Modeling soil erosion susceptibility using GIS-based different machine learning algorithms in monsoon dominated diversified landscape in India

Severe rainfall causes soil to erode, which has far-reaching effects on the natural environment and anthropogenic activity. It is regulated by the interplay of "lithology, orography, hydrography, land use, and vegetation " and is dependent on rainfall distribution (e.g., "intensity, duration, the cumulative per event "). More than 75 percent of India's yearly rainfall is brought on by the monsoons, making India a prime example of a country with a monsoonal climate. The monsoon season is the apex of all erosional rainfall. Therefore, changes in the frequency and volume of the monsoon's rain will result in corresponding shifts in the season's erosivity patterns. So, identifying vulnerable regions through proper modeling is one of the essential tasks in soil erosion research. In this study, the MaxEnt, RF, SVM, and SLR model has been considered for estimating erosion susceptibility. The overall result has been confirmed, and we found that the MaxEnt model is optimal to other models in this study in terms of performance. This kind of research is helpful to the decision-maker in determining whether or not to use these strategies for preventing soil erosion and achieving the desired results. The main novelty of this work is the use of an appropriate machine learning method to evaluate soil erosion susceptibility for the first time in the study area, taking into account the most significant feasible number of causal factors.