Application of SWAT, Random Forest and artificial neural network models for sediment yield estimation and prediction of gully erosion susceptible zones: study on Mayurakshi River Basin of Eastern India

Abstract

The study aims to estimate annual sediment yields of Mayurakshi River Basin using SWAT tool of ArcGIS. Climatic aspects, soil, land use and slope maps were integrated within SWAT tool to predict sediment yield capacity using MUSLE method. Alongside, the study attempts to identify spatial distribution of major gully erosion prone sites using ANN and RF models. Twelve Gully Erosion Conditioning Factors namely, elevation, curvature, aspect, runoff, TWI, slope, geology, stream frequency, rainfall erosivity, NDVI, LS-factor and LULC were selected. A gully erosion inventory map was prepared using 128 gully sites and divided into training (70%) and testing (30%) classes. Results obtained from the study shows that sediment yield capacity is very high in north-western and south-western parts of the basin due to high probability of gully erosion. Both ANN (ROC 0.96, Kappa 0.92) and RF (ROC 0.97, Kappa 0.94) models have high prediction accuracy and strong interrater reliability with MUSLE (0.60 and 0.75).

Keywords:

• Artificial neural network
• Mayurakshi river basin
• Musle
• Random Forest Swat

Disclosure statement

The authors declare that they have no conflict of interest.