Investigations of soil moisture inversion from polarimetric radar responses based on integral equation model

A polarimetric surface scattering model based on the integral equation model is used to study the dependence of backscattering coefficient on surface roughness and dielectric constant in a wide range of surface conditions as well as incident frequencies and angles. Then the sensitivity of the surface parameters to polarization states was investigated to aim at selection of optimum polarization angles which minimize the effects of surface roughness in order to extract the surface roughness and soil moisture. Because the inversion can be highly nonlinear, a dynamic learning neural network was adopted. When surface rms slope is less than 0.5, the cross-pol. response is negligible compared to co-pol. response. Useful data channels are thus selected from the co-pol. polarimetric responses as inputs to neural network for training and operation. It is found that the use of polarimetric data, particularly when only single angle-single frequency data available, increases the range moisture content and thus significantly enhances the capability of soil moisture retrieval from microwave remote sensing data