Geo-hydro-mechanical characterisation and 3D modelling of deep slow landslides

Slope instability phenomena represent a significant threat to public safety and socio-economic development, particularly in regions such as Italy, which are characterised by high landslide susceptibility. Amongst these, deep-seated slow-moving landslides constitute a category that is often underestimated because, unlike rapid events, they evolve over years or decades. Their activity, if neglected, gradually compromises the built environment and infrastructure. Consequently, an accurate diagnosis of such processes is essential for interpreting their mechanisms and designing effective mitigation measures. This research aims to advance the diagnosis of deep-seated slow-moving landslides in complex geo-hydro-mechanical (GHM) contexts, analysing the case study of the Pianello hillslope in Bovino, located in the Daunia Apennines. Bovino is characterised by heterogeneous turbiditic formations which, due to intrinsic and tectonic factors, exhibit poor mechanical properties and high piezometric levels even at great depths. These conditions act as predisposing factors for the instability of the landslide bodies. The methodology adopted during this research follows the Stage-Wise Methodology (SWM), enhanced by the resources of the National Research Centre in HPC, Big Data and Quantum Computing (CN HPC – Spoke 5). The Digital Twin of the Pianello hillslope, developed in a GIS environment and integrated with an interactive dashboard, has allowed for the storage and query of a vast amount of multidisciplinary data, including borehole logs, laboratory and in-situ tests, geophysical surveys, and long-term monitoring data. This tool facilitated a highly detailed geotechnical characterisation (hydraulic and mechanical), enabling an advancement in the diagnosis of active landslide mechanisms compared to the baseline established by previous research, resulting in a refined phenomenological interpretation. Geotechnical analysis allowed to characterize the recognised lithostratigraphic units and assessed the critical influence of mesostructural features, such as fissures and rocky inclusions, which significantly reduce the shear strength of the soil matrix. Numerical validation through Limit Equilibrium Method (LEM) back-analyses confirmed that the most critical landslide bodies exhibit mobilised friction angles closely approaching residual values. Furthermore, the study details a comprehensive conceptual and numerical model of the seepage regime of the hillslope. The integration of rock layers into the 3D model of the Digital Twin of the hillslope revealed their role as preferential flow paths. Three-dimensional finite element seepage analyses, performed using the MOOSE FE framework, validated the presence of high piezometric heads at great depths, identifying them as a primary internal predisposing factor for instability. The results demonstrated that 3D modelling is essential to accurately replicate the complex groundwater flow patterns that 2D analyses fail to capture. In conclusion, this research provides a robust framework for the diagnosis of deep-seated landslides in complex GHM contexts. A methodology for the parametric quantitative modelling of landslide processes using open-source software—which leverages the CN HPC computing infrastructure—has been developed. This approach can be replicated for other landslide processes and employed by stakeholders and professionals to study and monitor slow, deep-seated landslides, as well as to identify risk mitigation measures. The findings highlight the necessity of accounting for three-dimensional geometry and lithological heterogeneity in stability assessments. Future developments will involve hydro-mechanically coupled (HM) simulations to investigate the displacement field induced by the main landslide processes, ultimately supporting the design of more effective and sustainable mitigation strategies.