. Uppala, 2009: The ERA Interim archive:

Version 1.0. ERA Rep. Series, ECMWF, 16 pp. [Available Legates, D. R., and C. J. Willmott, 1990: Mean seasonal and spatial
online at http://ecmwf.int/publications/library/do/references/ variability in gauge-corrected, global precipitation. Int. J. Cli-
show?id=89203.] matol., 10, 111–127, doi:10.1002/joc.3370100202.
Betts, A. K., J. H. Ball, and P. Viterbo, 1999: Basin-scale surface Matsuura, K., and C. J. Willmott, 2009: Terrestrial air temperature:
water and energy budgets for the Mississippi from the 1900-2008 gridded monthly time series (version 2.01). Center
ECMWF reanalysis. J. Geophys. Res., 104 (D16), 19 293– for Climatic Research, University of Delaware, digital media.
19 306, doi:10.1029/1999JD900056. [Available online at http://climate.geog.udel.edu/;climate/
——, ——, and ——, 2003: Evaluation of the ERA-40 surface html_pages/download.html#P2009.]
water budget and surface temperature for the Mackenzie Mitchell, T. D., and P. D. Jones, 2005: An improved method of
River basin. J. Hydrometeor., 4, 1194–1211. constructing a database of monthly climate observations and
——, ——, ——, A. Dai, and J. Marengo, 2005: Hydrometeo- associated high-resolution grids. Int. J. Climatol., 25, 693–712,
rology of the Amazon in ERA-40. J. Hydrometeor., 6, doi:10.1002/joc.1181.
764–774. Mueller, B., M. Hirschi, and S. I. Seneviratne, 2010: New diagnostic
——, M. Köhler, and Y. Zhang, 2009: Comparison of river basin estimates of variations in terrestrial water storage based on
hydrometeorology in ERA-Interim and ERA-40 reanalyses ERA-Interim data. Hydrol. Processes, 25, 996–1008, doi:10.1002/
with observations. J. Geophys. Res., 114, D02101, doi:10.1029/ hyp.7652.
2008JD010761. Nicolas, J. P., and D. H. Bromwich, 2011: Precipitation changes in
Bosilovich, M. G., J. Chen, F. R. Robertson, and R. F. Adler, 2008: high southern latitudes from global reanalyses: A cautionary
Evaluation of global precipitation in reanalyses. J. Appl. Meteor. tale. Surv. Geophys., 32, 475–494, doi:10.1007/s10712-011-
Climatol., 47, 2279–2299. 9114-6.
——, F. R. Robertson, and J. Chen, 2011: Global energy and water Oki, T., and K. Shinjiro, 2006: Global hydrological cycles and world
budgets in MERRA. J. Climate, 24, 5721–5739. water resources. Science, 313, 1068–1072, doi:10.1126/science.
Chen, M., W. Shi, P. Xie, V. B. S. Silva, V. E. Kousky, R. W. 1128845.
Higgins, and J. E. Janowiak, 2008: Assessing objective Peixoto, J. P., and A. H. Oort, 1992: Physics of Climate. American
techniques for gauge-based analyses of global daily precipita- Institute of Physics, 520 pp.
tion. J. Geophys. Res., 113, D04110, doi:10.1029/2007JD Poccard, I., S. Janicot, and P. Camberlin, 2000: Comparison of
009132. rainfall structures between NCEP/NCAR reanalyses and ob-
Chido, G., and L. Haimberger, 2009: Interannual changes in mass served data over tropical Africa. Climate Dyn., 16, 897–915,
consistent energy budgets from ERA-Interim and satellite data. doi:10.1007/s003820000087.
J. Geophys. Res., 115, D02112, doi:10.1029/2009JD012049. Rasmusson, E. M., 1968: Atmospheric water vapor transport and
Fuchs, T., J. Rapp, F. Rubel, and B. Rudolf, 2001: Correction of the water balance of North America. II. Large-scale water
synoptic precipitation observations due to systematic measuring balance investigations. Mon. Wea. Rev., 96, 720–734.
errors with special regard to precipitation phases. Phys. Chem. Reichle, R. H., R. D. Kostera, G. J. M. D. Lannoy, B. A. Forman,
Earth, 26B, 689–693, doi:10.1016/S1464-1909(01)00070-3. Q. Liu, S. P. P. Mahanama, and A. Tour, 2011: Assessment and
Hagemann, S., K. Arpe, and L. Bengtsson, 2005: Validation of the enhancement of MERRA land surface hydrology estimates.
hydrological cycle of ERA-40. ERA Rep. Series, ECMWF, J. Climate, 24, 6322–6338.
42 pp. [Available online at http://www.ecmwf.int/publications/ Rienecker, M. M., and Coauthors, 2011: MERRA:–NASA’s
library/ecpublications/_pdf/era40/ERA40_PRS24.pdf.] Modern-Era Retrospective Analysis for Research and Appli-
Higgins, R. W., V. E. Kousky, V. B. S. Silva, E. Becker, and P. Xie, cations. J. Climate, 24, 3624–3648.
2010: Intercomparison of daily precipitation statistics over the Roads, J. O., S. C. Chen, A. K. Guetter, and K. P. Georgakaos,
United States in observations and in NCEP reanalysis prod- 1994: Large-scale aspects of the United States hydrologic cy-
ucts. J. Climate, 23, 4637–4650. cle. Bull. Amer. Meteor. Soc., 75, 1589–1610.
Jacobson, M. Z., 2007: Fundamentals of Atmospheric Modeling. ——, M. Kanamitsu, and R. E. Stewart, 2002: CSE water and en-
2nd ed. Cambridge University Press, 813 pp. ergy budgets in the NCEP–DOE Reanalysis II. J. Hydrome-
Janowiak, J. E., A. Gruber, C. R. Kondragunta, R. E. Livezey, and teor., 3, 227–248.
G. J. Huffman, 1997: A comparison of the NCEP–NCAR re- Robertson, F. R., M. G. Bosilovich, J. Chen, and T. L. Miller, 2011:
analysis precipitation and the GPCP rain gauge–satellite com- The effect of satellite observing system changes on MERRA
bined dataset with observational error considerations. water and energy fluxes. J. Climate, 24, 5197–5217.
J. Climate, 11, 2960–2979. Rudolf, B., and F. Rubel, 2005: Global precipitation. Observed
Jones, P. W., 1999: First- and second-order conservative remapping Global Climate, M. Hantel, Ed., Landolt-Börnstein: Numeri-
schemes for grids in spherical coordinates. Mon. Wea. Rev., cal Data and Functional Relationships in Science and Tech-
127, 2204–2210. nology, Vol. 6, Springer-Verlag, 11.01–11.53.
Jung, M., and Coauthors, 2010: Recent decline in the global land ——, and U. Schneider, 2005: Calculation of gridded precipitation
evapotranspiration trend due to limited moisture supply. Na- data for the global land-surface using in-situ gauge observa-
ture, 467, 951–954, doi:10.1038/nature09396. tions. Proc. Second Workshop of the Int. Precipitation Work-
Kalnay, E., and Coauthors, 1996: The NCEP/NCAR 40-Year Re- ing Group, Monterey, CA, IPWG, 231–247.
analysis Project. Bull. Amer. Meteor. Soc., 77, 437–471. Saha, S., and Coauthors, 2010: The NCEP Climate Forecast System
Kistler, R., and Coauthors, 2001: The NCEP–NCAR 50-Year Reanalysis. Bull. Amer. Meteor. Soc., 91, 1015–1057.
Reanalysis: Monthly means CD-ROM and documentation. Schneider, U., T. Fuchs, A. Meyer-Christoffer, and B. Rudolf,
Bull. Amer. Meteor. Soc., 82, 247–268. 2008: Global precipitation analysis products of the GPCC.
Kleidon, A., and S. Schymanski, 2008: Thermodynamics and op- Deutscher Wetterdienst Tech. Rep., 12 pp. [Available online at
timality of the water budget on land: A review. Geophys. Res. www.tam.cec.org/Atlas/Files/Precipitation/GPCC_intro_products_
Lett., 35, L20404, doi:10.1029/2008GL035393. 2008.pdf.]

Seneviratne, S. I., P. Viterbo, D. Lüthi, and C. Schär, 2004: In- ——, J. T. Fasullo, and J. Mackaro, 2011: Atmospheric moisture
ferring changes in terrestrial water storage using ERA-40 re- transports from ocean to land and global energy flows in re-
analysis data: The Mississippi River basin. J. Climate, 17, analyses. J. Climate, 24, 4907–4924.
2039–2057. Uppala, S. M., and Coauthors, 2005: The ERA-40 Re-Analysis.
Simmons, A., S. Uppala, D. Dee, and S. Kobayashi, 2006: ERA- Quart. J. Roy. Meteor. Soc., 131, 2961–3012.
Interim: New ECMWF reanalysis products from 1989 onwards. Van der Ent, R. J., H. H. G. Savenije, B. Schaefli, and S. C. Steele-
ECMWF Newsletter, No. 110, ECMWF, Reading, United King- Dunne, 2010: Origin and fate of atmospheric moisture over
dom, 25–35. [Available online at http://www.ecmwf.int/publications/ continents. Water Resour. Res., 46, W09525, doi:10.1029/
newsletters/pdf/110_rev.pdf.] 2010WR009127.
Sylla, M. B., E. Coppola, L. Mariotti, F. Giorgi, P. M. Ruti, Wang, W., P. Xie, S.-H. Yoo, Y. Xue, A. Kumar, and X. Wu, 2010:
A. Dell’Aquila, and X. Bi, 2010: Multiyear simulation of the An assessment of the surface climate in the NCEP climate
African climate using a regional climate model (RegCM3) with forecast system reanalysis. Climate Dyn., 37, 1601–1620,
the high resolution ERA-interim reanalysis. Climate Dyn., 35, doi:10.1007/s00382-010-0935-7.
231–247, doi:10.1007/s00382-009-0613-9. Willmott, C. J., C. M. Rowe, and Y. Mintz, 1985: Climatology of the
Szeto, K. K., H. Tran, M. D. MacKay, R. Crawford, and R. E. terrestrial seasonal water cycle. J. Climatol., 5, 589–606,
Stewart, 2008: The MAGS Water and Energy Budget Study. doi:10.1002/joc.3370050602.
J. Hydrometeor., 9, 96–115. Xue, Y., B. Huang, Z.-Z. Hu, A. Kumar, C. Wen, D. Behringer, and
Taylor, K. E., 2001: Summarizing multiple aspects of model perfor- S. Nadiga, 2010: An assessment of oceanic variability in the
mance in a single diagram. J. Geophys. Res., 106 (D7), 7183– NCEP climate forecast system reanalysis. Climate Dyn., 37,
7192, doi:10.1029/2000JD900719. 2511–2539, doi:10.1007/s00382-010-0954-4.
Trenberth, K. E., L. Smith, T. Qian, A. Dai, and J. Fasullo, 2007: Yeh, P. J.-F., and J. S. Famiglietti, 2008: Regional terrestrial water
Estimates of the global water budget and its annual cycle storage change and evapotranspiration from terrestrial and
using observational and model data. J. Hydrometeor., 8, atmospheric water balance computations. J. Geophys. Res.,
758–769. 113, D09108, doi:10.1029/2007JD009045.

Hydr JHM D 11 088 - 1

Uploaded by

Copyright:

Available Formats

Hydr JHM D 11 088 - 1

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Hydr JHM D 11 088 - 1

Uploaded by

Copyright:

Available Formats

VOLUME 13 JOURNAL OF HYDROMETEOROLOGY OCTOBER 2012

The Hydrological Cycle in Three State-of-the-Art Reanalyses:

(Manuscript received 21 July 2011, in final form 12 March 2012)

1. Introduction To bring these various observations into a consistent

Ó 2012 American Meteorological Society 1397

Unauthenticated | Downloaded 09/11/23 08:19 AM UTC

TABLE 1. Summary of the three reanalyses.

Interim MERRA CFSR

Variables Horizontal resolution Period Output times Version number

Unauthenticated | Downloaded 09/11/23 08:19 AM UTC

Unauthenticated | Downloaded 09/11/23 08:19 AM UTC

R ’ P 2 E, (3) moisture flux divergences are directly linked to the

Unauthenticated | Downloaded 09/11/23 08:19 AM UTC

The root-mean-square difference (radial distance be-

Unauthenticated | Downloaded 09/11/23 08:19 AM UTC

Unauthenticated | Downloaded 09/11/23 08:19 AM UTC

In general, the highest differences in spatial variability

Unauthenticated | Downloaded 09/11/23 08:19 AM UTC

Unauthenticated | Downloaded 09/11/23 08:19 AM UTC

Unauthenticated | Downloaded 09/11/23 08:19 AM UTC

Unauthenticated | Downloaded 09/11/23 08:19 AM UTC

Unauthenticated | Downloaded 09/11/23 08:19 AM UTC

Unauthenticated | Downloaded 09/11/23 08:19 AM UTC

Unauthenticated | Downloaded 09/11/23 08:19 AM UTC

Unauthenticated | Downloaded 09/11/23 08:19 AM UTC

Unauthenticated | Downloaded 09/11/23 08:19 AM UTC

TABLE 3. Mean global water cycle components over land and

Interim MERRA CFSR TB OKI

may be due to differences in the used land–sea mask or,

Unauthenticated | Downloaded 09/11/23 08:19 AM UTC

TABLE 4. Mean oceanic precipitation, evaporation, and P 2 E

moisture in the model. This agrees with the time evo-

Unauthenticated | Downloaded 09/11/23 08:19 AM UTC

storage change dS/dt may be significant during shorter

The global atmospheric water budgets $ Q (Fig. 15a) are

Unauthenticated | Downloaded 09/11/23 08:19 AM UTC

water cycle is not closed in CFSR). It should be noted,

Unauthenticated | Downloaded 09/11/23 08:19 AM UTC

0.3 mm day21 higher compared to HOAPS, the mean

Unauthenticated | Downloaded 09/11/23 08:19 AM UTC

Unauthenticated | Downloaded 09/11/23 08:19 AM UTC

Unauthenticated | Downloaded 09/11/23 08:19 AM UTC

Unauthenticated | Downloaded 09/11/23 08:19 AM UTC

You might also like