On the retrieval of soil moisture using multi parameter SAR data: Simulated results

The objective of the paper is to assess the feasibility of retrieving soil moisture over bare fields using multi parameter SAR data at C and L-band. The retrieval algorithm consists of a Neural Network (NN) trained by the Integral Equation (IEM) model based on single and multi-scale roughness description. In addition, an appropriate technique to improve the NN robustness vs measurement or model errors is applied. The algorithm performances in terms of attainable accuracy in the retrieved soil moisture content is assessed as a function of the different simulated SAR configurations. Moreover, the impact of calibration and model errors on the NN performances is investigated.