Soil salinity is one of the factors for land degradation, especially in the arid and semi-arid regions. In this paper, the typical salinity region in the upstream margin of Keriya River in Yutian County of Xinjiang was taken as the study object, EM38 sensor was used to in situ measure soil apparent electrical conductivity (ECa), WorldView-2 images were used to extract adjusted soil vegetation index (SAVI) under different conditions, and PLSR model derived from SAVI and ECa was setup to estimate soil salinization. The results showed that the correlation between SAVI and ECa was increased significantly from 0.30 to 0.5 when the adjusted parameter (L) increased from 0.1 to 1.0. The optimal model was established by using the combination of SAVI_1.0+B6+B7+B8, its determination coefficient (R²_P) was promoted by 0.11 compared with those of models derived from other variable combination, the validation coefficients were RMSE_C=0.77, R²_C=0.68, RMSE_P=0.79, R²_P=0.66, RPD=2.2. Therefore, the model derived from different variable combination can provide a fast and accurate method for monitoring soil salinization.