I had experience in estimating soil salinity using regression models. In general, highly saline soils were not well fitted by most models. One explanation may be the unique spectral features of these soils, such as the existence of hygroscopic salt (soil moisture interference) and salt crust (always highly reflective). The phenomenon is similar to the saturation of NDVI.

Thank you Xingwang, I am finding the same thing. Was just perpelexed with the early saturation of the salinity indices. Best regards

Yes, it is normal because soil salinity indices are only simple indicators for probable salt-affected soil but not for correct quantification. You must use a physical or semi-empirical model (calibration with EC measured at the laboratory) for accurate predection even for pure Sabkha with 500 dS/m.

Thanks Bannari for the comment, that would be interesting to try. Best regards