Machine Learning Approaches for Evaluating Water-Induced Soil Erosion and Its Vulnerability Factors in a Watershed

Soil erosion is a significant environmental concern that threatens agricultural activities, reduces soil fertility, and eventually impacts productivity. Assessing soil erosion is essential for effective planning and conservation initiatives in a basin or watershed. This study aims to assess water-induced soil erosion using machine learning techniques and identify key factors contributing to erosion vulnerability in a watershed. This study employed three advanced machine learning techniques: Random Forest (RF), k-Nearest Neighbors (kNN), and Extreme Gradient Boosting (XGBoost) - to analyze and forecast water-induced soil erosion patterns. The investigation identifies key factors contributing to soil erosion vulnerability by utilizing a comprehensive dataset derived from Digital Elevation Models, climatic records, and land use patterns. The models were trained on 80% of the data, while the other 20% were used for evaluation, resulting in an accuracy demonstrating their robustness in environmental modeling across various topographic features. The models were extensively assessed using various accuracy measures, including sensitivity, specificity, precision, and the Kappa coefficient. The Area under the Curve (AUC) values for the models were 87% for RF, 89% for kNN, and 91% for XGBoost, indicating high predictive performance. RF, kNN, and XGBoost models demonstrated high sensitivity values (0.9, 0.87, and 0.91, respectively) and specificity (0.9, 0.86, and 0.89). The Kappa index for the ML models was 0.80 for RF, 0.73 for kNN, and 0.80 for XGBoost. These metrics indicated that RF, kNN, and XGBoost are highly effective in identifying water-induced soil erosion in the research region. This study not only identifies critical sites susceptible to erosion but also provides a decision-support tool for evaluating soil erosion within the investigated area and similar riverine ecosystems.