Statistical Models of Sea Surface Salinity in the South China Sea Based on SMOS Satellite Data

Abstract

Study of sea surface salinity (SSS) plays an important role in the marine ecosystem, estimation of global ocean circulation and observation of fisheries, aquaculture, coral reef, and sea grass habitats. Three statistical methods applied without considering the physical effects of the input parameters are proposed to calculate SSS from soil moisture and ocean salinity (SMOS)-measured brightness temperature (TB) values and associated auxiliary data. Using these three statistical methods, named multiple linear regression (MLR) model, principal component regression (PCR) model, and quadratic polynomial regression (QPR) model, the first predictions of daily and monthly averaged SSS are made with $1 ^\circ\times1 ^\circ$ spatial resolution in the South China Sea (SCS, in the study area of 4°N-25 °N, 105°E-125°E) during the period between April and June 2013. Results are compared with the corresponding SMOS SSS products and Aquarius SSS products and validated using Argo measurements. Validation results show that the root-mean-squared error (RMSE) of the QPR model is around 0.46 practical salinity units (psu) compared to 0.58 psu for Aquarius daily SSS products. World Ocean Atlas (WOA13) SSS data are also used for validation in the SCS and the QPR model gives a 0.54-psu value of RMSE, which may be compared with 0.69 psu, 0.73 psu for SMOS and Aquarius Level-3 (L3) SSS products, respectively.