The aim of the present study is focusing on a validation and comparison among eight different physical models for soil salinity mapping in arid landscape. The considered models were developed for different geographic regions around the world, i.e. Latino- America (Mexico), Middle-East (Iraq), north and east Africa (Morocco and Ethiopia) and Asia (China). These models integrated different spectral bands and unlike mathematical functions in their conceptualization (stepwise, linear, second order, logarithmic, and exponential). Three main steps were considered. The Landsat-OLI image data was radiometrically standardized and the models were implemented to derive soil salinity maps. The field survey was organized during 4 days, two days before the OLI data acquisition, and a total of 100 soil samples were collected representing different salinity levels, and each sampling location was geographically localized using accurate GPS. The laboratory analysis was accomplished to derive electrical conductivity (EC-Lab) for validation purposes. Statistical analysis (p < 0.05) was applied between predicted salinity maps (EC-Predicted) and the measured ground truth (EC-Lab). The results obtained showed that predictive models based on VNIR bands and vegetation indices are inadequate for soil salinity prediction due to a serious signals confusion between the salt-crust and the soil optical properties in these spectral bands. The statistical tests revealed insignificant fits (R2 ≤ 0.41) with a very high prediction errors (R2 ≥ 0.41). While, the model based on second order polynomial function and integrating the SWIR bands provides results of best fitness in comparison to the ground truth, yielding an R2 of 0.97 and low overall RMSE of 13%.