Download the paper:
Abstract:
The efficiency of assimilating near-surface soil moisture retrievals from AMSR-E observations in a land data assimilation system (LDAS) is assessed using satellite rainfall forcing and two different satellite rainfall error models: a complex, multi-dimensional model (SREM2D) and the simpler model (CTRL) used in the NASA GEOS-5 LDAS. For the study domain of Oklahoma, LDAS soil moisture estimates improve over the satellite retrievals and the open-loop (no assimilation) land surface model estimates, exhibiting higher daily anomaly correlation coefficients (e.g., 0.36 in the open-loop, 0.38 in the AMSR-E, and 0.50 in LDAS for surface soil moisture). The LDAS soil moisture estimates also match the performance of a benchmark model simulation forced with high-quality radar precipitation. Compared to the CTRL rainfall error model, the more complex SREM2D exhibits only slight improvements in soil moisture estimates.