Water salinity is one of the most critical water properties which considerably affects the lives of marine flora and fauna. In this study, the water salinity of Lake Urmia was mapped using sentinel-2 Multispectral Images (MSI). A Support Vector Regression (SVR) was developed to predict the water salinity using sentinel-2 spectral bands and indices. Three main scenarios were considered when input features were used in the SVR model. In scenario 1, the SVR was fed by all the features generated from Sentinel-2 data, and in the other two scenarios, a Genetic Algorithm (GA) and a Sequential Feature Selection (SFS) were applied to select the optimum input features to be used in the SVR model. The results showed that the salinity of Lake Urmia was estimated with a relatively reliable accuracy using GA along with the SVR model, where the R2 of 65.7% and the Root Mean Square Error (RMSE) of 11.5 PSU were obtained when the results were compared with in-situ data. Overall, this study showed that Sentinel-2 provides valuable high spatial-temporal datasets for continuous monitoring of water salinity over Lake Urmia.