Remote Sensing Retrieval of Sea Surface Salinity Based on RBF Neural Network From SMAP Satellite

In order to address the problems of traditional Sea Surface Salinity remote sensing inversion methods, which have poor inversion accuracy and few influencing factors, this paper proposes a new retrieval method using SMAP L2C data, Argo data, and auxiliary data, with a focus on (160°E—120°W, 0°—30°N) in the Pacific Ocean. The proposed method considering more influence factors such as sea surface roughness and whitecap coverage, and firstly establishes an RBF bright temperature increment model using RBF neural network and modifies the brightness temperature of the flat sea surface, and then retrieves the SSS based on the Meissner-Wentz dielectric constant model. The validation results showed that the root mean square error of model predicted salinity and SMAP satellite salinity relative to Argo measured salinity were 0.4 and 0.5, respectively, and the mean absolute error were 0.3 and 0.4, respectively. Experiments show that the brightness temperature increment model established by RBF neural network can improve the accuracy of Sea Surface Salinity inversion, and has practical significance for Sea Surface Salinity inversion.