Continental-Scale Basin Water Storage Variation from Global and Dynamically Downscaled Atmospheric Water Budgets in Comparison with GRACE-Derived Observations

Benjamin Fersch Karlsruhe Institute of Technology (IMK-IFU), Garmisch-Partenkirchen, Germany

Search for other papers by Benjamin Fersch in
Current site
Google Scholar
PubMed
Close
,
Harald Kunstmann Karlsruhe Institute of Technology (IMK-IFU), Garmisch-Partenkirchen, Germany

Search for other papers by Harald Kunstmann in
Current site
Google Scholar
PubMed
Close
,
András Bárdossy Institute of Hydraulic Engineering (IWS), University of Stuttgart, Stuttgart, Germany

Search for other papers by András Bárdossy in
Current site
Google Scholar
PubMed
Close
,
Balaji Devaraju Institute of Geodesy (GIS), University of Stuttgart, Stuttgart, Germany

Search for other papers by Balaji Devaraju in
Current site
Google Scholar
PubMed
Close
, and
Nico Sneeuw Institute of Geodesy (GIS), University of Stuttgart, Stuttgart, Germany

Search for other papers by Nico Sneeuw in
Current site
Google Scholar
PubMed
Close
Full access

Abstract

Since 2002, the Gravity Recovery and Climate Experiment (GRACE) has provided gravity-derived observations of variations in the terrestrial water storage. Because of the lack of suitable direct observations of large-scale water storage changes, a validation of the GRACE observations remains difficult. An approach that allows the evaluation of terrestrial water storage variations from GRACE by a comparison with those derived from aerologic water budgets using the atmospheric moisture flux divergence is presented. In addition to reanalysis products from the European Centre for Medium-Range Weather Forecasts and the National Centers for Environmental Prediction, high-resolution regional atmospheric simulations were produced with the Weather Research and Forecast modeling system (WRF) and validated against globally gridded observational data of precipitation and 2-m temperature. The study encompasses six different climatic and hydrographic regions: the Amazon basin, the catchments of Lena and Yenisei, central Australia, the Sahara, the Chad depression, and the Niger. Atmospheric-related uncertainty bounds based on the range of the ensemble of estimated terrestrial water storage variations were computed using different configurations of the regional climate model WRF and different global reanalyses. Atmospheric-related uncertainty ranges with those originating from the GRACE products of GeoForschungsZentrum Potsdam, the Center for Space Research, and the Jet Propulsion Laboratory were also compared. It is shown that dynamically downscaled atmospheric fields are able to add value to global reanalyses, depending on the geographical location of the considered catchments. Global and downscaled atmospheric water budgets are in reasonable agreement (r ≈ 0.7 − 0.9) with GRACE-derived terrestrial mass variations. However, atmospheric- and satellite-based approaches show shortcomings for regions with small storage change rates (<20–25 mm month−1).

Corresponding author address: Benjamin Fersch, Karlsruhe Institute of Technology (KIT), Institute for Meteorology and Climate Research (IMK-IFU), Kreuzeckbahnstr. 19, 82467 Garmisch-Partenkirchen, Germany. E-mail: fersch@kit.edu

Abstract

Since 2002, the Gravity Recovery and Climate Experiment (GRACE) has provided gravity-derived observations of variations in the terrestrial water storage. Because of the lack of suitable direct observations of large-scale water storage changes, a validation of the GRACE observations remains difficult. An approach that allows the evaluation of terrestrial water storage variations from GRACE by a comparison with those derived from aerologic water budgets using the atmospheric moisture flux divergence is presented. In addition to reanalysis products from the European Centre for Medium-Range Weather Forecasts and the National Centers for Environmental Prediction, high-resolution regional atmospheric simulations were produced with the Weather Research and Forecast modeling system (WRF) and validated against globally gridded observational data of precipitation and 2-m temperature. The study encompasses six different climatic and hydrographic regions: the Amazon basin, the catchments of Lena and Yenisei, central Australia, the Sahara, the Chad depression, and the Niger. Atmospheric-related uncertainty bounds based on the range of the ensemble of estimated terrestrial water storage variations were computed using different configurations of the regional climate model WRF and different global reanalyses. Atmospheric-related uncertainty ranges with those originating from the GRACE products of GeoForschungsZentrum Potsdam, the Center for Space Research, and the Jet Propulsion Laboratory were also compared. It is shown that dynamically downscaled atmospheric fields are able to add value to global reanalyses, depending on the geographical location of the considered catchments. Global and downscaled atmospheric water budgets are in reasonable agreement (r ≈ 0.7 − 0.9) with GRACE-derived terrestrial mass variations. However, atmospheric- and satellite-based approaches show shortcomings for regions with small storage change rates (<20–25 mm month−1).

Corresponding author address: Benjamin Fersch, Karlsruhe Institute of Technology (KIT), Institute for Meteorology and Climate Research (IMK-IFU), Kreuzeckbahnstr. 19, 82467 Garmisch-Partenkirchen, Germany. E-mail: fersch@kit.edu

1. Introduction

On seasonal to monthly time scales, the assessment of the different quantities of the hydrological cycle on a global to continental extent is limited by the lack of sophisticated measuring methods (Trenberth et al. 2007) and particularly by the low spatial density of measuring networks. For river basins on the scale of 105 to 106 km2, solving the hydrological water balance equation
e1
relating precipitation P, evapotranspiration E, basin discharge R, and the water storage S, is often not possible. The most important reasons are 1) a limited hydrological and meteorological observation network density, 2) a lack of direct measurements for E and dS/dt at the appropriate scale, and 3) fully ungauged regions, where R is not even available. The satellite mission Gravity Recovery and Climate Experiment (GRACE) provides the opportunity to estimate large-scale variations of the mass changes in the Earth system. In hydrological research, the central question related to GRACE-derived terrestrial mass changes is the accuracy of GRACE water storage change estimates (Wahr et al. 1998).

GRACE cannot be classified as a direct observation method because it is only capable of observing the integral sum of all gravitational effects. Moreover, the GRACE estimates contain spurious patterns and correlated errors that also need to be corrected for by spatial filtering techniques (Swenson et al. 2003). A study by Winsemius et al. (2006) compares terrestrial water storage changes inferred from GRACE with the output of a calibrated hydrological model for the upper Zambezi River basin. The results indicate temporal inconsistencies between the two methods that are probably caused by GRACE artifacts. Contrarily, Werth and Güntner (2010) propose GRACE as a calibration reference for global hydrological models. A comprehensive overview of the applications for GRACE in hydrology studies is, for example, given by Schmidt et al. (2006) and Güntner et al. (2007).

The intent of this study is to determine the applicability of the GRACE-derived products for continental-scale water budget estimation. We use both global and regionally refined aerological water budgets to assess the uncertainty bounds. In turn, this allows us to draw conclusions with respect to the uncertainties of terrestrial water storage changes. Furthermore, we investigate how isotropic Gaussian filtering, which is used for processing the GRACE-derived estimates, influences the quantities of atmospheric water budgets. Finally, the derived uncertainty ranges are used for an intercomparison with the GRACE data.

The approach was introduced by Fersch et al. (2009), where the main focus was on the sensitivity of the dynamic downscaling approach. It was shown that both global and regional atmospheric moisture balance approaches contain considerable deviations that require an exclusion strategy, based on a sound validation with independent observations. The validation and the exclusion strategy are addressed in the current study. Furthermore, an estimation of the uncertainty of GRACE observations is determined using an ensemble of three different GRACE products.

2. Methods and data

a. GRACE

With GRACE, the spatiotemporal variations of the continental water storage body are derived by a continuous observation of the gravity field of Earth from space. The repetitive long-term measurements of the GRACE mission enable the detection of temporal mass variations with a spatial resolution of about 400 km every 30 days (Tapley et al. 2004). On a monthly time scale, the prevalent processes of mass variation on the globe are the redistribution of water within the hydrological cycle and the atmospheric circulation. As the gravity field is influenced by all mass variations on Earth and within the solar system, it is mandatory to remove all other mass change contributions to reveal the pure hydrological signal (Swenson et al. 2003). This de-aliasing procedure necessitates the incorporation of observations and models in order to reduce the signal for the specific components. Thereby, corresponding uncertainties are introduced. GRACE-derived observations of spatiotemporal variations of the terrestrial water storage body provide a method to close a central gap in large-scale hydrology. Apart from the included uncertainties, with GRACE-derived estimates of dS/dt it is possible to solve the water balance solely by using additional observations of P and R.

GRACE products of terrestrial water storage variation show some structural inconsistencies. Riegger et al. (2012) found that the GRACE products contain a number of outliers for certain hydrological basins and months. Moreover, the interbasin correlation patterns of GRACE deviated significantly from the hydrological observations. Knowing about these shortcomings raises the question about the value and applicability of GRACE in hydrology. As shown by Gao et al. (2010), over the western U.S. basins, the GRACE products tend to have a smaller range than expected from the large-scale water budget model Variable Infiltration Capacity (VIC). Unfortunately, the direct validation of dS/dt from GRACE is not possible. An indirect validation by means of the terrestrial water balance requires information about evapotranspiration (Werth and Avissar 2004). However, because of the high effort and complexity of measuring E, monthly time series of global- or continental-scale observations are not available. The consideration of the atmospheric-moisture-derived water budget provides an alternative to missing observations of E.

b. Atmospheric water budget

For continental-scale areas and at monthly to seasonal time scales, the spatially averaged water budget of the atmosphere can be considered as a proxy for evapotranspiration minus precipitation:
e2
with W and · Q denoting the atmospheric water storage and the net balance of moisture flux, respectively.
For the spatial and temporal scales considered, variations of the atmospheric water storage W can be assumed to be negligible (Peixoto and Oort 1992). For basins with observed discharge R, the atmospheric-moisture-budget-derived dS/dt can be compared with GRACE:
e3
In the absence of external discharge (e.g., desert environments), R is assumed to be zero. These regions allow a further simplification of the evaluation approach, as the GRACE storage variations and · Q become directly comparable. Atmospheric water budgets have been used in several hydrological and meteorological applications—for example, Berbery et al. (1996), Labraga et al. (2000), Roads et al. (2002), Hirschi et al. (2007), and Yeh and Famiglietti (2008).

Hirschi et al. (2006) compare the combined atmospheric–terrestrial water balance of the European Centre for Medium-Range Weather Forecasts (ECMWF) operational analysis with the GRACE products of three different data processing centers. The results suggest a general agreement of phase and amplitude. Moreover, it is concluded that GRACE is most precise for regions larger than 106 km2. Seitz et al. (2008) investigate hydrological extremes for central Europe with a regional, spherical wavelet-based GRACE solution with respect to the aerological water balance of the National Centers for Environmental Prediction–National Center for Atmospheric Research (NCEP–NCAR) reanalysis project (NNRP) with Global Runoff Data Centre (GRDC) discharge data. Similar to Hirschi et al. (2006), the phase and amplitude of the two approaches show reasonable agreement. A number of extreme conditions like the 2003 European summer drought can be seen from GRACE and the atmospheric approach. For other periods, differences of up to 70 mm month−1 are observed. Both studies do not address the uncertainties that are contained within the atmospheric analysis and reanalysis products. Two different atmospheric products were considered by Syed et al. (2007), where a combination of atmospheric moisture budgets and GRACE were used to estimate the pan-Arctic freshwater discharge. However, the quality of the atmospheric products was not analyzed with respect to independent observations—for example, precipitation. In Syed et al. (2009), the study was extended to other continental-scale river basins. Here, the two considered atmospheric analyses are also considered to be equivalent in terms of their quality.

Nevertheless, limitations of global reanalysis fields with respect to the hydrological cycle are addressed within several studies. For the ECMWF products, Hagemann et al. (2005) and Trenberth et al. (2007) report an overestimation of tropical precipitation over the oceans, an unbalanced global water budget, and an overestimation of evapotranspiration over many river catchments. Trenberth and Guillemot (1998) found that the NNRP reanalysis contains some severe shortcomings like an imbalance in the moisture budgets and that the model significantly deviates from the radiosonde data. Hence, it is a central issue of this study to assess the skill of the global atmospheric reanalyses before comparing them with GRACE.

The atmospheric water budgets of two prevalent global reanalysis products were used to provide an estimate of the uncertainty of the atmospheric water budget approach. The two reanalyses comprise the ECMWF Interim re-analysis (ERA-Interim) (Uppala et al. 2008) and the NNRP (Kalnay et al. 1996). Table 1 lists the main properties of the two products.

Table 1.

Properties of the global reanalysis products that are used for this study.

Table 1.

Because of the coarse spatial resolution of the global reanalysis models, one central research question of this study is to determine whether dynamic atmospheric downscaling can add value to the representation of water fluxes.

c. Dynamically downscaled atmospheric water budgets

Dynamical downscaling of global atmospheric fields represents a central methodology for a refined description of · Q. Regional atmospheric modeling allows for the consideration of physical processes in higher spatial and temporal resolution. Therefore, compared to the global atmospheric models, the downscaled atmospheric fields emerge with increased physical reality. In particular, the refined land surface and terrain information can lead to an improved representation of the moisture flux dynamics and budgets.

In our study, we use the Advanced Research version of the Weather Research and Forecast modeling system (ARW-WRF; Skamarock and Klemp 2008). The ARW-WRF supplies multiple interchangeable parameterization modules for the description of different physical compartments—for example, radiation, rainfall generation, or surface exchange processes. The results of a WRF simulation can be very sensitive to the chosen physical parameterizations (Borge et al. 2008; Awan et al. 2011). As shown by Fersch et al. (2009), the selection of these schemes can substantially affect the water budgets of ARW-WRF. In particular, the convective fraction of precipitation showed deviations, depending on the type of convective parameterization. However, the intensity varied for different climate regions. Furthermore, it was observed that the type of driving data affects the results in a significant way. With respect to an evaluation with GRACE, a thorough validation of the downscaled fields with independent observations is performed in advance. The validation approach is detailed in section 2d.

The general configuration of ARW-WRF is chosen according to the suggestions of Borge et al. (2008), Skamarock et al. (2008), and Wang et al. (2009). Table 2 lists the configuration of the regional atmospheric model selected for this study. A detailed description of the specific physical schemes can be found in Skamarock et al. (2008) and Wang et al. (2009).

Table 2.

Chosen ARW-WRF setup for the regional downscaling. A description of the different schemes can be found in Skamarock et al. (2008).

Table 2.

The horizontal resolution is selected with 30 × 30 km2. All relevant output fields are archived at 0000, 0600, 1200, 1800 UTC and monthly means are derived thereof. The regional model simulations cover the period 2001–06.

Vivoni et al. (2009) showed that the initialization of the soil moisture field can significantly affect the precipitation characteristics in a WRF simulation. Over longer periods from several months to years this effect becomes leveled out. Therefore, in this study, the first 2 years (2001/02) account for the spinup of the Noah land surface model (LSM). For the Siberian domain, the start time is shifted toward May 2001 so that an initialization of the snow layer is not mandatory.

The quality of the global driving data and the regional simulations is analyzed using independent observations of precipitation and temperature. The analysis includes the spatial patterns and the basin-aggregated time series. Those configurations of the regional model that lead to reasonable results are then considered for the comparison with GRACE and the estimation of uncertainty bounds for the atmospheric water budgets.

Downscaled reanalyses of ERA-Interim and NNRP will be abbreviated with EI and NR, respectively. BMJ and KF denote the convective parameterization schemes that are applied here: Betts–Miller–Janjić (Janjić 2000) and Kain–Fritsch (Kain 2004), respectively. The two schemes describe the formation of convective clouds and precipitation with different conceptual models. With the Betts–Miller–Janjić scheme, the thermal and moisture structure of the model are adjusted to a specified reference profile. The amount of precipitation results from the difference between the actual and the reference state. The Kain–Fritsch scheme relies on the convective available potential energy (CAPE). The precipitation is computed from the vertical moisture fluxes and a measure of precipitation efficiency. Compared to BMJ, increased physical detail is contained in KF. A more precise description on the conceptual differences of the two cumulus schemes can be found in Wang and Seaman (1997). Four-dimensional data assimilation (FDDA) refers to the constraining of some of the prognostic variables of the regional atmospheric model toward the global driving fields. If enabled, FDDA is only applied for wind, temperature, and humidity and only levels above the planetary boundary layer. The SST flag describes a time variable input for sea surface temperature. A more comprehensive discussion on the downscaling approach can be found in Fersch (2011).

d. Validation of the atmospheric water budget approach

Prior to the comparison of the GRACE data with the atmospheric water budget approach, it is mandatory to prove the reasonableness of the global and downscaled atmospheric products by validating them with independent observations. The comparison comprises spatial deviation patterns and basin-aggregated time series. For the spatial pattern analysis, deviation maps of annual means are prepared. The statistical coherence for the basin-aggregated time series is investigated using a diagram type proposed by Taylor (2001). All comparisons are based on a 0.5° × 0.5° mesh, and differing products are regridded and interpolated.

This validation allows for the identification of reasonable global reanalyses and regional model setups that will be used for the comparison with GRACE. The time series for the best-performing atmospheric products are then plotted along with the collective of three prevalent GRACE products—namely, GeoForschungsZentrum Postdam (GFZ); the Center for Space Research, University of Texas at Austin (CSR); and the National Aeronautics and Space Administration (NASA) Jet Propulsion Laboratory, Pasadena (JPL) (see section 2f).

Precipitation datasets are taken from the Global Precipitation Climatology Centre (GPCC). The full data reanalysis product (version 4) (Schneider et al. 2008) is used.

The locations of the gauges contributing to the interpolation are often distributed very heterogeneously with some regions being only sparsely sampled. Some alternative products exclusively ingest information from additional stations. To assess typical uncertainties for the precipitation product, GPCC is compared to other global observations. They constitute the gridded precipitation time series from the University of Delaware (DELP; Matsuura and Willmot 2009); the precipitation dataset Climatic Research Unit Time Series version 3.0 (CRU TS3.0) from the Climatic Research Unit, University of East Anglia (CRUP; Jones and Harris 2008; Mitchell and Jones 2005); and the Global Precipitation Climatology Project (GPCP; Adler et al. 2003). The deviations among these products provide an estimate of the reliability of these datasets. For validation of the global and regional temperature fields with gridded observations, the 2-m temperature field CRU TS3.0 (CRUT; Mitchell and Jones 2005; Jones and Harris 2008) is chosen. The global and regional atmospheric water budgets are defined as reasonable if their monthly precipitation and temperature patterns and basin-aggregated time series show no larger deviations than the different observations products themselves. If this is not the case, the respective global or regional water budget is no longer included in the evaluation of GRACE in this study.

e. Discharge observations

We use monthly discharge data from the GRDC; the United States Army Corps of Engineers (USACE); the Environmental Research Observatory on the Rivers of the Amazon basin (ORE HYBAM); the Department of Water, Land and Biodiversity Conservation from the Government of Australia; and from the Department of Water Affairs and Forestry of the Republic of South Africa. For the regions that are selected within this study, discharge data have been available mainly for the period 2000–07.

f. GRACE-derived products of water storage change

As for all geophysical observation systems, the maximum attainable information density of GRACE is determined by the spatiotemporal resolution of the sampling. Tapley et al. (2004) mentioned a spatial resolution of up to 400 km for monthly GRACE products. Thus, obtaining accurate estimates the region considered requires a minimum basin scale of several hundred km and gravitational effects that exceed the observational errors of GRACE (Swenson et al. 2003). The GRACE products are provided as a set of monthly spherical harmonic coefficients (see, e.g., Wahr et al. 1998). The monthly solutions are converted to equivalent water heights on a 0.5° × 0.5° mesh. Basin-aggregated time series are derived through area-weighted averaging. The three-point midpoint formula is used to obtain the time derivative (dM/dt) for a specific month.

To harmonize the monthly time series of the hydrological flux variables ( · Q, R) with the GRACE data processing, it is necessary to apply a time filter. Therefore, a running central mean approach according to Swenson and Wahr (2006a) and Landerer et al. (2010) is followed. The basin-aggregated time series of water storage change are computed with
e4
The brackets denote the monthly mean values of the previous (t − 1), the current (t), and the subsequent (t + 1) month.

1) Data collective

Monthly GRACE products of release 4 are provided by multiple data centers. These include GFZ (Flechtner 2007), CSR (Bettadpur 2007), and JPL (Watkins and Yuan 2007).

The three data centers share most of the geophysical de-aliasing models but use different algorithms for the processing. Therefore, variations between the products, we think, can be used as an estimate of the sensitivity of GRACE. Such an ensemble approach is, for example, used in Gao et al. (2010). Of course, this kind of error estimate does not include uncertainties common to all background models. Such information, and thus uncertainty bounds, for the single GRACE products is not available from the data centers. A study by Wahr et al. (2006) quantified errors for the JPL GRACE water storage change product with a range from 8 to 22 mm month−1, depending on latitude and varying with time.

2) Noise reduction, signal attenuation, and leakage effect

To mitigate the effect of the meridional stripe patterns contained in the GRACE solutions, we apply a decorrelation filter as proposed by Swenson and Wahr (2006b) and an additional spatial smoothing with an isotropic Gaussian filter with a half width of 500 km (Rodell et al. 2004a,b; Swenson and Wahr 2006b; Duan et al. 2009). By using a filter radius of 800 km, Klees et al. (2007) and Chen et al. (2007) showed that for continental-scale river basins like the Amazon, Mississippi, Ganges, and Congo, the amplitude of the annual bias ranges between 20% and 35% of the corresponding water storage variation.

One method to address this problem is the introduction of a gain factor to restore the amplitudes of the GRACE time series—for example, Landerer et al. (2010). However, the gain factors obtained from the filtering of a global water storage model are rather small. We think that by scaling a filtered field with a gain factor, one only bloats up the spatial scales retained by the filter and does not recapture the signal present in the filtered-out spatial scales. Here, no additional scaling is made for the GRACE-derived time series. Instead, the filter, which is applied to GRACE data, is also used for the atmospheric moisture budget fields. The reason for that is twofold: first and foremost it brings both the datasets to the same resolution, and secondly, because of the properties of the filter operator, there is exchange of mass from adjacent catchments and also oceans.

g. Study regions

To cover different climatic conditions, four globally distributed domains were selected: the Amazon tropics, central and western Africa, Siberia, and Australia. Because of the coarse resolution of the GRACE products, only hydrological basins with an area >106 km2 are chosen for a comparison. In addition to that, the availability of discharge data is mandatory, unless no runoff leaves the catchment boundaries. Figure 1 shows the different domains and the river basins that are selected for the comparison with GRACE in this study.

Fig. 1.
Fig. 1.

Catchment boundaries and distribution of the selected study regions.

Citation: Journal of Hydrometeorology 13, 5; 10.1175/JHM-D-11-0143.1

For the global atmospheric and GRACE datasets, the boundaries for the study regions are derived from a 0.5° × 0.5° global digital elevation model. For the regional downscaled fields, the catchment borders were adapted with respect to the increased detail of the height information.

h. Data standardization and aggregation

For validation, all WRF simulations are interpolated from 30 km × 30 km to 0.5° × 0.5°. The coarse NNRP reanalysis is converted to 0.5° × 0.5° using nearest-neighbor regridding. Fields of the ERA-Interim reanalysis and all observation datasets are already available in 0.5° × 0.5°. For the derivation of the basin-aggregated time series, the spatial averaging is performed with the respective boundary mask for the global and downscaled fields.

3. Results and discussion

a. Validation of the regional atmospheric simulations

Figure 2 compares the statistical properties of the catchment-aggregated temperature and precipitation time series (2003–06) for the Amazon River basin. For precipitation, it is seen that, except for the global reanalyses (E, R), and the WRF configuration (F), a significant periodical overestimation exists. However, for all downscaling results, the correlation coefficients remain within the range of the global reanalyses. The different regional simulations with Interim driving data result in decreased correlation, except for F where only the standard deviation is slightly higher for GPCC. Interim is biased by 8–30 mm month−1 with respect to the gridded observations. The EI SST+KF+FDDA (F) configuration yields a dry bias ranging from −26 to −4 mm month−1. Hence, this configuration of WRF leads to reasonable results within the uncertainty range of the evaluation datasets. NNRP correlates less with the observations. The same is found for all NNRP-driven regional simulations. Therefore, none of these realizations are considered for the comparison with GRACE. For temperature, most of the time series agree closely with the reference data (CRUT). Figure 3 shows the spatial deviation patterns of the time-averaged 2-m temperature of the global reanalysis and the regional simulation with the EI SST+KF+FDDA configuration (F). Obviously, ERA-Interim contains a cold bias over the whole domain (−1.2°C). Here, for most of the basin area, the regional simulation adds skill to the global reanalysis by reducing the bias (−0.2°C). Toward the south, a tendency for overestimation exists. This mismatch is reflected by a different standard deviation in the Taylor diagram (Fig. 2f). Consequently, for the Amazon basin, ERA-Interim and EI SST+KF+FDDA depict the best matching realizations for the atmospheric water budget and are therefore selected for the comparison with GRACE.

Fig. 2.
Fig. 2.

Taylor diagram for the Amazon River basin. The black triangles represent the standard deviations of the reference time series. The distance of the different datasets from the origin denotes the standard deviation, the angle between the x axis and a specific dataset is proportional to the correlation with the reference time series, and the gray circular sectors are the centered RMS differences between the different datasets and the reference time series.

Citation: Journal of Hydrometeorology 13, 5; 10.1175/JHM-D-11-0143.1

Fig. 3.
Fig. 3.

Amazon domain: 2-m temperature difference with respect to CRUT: 2005 annual mean deviation for (a) the global ERA-Interim reanalysis and (b) the dynamical downscaling; 2003–06 mean deviation for (c) ERA-Interim and (d) the downscaling.

Citation: Journal of Hydrometeorology 13, 5; 10.1175/JHM-D-11-0143.1

For Yenisei and Lena, the results for the dynamical downscaling are not very sensitive to the configuration of the regional atmospheric model (Fig. 4, top). For precipitation, the correlation of regional simulations and observations is only in the range of 0.6–0.7. However, as shown in Table 3 the bias of the global reanalyses could be reduced. For the comparison with GRACE both global reanalyses and the EI SST+KF downscaling are selected.

Fig. 4.
Fig. 4.

Taylor diagrams for (top) Yenisei/Lena and (bottom) Lake Chad.

Citation: Journal of Hydrometeorology 13, 5; 10.1175/JHM-D-11-0143.1

Table 3.

Bias in mm month−1 for monthly time series of precipitation (GPCC) and storage change (GRACE GFZ) of global reanalyses and regional simulations (January 2003–December 2006). Interim and NNRP are the global reanalyses, and EI and NR represent different configurations of regional simulations using WRF.

Table 3.

The atmospheric models lead to similar results for the Lake Chad basin and for the Niger (see Fig. 4, bottom). For the global reanalyses, ERA-Interim gives the best coherence. Contrarily, for the dynamical downscaling, the NNRP-driven WRF with the BMJ cumulus scheme yields the best performance. In the Taylor diagram strong coherence is also present for the constant sea surface temperature configuration with NNRP driving (NR KF) but this configuration includes a higher bias in terms of precipitation (Table 3). Therefore, Interim and NR SST+BMJ are selected for the comparison with GRACE.

The Sahara basin has only a few meteorological stations that contribute to the gridded observation products. Therefore, it is difficult to find a conclusion for the validation of the atmospheric fields. The correlations are mainly below 0.5. The Interim- and NNRP-driven regional simulations perform similarly for P and T. NNRP contains a cold and dry bias of −1.9°C and −5.4 mm month−1. For the atmospheric uncertainty bounds, Interim and EI SST+BMJ are selected.

Likewise, for the central Australian plain, also the BMJ cumulus scheme leads to the most reasonable results. Although the bias for NR KF is better than for NR SST+BMJ, the realization should be rejected as the correlation with the precipitation observations reaches only 0.5 compared to 0.75 for EI SST+BMJ. The global reanalyses are much closer to the observations than the regional simulations as they tend to overestimate summer precipitation.

Altogether, the validation indicates the potential for improvement of the global reanalyses by dynamical downscaling. While the correlation of the basin-averaged time series does not change much, skill is often added with respect to the bias. Many configurations of the regional atmospheric model have to be rejected because of their insufficient representation of the water budget.

b. Evaluation of GRACE

1) Amazon basin

Figure 5 displays the water storage variations that were derived from the atmospheric water budget approach. The vertical bars depict the range in the GRACE products of GFZ, CSR, and JPL. According to the validation of the atmospheric approach, the global reanalysis of ERA-Interim (E) and the EI SST+KF+FDDA configuration (F) (Fig. 2) of the regional downscaling are selected for the determination of atmospheric uncertainty bounds with respect to the evaluation with GRACE. The global NNRP reanalysis (R) is rejected because of the poor performance with respect to precipitation.

Fig. 5.
Fig. 5.

Monthly water storage change for the Amazon River basin. Solid line: global reanalysis, ERA-Interim. Dotted line: regional downscaling with WRF. G500: smoothed with 500-km Gaussian filter. Filled region: range from different GRACE products (GFZ, CSR, and JPL).

Citation: Journal of Hydrometeorology 13, 5; 10.1175/JHM-D-11-0143.1

The moisture-flux-derived storage change rates indicate dryer conditions as it is the case for GRACE. The selected regional downscaling (F) is very similar to its global counterpart. During the winter season, the regional downscaling yields lower storage input rates.

From the beginning to the peak of the rainy season, GRACE fits well to the computed storage change rates (r ≈ 0.9). However, for several periods, GRACE shows a wide range of variations while the atmospheric water budgets have rather small uncertainty bounds.

For the different GRACE products, the 2003–06 mean rates of storage change are close to zero (–1.1 to 0.8 mm month−1). However, the long-term storage change is assumed to be balanced, but the global and downscaled ERA-Interim reanalyses yield a negative bias of −9.8 to −24.35 mm month−1 for the unfiltered fields and 17.6 to −21.6 mm month−1 for the Gaussian filtered fields, respectively. For the global reanalysis, the spatial filtering affects the storage input rate stronger than the output rates. This leads to an increase of the dry bias. The effect could be caused by the different spatial and temporal characteristics of precipitation and evapotranspiration. Precipitation concentrates often on small spatial extents while evapotranspiration is more uniform over wide areas. The low pass filter affects an inhomogeneous field stronger than a homogeneous one. In general, the application of the Gaussian filter with a 500-km radius increases the coherence between GRACE and the regionally downscaled storage change rates. In particular, the peak rates are damped toward GRACE.

2) Combined basins of Yenisei and Lena

As Fig. 6 shows, for the combined catchments of Yenisei and Lena, the uncertainty of the moisture-flux-derived water storage change rates is significantly smaller than for the Amazon River. The results of the global and regional atmospheric water budgets are very uniform. Even if different driving data or different configurations of the regional model are used, only minor variations in the time series are observed. On the contrary, the storage change rates derived from NNRP and Interim show a much higher range. In particular, for the August–October months, NNRP leads to elevated amounts of storage depletion with respect to ERA-Interim. For Yenisei and Lena, the validation for precipitation (Fig. 4) revealed poor performance of the global NNRP reanalysis. Thus, this dataset is not considered for the definition of the atmospheric uncertainty bounds. In Fig. 6, the time series for NNRP is included to provide some idea of the amplitude of the deviation.

Fig. 6.
Fig. 6.

Monthly water storage variations for the combined river catchments of Lena and Yenisei. Solid line: global reanalysis, ERA-Interim. Dotted line: regional downscaling with WRF. G500: smoothed with 500-km Gaussian filter. NNRP G500 for comparison. Filled region: range from different GRACE products (GFZ, CSR, and JPL).

Citation: Journal of Hydrometeorology 13, 5; 10.1175/JHM-D-11-0143.1

The range of the three different GRACE products (CSR, GFZ, and JPL) varies over time between 10 and 40 mm month−1. The satellite observations resemble the global and regional atmospheric models in principal but with higher uncertainties. The correlation coefficient yields 0.77. For some periods, GRACE suggests reduced diminishing rates. For the winter season of 2005 and 2006, the uncertainty bounds of the atmospheric approach are significantly smaller than the range of the GRACE datasets. If only the months July–January are considered, the correlation of GRACE with the aerological water budget increases to r ≈ 0.86.

For the Siberian domain, the evaluation of GRACE yields a relatively high range as compared to the reanalyses. Apart from certain individual months, spatial filtering does not have a remarkable influence on the atmospheric water budget.

3) Lake Chad basin

The results for the Lake Chad basin are illustrated in Fig. 7.

Fig. 7.
Fig. 7.

Monthly water storage variations for the African domain. Discharge measurements were not available for the Niger after 2005.

Citation: Journal of Hydrometeorology 13, 5; 10.1175/JHM-D-11-0143.1

According to GRACE and the global reanalyses, the monthly storage change rates oscillate from −50 to 50 mm month−1. For the Chad basin, the regional downscaling of NNRP is selected as it outperforms the ERA-Interim scenario. The global reanalysis of ERA-Interim is more reasonable in terms of precipitation. Thus, for the comparison with GRACE, the atmospheric uncertainty arising from the global Interim and the downscaled NNRP fields is considered. The CSR and JPL GRACE time series are very congruent. The GFZ GRACE solution contains a stronger amplitude and is responsible for the peak in 2004. The overall uncertainty bounds for GRACE range between 10 and 50 mm month−1. The global reanalyses have their largest variability in early summer. Most of the deviations are caused by the NNRP time series. For GRACE, the minimum and maximum peaks in 2004 are clearly out of the atmospheric uncertainty bounds. For the specific months, including July–October 2004 and also December 2006, GRACE follows a repeating orbit track. This leads to a coarser spatial sampling rate and thus increased uncertainties (Flechtner 2011).

As for the Siberian domain, spatial smoothing has only little to no impact on the basin-aggregated time series.

4) Niger basin

The time series for the Niger catchment of Fig. 7 is limited to the period of 2003–05 because discharge data is not available for the subsequent years.

Concerning the global reanalysis fields, ERA-Interim resembles the GFZ and CSR GRACE products best. NNRP is not included in the atmospheric uncertainty bounds as significant limitation was revealed when simulated precipitation was compared to independent GPCC data. As the ERA-Interim evaluation with observed precipitation shows a disagreement for the regional downscaling, the summer months, which show major differences to independent GPCC data, should not be considered for the comparison with GRACE. For the remaining time periods, the uncertainty bounds are comparable to the range of the different GRACE products. However, the rates for GRACE are mostly below those of the reanalyses. The uncertainty bound for the different GRACE products varies between 2 mm month−1 in winter and spring and 40 mm month−1 during the rainy season.

As with the Lake Chad basin, the spatial filter does not significantly affect the global reanalysis and the reasonable periods of the downscaling. However the general effect is a gain in coherence with the GRACE-derived estimates of dS/dt.

5) Sahara basin

For the Sahara basin, the variation in water storage is usually very small. As shown in Fig. 7, for the three GRACE products, the overall span is −20 to 40 mm month−1. GRACE is approximately centered around zero and defined by a 10–20-mm uncertainty range. The global reanalyses show a larger and more variable span. The global time series of NNRP is omitted as no coherence exists with the precipitation observations. As depicted in Fig. 8, the different GRACE products show significant differences. For GFZ and JPL the time series correlation coefficient is around zero, and for GFZ and CSR a value of 0.6 is observed. The standard deviation is similar for both CSR and GFZ. It seems that the JPL solution is inaccurate for regions where only small mass variations occur. Therefore, it is assumed that only CSR and GFZ provide reasonable water storage dynamics for the Sahara basin. Altogether, the GRACE estimates and the atmospheric uncertainty bounds show a similar extent. Thus, for such small storage change rates, both approaches provide a common level of uncertainty.

Fig. 8.
Fig. 8.

Time series of the different GRACE datasets for the Sahar basin (Fig. 7).

Citation: Journal of Hydrometeorology 13, 5; 10.1175/JHM-D-11-0143.1

Spatial smoothing significantly alters the global ERA-Interim time series and introduces a dry bias. For the downscaled reanalysis, the effects are only minuscule.

6) Australian basin

For the central Australian basin, the typical variations in water storage are in the same range to those of the Sahara basin. Figure 9 shows the time series for GRACE and the best-performing global (Interim) and dynamically downscaled reanalyses (EI SST+KF+FDDA). In addition, the NNRP time series is also plotted. EI SST+KF+FDDA yields similar performance to NR SST+BMJ for the correlation of P. However, the respective bias for EI SST+KF+FDDA is smaller.

Fig. 9.
Fig. 9.

Monthly water storage variations for the central Australian basin.

Citation: Journal of Hydrometeorology 13, 5; 10.1175/JHM-D-11-0143.1

Storage recharge takes place, usually, from December to February or March. For the rest of the year, the storage change rate is negative. GRACE varies between −28 and +40 mm month−1 with an interproduct uncertainty range of 2–27 mm month−1 for the 2003–06 time series of the GRACE ensemble.

The comparison with precipitation observations confirms that for the summer recharge periods around December and January, the regional downscaling tends to overestimate the water storage input for the northern part of Australia. The overestimation patterns of the regional simulations differ in their extension toward the south and affect the basin-aggregated water budget for the central plane. By constraining the regional model through nudging it toward the global model the water budgets improve. It seems that some important large-scale features are not properly resembled in the regional simulation leading to a wrong representation of summer precipitation.

As with the Sahara basin, the uncertainty bounds of the atmospheric water budget approach exceed the range of the GRACE estimates for southern summer and equal the range of the GRACE ensemble during winter.

7) Global comparison

Figure 10 presents the statistical properties of the global GRACE-derived estimates of dS/dt and of the reanalysis approach. The comparison is restricted to 33 regions that are larger than 100 000 km2 and have data for R available. Furthermore, the global reanalyses are analyzed before and after the application of the 500-km Gaussian filter. Table 4 gives an overview of the included hydrological basins.

Fig. 10.
Fig. 10.

Box and whisker plots for dS/dt from the three GRACE products and the global atmospheric reanalyses for the basins listed in Table 4. G500 denotes the application of the 500-km Gaussian filter. The bottoms and tops of the rectangles represent the 25th and 75th percentile and the horizontal bands near the center are the medians of the different datasets. The circles below and above the boxes represent outliers with a value larger or smaller than 1.5 times the interquartile range of the highest and lowest quartile, respectively.

Citation: Journal of Hydrometeorology 13, 5; 10.1175/JHM-D-11-0143.1

Table 4.

List of hydrological basins used for the statistical comparison shown in Fig. 10. The three sections denote rivers that flow into an ocean, desert basins, and endorheic basins, respectively. The area information depicts the size of the regions as they have been delineated for this study.

Table 4.

As could be expected for long-term means of terrestrial water storage change rates, the GRACE products are symmetric around zero. While this applies also to the unfiltered ERA-Interim reanalysis, NNRP shows a dry tendency. Many more outliers are further observed for NNRP. The spatial filtering of the atmospheric water budget fields also tend toward dryness. In general, the atmospheric approach contains an increased amount of statistical outliers with respect to the satellite data. JPL deviates remarkably from the statistical characteristics of CSR and GFZ. As the underlying background models are identical, it is assumed that the differences among the products are related to the specific estimation strategies of the three data centers.

4. Conclusions

Dynamic downscaling offers a sophisticated method for refining and improving global atmospheric fields. The study showed that skill is gained rather for the bias than for the basin-aggregated time series when compared to global observation datasets. With respect to the NCAR–NCEP reanalysis (NNRP), substantial improvement was seen in terms of the precipitation bias and also for the coherence with the GRACE products. With respect to global atmospheric fields, ECMWF ERA-Interim performed better than NNRP when validating with independent observations of temperature and precipitation. Concerning the driving of ARW-WRF, NNRP yielded the best results for the dry and arid study regions of Lake Chad and Niger while Interim was more suitable for the tropical Amazon region. Except for the Amazon, the Betts–Miller–Janjić cumulus parameterization led to the best results, independent from the global atmospheric driving used. It was shown that, after a thorough evaluation with independent observations, the atmospheric water budget approach is able to provide valuable constraints and uncertainty bounds for a further evaluation with GRACE data.

The comparison of GRACE and the atmospheric-water-budget-derived water storage changes revealed systematic deviations of differing extents, depending on the geographic region and the considered season. It was shown that both the atmospheric water budget approach and GRACE-derived water storage changes are in good agreement for the Amazon basin and also during certain periods for the other study regions. For the combined Yenisei and Lena catchment, the atmospheric bounds are much smaller than the range stretched by the three GRACE solutions. Where both methods match, the true water storage change can be assumed to follow the estimates from GRACE and the atmospheric moisture budgets. On the other hand, both the atmospheric and the satellite-based approach have limitations for regions with small storage change rates such as for the Sahara or the central Australian basin. The presented time series often show deviations in terms of phase shifts and differing amplitudes. However, some coherence is also seen for storage change rates that are below the detection threshold as described in Wahr et al. (2006) with 25–27 mm month−1 for low-latitude regions.

Spatial filtering of the atmospheric moisture budget fields normally leads to a gain in coherence with GRACE. This is especially seen when the water storage change rates are very high, like for the Amazon basin. When the storage variation is low, the Gaussian smoothing can lead to an increased bias that might be due to a higher relative impact of the leakage effect. Despite the problems with the spatial filtering of the moisture divergence field, we think that its application improves the coherence between the datasets and therefore we suggest this method for comparisons of GRACE with the atmospheric water budget approach.

For water budget studies, GRACE offers an important innovative observation instrument. The comparison with the atmospheric water budgets revealed shortcomings and potential for improvement on both sides. A follow-on mission for GRACE should feature a significantly increased resolution in order to foster mesoscale applications.

Acknowledgments

This work is supported by the DFG (Deutsche Forschungsgemeinschaft KU 2090/1–2) and contributes to the Special Priority Program Mass Transport and Mass Distribution in the System Earth (www.massentransporte.de). The authors thank the anonymous reviewers for their helpful comments and advice. Furthermore, we acknowledge ECMWF and NCEP for providing the global atmospheric reanalysis fields. Special thanks go to Gianpaolo Balsamo and Pedro Viterbo from ECMWF who performed the computation of the moisture convergence fields for ERA-Interim. We are also thankful to Henry Kindt, and Johannes Riegger from IWS Stuttgart, for the collection and preparation of the runoff time series, and to Christof Lorenz and Richard Foreman from KIT/IMK-IFU for proofreading and intensive discussions. This study benefited from open source software like the WRF model (www.mmm.ucar.edu/wrf/users), R, GMT, and Kile.

REFERENCES

  • Adler, R. F., and Coauthors, 2003: The Version-2 Global Precipitation Climatology Project (GPCP) Monthly Precipitation Analysis (1979–present). J. Hydrometeor., 4, 11471167.

    • Search Google Scholar
    • Export Citation
  • Awan, N. K., Truhetz H. , and Gobiet A. , 2011: Parameterization-induced error characteristics of MM5 and WRF operated in climate mode over the alpine region: An ensemble-based analysis. J. Climate, 24, 31073123.

    • Search Google Scholar
    • Export Citation
  • Berbery, E. H., Rasmusson E. M. , and Mitchell K. E. , 1996: Studies of North American continental-scale hydrology using Eta model forecast products. J. Geophys. Res., 101 (D3), 73057319.

    • Search Google Scholar
    • Export Citation
  • Bettadpur, S., 2007: CSR level-2 processing standards document for level-2 product release 0004, 17 pp. [Available online at http://isdc.gfz-potsdam.de/index.php?name=UpDownload&req=getit&lid=405.]

  • Borge, R., Alexandrov V. , José del Vas J. , Lumbreras J. , and Rodríguez E. , 2008: A comprehensive sensitivity analysis of the WRF model for air quality applications over the Iberian Peninsula. Atmos. Environ., 42, 85608574, doi:10.1016/j.atmosenv.2008.08.032.

    • Search Google Scholar
    • Export Citation
  • Chen, J. L., Wilson C. R. , Famiglietti J. S. , and Rodell M. , 2007: Attenuation effect on seasonal basin-scale water storage changes from GRACE time-variable gravity. J. Geod., 81, 237245, doi:10.1007/s00190-006-0104-2.

    • Search Google Scholar
    • Export Citation
  • Duan, X. J., Guo J. Y. , Shum C. K. , and van der Wal W. , 2009: On the postprocessing removal of correlated errors in GRACE temporal gravity field solutions. J. Geod., 83, 10951106, doi:10.1007/s00190-009-0327-0.

    • Search Google Scholar
    • Export Citation
  • Fersch, B., 2011: Large scale water balance estimation from downscaled atmospheric moisture budgets and evaluation with global climatological data sets and the GRACE spaceborne gravimetry. Ph.D. dissertation, University of Augsburg, 117 pp.

  • Fersch, B., Kunstmann H. , Sneeuw N. , and Devaraju B. , 2009: Large-scale water balance estimations through regional atmospheric moisture flux modelling and comparison to GRACE signals. Proc. Symp. HS.2, Hyderabad, India, IAHS and IAH, 211–219. [Available online at http://iahs.info/redbooks/333.htm.]

  • Flechtner, F., 2007: GFZ level-2 processing standards document for level-2 product release 0004. GRACE GFZ L-2 Proc Standards Doc. RL 0004, 19 pp. [Available online at http://isdc.gfz-potsdam.de/index.php?name=UpDownload&req=getit&lid=401.]

  • Flechtner, F., 2011: Release notes for GFZ RL04 GRACE L2 products. GFZ Potsdam Tech. Information, 8 pp.

  • Gao, H., Tang Q. , Ferguson C. R. , Wood E. F. , and Lettenmaier D. P. , 2010: Estimating the water budget of major US river basins via remote sensing. Int. J. Remote Sens., 31, 39553978, doi:10.1080/01431161.2010.483488.

    • Search Google Scholar
    • Export Citation
  • Güntner, A., Schmidt R. , and Döll P. , 2007: Supporting large-scale hydrogeological monitoring and modelling by time-variable gravity data. Hydrogeol. J., 15, 167170, doi:10.1007/s10040-006-0089-1.

    • Search Google Scholar
    • Export Citation
  • Hagemann, S., Arpe K. , and Bengtsson L. , 2005: Validation of the hydrological cycle of ERA-40. European Centre for Medium Range Weather Forecasts ERA-40 Project Rep. Series 24, 42 pp.

  • Hirschi, M., Viterbo P. , and Seneviratne S. I. , 2006: Basin-scale water-balance estimates of terrestrial water storage variations from ECMWF operational forecast analysis. Geophys. Res. Lett., 33, L21401, doi:10.1029/2006GL027659.

    • Search Google Scholar
    • Export Citation
  • Hirschi, M., Seneviratne S. I. , Hagemann S. , and Schär C. , 2007: Analysis of seasonal terrestrial water storage variations in regional climate simulations over Europe. J. Geophys. Res., 112, D22109, doi:10.1029/2006JD008338.

    • Search Google Scholar
    • Export Citation
  • Janjić, Z. I., 2000: Comments on “Development and evaluation of a convection scheme for use in climate models.” J. Atmos. Sci., 57, 36863686.

    • Search Google Scholar
    • Export Citation
  • Jones, P., and Harris I. , cited 2008: CRU datasets—CRU TS time-series. [Available online at http://badc.nerc.ac.uk/view/badc.nerc.ac.uk%5F%5FATOM%5F%5Fdataent%5F1256223773328276.]

  • Kain, J. S., 2004: The Kain–Fritsch convective parameterization: An update. J. Appl. Meteor., 43, 170181.

  • Kalnay, E., and Coauthors, 1996: The NCEP/NCAR 40-Year Reanalysis Project. Bull. Amer. Meteor. Soc., 77, 437471.

  • Klees, R., Zapreeva E. A. , Winsemius H. C. , and Savenije H. H. G. , 2007: The bias in GRACE estimates of continental water storage variations. Hydrol. Earth Syst. Sci., 11, 12271241. [Available online at http://www.hydrol-earth-syst-sci.net/11/1227/2007/.]

    • Search Google Scholar
    • Export Citation
  • Labraga, J. C., Frumento O. , and Lopez M. , 2000: The atmospheric water vapor cycle in South America and the tropospheric circulation. J. Climate, 13, 18991915.

    • Search Google Scholar
    • Export Citation
  • Landerer, F. W., Dickey J. O. , and Güntner A. , 2010: Terrestrial water budget of the Eurasian pan-Arctic from GRACE satellite measurements during 2003–2009. J. Geophys. Res., 115, D23115, doi:10.1029/2010JD014584.

    • Search Google Scholar
    • Export Citation
  • Matsuura, K., and Willmot C. J. , cited 2009: Terrestrial precipitation: 1900-2008 gridded monthly time series (version 2.01). [Available online at http://climate.geog.udel.edu/~climate/html_pages/archive.html.]

  • Mitchell, T. D., and Jones P. D. , 2005: An improved method of constructing a database of monthly climate observations and associated high-resolution grids. Int. J. Climatol., 25, 693712.

    • Search Google Scholar
    • Export Citation
  • Peixoto, J. P., and Oort A. H. , 1992: Balance equations for water. Physics of Climate, T. von Foerster, M. Hennelly, and A. Prince, Eds., American Institute of Physics, 274–277.

  • Riegger, J., Tourian M. , Devaraju B. , and Sneeuw N. , 2012: Analysis of GRACE uncertainties by hydrological and hydro-meteorological observations. J. Geodyn., 59–60, 1627, doi:10.1016/j.jog.2012.02.001.

    • Search Google Scholar
    • Export Citation
  • Roads, J., Kanamitsu M. , and Stewart R. , 2002: CSE water and energy budgets in the NCEP–DOE Reanalysis II. J. Hydrometeor., 3, 227248.

    • Search Google Scholar
    • Export Citation
  • Rodell, M., Famiglietti J. S. , Chen J. , Seneviratne S. I. , Viterbo P. , Holl S. , and Wilson C. R. , 2004a: Basin scale estimates of evapotranspiration using GRACE and other observations. Geophys. Res. Lett., 31, L20504, doi:10.1029/2004GL020873.

    • Search Google Scholar
    • Export Citation
  • Rodell, M., and Coauthors, 2004b: The global land data assimilation system. Bull. Amer. Meteor. Soc., 85, 381394.

  • Schmidt, R., and Coauthors, 2006: GRACE observations of changes in continental water storage. Global Planet. Change, 50 (1–2), 112126, doi:10.1016/j.gloplacha.2004.11.018.

    • Search Google Scholar
    • Export Citation
  • Schneider, U., Fuchs T. , Meyer-Christoffer A. , and Rudolf B. , cited 2008: Global precipitation analysis products. [Available online at ftp://ftp-anon.dwd.de/pub/data/gpcc/html/fulldata_v6_download.html.]

  • Seitz, F., Schmidt M. , and Shum C. K. , 2008: Signals of extreme weather conditions in Central Europe in GRACE 4-D hydrological mass variations. Earth Planet. Sci. Lett., 268, 165170, doi:10.1016/j.epsl.2008.01.001.

    • Search Google Scholar
    • Export Citation
  • Skamarock, W. C., and Klemp J. B. , 2008: A time-split nonhydrostatic atmospheric model for weather research and forecasting applications. J. Comput. Phys., 227, 34653485, doi:10.1016/j.jcp.2007.01.037.

    • Search Google Scholar
    • Export Citation
  • Skamarock, W. C., Klemp J. B. , Dudhia J. , Gill D. O. , Barker D. M. , Wang W. , and Powers J. G. , 2008: A description of the Advanced Research WRF version 3. NCAR Tech. Note NCAR/TN-468+STR, 113 pp.

  • Swenson, S., and Wahr J. , 2006a: Estimating large-scale precipitation minus evapotranspiration from GRACE satellite gravity measurements. J. Hydrometeor., 7, 252270.

    • Search Google Scholar
    • Export Citation
  • Swenson, S., and Wahr J. , 2006b: Post-processing removal of correlated errors in GRACE data. Geophys. Res. Lett., 33, L08492, doi:10.1029/2005GL025285.

    • Search Google Scholar
    • Export Citation
  • Swenson, S., Wahr J. , and Milly P. C. D. , 2003: Estimated accuracies of regional water storage variations inferred from the Gravity Recovery and Climate Experiment (GRACE). Water Resour. Res., 39, 1223, doi:10.1029/2002WR001808.

    • Search Google Scholar
    • Export Citation
  • Syed, T. H., Famiglietti J. S. , Zlotnicki V. , and Rodell M. , 2007: Contemporary estimates of Pan-Arctic freshwater discharge from GRACE and reanalysis. Geophys. Res. Lett., 34, L19404, doi:10.1029/2007GL031254.

    • Search Google Scholar
    • Export Citation
  • Syed, T. H., Famiglietti J. S. , and Chambers D. P. , 2009: GRACE-based estimates of terrestrial freshwater discharge from basin to continental scales. J. Hydrometeor., 10, 2240.

    • Search Google Scholar
    • Export Citation
  • Tapley, B. D., Bettadpur S. , Watkins M. , and Reigber C. , 2004: The gravity recovery and climate experiment: Mission overview and early results. Geophys. Res. Lett., 31, L09607, doi:10.1029/2004GL019920.

    • Search Google Scholar
    • Export Citation
  • Taylor, K. E., 2001: Summarizing multiple aspects of model performance in a single diagram. J. Geophys. Res., 106 (D7), 71837192.

  • Trenberth, K. E., and Guillemot C. J. , 1998: Evaluation of the atmospheric moisture and hydrological cycle in the NCEP/NCAR reanalyses. Climate Dyn., 14, 213231, doi:10.1007/s003820050219.

    • Search Google Scholar
    • Export Citation
  • Trenberth, K. E., Smith L. , Qian T. , Dai A. , and Fasullo J. , 2007: Estimates of the global water budget and its annual cycle using observational and model data. J. Hydrometeor., 8, 758769.

    • Search Google Scholar
    • Export Citation
  • Uppala, S., Dee D. , Kobayashi S. , Berrisford P. , and Simmons A. , 2008: Towards a climate data assimilation system: Status update of ERA-Interim. ECMWF Newsletter, No. 115, ECMWF, Reading, United Kingdom, 12–18.

  • Vivoni, E. R., Tai K. , and Gochis D. J. , 2009: Effects of initial soil moisture on rainfall generation and subsequent hydrologic response during the North American monsoon. J. Hydrometeor, 10, 644664.

    • Search Google Scholar
    • Export Citation
  • Wahr, J., Molenaar M. , and Bryan F. , 1998: Time variability of the Earth’s gravity field: Hydrological and oceanic effects and their possible detection using GRACE. J. Geophys. Res., 103 (B12), 30 20530 229.

    • Search Google Scholar
    • Export Citation
  • Wahr, J., Swenson S. , and Velicogna I. , 2006: Accuracy of GRACE mass estimates. Geophys. Res. Lett., 33, L06401, doi:10.1029/2005GL025305.

    • Search Google Scholar
    • Export Citation
  • Wang, W., and Seaman N. L. , 1997: A comparison study of convective parameterization schemes in a mesoscale model. Mon. Wea. Rev., 125, 252278.

    • Search Google Scholar
    • Export Citation
  • Wang, W., and Coauthors, 2009: WRF ARW version 3 modeling system user’s guide, April 2009. National Center for Atmospheric Research, 310 pp.

  • Watkins, M. M., and Yuan D.-N. , 2007: JPL level-2 processing standards document for level-2 product release 04. GRACE JPL L-2 Proc Standards Doc. RL 04, 16 pp. [Available online at http://isdc.gfz-potsdam.de/index.php?name=UpDownload&req=getit&lid=400.]

  • Werth, D., and Avissar R. , 2004: The regional evapotranspiration of the Amazon. J. Hydrometeor., 5, 100109.

  • Werth, S., and Güntner A. , 2010: Calibration analysis for water storage variability of the global hydrological model WGHM. Hydrol. Earth Syst. Sci., 14, 5978, doi:10.5194/hess-14-59-2010.

    • Search Google Scholar
    • Export Citation
  • Winsemius, H. C., Savenije H. H. G. , van de Giesen N. C. , van den Hurk B. J. J. M. , Zapreeva E. A. , and Klees R. , 2006: Assessment of Gravity Recovery and Climate Experiment (GRACE) temporal signature over the upper Zambezi. Water Resour. Res., 42, W12201, doi:10.1029/2006WR005192.

    • Search Google Scholar
    • Export Citation
  • Yeh, P. J.-F., and Famiglietti J. S. , 2008: Regional terrestrial water storage change and evapotranspiration from terrestrial and atmospheric water balance computations. J. Geophys. Res., 113, D09108, doi:10.1029/2007JD009045.

    • Search Google Scholar
    • Export Citation
Save
  • Adler, R. F., and Coauthors, 2003: The Version-2 Global Precipitation Climatology Project (GPCP) Monthly Precipitation Analysis (1979–present). J. Hydrometeor., 4, 11471167.

    • Search Google Scholar
    • Export Citation
  • Awan, N. K., Truhetz H. , and Gobiet A. , 2011: Parameterization-induced error characteristics of MM5 and WRF operated in climate mode over the alpine region: An ensemble-based analysis. J. Climate, 24, 31073123.

    • Search Google Scholar
    • Export Citation
  • Berbery, E. H., Rasmusson E. M. , and Mitchell K. E. , 1996: Studies of North American continental-scale hydrology using Eta model forecast products. J. Geophys. Res., 101 (D3), 73057319.

    • Search Google Scholar
    • Export Citation
  • Bettadpur, S., 2007: CSR level-2 processing standards document for level-2 product release 0004, 17 pp. [Available online at http://isdc.gfz-potsdam.de/index.php?name=UpDownload&req=getit&lid=405.]

  • Borge, R., Alexandrov V. , José del Vas J. , Lumbreras J. , and Rodríguez E. , 2008: A comprehensive sensitivity analysis of the WRF model for air quality applications over the Iberian Peninsula. Atmos. Environ., 42, 85608574, doi:10.1016/j.atmosenv.2008.08.032.

    • Search Google Scholar
    • Export Citation
  • Chen, J. L., Wilson C. R. , Famiglietti J. S. , and Rodell M. , 2007: Attenuation effect on seasonal basin-scale water storage changes from GRACE time-variable gravity. J. Geod., 81, 237245, doi:10.1007/s00190-006-0104-2.

    • Search Google Scholar
    • Export Citation
  • Duan, X. J., Guo J. Y. , Shum C. K. , and van der Wal W. , 2009: On the postprocessing removal of correlated errors in GRACE temporal gravity field solutions. J. Geod., 83, 10951106, doi:10.1007/s00190-009-0327-0.

    • Search Google Scholar
    • Export Citation
  • Fersch, B., 2011: Large scale water balance estimation from downscaled atmospheric moisture budgets and evaluation with global climatological data sets and the GRACE spaceborne gravimetry. Ph.D. dissertation, University of Augsburg, 117 pp.

  • Fersch, B., Kunstmann H. , Sneeuw N. , and Devaraju B. , 2009: Large-scale water balance estimations through regional atmospheric moisture flux modelling and comparison to GRACE signals. Proc. Symp. HS.2, Hyderabad, India, IAHS and IAH, 211–219. [Available online at http://iahs.info/redbooks/333.htm.]

  • Flechtner, F., 2007: GFZ level-2 processing standards document for level-2 product release 0004. GRACE GFZ L-2 Proc Standards Doc. RL 0004, 19 pp. [Available online at http://isdc.gfz-potsdam.de/index.php?name=UpDownload&req=getit&lid=401.]

  • Flechtner, F., 2011: Release notes for GFZ RL04 GRACE L2 products. GFZ Potsdam Tech. Information, 8 pp.

  • Gao, H., Tang Q. , Ferguson C. R. , Wood E. F. , and Lettenmaier D. P. , 2010: Estimating the water budget of major US river basins via remote sensing. Int. J. Remote Sens., 31, 39553978, doi:10.1080/01431161.2010.483488.

    • Search Google Scholar
    • Export Citation
  • Güntner, A., Schmidt R. , and Döll P. , 2007: Supporting large-scale hydrogeological monitoring and modelling by time-variable gravity data. Hydrogeol. J., 15, 167170, doi:10.1007/s10040-006-0089-1.

    • Search Google Scholar
    • Export Citation
  • Hagemann, S., Arpe K. , and Bengtsson L. , 2005: Validation of the hydrological cycle of ERA-40. European Centre for Medium Range Weather Forecasts ERA-40 Project Rep. Series 24, 42 pp.

  • Hirschi, M., Viterbo P. , and Seneviratne S. I. , 2006: Basin-scale water-balance estimates of terrestrial water storage variations from ECMWF operational forecast analysis. Geophys. Res. Lett., 33, L21401, doi:10.1029/2006GL027659.

    • Search Google Scholar
    • Export Citation
  • Hirschi, M., Seneviratne S. I. , Hagemann S. , and Schär C. , 2007: Analysis of seasonal terrestrial water storage variations in regional climate simulations over Europe. J. Geophys. Res., 112, D22109, doi:10.1029/2006JD008338.

    • Search Google Scholar
    • Export Citation
  • Janjić, Z. I., 2000: Comments on “Development and evaluation of a convection scheme for use in climate models.” J. Atmos. Sci., 57, 36863686.

    • Search Google Scholar
    • Export Citation
  • Jones, P., and Harris I. , cited 2008: CRU datasets—CRU TS time-series. [Available online at http://badc.nerc.ac.uk/view/badc.nerc.ac.uk%5F%5FATOM%5F%5Fdataent%5F1256223773328276.]

  • Kain, J. S., 2004: The Kain–Fritsch convective parameterization: An update. J. Appl. Meteor., 43, 170181.

  • Kalnay, E., and Coauthors, 1996: The NCEP/NCAR 40-Year Reanalysis Project. Bull. Amer. Meteor. Soc., 77, 437471.

  • Klees, R., Zapreeva E. A. , Winsemius H. C. , and Savenije H. H. G. , 2007: The bias in GRACE estimates of continental water storage variations. Hydrol. Earth Syst. Sci., 11, 12271241. [Available online at http://www.hydrol-earth-syst-sci.net/11/1227/2007/.]

    • Search Google Scholar
    • Export Citation
  • Labraga, J. C., Frumento O. , and Lopez M. , 2000: The atmospheric water vapor cycle in South America and the tropospheric circulation. J. Climate, 13, 18991915.

    • Search Google Scholar
    • Export Citation
  • Landerer, F. W., Dickey J. O. , and Güntner A. , 2010: Terrestrial water budget of the Eurasian pan-Arctic from GRACE satellite measurements during 2003–2009. J. Geophys. Res., 115, D23115, doi:10.1029/2010JD014584.

    • Search Google Scholar
    • Export Citation
  • Matsuura, K., and Willmot C. J. , cited 2009: Terrestrial precipitation: 1900-2008 gridded monthly time series (version 2.01). [Available online at http://climate.geog.udel.edu/~climate/html_pages/archive.html.]

  • Mitchell, T. D., and Jones P. D. , 2005: An improved method of constructing a database of monthly climate observations and associated high-resolution grids. Int. J. Climatol., 25, 693712.

    • Search Google Scholar
    • Export Citation
  • Peixoto, J. P., and Oort A. H. , 1992: Balance equations for water. Physics of Climate, T. von Foerster, M. Hennelly, and A. Prince, Eds., American Institute of Physics, 274–277.

  • Riegger, J., Tourian M. , Devaraju B. , and Sneeuw N. , 2012: Analysis of GRACE uncertainties by hydrological and hydro-meteorological observations. J. Geodyn., 59–60, 1627, doi:10.1016/j.jog.2012.02.001.

    • Search Google Scholar
    • Export Citation
  • Roads, J., Kanamitsu M. , and Stewart R. , 2002: CSE water and energy budgets in the NCEP–DOE Reanalysis II. J. Hydrometeor., 3, 227248.

    • Search Google Scholar
    • Export Citation
  • Rodell, M., Famiglietti J. S. , Chen J. , Seneviratne S. I. , Viterbo P. , Holl S. , and Wilson C. R. , 2004a: Basin scale estimates of evapotranspiration using GRACE and other observations. Geophys. Res. Lett., 31, L20504, doi:10.1029/2004GL020873.

    • Search Google Scholar
    • Export Citation
  • Rodell, M., and Coauthors, 2004b: The global land data assimilation system. Bull. Amer. Meteor. Soc., 85, 381394.

  • Schmidt, R., and Coauthors, 2006: GRACE observations of changes in continental water storage. Global Planet. Change, 50 (1–2), 112126, doi:10.1016/j.gloplacha.2004.11.018.

    • Search Google Scholar
    • Export Citation
  • Schneider, U., Fuchs T. , Meyer-Christoffer A. , and Rudolf B. , cited 2008: Global precipitation analysis products. [Available online at ftp://ftp-anon.dwd.de/pub/data/gpcc/html/fulldata_v6_download.html.]

  • Seitz, F., Schmidt M. , and Shum C. K. , 2008: Signals of extreme weather conditions in Central Europe in GRACE 4-D hydrological mass variations. Earth Planet. Sci. Lett., 268, 165170, doi:10.1016/j.epsl.2008.01.001.

    • Search Google Scholar
    • Export Citation
  • Skamarock, W. C., and Klemp J. B. , 2008: A time-split nonhydrostatic atmospheric model for weather research and forecasting applications. J. Comput. Phys., 227, 34653485, doi:10.1016/j.jcp.2007.01.037.

    • Search Google Scholar
    • Export Citation
  • Skamarock, W. C., Klemp J. B. , Dudhia J. , Gill D. O. , Barker D. M. , Wang W. , and Powers J. G. , 2008: A description of the Advanced Research WRF version 3. NCAR Tech. Note NCAR/TN-468+STR, 113 pp.

  • Swenson, S., and Wahr J. , 2006a: Estimating large-scale precipitation minus evapotranspiration from GRACE satellite gravity measurements. J. Hydrometeor., 7, 252270.

    • Search Google Scholar
    • Export Citation
  • Swenson, S., and Wahr J. , 2006b: Post-processing removal of correlated errors in GRACE data. Geophys. Res. Lett., 33, L08492, doi:10.1029/2005GL025285.

    • Search Google Scholar
    • Export Citation
  • Swenson, S., Wahr J. , and Milly P. C. D. , 2003: Estimated accuracies of regional water storage variations inferred from the Gravity Recovery and Climate Experiment (GRACE). Water Resour. Res., 39, 1223, doi:10.1029/2002WR001808.

    • Search Google Scholar
    • Export Citation
  • Syed, T. H., Famiglietti J. S. , Zlotnicki V. , and Rodell M. , 2007: Contemporary estimates of Pan-Arctic freshwater discharge from GRACE and reanalysis. Geophys. Res. Lett., 34, L19404, doi:10.1029/2007GL031254.

    • Search Google Scholar
    • Export Citation
  • Syed, T. H., Famiglietti J. S. , and Chambers D. P. , 2009: GRACE-based estimates of terrestrial freshwater discharge from basin to continental scales. J. Hydrometeor., 10, 2240.

    • Search Google Scholar
    • Export Citation
  • Tapley, B. D., Bettadpur S. , Watkins M. , and Reigber C. , 2004: The gravity recovery and climate experiment: Mission overview and early results. Geophys. Res. Lett., 31, L09607, doi:10.1029/2004GL019920.

    • Search Google Scholar
    • Export Citation
  • Taylor, K. E., 2001: Summarizing multiple aspects of model performance in a single diagram. J. Geophys. Res., 106 (D7), 71837192.

  • Trenberth, K. E., and Guillemot C. J. , 1998: Evaluation of the atmospheric moisture and hydrological cycle in the NCEP/NCAR reanalyses. Climate Dyn., 14, 213231, doi:10.1007/s003820050219.

    • Search Google Scholar
    • Export Citation
  • Trenberth, K. E., Smith L. , Qian T. , Dai A. , and Fasullo J. , 2007: Estimates of the global water budget and its annual cycle using observational and model data. J. Hydrometeor., 8, 758769.

    • Search Google Scholar
    • Export Citation
  • Uppala, S., Dee D. , Kobayashi S. , Berrisford P. , and Simmons A. , 2008: Towards a climate data assimilation system: Status update of ERA-Interim. ECMWF Newsletter, No. 115, ECMWF, Reading, United Kingdom, 12–18.

  • Vivoni, E. R., Tai K. , and Gochis D. J. , 2009: Effects of initial soil moisture on rainfall generation and subsequent hydrologic response during the North American monsoon. J. Hydrometeor, 10, 644664.

    • Search Google Scholar
    • Export Citation
  • Wahr, J., Molenaar M. , and Bryan F. , 1998: Time variability of the Earth’s gravity field: Hydrological and oceanic effects and their possible detection using GRACE. J. Geophys. Res., 103 (B12), 30 20530 229.

    • Search Google Scholar
    • Export Citation
  • Wahr, J., Swenson S. , and Velicogna I. , 2006: Accuracy of GRACE mass estimates. Geophys. Res. Lett., 33, L06401, doi:10.1029/2005GL025305.

    • Search Google Scholar
    • Export Citation
  • Wang, W., and Seaman N. L. , 1997: A comparison study of convective parameterization schemes in a mesoscale model. Mon. Wea. Rev., 125, 252278.

    • Search Google Scholar
    • Export Citation
  • Wang, W., and Coauthors, 2009: WRF ARW version 3 modeling system user’s guide, April 2009. National Center for Atmospheric Research, 310 pp.

  • Watkins, M. M., and Yuan D.-N. , 2007: JPL level-2 processing standards document for level-2 product release 04. GRACE JPL L-2 Proc Standards Doc. RL 04, 16 pp. [Available online at http://isdc.gfz-potsdam.de/index.php?name=UpDownload&req=getit&lid=400.]

  • Werth, D., and Avissar R. , 2004: The regional evapotranspiration of the Amazon. J. Hydrometeor., 5, 100109.

  • Werth, S., and Güntner A. , 2010: Calibration analysis for water storage variability of the global hydrological model WGHM. Hydrol. Earth Syst. Sci., 14, 5978, doi:10.5194/hess-14-59-2010.

    • Search Google Scholar
    • Export Citation
  • Winsemius, H. C., Savenije H. H. G. , van de Giesen N. C. , van den Hurk B. J. J. M. , Zapreeva E. A. , and Klees R. , 2006: Assessment of Gravity Recovery and Climate Experiment (GRACE) temporal signature over the upper Zambezi. Water Resour. Res., 42, W12201, doi:10.1029/2006WR005192.

    • Search Google Scholar
    • Export Citation
  • Yeh, P. J.-F., and Famiglietti J. S. , 2008: Regional terrestrial water storage change and evapotranspiration from terrestrial and atmospheric water balance computations. J. Geophys. Res., 113, D09108, doi:10.1029/2007JD009045.

    • Search Google Scholar
    • Export Citation
  • Fig. 1.

    Catchment boundaries and distribution of the selected study regions.

  • Fig. 2.

    Taylor diagram for the Amazon River basin. The black triangles represent the standard deviations of the reference time series. The distance of the different datasets from the origin denotes the standard deviation, the angle between the x axis and a specific dataset is proportional to the correlation with the reference time series, and the gray circular sectors are the centered RMS differences between the different datasets and the reference time series.

  • Fig. 3.

    Amazon domain: 2-m temperature difference with respect to CRUT: 2005 annual mean deviation for (a) the global ERA-Interim reanalysis and (b) the dynamical downscaling; 2003–06 mean deviation for (c) ERA-Interim and (d) the downscaling.

  • Fig. 4.

    Taylor diagrams for (top) Yenisei/Lena and (bottom) Lake Chad.

  • Fig. 5.

    Monthly water storage change for the Amazon River basin. Solid line: global reanalysis, ERA-Interim. Dotted line: regional downscaling with WRF. G500: smoothed with 500-km Gaussian filter. Filled region: range from different GRACE products (GFZ, CSR, and JPL).

  • Fig. 6.

    Monthly water storage variations for the combined river catchments of Lena and Yenisei. Solid line: global reanalysis, ERA-Interim. Dotted line: regional downscaling with WRF. G500: smoothed with 500-km Gaussian filter. NNRP G500 for comparison. Filled region: range from different GRACE products (GFZ, CSR, and JPL).

  • Fig. 7.

    Monthly water storage variations for the African domain. Discharge measurements were not available for the Niger after 2005.

  • Fig. 8.

    Time series of the different GRACE datasets for the Sahar basin (Fig. 7).

  • Fig. 9.

    Monthly water storage variations for the central Australian basin.

  • Fig. 10.

    Box and whisker plots for dS/dt from the three GRACE products and the global atmospheric reanalyses for the basins listed in Table 4. G500 denotes the application of the 500-km Gaussian filter. The bottoms and tops of the rectangles represent the 25th and 75th percentile and the horizontal bands near the center are the medians of the different datasets. The circles below and above the boxes represent outliers with a value larger or smaller than 1.5 times the interquartile range of the highest and lowest quartile, respectively.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 744 411 1
PDF Downloads 167 47 0