Landslide Susceptibility Maps In Alpine Areas Subjected To Climate Change By Using A Machine Learning-Based Approach

The increasing frequency and intensity of landslides is a clear manifestation of climate change, particularly in mountainous regions such as the Northwestern Italian Alps. In these areas, heavy precipitation and temperature fluctuations, along with permafrost degradation, are major drivers of slope instability. This study aims to evaluate the future evolution of landslide susceptibility under climate change and its implications for critical infrastructure, focusing on high-voltage energy lines. An Extreme Gradient Boosting (XGBoost) model was applied to generate susceptibility maps using a combination of static and dynamic conditioning factors. A dataset of 728 spatially distributed points, comprising both landslide and non-landslide events, was used for model training and validation. To account for climate change impacts, downscaled projections from the Coupled Model Intercomparison Project Phase 6 (CMIP6) were used, incorporating data from global climate models under the high-emission scenario SSP5-8.5 for the period 2021–2040. Model performance was evaluated using the Area Under the ROC Curve (AUC), with values ranging from 0.92 to 0.99 across all months, indicating high predictive accuracy. A comparison of current (2003–2022) and future (2020–2040) susceptibility maps reveals a significant increase in areas classified as “Very High” susceptibility, rising from 9.96% to 12.85%, while “Very Low” susceptibility areas decrease from 17% to 13%. In addition, the exposure of high-voltage energy infrastructure was analyzed. The spatial overlap between energy lines and susceptibility classes shows that the proportion of energy lines located in “Very High” susceptibility zones is expected to increase from 13% to 17%, raising concerns for future infrastructure vulnerability and planning.