Identifying river bank erosion potential zones through geo-spatial and binary logistic regression modeling approach: a case study of river Ganga in Malda district (India)

This study aims to evaluate the causative factors for high bank erosion probability along the left bank of Ganga river in Malda district using binary logistic regression model. The bank erosion at the outer bend of Ganga in Manikchak and Kaliachak-II blocks during the recession of flood water in Ganga poses serious threats to the inhabitants of Diara since the construction of the Farakka barrage. The constriction slowly started a problem of water pilling at the up-stream of the barrage and extended up to the Bhutni Island (40 km up stream). The seepage mechanism allows the entry of rising flood water to the banks and again released when water level recedes gradually and causes bank slumping. In recent monsoon (2020), Gopalpur, Jot Bhabani Dharampur gram panchayats of Manikchak block are heavily affected by bank erosion. A total of nine causative factors are selected as predictor variables in binary logistic regression model categorized broadly as vegetation, water and moisture indices, proximity based on river channel and settlement, soil characteristics and land use-cover classes. The omnibus test of model coefficient gives the likelihood ratio (224.433) for the overall model fitting (p value 0.0). The model predicts correctly 254 sites as low bank erosion (LBE) and 111 sites as High bank erosion (HBE) category with 84.4% and 67.3% accuracy. The soil bearing capacity significantly expresses highest odds (87.6%) of telling the high probability of bank erosion. The model produces an accuracy of up to 87.4%.