Soil Erosion Susceptibility Mapping with the Application of Logistic Regression and Artificial Neural Network

Abstract

At present, the multi-criteria-based predictive mapping of geoenvironmental processes is a considerable research concern in the domain of earth sciences. This study aimed to predict the probability of occurrence or nonoccurrence of soil erosion at any given site based on a range of factors active therein. The GPS survey of 412 observation sites recorded the soil erosion status of Purulia district in the state of West Bengal in India. An algorithm of binary logistic regression and the artificial neural network was used to build an erosion susceptibility model by using 70% dataset of the sample sites. The model performance was assessed with the remaining 30% of the datasets. After a successful validation, the model equation was used to prepare a susceptibility map for the study area. The model was found to be valid for the given study area; however, it can be used for preparing the susceptibility map of water-induced soil erosion at any other area with similar geoenvironmental conditions. The map will help in framing the comprehensive policies for scientific land-use planning and area-specific soil conservation practices. In addition, the susceptibility map may be used as a base map for designing income, employment, and livelihood planning for the agro-based areas and serves as a decision support to the priority-based allocation of funds to different micro-level administrative units for government-sponsored schemes related to agriculture, irrigation, and soil and land management.