Highways serve as crucial links between regions, enhancing transportation and facilitating economic activities within communities. However, strategic infrastructure such as bridges and tunnels traversing steep mountainous regions is often susceptible to landslides, resulting in damages and casualties. The geological conditions in the Calabria region of south Italy frequently led to landslide disasters, posing a significant threat to regular operations. Accurate and rapid landslide susceptibility mapping (LSM) is essential for analyzing and evaluating the degree of landslide susceptibility, enabling the prediction of severe risks that could affect strategic infrastructure such as highways. The LSM based on machine learning (ML) models exhibits significantly higher accuracy than those using traditional expert knowledge and conventional mathematical statistics models. The proper and reasonable selection of non-landslide samples in ML models significantly enhances the prediction accuracy and reliability of the regional LSM. Therefore, our work is to analyze the spatial distribution pattern between predisposing factors and landslide occurrences in the selected area. Subsequently, we utilize progressive tree-based ML models, such as decision tree (DT), random forest (RF), gradient boosting decision tree (GBDT), and their information value method (IVM)-enhanced models (i.e., IVM-DT, IVM-RF, and IVM-GBDT), to conduct the LSM. These techniques generated a classification in terms of very low, low, medium, high, and very high landslide susceptibility zones. The receiver operating characteristic (ROC) curve will be used to quantitatively evaluate and compare the prediction accuracy of these models. The outcomes of this work have the potential to facilitate rapid and precise evaluations for effective hazard management and provide theoretical guidance for the maintenance and normal operation of bridges, tunnels, and other susceptible structures along highways in the Calabria region.
Enhanced Landslide Susceptibility Mapping Along Highways Using Progressive Tree-Based Ensemble Models with Optimal Non-Landslides Selection
Luca Simoni;Lorenzo Brezzi
2025
Abstract
Highways serve as crucial links between regions, enhancing transportation and facilitating economic activities within communities. However, strategic infrastructure such as bridges and tunnels traversing steep mountainous regions is often susceptible to landslides, resulting in damages and casualties. The geological conditions in the Calabria region of south Italy frequently led to landslide disasters, posing a significant threat to regular operations. Accurate and rapid landslide susceptibility mapping (LSM) is essential for analyzing and evaluating the degree of landslide susceptibility, enabling the prediction of severe risks that could affect strategic infrastructure such as highways. The LSM based on machine learning (ML) models exhibits significantly higher accuracy than those using traditional expert knowledge and conventional mathematical statistics models. The proper and reasonable selection of non-landslide samples in ML models significantly enhances the prediction accuracy and reliability of the regional LSM. Therefore, our work is to analyze the spatial distribution pattern between predisposing factors and landslide occurrences in the selected area. Subsequently, we utilize progressive tree-based ML models, such as decision tree (DT), random forest (RF), gradient boosting decision tree (GBDT), and their information value method (IVM)-enhanced models (i.e., IVM-DT, IVM-RF, and IVM-GBDT), to conduct the LSM. These techniques generated a classification in terms of very low, low, medium, high, and very high landslide susceptibility zones. The receiver operating characteristic (ROC) curve will be used to quantitatively evaluate and compare the prediction accuracy of these models. The outcomes of this work have the potential to facilitate rapid and precise evaluations for effective hazard management and provide theoretical guidance for the maintenance and normal operation of bridges, tunnels, and other susceptible structures along highways in the Calabria region.
Pubblicazioni consigliate
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
