Determining Rainfall Thresholds for Landslide Prediction: A Case Study on the Hills of Assam

This study aims to estimate rainfall thresholds for landslide prediction in Assam's hilly regions. Historical landslide and rainfall data from monitoring stations in Assam, including Guwahati and Haflong, were collected from the Indian Meteorological Department and NASA. Statistical analysis techniques were applied to establish empirical relationships between rainfall and landslide events, including rainfall duration analysis and cumulative rainfall duration thresholds that indicate potential landslide occurrence. Initial estimates focused on Assam's overall thresholds, followed by local thresholds for Guwahati and Haflong. Moreover, to build a proper model for solving these issues, genetic programming (GP) was implemented based on the tree expressions for Assam, Guwahati, and Haflong, respectively. The GP model demonstrated strong performance in estimating landslide movements, with coefficients of determination \(\:\left({R}^{2}\right)\) of 0.666385, 0.838053 and 0.963469 for Assam, Guwahati, and Haflong, respectively, indicating a high level of accuracy in predicting real values. The analysis revealed significant correlations between rainfall characteristics and landslide events in Assam.