The Terrestrial System Modeling Platform (TerrSysMP) is a massively parallel, fully integrated soil-vegetation-atmosphere modeling system. However, models are built on a simplified representation of reality, which leads to limited predicting skills of simulated water storages and fluxes. The assimilation of total water storage anomalies from the satellite mission GRACE allows for reducing uncertainties of hydrological model output, while simultaneously disaggregating coarse GRACE data spatially and temporally. My research addresses the optimization of the assimilation strategy for integrating GRACE data into the land-surface component of TerrSysMP over Europe. Main challenges arise through the correlated error structure of GRACE observations, the representation of model errors through appropriate ensembles, and the different spatial and temporal resolution of model and observations.