Soil salinity mapping using spectral index and Landsat 8 in arid zones

This study was implemented to prepare a model for soil salinity mapping using Landsat5 images in several provinces including Bushehr, Semnan, Fars, Kerman and Hormozgan. At the beginning, 50-100 samples from soil surface were taken and sent to the Laboratory. Then in order to evaluate and identify soil salinity, TM Landsat satellite images and statistical models combined with satellite`s spectral indices were used. After evaluating the accuracy of statistical models using test points, the best model for the study area was selected and the salinity maps were developed based on the regression model. The results showed a significant relationship between soil salinity and spectral index. In Fars province, NDSI salinity index had the highest correlation with soil electrical conductivity (0.35) with a regression coefficient of 66% and RMSE and MBE statistics of 2.58 and 0.66, respectively. In Kerman province, the tasseled cap three index had the highest correlation with soil electrical conductivity (0.47) with a regression method coefficient of 65%, and RMSE and MBE of 10.3 and 0.51, respectively. In Hormozgan province, the results showed high correlation with soil salinity indicators SI2 level of 72 percent. Stepwise method with R-square of 0.518 was selected for the Hormozgan province whre the RMSE and MBE were reported to be 2.5 and -0.35, respectively. Also in Semnan province, 5 and 7 bands of Landsat showed the highest correlation with soil electrical conductivity (respectively 0.65 and 0.75). By using stepwise regression, the linear relation with R-square of 0.6 was obtained, and RMSE and MBE values were reported to be 2.83 and -0.81, respectively.