Evaluation and Preparation of Soil Salinity Map Using Vegetation Indicators and Sentinel-2 and Landsat-8 Multispectral Images in Salt Marsh Qazvin Plain

In this research, 23 soil samples with specific geographical characteristics were collected to investigate and monitor salinity changes in the region. Using the Sentinel-2 and Landsat-8 sensors, seven vegetation cover indices and five salinity indices were examined and evaluated in the GEE environment, resulting in a total of 240 outputs from the two sensors. To assess the modeled values, several statistical indices including root mean square error (RMSE), coefficient of determination (R2), normalized root mean square error (NRMSE), and percent bias (PBIAS) were utilized. The results indicated that the SI-2 index exhibited the highest correlation with the measured salinity values in the region, with an R2 value of 0.91, demonstrating its accuracy in estimating salinity levels. In the next step, a multiple regression model was employed to investigate the mean values of measured ECe (electrical conductivity of the saturation extract) and the vegetation indices GDVI (Green Difference Vegetation Index) and CRSI (Crop Salt Stress Index) obtained from the Sentinel-2 sensor, which showed the highest correlation with the salinity data. The results demonstrated that the two-variable regression model achieved a satisfactory accuracy with an R2 value of 0.84 and a PBIAS value of 0.01 in producing a salinity map of the area. Therefore, this model can be utilized as a cost-effective approach for salinity mapping in the region with minimal ground-based data. Furthermore, the investigation of the impact of constructing a barrier drain in the area revealed that the construction of a barrier drain within a distance of 250 meters had a significant effect of approximately 40 percent in controlling salinity. It was able to prevent a substantial increase in salinity levels in the region. Therefore, if a barrier drain is not constructed in the area, salinity progression in the upstream agricultural lands could significantly escalate.