Current Status and Future Projection of Soil Erosion Based on RUSLE, Landscape Ecology and Deep Learning Technologies: Observations from a Tropical River Basin in Eastern India

Abstract

Abstract Soil erosion has contributed to loss of enormous amounts of top soil worldwide. Since the exact quantification of soil erosion is impossible, numerous researchers across the world have used prediction-based models (such as Revised Universal Soil Loss Equation, RUSLE) for assessing the temporal context of soil erosion at the catchment-scale. This paper has tried to integrate the RUSLE-based empirical soil erosion model and landscape ecology for the soils of a tropical river basin in Eastern India. It is observed that more than 60% of the areas in the studied basin are presently witnessing erosion greater than 11.2 tons/ha/year, which is above the tolerable limit as proposed by Food and Agricultural Organization (FAO). The process was applied for 2011 and 2021 and it was observed that soil erosion was augmented by about 6% during this period. Landscape ecological metrices reveal that the patches of high erosion are getting clustered and coalesced and becoming larger in areal extent, especially in the upper and middle domains of the studied basin. This paper, with the help of the soil erosion status of 2011 and 2021, has tried to predict the future scenario of soil erosion in the next five decades (2021 – 2071) with the help of the Artificial Neural Network, a popular deep learning technology. It is found that if erosion continues at the present rate, the patches may increase in extent by about 50% in the next five decades, which is detrimental. Finally, it is recognized that due to the lower clay content (< 30%) in the upper and middle domains of the basin, the study suggests the use of plot-scale mulching technique as an efficient measure to combat soil erosion in the region.