Salinity is an important physical oceanographic parameter which influences many processes in the ocean. A multi-linear regression model and artificial neural network are developed for sea surface salinity and tested against in situ measurements in the Bohai Sea. The models are validated by in situ data. The root mean square errors are 0.858 psu and 0.689 psu (psu stands for practical salinity unit) with the correlation coefficients of R2=0.81 and R2=0.82, respectively. Then the model is applied to MERIS data to derive sea surface salinity map which can generally characterize the spatial pattern of sea surface salinity of the Bohai Sea. Both models are suitable for multi-spectral remote sensing data.