Assessing the Influence of Spatially Correlated Errors on Assimilation of GRACE-derived Total Water Storage Anomalies into a Global Hydrological Model

Abstract:

Monthly total water storage anomaly (TWSA) data from the Gravity Recovery and Climate Experiment (GRACE) mission have been used in a number of studies to improve modelled hydrological water states via data assimilation or parameter calibration. To this end, data assimilation is usually implemented through ensemble Kalman filter/smoother (EnKF/S) approaches, in which the modelled water states are updated by column-integrated TWSA with the assumption that the errors are uncorrelated. However, due to the peculiar GRACE orbital geometry, measurement principle and processing chain, TWSA fields exhibit spatially correlated errors at all spatial scales, even after applying decorrelation filters and after aggregation. Two synthetic experiments are performed in this study to assess the influence of the spatially correlated errors of GRACE TWSA within an evolving data assimilation and calibration framework. Here, an EnKF method (Eicker et al., 2014) is used to adjust the water states of the Global Hydrology WaterGAP Model (WGHM) towards synthetic GRACE-like TWSA and, simultaneously, calibrate model parameters. 'True' and 'uncertain' model states are defined by interchanging the climate forcing input fields of WGHM. Then, synthetic TWSA data are contaminated by adding (i) spatially uncorrelated Gaussian random noise, and (ii) spatially correlated GRACE-like errors simulated by projecting correlated potential coefficient errors to (sub-) basin means of TWSA. Both synthetic data sets (TWSA signal + errors) are introduced in the EnKF, and the influence on the updated water storage values, fluxes and calibration parameters is assessed by comparing with the simulated 'truth'. Our results indicate significant influence of the spatially correlated errors mainly over small (sub-) basins with north-south spatial extensions.