Assimilation of GRACE Derived Terrestrial Water Storage into a Hydrological Land Surface Model

Matthew Rodell, Isabelle Velicogna, Jeffrey P. Walker and Rolf H. Reichle

The Gravity Recovery and Climate Experiment (GRACE) provides highly accurate satellite observations of Earth’s gravity field, from which will be inferred changes in terrestrial water storage at regional scales on a monthly basis. Thus satellite gravimetry may prove to be extremely beneficial for hydrological, meteorological, and climatological research and applications. However, the nature of the derived hydrological fields, in particular their low spatial and temporal resolutions and lack of information on the vertical distribution of water, will necessitate their disaggregation in space and time, using auxiliary knowledge, in order to maximize their value. One potential downscaling approach is to constrain a land surface model (LSM) with data derived from GRACE. LSMs embody our current state of knowledge of hydrological processes, observations provide a basis in reality, and data assimilation can be used to synthesize the two. In this study we used an ensemble Kalman filter to assimilate simulated GRACE derived terrestrial water storage data into the Mosaic LSM. To simplify this initial experiment we limited the region to a 3x3 model grid with all water stored in three soil layers. Errors due to uncertainty in the GRACE data, atmospheric mass, and “leakage” across the boundaries of the study region were considered, but temporal “aliasing” due water storage variations at frequencies higher than the GRACE overpass rate and certain other complicating factors were ignored, to be revisited in future.