PSI - Issue 84

Alessandro Brunetti et al. / Procedia Structural Integrity 84 (2026) 97–102

100

2.2. Landslide Susceptibility

Building on the susceptibility analysis presented in Argyroudis et al. (2020), this study introduces a more detailed approach to characterize landslide hazards affecting linear infrastructure. The methodology refines both the input data structure and the classification logic, with particular attention to the kinematic behavior of slope movements. Susceptibility is evaluated as the intrinsic propensity of terrain to generate landslides, based on factors such as slope angle, lithology, land use, and morphometry. As in Argyroudis et al. (2020), machine learning algorithms trained on available inventories are employed to model landslide density and generate susceptibility maps. Persistent Scatterers (PS) from satellite InSAR data are incorporated more systematically than in Argyroudis et al. (2020), serving as a proxy for activity status. These data support the validation and enhancement of susceptibility outputs by highlighting zones with active deformation. Figure 2 represents the combination of susceptibility classes and PS velocities, considering a threshold of 2.5 mm/years.

Fig. 2. Confusion matrix for the combination of slow-moving landslide susceptibility with PS-InSAR displacement data.

2.3. CdA Classification The initial classification undergoes systematic adjustment through model reliability assessment and mitigation measures evaluation. Model reliability considers validation metrics and spatial data coverage (quantity and consistency of data), while mitigation status accounts for monitoring systems and stabilization works. The final CdA classification provides standardized attention classes for infrastructure management based on the Italian guidelines. Linear infrastructure assessment employs an upslope buffer analysis, implemented through tailored algorithms designed to identify areas potentially affecting roads via landslide runout. Hazard statistics extraction within these buffers enables road segment classification and priority ranking for intervention planning. 2.4. Summary Layer A key innovation introduced in this study is the development of a standardized summary geospatial layer that consolidates landslide hazard assessment outputs following the Italian guidelines into a decision support product. This synthesis layer serves as a comprehensive tool for identifying high-hazard zones along linear infrastructure, prioritizing them for monitoring, maintenance, or intervention. The summary layer integrates the results from the P Factor calculation, CdA classification, and infrastructure specific hazard analysis (Figure 3). These assessments are spatially combined through a rule-based approach to assign a composite hazard class/score to each infrastructure segment. When applied to bridge assets, the synthesis supports an immediate portfolio view of landslide hazard classes and a clear prioritization backlog (Figure 4): out of 183 test bridges, 67 (37%) are classified as Low (L), 50 (27%) as Medium-Low (ML), 19 (10%) as Medium (M), 10 (5%) as Medium-High (MH), and 7 (4%) as High (H). Overall, 17 bridges (9%) fall into the MH–H classes, effectively flagging the subset requiring the highest operational focus. The results are made available within a GIS environment, enabling interactive visualization of the hazard assessments for each segment of the infrastructure network. By visualizing susceptibility through attention classes, decision-makers can easily identify critical areas where hazard severity justifies immediate action according to