Modelling of gully erosion risk using new ensemble of conditional probability and index of entropy in Jainti River basin of Chotanagpur Plateau Fringe Area, India

Abstract

Gully erosion, in the fringe areas of the Chotanagpur, is one of the important environmental issues from the perspective of agricultural sustainability, economic framework and land-resource management. The spatial gully erosion risk map (SGERM) can afford to identify the locales sensitive to gully erosion hazard. The present work intends toward modelling spatial gully erosion risk using the new ensemble model in a plateau fringe semi-agricultural watershed of Jainti River (542.69 km2) in Jharkhand, India. Coupling field operation (gully boundary and area measurement) and geostatistical models, the study was accomplished. Field gullies were split into training and validation sets using random partitioning method. Selected predisposing variables for gully erosion based on physical appearances of this basin and literature survey are slope steepness, length-slope factor, soil type, geomorphology, rainfall erosivity, distance from river, stream power index, topographical position index, topographical wetness index, altitude, slope aspect, land use/land cover, normalised difference vegetation index and geology. To prepare the statistical database, information regarding the number of pixel of individual parameter was extracted using GIS software and for tracing the response of each variable for gully erosion, entropy values were calculated against each of them using IoE. While the probability of gullying within each sub-class of each variable was computed using the CP model. In case of ensemble of CP-IoE, variable weights were assigned through IoE and weights to the sub-classes were assigned through CP. The results of CP, IoE and ensemble models that quantified 10.58%, 7.73% and 8.25% areas of the basin were under the very high-risk category respectively. Based upon the results of ROC curves, the predicting accuracy of ensemble model with the highest area under curve (AUC) is better in comparison to individual CP and IoE models.