An empirical model of radar backscattering on multiscale bare soil surfaces

The electromagnetic response of a surface to a radar incident wave is highly dependent on its roughness and moisture content. Soil moisture is an important measure for crop yield prediction, mainly for countries at risk of drought. In addition, the humidity control allows a quick detection of dewatering agricultural areas. Radar remote sensing of soil moisture is also an essential tool in flood prediction by indicating soil water saturation. The existent relationship that links radar backscatter and soil erosion is the surface roughness. Soil erosion presents a significant threat against the health and productivity of agricultural resources. SAR imagery is useful for monitoring soil erosion over large areas. The objectives of this work concern the characterization of rough surfaces in order to extract its geophysical parameters as well as the modeling of the influence of the SAR resolution (in C-band) on the electromagnetic response of a rough surface. Moreover, an adapted multilayered neural network inversion procedure trained by a back propagation learning rule has been carried out for the purpose of retrieving these physical parameters. Indeed, radar backscatter models do not always reliably simulate the behaviour of the radar signal, which makes the inversion processes quite difficult. An adapted multiscale small perturbations model, and the Dubois Model, have been used to describe radar backscattering response of semiarid surfaces for both horizontal and vertical co-polarizations.