Correction of Satellite Sea Surface Salinity Products Using Ensemble Learning Method

Although salinity satellites can provide high-resolution global sea surface (SSS) data, the satellite data still display large errors close to coast. In this paper, a nonlinear empirical method based on random forest is proposed correct two Soil Moisture and Ocean Salinity (SMOS) L3 products in tropical Indian Ocean, including SMOS BEC CATDS data. The agreement between in-situ corrected better than that original root-mean-square deviation (RMSD) of SSS decreased from 0.366 0.275 0.367 0.255 for CATDS, respectively. effect correction model was Arabian Sea Bay Bengal. RMSD (CATDS) reduced 0.44 (0.48) 0.276 (0.269), correlation coefficient increased 0.915 0.741(0.801) while improved less 0.02 cross-validation results highlight robustness effectiveness model. Additionally, effects different features are discussed demonstrate vital role geographical information outperformed other machine-learning methods with respect coefficient.