Automated Identification of Landslide-Prone Areas in Southern Italy: A Case Study from Caiazzo

Landslides pose a significant threat to community safety globally, with Italy being particularly vulnerable. In the Campania Region (Southern Italy), nearly all municipalities are classified as high geo-hydrological risk areas, necessitating focused attention on these natural hazards. From a geological point of view, the Campania Region is characterised by a high complexity, presenting lithologies affected by both rapid (debris flow) and slow (earthflow) landslides, almost all of which are triggered by rainfall, sometimes by earthquakes. This concern is underscored by requests from rail transport authorities in Campania to enhance monitoring systems to identify landslide-prone areas that may impact railway operations. This study investigates the use of unsupervised machine learning techniques for the automatic identification of landslide-prone areas in the western region of Caiazzo, Caserta (Southern Italy). The research addresses the frequent disruptions of the Naples-Caiazzo-Piedimonte Matese railway line due to severe hydrogeological instability. An automatic procedure was developed to identify areas at higher risk, utilizing a dataset comprising 12 geomorphological parameters relevant to landslide susceptibility. The analysis involved dimensionality reduction through principal component analysis and clustering using the K-Means algorithm. The clustering results segmented the area into twelve zones, highlighting three critical zones with the highest landslide risk. Comparison with a landslide inventory map indicated that most triggering points fell within these clusters, offering valuable insights for targeted monitoring and risk management strategies.