University of Tehran Journal of Range and Watershed Managment 5044-2008 78 4 2026 01 01 Assessment of Spatial Distribution of Surface and Subsurface Soil Salinity Using a Digital Mapping Approach (A Case Study in Dehloran County) Assessment of Spatial Distribution of Surface and Subsurface Soil Salinity Using a Digital Mapping Approach (A Case Study in Dehloran County) 513 528 105464 10.22059/jrwm.2025.395192.1832 FA Sahereh Safarlaki Department of Soil Science and Engineering, College of Agriculture, Bu-Ali Sina University, Hamadan, Iran 0009-0002-7017-0710 Azadeh Safadoust Department of Soil Science and Engineering, College of Agriculture, Bu-Ali Sina University, Hamadan, Iran 0000-0002-1535-7833 Mahmood Rostaminia Soil and Water Department, Agriculture Faculty, Ilam University, Ilam, Iran 0000-0002-8921-5627 Seyedeh Bahareh Azimi Department of Environment, Research Group of Environmental Assessment and Risk, Research Center for Environment and Sustainable Development (RCESD), Tehran, Iran Journal Article 2025 05 16 Accurate spatial data on soil property distribution is crucial for monitoring of land resources, informed management practices, and robust environmental modeling, especially in arid and semi-arid regions. This study aimed to develop a spatial prediction model for soil salinity in the Meymeh Plain, Dehloran Province. The Random Forest (RF) algorithm was employed to investigate spatial variations in soil salinity within the surface (0–30 cm) and subsurface (30–60 cm) soil layers. Soil samples were collected from 100 sites, analyzed for electrical conductivity (EC), and the spatial variability of soil salinity was modeled using random forest (RF) analysis. Seven environmental variables of Greenery, Diffuse Radiation, Valley Bottom Flatness Index, Normalized Difference Vegetation Index, Salinity Index, Wind Direction Index, and Brightness were selected based on the Variance Inflation Factor, including parameters from a digital elevation model and Sentinel-2 satellite reflectance data. The model used 80% of the data for calibration and 20% for validation, with performance assessed through root mean square error (RMSE), coefficient of determination (R²), and concordance correlation coefficient (CCC). The RF model showed high prediction accuracy for surface EC and relatively acceptable results for subsurface layers. The R² for the surface layer was 0.92, and for the subsurface layer was 0.37; the RMSE for the surface and subsurface layers was 0.22; and the CCC for the surface layer was 0.82 and for the subsurface layer was 0.97. Overall, topographic derivatives demonstrated a greater influence on predicting soil salinity in both surface and subsurface layers compared to remote sensing data. The multi-resolution valley bottom flatness index with high spatial resolution was identified as the most important predictor of soil salinity, highlighting the impact of topographic factors in the study area. Accurate spatial data on soil property distribution is crucial for monitoring of land resources, informed management practices, and robust environmental modeling, especially in arid and semi-arid regions. This study aimed to develop a spatial prediction model for soil salinity in the Meymeh Plain, Dehloran Province. The Random Forest (RF) algorithm was employed to investigate spatial variations in soil salinity within the surface (0–30 cm) and subsurface (30–60 cm) soil layers. Soil samples were collected from 100 sites, analyzed for electrical conductivity (EC), and the spatial variability of soil salinity was modeled using random forest (RF) analysis. Seven environmental variables of Greenery, Diffuse Radiation, Valley Bottom Flatness Index, Normalized Difference Vegetation Index, Salinity Index, Wind Direction Index, and Brightness were selected based on the Variance Inflation Factor, including parameters from a digital elevation model and Sentinel-2 satellite reflectance data. The model used 80% of the data for calibration and 20% for validation, with performance assessed through root mean square error (RMSE), coefficient of determination (R²), and concordance correlation coefficient (CCC). The RF model showed high prediction accuracy for surface EC and relatively acceptable results for subsurface layers. The R² for the surface layer was 0.92, and for the subsurface layer was 0.37; the RMSE for the surface and subsurface layers was 0.22; and the CCC for the surface layer was 0.82 and for the subsurface layer was 0.97. Overall, topographic derivatives demonstrated a greater influence on predicting soil salinity in both surface and subsurface layers compared to remote sensing data. The multi-resolution valley bottom flatness index with high spatial resolution was identified as the most important predictor of soil salinity, highlighting the impact of topographic factors in the study area. https://jrwm.ut.ac.ir/article_105464_1799f4ede88253d78a16787016b88d54.pdf