PSI - Issue 84

Nicola Perilli et al. / Procedia Structural Integrity 84 (2026) 813–820

816

Figure 1. Workflow including Tasks and Sub-tasks supporting inspector during the back-office and field data collection (from Perilli et al. 2024b) 3. Reinforcing and Optimizing of Landslide-Brige Inspections for CdA Assessment As anticipated, the proposed activities, which ensure a coherent and integrated acquisition, analysis, verification, and assessment of landslide CdA, are organized into the following three operational components. 3.1 Preparation of Thematic Layers and Landslide Susceptibility Model Generation The first component integrates three activities. The first activity consists of producing ten thematic layers (TLs) using satellite imagery and secondary datasets based on the primary Conditioning Factors (CFs) controlling slope instability and landslide susceptibility (Tab. 4). These TLs describe geomorphological and lithological setting and capture the hydrographic/hydrological, environmental and anthropogenic factors influencing slope stability. They form the foundational dataset for subsequent analysis, and support the periodic compilation of the Field Survey Form during the inspections. The second activity consists of creating a georeferenced landslide inventory using the IFFI database, together with a structured attribute table including key parameters such as activity state, estimated area, and mitigation works. The inventory serves as the training and validation dataset for slope instability and susceptibility model development, while the attribute table supports the estimation of primary and secondary parameters. The third activity involves generating Slope Instability Models (SIM/ 4 ) and Landslide Susceptibility Models (LSM/ 7 , LSM/ 10 ). Models are developed using three machine-learning algorithms (Random Forest, Logistic Regression and Support Vector Machine), applying three distinct clusters of conditioning factors with 4, 7, and 10 CFs (Tab. 4)) These outputs enable identification of potentially unstable slopes at a regional scale. 3.2. Back-Office Analysis and Field Data Collection Based on the produced ten thematic layers and the generated slope instability and susceptibility models, the second