Improving soil moisture retrievals from a physically-based radiative transfer model

Near surface soil moisture is being estimated from space-borne passive microwave observations through inverting a physically-based radiative transfer model (RTM), the land surface microwave emission model (LSMEM) at Princeton University for the past several years. The existing retrieval scheme utilizes only the horizontal (H) polarization measurement from a single channel (10.65. GHz). This physically-based approach requires a relatively large number of parameters, and it generally suffers from large biases/errors due to the difficulty in determining the correct parameters. This study characterizes these errors in order to improve the retrieval performance. Through model sensitivity analysis, this study finds that a dual polarization approach (using both horizontal and vertical polarizations) is needed to infer the correct vegetation opacity and correct polarization mixing measured by the space-borne sensor. Revisions are then made to the LSMEM formulations and soil moisture retrieval algorithm by 1) combining two vegetation parameters and one roughness parameter into one effective vegetation optical depth (VOD) value; and 2) providing an additional model equation that estimates the effective VOD from both polarizations and an initial guess of soil moisture value. The new retrieval algorithm is implemented to produce a daily 0.25° gridded soil moisture dataset based on observations from the Advanced Microwave Scanning Radiometer-EOS (AMSR-E). Validations are performed globally against land surface model simulations and at local/point scale against in-situ data within the continental United States. The new retrievals are shown to have good and robust performance over most parts of the world in terms of reproducing the spatial and temporal dynamics of soil moisture.