N. Sneeuw¹, A. Bardossy², J. Riegger², H. Kunstmann³, B. Devaraju¹, M.J. Tourian², B. Fersch³, C. Lorenz³

¹ Institute of Geodesy, Universität Stuttgart;

² Institut für Wasserbau, Universität Stuttgart;

³ Institut für Meteorologie und Klimaforschung (IMK-IFU), Karlsruhe Institute of Technology.

GRACE gravity measurements provide a direct measure of water storage changes over continents. As such, it enables---for the first time---to close the continental water balance on large scales, and a direct determination of actual evapotranspiration---the unknown component of water balance---from terrestrial precipitation and run-off measurements on large scales. Atmospheric moisture flux offers another independent way of determining water storage changes, where there is no need for evapotranspiration information. This allows for a mutual inter-comparison of data from three independent disciplines and an evaluation of hydrological and atmospheric models.

Figure 1: Conceptual framework of the project and the activities involved in different phases of the project.

The overall objectives of the project are

1. the direct analysis of large-scale water balances, and

2. the quantification of related uncertainties for large-scale catchment areas in different climatic zones.

In order to achieve consistent water balances, the mass change rates from GRACE, hydrology, and hydrometeorology have to be evaluated with respect to natural fluctuations and intrinsic errors. Statistical investigations are needed to characterize the respective contributions.

The time-series of storage change in the catchment was computed by taking area-weighted average of the catchment pixels. The ensuing analysis showed that the type of filtering did not matter for large catchments, but for small catchments different filters produces different storage change estimates. Further, an analysis of the errors based on catchment size showed that for catchments (> 250,000 km2) there is a constant error, but for larger catchments the error decreases with increasing area.

Figure 2: Time-series of catchment averages for a small (Aravalli) and a large (Ob) catchment. The right-most panel shows the dependency of error behaviour on the size of the catchment. The catchments are arranged in ascending order, and the errors seem to decrease with the size of catchment after a certain point. This point of inflection coincides with 250,000 sq. km

Vertically integrated moisture flux divergence (VIMFD) provides an independent way of evaluating GRACE mass storage changes (Fig. 1). VIMFD is typically computed from atmospheric reanalysis models. A further refinement is obtained by the application of the regional atmospheric model WRF (Weather Research and Forecast modeling system). These models produce temperature, precipitation, and other datasets in addition to VIMFD. Therefore, prior to evaluating GRACE data, the models are validated with independent datasets of precipitation and temperature. The validated atmospheric fields provide constraints for the terrestrial water storage variations. Figure 3 shows the uncertainty bounds for terrestrial water storage change estimates as derived from the global reanalyses and regional downscaling with WRF. For many periods it is seen that the range in GRACE solutions is larger than for the atmospheric water budget approach. This provides valuable constraints for the evaluation of GRACE basin time series.

Figure 3: Time-series of storage change estimates from global and regional downscaled atmospheric models, and uncertainty bounds from different GRACE products (GFZ, CSR, JPL) for the Amazon basin.

When quantifying uncertainties, positive and negative peaks are detected in GRACE aggregated time series that do not correspond to hydrological or atmospheric signals. These peaks must be interpreted as outliers, which carry the danger of signal degradation. An algorithm is developed to identify outliers and replace them with appropriate values. The procedure of outlier detection is verified by evaluating catchment averaged signals with ground truth from hydrology and atmospheric signals. The results show improvement in the correlation of versus atmospheric and hydrological signals over 255 catchments. Also, the noise level is significantly reduced.

Figure 4: a) Mass deviation (mm), dM/dt (mass derivative) from GFZ, P-R and -VIMFD-R (VIMFD - Verically integrated moisture flux divergence), over Okavango catchment. b) Computed evapotranspiration by P-R-dM/dt over the Okavango catchment. Evapotranspiration computed from GRACE is negative, which cannot be true as evapotranspiration is always positive.

1. Improved GRACE storage change estimates consistent with observed hydrology and modelled hydro-meteorology data.

2. Closing terrestrial water balance gauged as well as ungauged catchments.

3. Consistent estimation of evapotranspiration over gauged and ungauged cacthments.

4. Quantifying run-off from altimetry data.

5. Quantifying uncertainty bounds for all the datasets involved.

6. Determination of run-off for ungauged catchments.

Refereed conference proceedings

Devaraju B. and Sneeuw N. (2009): Performance analysis of isotropic spherical harmonic spectral windows. In: Prooceedings of the VII Hotine-Marussi Symposium on Theoretical Geodesy, Rome, Italy (6.-10.7.2009), accepted

Devaraju B., Sneeuw N., Kindt H. and Riegger J. (2008): Estimating GRACE monthly water storage change consistent with hydrology. In: Proceedings of the IAG symposium on Gravity, Geoid, and Earth Observation 2008. Chania, Crete, Greece (23.-27.6.2008)

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., in: New Approaches to Hydrological Prediction in Data-Sparse Regions, IAHS publ. 333

Iran Pour S. and Sneeuw N. (2009): Properties and applications of EOF-based filtering of GRACE solutions. In: Proceedings of the VII Hotine-Marussi Symposium on Theoretical Geodesy, Rome, Italy (6.-10.7.2009)

Conference presentations and posters

Devaraju B., Sneeuw N., Tourian M.J., Riegger J., Fersch B. and Kunstmann H. (2010): Assimilating GRACE, hydrology and hydro-meteorology datasets for estimating monthly water storage changes over continents. Geophysical Research Abstracts, vol. 12: EGU2010-11171, 2010. EGU General Assembly, Vienna, Austria (2.-7.5.2010).

Devaraju B., Tourian M.J., Sneeuw N. and Riegger J. (2009): Inter-catchment correlation estimates from filtered GRACE monthly solutions. Geodätische Woche Karlsruhe (22.-24.9.2009)

Devaraju B. and Sneeuw N. (2008): Impact on filtering on the spatial covariance structure of GRACE. Geodätische Woche Bremen (30.9.-2.10.2008)

Devaraju B., Sneeuw N., Kindt H. and Riegger R. (2008): Estimating GRACE monthly water storage change consistent with hydrology. IAG International Symposium on Gravity, Geoid and Earth Observations (GGEO 2008), Chania, Crete, Greece (23.-27.6.2008) (Poster)

Devaraju B., Sneeuw N., Kindt H., Riegger J. and Lorenz C. (2008): Constraining monthly GRACE solutions with hydrological mass estimates. Geophysical Research Abstracts, vol. 10: EGU2008-A-00761. EGU General Assembly, Vienna, Austria (13.-18.4.2008) (Poster)

Devaraju B. and Sneeuw N. (2007): Stochastic averaging of GRACE data. Geodätische Woche 2007, Leipzig, (25.-27.9.2007) (Poster)

Fersch B. and Kunstmann H. (2010), Water balance estimation for large scale basins from regional atmospheric moisture budgets and comparison to GRACE. EGU General Assembly 2010, European Geophysical Union, Vienna, EGU2008-A-06976.

Fersch B., Kunstmann H., Sneeuw N. and Devaraju B. (2009): Continental scale atmospheric and terrestrial water budget modelling and comparison to GRACE. EGU General Assembly 2010, European Geophysical Union,Vienna, EGU2009-10219.

Fersch B., Kunstmann H. and Bárdossy A. (2008): Regional atmospheric mass and moisture flux modeling and comparison to GRACE signals. EGU General Assembly 2010, European Geophysical Union, Vienna, EGU2008-A-06976.

Fersch B. (2008): Implementation of a routing scheme for the regional atmospheric model WRF. 12. 12th Workshop on large-scale hydrological modeling. Kiel, Germany.

Iran Pour S., Bentel K. and Sneeuw N. (2009): EOF-based filtering of GRACE gravity field solutions: A comparison between spectral and spatial approaches. Geophysical Research Abstracts, vol. 11: EGU2009-2853-1. EGU General Assembly, Vienna, Austria (19.-24.4.2009) (Poster)

Iran Pour S., Sneeuw N. and Devaraju B. (2009): The EOF-based filtering of GRACE solutions - Properties and applications. Geodätische Woche Karlsruhe (22.-24.9.2009)

Kindt H., Riegger J., Bárdossy A., Devaraju B. and Sneeuw N. (2008): Evaluation of GRACE Measurements with Hydrologic Data . In EGU General Assembly 2010, volume 10 of Geophysical Research Abstracts, Vienna. European Geophysical Union

Lorenz C, Devaraju B. and Sneeuw N. (2009): On the computation of a reliable signal covariance for the stochastic filtering of time-variable gravity field from GRACE. Geophysical Research Abstracts, vol. 11: EGU2009-4677. EGU General Assembly, Vienna, Austria (19.-24.4.2009)

Riegger J., Kindt H., Bárdossy A., Sneeuw N. and Devaraju B. (2008): Möglichkeiten zur Charakterisierung von großskaligen Grundwasserressourcen durch Satelliten-Gravimetrie. In FH-DGG Konferenz 2008, Germany

Tourian M.J., Riegger J., Devaraju B. and Sneeuw N. (2010a): Investigation on Inter Cell Correlations of GRACE monthly solutions over the globe . In EGU General Assembly 2010, volume 12 of Geophysical Research Abstracts, Vienna. European Geophysical Union

Tourian M.J., Sneeuw N. and Riegger J. (2010c): Feasibility study of extracting runoff data from satellite altimetry over continental surface waters. In EGU General Assembly 2010, volume 12 of Geophysical Research Abstracts, Vienna. European Geophysical Union

Reports, theses

Bentel K (2009): Empirical orthogonal function analysis of GRACE gravity data. Diploma Thesis, Institute of Geodesy, University of Stuttgart

Lorenz C (2009): Applying stochastic constraints on time-variable GRACE data. Diploma Thesis, Institute of Geodesy, University of Stuttgart

From l to r: Nico Sneeuw, Mohammad Tourian, Balaji Devaraju, Harald Kunstmann, Benjamin Fersch and Johannes Riegger