Soil salinity assessment in Southern Kazakhstan using remote sensing data

Soil salinity maps were made based on regression analysis between field data and satellite images (Landsat 8 OLI, KazEOSat). We used the algorithm of stepwise regression analysis with inclusion of new variables. Spectral bands of satellite images, different vegetation indices, the ratio of satellite bands were used as predictors. Regression models derived from satellite images KazEOSat have average values of R² for fields with cotton at a depth of 0–20 cm (0,53) and 50–100 cm (0,67). Regression models obtained from Landsat 8 characterized by high value of R² (>0,8) on the territory of the second, third and fourth regions. The difference in results is due to the fact that Landsat images have more generalized images with smooth spatial heterogeneity then high-resolution images. Fields grouping on the basis of crops and location proximity allows us to get regression models with high values of R².