Identification of soil erosion-susceptible areas using fuzzy logic and analytical hierarchy process modeling in an agricultural watershed of Burdwan district, India

Soil erosion is a natural process; it adversely impacts natural resources, agricultural activities, ecological systems, and environmental quality as it degrades landscapes and water quality, disrupts ecosystems, and intensifies hazards. Management strategies are needed that protect soil erosion in agricultural watersheds to achieve the sustainable land-use planning. This study maps soil erosion susceptibility using two GIS-based machine-learning approaches—analytical hierarchy process (AHP) and fuzzy logic modeling in the Kunur River Basin, West Bengal, India. Fifteen soil erosion conditioning variables were integrated with the modeling methods, remote sensing data, and GIS analysis. The relative importance of the conditioning variables was assessed for their capacities to predict susceptibility of locations to soil erosion. The soil erosion susceptibility maps generated from the two models used 70% of surveyed soil erosion sites. These models’ maps were validated with the characteristics of the remaining 30% of the soil erosion sites to produce a receiver operating characteristics curve. The results indicated that the fuzzy logic model has the higher prediction accuracy; the area under the curve (AUC) value was 91.4%. The AUC value of the AHP model was 89.7%. Both models indicated that study area contains regions of high to severe soil erosion susceptibility. Logistic regression was used to discern the variables’ importance in the assessment. Relief, NDVI, distance from a river, rainfall erosivity, and soil types were the most important variables. TWI, SPI, aspect, and a sediment transportation index were of least importance. Fuzzy-logic-generated SESMs can be effective tools to guide protective actions and land managers’ measures during the primary stages of soil erosion to control the development of soil degradation.