Abstract:
This study assesses the impact of satellite-rainfall error structure on the efficiency of assimilating soil moisture in a land data assimilation system. Specifically, the study contrasts a multi-dimensional satellite rainfall error model (SREM2D) to the standard rainfall error model used to generate rainfall ensembles as part of the Land Data Assimilation System developed at the NASA Global Modeling and Assimilation Office (NASA-LDAS). The study is conducted in the Oklahoma region, which presents a good coverage by weather radars and multi-year satellite rainfall products. We used high-resolution satellite rainfall fields derived from the NOAA CMORPH global satellite product and rain gauge-calibrated radar rainfall fields (considered as reference rainfall). The soil moisture simulations, produced by forcing the land surface model with the radar unperturbed rainfall, are used in the LDAS as synthetic observations of near surface soil moisture, which are produced using a random field generator representing satellite retrieval error of near surface soil moisture. The land surface model that has been chosen is the NASA Catchment Land Surface Model (CLSM; Koster et al., 2000) and the data assimilation framework is the system used at the NASA GMAO (Reichle et al., 2007) that utilizes the ensemble Kalman filter (EnKF). Comparisons of assimilation experiments against simulations of soil moisture generated by the land surface model show an improvement in the error statistics, with higher anomaly correlation coefficients and lower root mean squared errors and biases. In terms of the rainfall error modeling, we show that the multi-dimensional error-structure in SREM2D generates rainfall replicates with higher variability that better envelope the reference rainfall than those generated by the NASA-LDAS error model. Investigation is underway to show the impact of this enhanced rainfall error model complexity on the effectiveness of near surface soil moisture assimilation.