PSI - Issue 84

Erica Cernuto et al. / Procedia Structural Integrity 84 (2026) 1167–1174

1170

grain size analyses, and the landslide was modelled as a roto-translational mechanism evolving along a defined sliding surface. Two soil volumes were included in the model (Fig. 2b): the stable soil and the landslide, which share the same properties except for shear strength, calibrated through iterative back-analysis . Strength values of c′=30 kPa and φ′=35° were assigned to the stable soil, while c′=28 kPa and φ′=30° for the landslide. The sliding surface was represented by an interface with residual param eters (c′=1 kPa, φ′=18°) to simulate movement along the failure surface. Initial values were taken from geotechnical literature (Terzaghi et al., 1996) and subsequently refined through a parametric study to ensure consistency with field evidence. The Hardening Soil constitutive model was assigned to all materials, as it simulates the non-linear behaviour of soil and the variation in stiffness as a function of strain level. Stiffness was defined using three parameters: the secant modulus 5 0 =30.00 ∙ 10 3 kN/m 2 ,the oedometer modulus =36.01 ∙ 10 3 kN/m 2 , and the unloading/reloading modulus =110.80 ∙ 10 3 kN/m 2 , estimated from geotechnical characteristics of the site and empirical correlations from literature. Additionally, the model includes the parameter m =0.5, which accounts for stiffness dependency on effective stress and was adopted following the values proposed by Janbu (1963) and Wu and Tung (2020) (for further details, see Cernuto et al., 2025). The bridge was incorporated using the geometric information introduced previously (Fig. 2a). To focus on soil–foundation interaction, the superstructure was simplified: piers were modelled as linear-elastic beam elements and the deck as a simply supported beam. Groundwater conditions were introduced through a piezometric surface located 2.5 m below ground level. The InSAR analysis was carried out using Sentinel-1 data processed within the European Ground Motion Service (EGMS). Among the available product levels, the Calibrated dataset was selected because it provides LOS deformation velocities referenced to a terrestrial frame through GNSS calibration and is available in both ascending and descending geometries. The Ortho product, although providing vertical and horizontal components directly, was not used due to its coarse 100 m resolution and its ETRS89-LAEA reference system (East, North, Vertical), which is not optimal for representing actual displacement, as it is not linked to the orientation of the deformation phenomenon. The analysed dataset covers the period 2018–2022, including ascending and descending geometries. Since interferometric techniques detect only the displacement component along the LOS, significant horizontal movements may be underestimated. To overcome this limitation, a post-processing procedure was adopted to reconstruct the actual displacement components, which requires spatial and temporal correspondence between PS acquired from the two orbital geometries. As this condition is rarely met in practice, spatial and temporal interpolation was applied.

a

c

b

d

Fig. 3. Combination schema of the two datasets: (a) resampling grid, (b) ascending synthetic PS, (c) descending synthetic PS, (d) combined synthetic PS (adapted from Cernuto et al. 2025).

Specifically, the PS datasets were resampled on a regular grid with a 25 m cell size (Fig. 3a), selected to ensure a sufficient number of PS per cell despite the heterogeneous spatial distribution, particularly the lower density within the vegetated landslide area. Due to the limited availability of dual-geometry data downstream, the analysis focused on the upstream portion of the slope, where the interaction with the infrastructure occurs. The grid size was defined after testing different approaches, including the 7 m pairing buffer proposed by Notti et al. (2010) and the 50 m cell