A Semiphysical Microwave Surface Emission Model for Soil Moisture Retrieval

This study proposes a microwave surface emission model for soil moisture retrieval using radiometer data based on today´s most widely used physical model, i.e., advanced integral equation model (AIEM). Soil roughness and moisture effects are easily yet accurately decoupled in the proposed model. In the field case study, the total least squares method, instead of the least squares (LS) method, is applied for the first time in soil moisture retrieval to solve the error in variable linear equation set to further reduce the estimation error. Validated by the Soil Moisture Experiment 2003 campaign data in Oklahoma, the root mean square error (RMSE) and R² of volumetric soil moisture varies from 1.5% to 4.2% and 0.92 to 0.43 at L/C/X bands and 40/55° incidence angles. Compared with previous studies, the proposed model has several new features: 1) it is location independent since the model is derived through reproducing the behavior of the AIEM; 2) its high fidelity to AIEM significantly improves the accuracy, whereas its linearity makes it easy to invert; and 3) the soil moisture retrieval based on the proposed model requires no prior knowledge of soil roughness in the scenario of the demonstrated case study. The L-band/V-polarization radiometer data yield the best retrieval result with an RMSE of 1.5% and R² of 0.92, whereas increasing frequency increases the error because the sensitivity of emissivity to ground soil moisture decreases, and the valid roughness region, i.e., kh_RMS <; 3, of the AIEM narrows. Furthermore, the model can be readily extended to broader regions than the investigated case study on field scale in this paper by nesting the model in the τ - ω model and using satellite data from SMOS or SMAP.