Landslide hazard mapping using ensemble machine learning algorithm in Ba Be lake basin

Tóm tắt

The paper presents landslide hazard mapping for Ba Be lake basin using four machine learning models namely: Logistic Regression (LR), Random Forest (RF), Gradient Boosting (GBC) and Xgboost (XGB). Based on field surveys, literature review and available data, ten indicators were used to derive the landslide hazard, including: slope, soil, plan curvature, normalized difference vegetation index (NDVI), topography, geomorphons, distance from roads, distance from rivers, density of streams and rainfall accumulation. These indicators were arranged in grid cells. The Receiver Operating Curves (ROC) and Area Under Curve (AUC) were used to validate the modes. The results of the analysis showed that the RF and GBC models had the highest predictive ability (AUC = 0.88), followed by the XGB models with AUC = 0.86 and the last one is LR with AUC = 0.78. The results could be useful for planners in general land use planning and management