Extended analysis of SMOS salinity retrieval by using support vector regression (SVR)

The performances of a novel salinity retrieval strategy by means of Support Vector Regression (SVR) have been studied. The SMOS satellite brightness temperatures measurements and additional auxiliary parameters have been co-located with salinity data collected by Argo buoys, which represented the ground-truth to be matched by the algorithm. Salinity fields estimated by the SVR with a combination of features and at two polarizations are in good agreement with the ground-truth data. A critical additional feature that was studied was the TB/emissivity incidence angle dependency, analyzing the possible performances improvements due to the multi-angular information content. The robustness and versatility of this approach are under further assessment over wider areas and time lags, and in various combinations of SMOS features.