A new SMOS sea surface salinity retrieval method

Sea surface roughness and foam have great influence on the retrieval accuracy of sea surface salinity (SSS) from satellite measured L-band brightness temperature (TB). A lot of work need to be done to improve the SSS retrieval accuracy especially using the forward model to reduce the error induced by sea surface roughness and foam. In this paper, seven factors, whitecap coverage, significant wave height (SWH), wave steepness, wavelength, sea surface temperature (SST), foam thickness, Water droplets density, their second-order items, and interaction items are selected as our model input candidates, and the Least Absolute Shrinkage and Selection Operation (LASSO) method is used for critical factor selection. As a result, two factors, SST and SWH, are carefully picked up to establish the new model to calculate the TB variants with quadratic curve regression formulas. The model shows that the mean absolute error (MAE) is 1.09psu and the root mean square (RMSE) is 1.70psu compared with Argo SSS data, in the meanwhile, the MEA of SMOS L2 SSS is 1.69psu and the RMSE is 2.48psu respectively. Results show that the new model can improve the retrieval accuracy for SMOS SSS measurements.