Towards Sustainable Heritage Conservation: A Hybrid Landslide Susceptibility Mapping Framework in Japan’s UNESCO Mountain Villages

A hybrid landslide susceptibility mapping framework significantly enhances the accuracy of predicting landslides in Japan's UNESCO World Heritage sites, Shirakawa-gō and Gokayama. This approach integrates probabilistic slope stability modeling with ensemble learning techniques like XGBoost, CatBoost, and LightGBM, achieving AUC gains of 0.072, 0.066, and 0.063 respectively compared to buffer-based methods. The framework also incorporates future precipitation projections from ARIMA time-series modeling, providing a robust tool for sustainable risk management. SHAP analysis highlights the importance of physically meaningful features such as rainfall in landslide susceptibility. Additionally, the study evaluates the exposure of historic Gasshō-style houses and other buildings to varying levels of landslide susceptibility, offering actionable insights for local planning and heritage conservation.

Bassem, A., Shokry, H., Kanae, S., & Sharaan, M. (2025). Towards Sustainable Heritage Conservation: A Hybrid Landslide Susceptibility Mapping Framework in Japan’s UNESCO Mountain Villages. Sustainability, 18(1), 237. https://doi.org/10.3390/su18010237