1. Assessment of Basin-scale Terrestrial Water Storage Variations from Reprocessed GRACE Gravity Fields for Climate Model Validation L. Zhang, H. Dobslaw, F. Flechtner, M. Thomas German Research Centre for Geosciences, GFZ, Germany Reliable predictions for the evolution of the climate on time-scales from seasons up to a few years are of utmost importance for a number of economic and social questions. Besides various oceanic properties, soil moisture is believed to be a key quantity for skillful predictions (Meehl et al. 2009). The Gravity Recovery and Climate Experiment (GRACE) is now widely accepted to provide valuable insights into the terrestrial water storage dynamics of the largest discharge basins. In this contribution, the mission's observations are therefore tested for their ability to contribute to the validation of coupled climate model hindcast experiments that are available from the CMIP5 inter-comparison project. In order to do so, we calculated basin-scale mass anomalies for a number of basins that include Amazon Danube and Columbia, from the newly released 05 GRACE gravity Before calculating basin- scale averages, degree-1 Stokes coefficients were included from a combination of GRACE data and simulated ocean bottom pressure (Swenson et al. 2008), and correlated errors were largely reduced by applying an anisotropic filter introduced by Kusche (2007). Finally, basin averages were calculated by a spatiospectral concentration(SP) method based on Slepian basis functions (e.g., Simmons et al. 2006). This method amplifies the signal-to-noise ratio over the particular geographical region and is particularly useful for the assessment of smaller basins. With newly released GRACE data and this processing method, we would also like to investigate how the regional hydrological estimates might be improved. ensed with the satellite gravity mission GRACE provide a promising way to broaden our observational basis. Besides monthly mean gravity fields that provide ocean bottom pressure variations averaged over 30 days, several alternative GRACE products with higher temporal resolution have been developed during the most recent years. These include monthly mean gravity fields from GFZ, CSR and JPL, 10-day solutions from GRGS Toulouse, as well as constrained daily solutions from the University of Bonn which have been obtained GRACE GFZ Release 05 data(2005—2010) updated Level-1B instrument data Improved background models modified processing standards  Degree1 terms added: based on the method of Swenson et al. (2008) with RL05 GRACE gravity data combined with RL05 Atmosphere-Ocean Dealiasing OBP from the GAD files  Destriping method: DDK3 (Approximate decorrelation method from Kusche et al. (2009)  Basin function: spatiospectral concentration method based on Slepian basis functions from Simmons et al. (2006) www.gfz-potsdam.de AmazonDanubeColumbia Unit(cm)BiasLeakageBiasLeakageBiasLeakage Gaussian051.881.074.072.954.052.37 SP(DDK3)0.540.301.790.312.120.35 Bias and leakage from GLDAS Annual variation comparison between GRACE (Gaussian/SP(DDK3)) and LSDM Abstract Datasets Processing Methods + Basin function Synthetic test(GLDAS ) Comparison Conclusion  The maximum errors of the water equivalent height estimation caused by missing the degree-1 Coefficients can reach 13mm.  For the GFZ RL05 GRACE data, the difference between the three versions of decorrelation filters with different filter radius is not very big at large basins, while at small basins, result with DDK3 show better agreement with LSDM model.  Spatiospectral concentration performs better than Gaussian averaging method to estimate the basin average values, especially at smaller basins  The total water storage estimated from GRACE show good agreements with the results from LSDM model at the chosen basins, which shows the ability of GRACE to detect the water storage variations at small basins with the processing method applied here. liangjing.zhang@gfz-potsdam.de Amazon Danube Processing method Gauss05: Processed with the GFZ RL 05 GRACE data (following the same) and basin function of Gaussian average method SP05(DDK3): Processed with basin function of SP method and destriping method of Kusche DDK3 SP05(DDK2): Processed with basin function of SP method and destriping method of Kusche DDK2 SP05(DDK1): Processed with basin function of SP method and destriping method of Kusche DDK1 LSDM: The basin average estimated directly from LSDM model Legends illustration Columbia AmazonDanubeColumbia Amplitude(cm) LSDM17.34-18.176.107.7711.086.64 SP(DDK3)19.52-23.066.198.711.394.18 Gauss0517.49-23.045.589.478.171.90 SP05(DDK3)SP05(DDK2)SP05(DDK1) Gauss05 LSDM