Results Analysis of Coastal Regions Sea Surface Salinity Retrieval from Aquarius Mission Using Deep Neural Network.

Sea surface salinity (SSS) is of great significance for studying the global water cycle and climate change. Aquarius is a satellite dedicated to measuring SSS from space. It equipped the active/passive instrument to combination measure SSS from space. This paper proposes an algorithm using machine learning of depth neural network (DNN) to retrieve SSS in coastal regions based on the Aquarius V5 Level-2 (L2) Data Product. The retrieval results are compared with HYCOM SSS and Scripps Institution of Oceanography Argo salinity (Scripps SSS). Compared with HYCOM SSS and Scripps SSS, the average root square error (RMSE) of retrieved SSS for three incident angles are 0.39psu and 0.40psu, and are 1.64psu and 1.87psu for Aquarius SSS product. The results show that compared with the physical retrieval algorithm, the DNN has better accuracy for SSS retrieval in coastal regions.