Inversion of surface scattering models: comparing criteria and algorithms to estimate bare soil parameters

Estimates bare soil geophysical parameters (i.e. standard deviation of roughness height s, length of correlation l and soil moisture m_v) using polarimetric SAR data. Two direct models have been implemented to simulate the radar measurements and were introduced in an inversion scheme. The methods have been tested on simulated data where speckle effects and model errors were taken into account. A correlation between the roughness parameters, as it can be expected in real measures, has been imposed in the simulation in order to evaluate its effect on the estimation accuracy. Different inversion algorithms based on a Bayesian approach, developed with simple optimization-integration procedures have been compared to neural networks. The comparison has assessed the accuracy achievable by different radar system configurations. Polarimetric data acquired during MAC Europe and SIR-C campaigns, over selected bare soil fields, has been processed to show preliminary estimation results.