Landslide Susceptibility Mapping Using Machine Learning Algorithm: A Case Study Along Karakoram Highway (KKH), Pakistan

The China-Pakistan Karakoram Highway links China to South Asia and the Middle East through Pakistan. Rockfall and debris flows are dangerous geological risks on the main route, and they often disrupt traffic and result in fatalities. As a result, the landslide susceptibility map (LSM) evolution along the highway could make it safer to drive. In this article, remote sensing data are combined with machine learning algorithms such as Random Forest (RF), Extreme Gradient Boosting (XGBoost), and k-Nearest Neighbors (KNN) to develop the LSM. Initially, 274 landslide locations we determined and mapped in ArcGIS software and randomly divided into a ratio of 8/2. Secondly, ten landslide susceptibility factors were developed using satellite imagery and different topographical and geological maps. Finally, the receiver operating characteristic (ROC) curve, the area under the ROC curve (AUC) value, was used to estimate the model's effectiveness. Our consequences showed that, for three models, the RF, XGBoost, and KNN models, as well as slope, elevation, and distance from the river parameters, had the maximum influence upon landslide sensitivity. Accordingly, the prediction rates are 83.5%, 82.7%, and 80.7% for RF, XGBoost, and KNN. Furthermore, the RF method has better efficiency as compared to other models on the base of AUC. Our findings show that all three machine learning algorithms positively impact, and the results may assist the highway's safe operations.

Automated summary

This study combines remote sensing data with machine learning algorithms to develop a landslide susceptibility map for the China-Pakistan Karakoram Highway. The results show that the Random Forest model is the most effective, and factors such as slope, elevation, and distance from the river have the greatest impact on landslide sensitivity.