Modeling the backscattering coefficient of random rough surfaces to improve soil moisture retrieval from polarimetric SAR data

FRANCO, M.; VILLA, JULIAN; GRINGS, FRANCISCO; BARBER, MATIAS

Cambridge

Encuentro; 15th Specialist Meeting on Microwave Radiometry and Remote Sensing of the Environment (MicroRad 2018); 2018

p { margin-bottom: 0.1in; line-height: 120%; }a:link { }Thescattering properties of surfaces are of fundamental importance togenerate retrieval algorithms in the context of remote sensing (atypical example is soil moisture retrieval from SAR data). Due to theinherent stochastic nature of natural surfaces, the scatteringfunction of these kind of targets is usually computed through eitherperturbative or approximated methods. From the latter, one of mostcommonly used is the Kirchhoff -or tangent plane-  approximation(KA). Kirchhoff approximation is widely used because its ease ofimplementation. This holds if the induced currents on the surface arecomputed using stationary phase approximation. However, under suchapproach, null cross-polarization is achieved in the backscatteringcondition, which is not supported by observations. This isparticularly relevant in the context of SAR polarimetry, since asurface modeled as having zero cross-polarization is associated witha rank-reduced coherency matrix and zero entropy values. Therefore,in order to fully exploit polarimetric SAR data, it is paramount topropose a model for the scattering of natural surfaces characterizedby a non-zero cross-polarization response.Inthis paper, we propose a scattering model for random rough surfacesbased on KA, that predicts non-zero cross-polarization inbackscattering. This is achieved by computing the overall scatteringthrough an ensemble of periodic surfaces with both random amplitudesand frequencies. The advantage of using this scheme is that forperiodic surfaces KA can be implemented without any furtherapproximation (such as phase stationary) which imply non zerocross-pol in backscattering. Therefore, the full-rank coherencymatrix of the random surface can be modeled and surface?spolarimetric features can be studied from a theoretical point of viewand compared with available observations. In this paper, we comparedsimulations based on the proposed model with observations from UAVSARdata, in order to evaluate to which degree the proposed approach isable to explain polarimetric features observed for real surfaces.