Ensemble of regressors for soil moisture retrieval in agricultural fields

This paper presents an approach to improve the capability to retrieve soil moisture information from SAR data. More in details the proposed approach consider different inversion approaches and outlines a procedure how to combine the results derived from these regressors with the main aim to improve the accuracy in the estimation of the target variables. The approach was tested in the case of fully polarimetric AirSAR images acquired over agricultural fields covered with soybean and corn crops. The single regressors were an empirical and a Bayesian approach. The approaches were applied to C and L band images and also to a combination of both frequencies. The results indicate that when the retrieved information from the regressors are properly combined based on the select figures of merit such as R² and RMSE the accuracy can improve up to around 30%.