Landslides are commonly occurring global geological hazards, with negative and far-reaching consequences for human life, and the economic and natural environment. This study analyzes the risk, vulnerability, and resilience of several landslide-prone areas in eight countries based on community perception and attitude related to landslide events through an online survey questionnaire that includes four sections: (1) natural disaster and landslide susceptibility, (2) knowledge and awareness of the causes of landslides, (3) psychological effects of landslide events, and (4) sociodemographic and socioeconomic characteristics. To gain an in-depth understanding of the landslide risk and sustainability, three state-of-the-art machine learning (ML) approaches, Decision Tree (DT), Xtreme Gradient Boosting (XGB), and Random Forest (RF), and a statistical approach, Logistic Regression (LR), were used to investigate and compare the performance of the ML models. The results suggested that indirect losses caused by landslides significantly affect the communities across the globe both psychologically and economically. Furthermore, heavy rain is the most relevant event that was perceived by the communities to trigger a potential landslide, while major civil infrastructure systems, including building and road damages, were found as the main threats to the community due to landslides. The most significant factors related to landslide risk based on the Gini coefficient were found to be land classification, environmental factors, and building material. In addition, based on LR, land classification, landslide health impact, environmental factors, and emergency response were significantly related to landslide risk perception. In summary, the XGB model is feasible in the assessment of landslide susceptibility; the prediction performance is also robust and produced better prediction capacities compared to other ML models. This research can considerably assist stakeholders in making better decisions toward effective landslide hazard management. Future research should focus on expanding this study to other critical landslide-prone communities by involving more diverse population.
- Landslide risk perception
- Landslide socio-economic impacts
- Logistic regression
- Machine learning