Integration of Machine Learning Models with Game Theory for Understanding Water-Induced Soil Erosion in an Urban Watershed

In recent years, soil erosionSoil erosion has become a growing global concern, exacerbated by factors like abrupt climate shifts, natural resource exploitation, and soil quality degradation, with urban areas facing significant challenges. The Guwahati urban watershed, located in northeastern India, is one such urban region grappling with water-induced soil loss. To address this environmental issue, this study employs a range of models and techniques. Firstly, the Revised Universal Soil Loss EquationRevised universal soil loss equation (RUSLE) model is utilized to estimate potential soil erosion in the area. To enhance the accuracy of erosion prediction, a finely tuned random forestRandom forest (RF) machine learning model is deployed. Additionally, a Deep Neural NetworkDeep neural network (DNN)-based game theoryGame theory model sheds light on feature contributions to soil erosionSoil erosion. The findings of the RUSLE model categorize the six sub-watersheds into varying risk zones, highlighting areas of high susceptibility to erosion, particularly Bharalu, Silsako, and Foreshore. Furthermore, after rigorous optimization, the RF model achieved impressive accuracy, with a root mean square errorRoot mean square error (RMSE) of 1.66 and mean absolute error (MAE) of 1.1, indicating reliability. Precipitation emerges as the primary influencer of erosion, emphasizing the importance of monitoring precipitation patterns and intensity. Also, the DNNDeep neural network-based game theoryGame theory highlights that factors such as basin shape (Ff) and area significantly impact soil erosionSoil erosion. The heightened sensitivity to rainfall, Shannon Diversity IndexShannon diversity index (SHDI), and patch cohesion indexPatch cohesion index (COHES) emphasize the importance of continuous monitoring of these parameters. These findings provide valuable insights into the dynamics of soil erosion in the study area, emphasizing the need for effective management strategies.