Salinity retrieval from SMOS brightness temperatures

The neural network methodology is applied to the sea surface salinity retrieval from SMOS brightness temperatures. The direct model for simulating the brightness temperatures is the Small Slope Approximation model (SSA). Different cases are compared to analyze the retrieval quality. The effect of a bias on the brightness temperatures and of the instrumental accuracy expected on the SMOS measurements are evaluated. The inversion algorithm is improved when adding ancillary parameters (sea surface temperature, wind speed and a priori salinity).