Digital mapping of soil erodibility: A case study of the Ravang watershed, southern Iran

Abstract The Universal Soil Loss Equation incorporates soil erodibility as a key parameter for erosion quantification. This study focused on mapping soil erodibility patterns and identifying the primary factors influencing its spatial distribution within the Ravang watershed, located in southern Iran's Hormozgan Province. To ensure representative and stratified spatial coverage, 100 sites within the study area were selected for soil samples using the conditioned Latin hypercube sampling method. Spatial modeling of soil erodibility was performed by assessing variables such as organic carbon content, soil texture, structure, permeability, and erodibility, and applying random forest and boosted regression trees algorithms. The mean soil erodibility in the study area was 0.27 t·ha·h/(ha·MJ·mm). The results indicated comparable accuracy between both methods. Variable importance analysis revealed that maps of very fine sand, medium sand, and total sand content were the most significant predictors of soil erodibility distribution. Furthermore, incorporating soil texture fraction maps enhances prediction accuracy in soil erodibility modeling. The highest soil erodibility rates were identified in the southern and southwestern portions of the Ravang watershed through spatial mapping. Soil erosion mapping provides critical data to prioritize areas for erosion control interventions, helping to mitigate land degradation in vulnerable regions similar to the southern Iranian study area. To achieve enhanced spatial accuracy in digital soil erodibility mapping, we recommend incorporating soil texture fraction maps as essential input variables in comparable studies, given their demonstrated importance in optimizing predictive model performance.