Model-based methods for soil moisture estimations from SAR Data

In this paper, two model-based methods for the soil moisture retrieval from SAR data are investigated. These methods implicitly consider the physical theory relating the direct relationships between geophysical parameters and SAR measurements and, moreover, can incorporate a priori information to make the parameter estimation more accurate. Given that the inverse problem of recovering soil moisture from SAR observations doesn't have a unique solution, the proposed methods perform a probabilistic estimation of such parameter, finding solutions representative of an unknown probabilistic distribution such as the mean or the most probable solutions. The methods are a Neural Network based-methods and a Mixture Model method. The difference of the solution found by these methods are discussed. Moreover, a simulations about soil moisture estimations from ERS and ENVISAT ASAR data are presented.