A Bayesian methodology for soil parameters retrieval from sar images

MATÍAS BARBER; MARCELA PISCITELLI; PABLO PERNA; CINTIA BRUSCANTINI; FRANCISCO GRINGS; JULIO JACOBO BERLLES; HAYDEE KARSZENBAUM

Vancouver

Simposio; International Geoscience & Remote Sensing Symposium; 2011

Soil moisture retrieval from SAR data presents two main sources ofuncertainty: terrain heterogeneity and speckle noise. In this paper,these issues will be addressed by using a Bayesian approach. Sucha Bayesian approach (1) needs only a forward model (no retrievalmodel required), (2) gives the optimal unbiased estimator for thesoil moisture and its error and (3) can include as many error sourcesas required. Through numerical simulations, a standard Oh retrievalprocedure and the Bayesian approach were tested for different number of looks (n = 3 and n = 64). The results indicate that for a large number of looks the region of validity of both approaches are similar. Furthermore, contrary to the Oh model retrieval procedure which is only valid in a bounded region of the (hh, vv, hv)-space, the Bayesian approach gives an estimation of soil moisture and its error for any combination of hh, vv and hv, so enlarging the region where the retrieval is possible.