A Parallel Ensemble Kalman Filter for Four-dimensional Land Data Assimilation

Successful weather and climate forecasts may depend on the accurate initialization of the land surface states. To obtain land initial conditions, land model integrations and observations of land surface states can be merged using advanced data assimilation algorithms such as the Kalman filter. We present an ensemble-based four-dimensional assimilation algorithm that can ingest land surface observations into the Catchment Model of the NASA Seasonal-to-Interannual Prediction Project (NSIPP). The assimilation algorithm is fully parallelized. Ensemble integration between update times is inherently parallel. Sub-regions of the domain are updated in parallel using covariance localization (or compact support) techniques. A short overview of the algorithm is given. We also present preliminary results of retrospective soil moisture estimates that have been derived by assimilating soil moisture retrievals from the space-borne Scanning Multichannel Microwave Radiometer (SMMR) into the NSIPP Catchment Model.