PSI - Issue 84

Francesca Ceccato et al. / Procedia Structural Integrity 84 (2026) 599–606

605

Figure 4 shows the CILSI map in which each bridge is associated with a landslide susceptibility class. Table 2 summarizes the percentage distribution of bridges across different susceptibility classes for rapid and slow-moving landslide processes. The results show that the majority of bridge infrastructure in the Veneto region is located within low-susceptibility areas, particularly for slow-moving landslides the value is 83.72% and 70.23% for rapid. This reflects the concentration of transportation infrastructure in relatively stable, low-relief terrain. About 3% of bridges falls in high and very-high hazard classes and therefore deserve more attention. 3.1. Limitations and Future Developments The integration of continuous landslide susceptibility (LS) with the Infrastructure Spatial Index (ISI) through the Combined Infrastructure–Landslide Susceptibility Index (CILSI) provides a scalable and spatially explicit framework for supporting the regional assessment of landslide risk to bridges. The proposed approach enables the identification and prioritization of potentially exposed infrastructure; however, the results should be interpreted as indicative and form part of an ongoing research effort. The analysis is conducted at a regional scale and is based on pixel-level susceptibility modelling combined with an 800 m circular influence area of the infrastructure. Consequently, the resulting susceptibility metrics represent relative likelihoods rather than absolute probabilities of landslide occurrence and are intended for screening and prioritization purposes, not for site-specific risk evaluation. Temporal variability in triggering factors and bridge-specific vulnerability characteristics is not explicitly addressed in the current implementation. Uncertainties in landslide susceptibility mapping primarily stem from the quality, completeness, and spatial distribution of the training data derived from available landslide inventories. Limitations related to inventory completeness, landslide typology, and the lack of information on landslide volume may affect model performance and predictive capability. Future work will explore advanced strategies, such as transfer learning, to improve model robustness and generalization across different physiographic and geological settings. Additional uncertainty arises from the choice of the spatial scale adopted for the ISI. In this study, a single ISI configuration is applied; however, the optimal spatial extent is likely dependent on landslide type and local morphological conditions. A systematic multi-scale analysis, with ISI dimensions tailored to different landslide mechanisms and terrain contexts, would further strengthen the methodology. The proposed framework primarily addresses the hazard and exposure components of landslide risk. Bridge specific vulnerability attributes, including structural typology, foundation conditions, age, and maintenance status, are not currently incorporated and are essential for a comprehensive risk assessment. Nevertheless, the methodology provides a consistent, reproducible basis for refining landslide-related attention classes for bridges, in alignment with Italian guidelines. Future developments will focus on integrating bridge vulnerability models, dynamic triggering variables, higher-resolution datasets, and validation against historical damage and impact records to enhance the robustness and applicability of the framework. 4. Conclusion This study developed a regional-scale framework to support the screening and prioritization of bridges exposed to landslide processes in the Veneto region, based on landslide susceptibility maps derived from Random Forest modelling. Separate models were implemented for rapid and slow-moving landslides, achieving strong predictive performance, with AUC–ROC values of 0.96 and 0.93, respectively. The results highlight distinct spatial patterns and controlling factors for different landslide mechanisms: rapid landslides are predominantly governed by topographic and morphological variables, whereas slow-moving instabilities show a stronger dependence on soil properties and hydrological conditions. By integrating landslide susceptibility with an Infrastructure Spatial Index (ISI), the proposed Combined Infrastructure Landslide Susceptibility Index (CILSI) enables a consistent comparison of bridge exposure at the network scale. Application to the Veneto Region indicates that the large majority of bridges fall within low CILSI classes, while only a limited fraction is associated with high susceptibility conditions (1.4% for rapid landslides and 0.93% for slow-moving processes). These results confirm the suitability of the framework for network-level screening, allowing potentially critical bridges to be identified efficiently.