Global retrieval of surface soil moisture using L-band SMAP SAR data and its validation

Surface soil moisture retrievals using radar observations have been challenging due to the strong effects by surface roughness and vegetation scattering. Physically-based forward models for radar scattering are inverted using time-series retrieval algorithm to systematically correct for the effect of the roughness and vegetation. The retrievals are performed for a top 5-cm layer soil moisture at 3-km spatial resolution using the L-band radar data acquired by the Soil Moisture Active Passive (SMAP) satellite globally every three days from mid-April to early July, 2015. These were assessed over 13 rigorously-chosen core validation sites covering a wide range of biomass types and amounts and soil conditions. The soil moisture retrieval performance had an accuracy of approaching the goal of 0.063 m3/m3 unbiased-RMSE (root mean square error), a near zero bias, and a correlation of 0.56. The successful retrieval demonstrates that the physically-based retrieval method is capable of characterizing soil moisture over diverse conditions of soil moisture, surface roughness, and vegetation on a global scale.