This study assesses the impact of satellite-rainfall error structure on the efficiency of assimilating near-surface soil moisture observations. Specifically, the study contrasts a multi-dimensional satellite rainfall error model to a simpler rainfall error model currently used to generate rainfall ensembles as part of the ensemble-based Land Data Assimilation System (LDAS) developed at NASA Global Modeling and Assimilation Office. The study is conducted in the Oklahoma region using rainfall data from NOAA's multi-satellite global rainfall product (CMORPH) and the NWS rain gauge-calibrated radar rainfall product (WSR-88D) representing the ‘uncertain' and ‘reference' model rainfall data forcing, respectively. Soil moisture simulations, obtained by forcing the Catchment Land Surface Model (CLSM) with ‘reference' rainfall, are randomly perturbed to represent satellite retrieval uncertainty, and then used in LDAS as synthetic near surface soil moisture observations. The assimilation estimates show improved performance metrics, exhibiting higher anomaly correlation coefficients (e.g. ~0.79 in open-loop experiments, ~0.89 in assimilation experiments) and lower root mean square errors (e.g. ~0.036 m3/m3 in open-loop experiments, ~0.027 m3/m3 in assimilation experiments). The more elaborate rainfall error model in the assimilation system leads to moderately improved assimilation estimates. Investigation is underway to verify the above findings assimilating actual satellite soil moisture data, i.e. the AMSR-E Land Parameter Retrieval Model (LPRM) product.