Machine learning applied to natural hazards: Risk of rockfall in the Oued M’goun

Abstract

Morocco faces natural catastrophes due to terrain movement, particularly in mountainous areas with instability of movement and falsification. This study focuses on terrain movement risks, specifically the risk of terrain ejection in the M’goun Basin, located in the southern part of the Haut Atlas. The study uses machine learning techniques to simulate the risk of the ejection of rocks, aiming to create a predictive map of this risk using statistical calculations. The study aims to better understand and mitigate these risks.The present study is designed to elaborate the rockfall risk map in the Oued M' goune watershed, and the "ROC" curve for the validation of the rockfall risk map. the rockfall risk map produced by running the XGB Algorithm in the R studio program, indicates that the degree of risk is higher in the downstream section, with a good prediction accuracy of 62% using the ROC curve.