Integrated Application of Remote Sensing and Spatial Statistical Models to the Identification of Soil Salinity: A Case Study from Garmsar Plain, Iran

Document Type : Original Articles


1 PhD Graduate, Department of Soil Science, Faculty of Agriculture, Tarbiat Modares University, Tehran (Iran).

2 Professor, Department of Soil Science, Faculty of Agriculture, Tarbiat Modares University, Tehran (Iran)

3 Assistant Professor Department of Earth Science, Faculty of ITC, Twente University (Netherlands).


Soil salinity expansion is an environmental challenge particularly in arid and semi arid regions. In order to evaluate the progressing extent of soil salinity in relation with natural and human-induced conditions, a study was conducted using the Landsat TM imagery. The present study was conducted in the Garmsar area to the East of Tehran. A total of 288 soil samples were analyzed to determine the relationship between the spectral reflectance and Electrical Conductivity (EC), as salinity indicator. Multiple regression analysis and Ordinary Least Square regression (OLS) were used to examine the relationships between EC and derived spectral to generate several models. In the case of derived spectral, mid-infrared band (TM Band-7), visible band (Band-1), Tasseled cap3 (Wetness index) and PCA2 (Principal Component Analysis) were found to be most correlated with the observed EC values of the surface layer of the soil, at 99% confidence level. The accuracy of the prediction model was tested using a validation set of 52 soil samples in Eyvanekey plain, close to study area where the environmental circumstance consist of similar properties. RMSE and MAE were used to evaluate the performance of the map prediction quality. Results showed that the appropriate model could predict the soil salinity with precision of 4.1 and 0.49 dS m-1, respectively. The predicted salinity ranged from 0dS/m to 110dS/m. Therefore, the EC estimations were suitable to generate soil salinity map. Sensitivity analysis was tested on applied parameters that showed Band-1 and Band-7 were 3 and 2 times more than sensitive rather than other parameters respectively. The results are promising and certainly useful for soil salinity prediction.