Forest Cover and Landslide Susceptibility Assessment Using a Machine Learning Approach in Northern Midland and Mountainous Region of Vietnam

Landslides are a major geo-environmental hazard in Vietnam’s midland and mountainous regions, further intensified by land-use pressures and climate change. This study investigated the influence of forest cover on landslide susceptibility in Cau River Watershed. A forest status map was constructed using inventory and field data by the K-Nearest Neighbors (KNN) algorithm, while landslide susceptibility was modeled using historical events and nine conditioning factors through a hybrid machine learning approach integrating Random Forest (RF), Multilayer Perceptron (MLP) and KNN. The proposed hybrid model achieved an overall accuracy of 85.33%, demonstrating its robustness in susceptibility prediction. Results indicated that natural and native-species forests significantly reduce landslide density and susceptibility relative to non-forested areas and exotic plantations. These findings highlight the critical role of forest structure and species composition in stabilizing slopes. The study provides evidence-based insights to guide adaptive land management, forest policy, and regional strategies for climate resilience and sustainable development.