Application of weights-of-evidence (WoE) and evidential belief function (EBF) models for the delineation of soil erosion vulnerable zones: a study on Pathro river basin, Jharkhand, India

Abstract

Soil erosion is a natural geomorphic process and it significantly threatens agriculture, natural resources and environment. The paper seeks to produce soil erosion susceptibility map (SESM) using weights-of-evidence (WoE) and evidential belief function (EBF) models based on geographic information system (GIS) in Pathro river basin, Jharkhand, India. In doing that, the research has taken into consideration a total number of 430 soil erosion patches in highly erosion prone areas in Pathro river basin. Within 430 cases, 300 (69.8%) are randomly selected for training purpose, while the remaining 130 (30.2%) are used for validating the model. It could be mentioned here that both the models have taken into consideration 14 soil erosion conditioning factors, which are derived from the spatial geodatabases. These soil erosion conditioning factors typically include slope, slope aspect, altitude, NDVI, curvature, distance from river, land use and land cover, soil type, rainfall erosivity, stream power index, slope length, distance from road, distance from lineament and lineament density. At the same time, use of WoE and EBF models on a GIS platform would also help one to estimate the soil erosion susceptibility value (SESV) for each pixel. At the same time, by drawing on a suitable classification technique, the paper has also prepared the soil erosion susceptibility map (SESM). Finally, the receiver operating characteristic (ROC) curves have also been prepared for both the soil erosion susceptibility models, mentioned above. The paper also calculates the area under the curve (AUC) for the verification and accuracy assessment of these two models. The result seems to suggest that the SESM, thus, prepared, drawing on both the models, has a high prediction accuracy of 89.80% for the WoE model. Similarly, the AUC for the EBF model has shown 91.80% prediction accuracy. Taken together, it could be argued that both the models have performed good prediction accuracy. In terms of the immediate value, these SESMs, produced with a high success rate and accuracy, would be of some interest and use for planners and policy makers to induct remedial measures in the erosion prone areas.