DEM-assisted in-season soil moisture estimation based on normalized Sentinel-1 SAR imagery

Abstract

Soil moisture is a crucial in-field variable used in many applications. Soil moisture might be measured in the field using soil sensors and can be estimated via satellite imagery. The present study proposes an innovative SAR-based method that significantly improves the accuracy of soil moisture estimation and does not require field-measured data. The method is based on the previously developed SAR local incidence angle normalization method and utilizes a newly developed equation, which takes a digital elevation model DEM into account. The volumetric water content (VWC) measurements were recorded at depths of 20 and 46 cm on 10 alfalfa fields in the US by 37 soil sensors. Recorded VWC data was correlated to the average field values of SAR imagery processed by the proposed method. The developed models have the following statistical performance: R2 = 0.5616 with RMSE = 3.9758 for VWC at 20 cm and R2 = 0.4247 with RMSE = 4.0133 for VWC at 46 cm. In both cases, the improvement of R2 of models based on the proposed method over models based on SAR imagery, which were not processed by the new method, was significant.