Sea Surface Salinity Retrieval from Aquarius in the South China Sea Using Machine Learning Algorithm

Aquarius is a satellite mission to measure sea surface salinity (SSS) from space using a combined passive/active L-band instrument. It has mapped SSS nearly 4 years since its launch in 2011. In this paper, four Machine Learning Algorithms applied without considering the physical effects of the forward model are proposed to retrieve SSS from Aquarius measurements and associated auxiliary data. The results are compared with multisource SSS data, including HYCOM SSS, Aquarius Remote Sensing Systems (RSS) SSS products, Aquarius Combined Active-Passive (CAP) SSS products, and the interpolated monthly Scripps Institution of Oceanography Argo salinity (Scripps SSS) in the South China Sea. The analysis shows that the performance of Deep Neural Network (DNN) algorithm is better than the other machine algorithms. Compared to HYCOM and Scripps SSS, the root-mean-squared error (RMSE) of the DNN algorithm is smaller than Aquarius RSS and CAP SSS products in the South China Sea.