Integrating passive seismic monitoring and remote sensing as a tool for landslide risk prediction

Landslide risk mitigation is a complex task due to the multiplicity of factors that can induce slope instability. The use of tools for defining quantitative scenarios represents a robust approach to risk mitigation. Their application requires an in-depth analysis of factors that can be classified as preparatory or triggering (in this study, rainfall, earthquakes, and soil moisture) and provides an effective strategy for detecting and monitoring seasonal ground instability, while also serving as a valuable support for calibrating numerical models aimed at predicting multi-hazard scenarios. The selected case study is the San Vito Romano landslide (RM), a village located about 50 km east of Rome. The most recent part of the settlement is built on an active earth-slide covering approximately 1 km², characterized by a roto-translational mechanism and slow but continuous kinematics. The landslide body consists of siliciclastic deposits of the Laga Flysch Formation (Upper Tortonian), composed of alternating clayey marls and marly sandstones, which make the slope particularly prone to instability. The last significant reactivation occurred in 2011, following intense rainfall, causing widespread fractures in buildings and the evacuation of an entire apartment block. This phenomenon is leading not only to a progressive demographic decline but also to a reduction in tourism, one of the main pillars of the local economy. For the analysis of preparatory and/or triggering factors, a passive seismic monitoring network has been operating since February 2024, continuously recording ambient seismic noise. Four three-component velocimeters with a sampling frequency of 250 Hz have been installed: two within the landslide body and two outside. This monitoring allows for the continuous estimation of landslide mobility indices (LSMI), such as variations in natural period (dT/T), peak polarization (dP/P), and Rayleigh wave velocity (dV/V) over time. Surface wave velocity is measured through ambient noise seismic interferometry, which, by cross-correlating the recorded signals, allows the detection of velocity changes between monitoring stations. The temporal evolution of these parameters provides continuous insight into variations in stiffness, amplification, and non-linear elastic properties of the soil, helping to assess the slope’s state of preparedness for potential instability. The LIDAR model, has provided a detailed geomorphological layout of the area, highlighting significant gravitational features. In parallel, within the framework of satellite techniques, the DInSAR approach using the SBAS (Small Baseline Subset) methodology has been adopted, an interferometric technique that produces maps of average deformation velocity and time series of point displacements. Analysis of data from 2022 to the present has revealed an increase in displacement velocity in certain sectors of the landslide. The ultimate goal of the project is to develop a quantitative tool capable of integrating and correlating rainfall data, seismic parameters, and LSMI indices with satellite displacement measurements. These data will be used to train a machine learning model that, starting from a constant set of inputs with variable values (such as cumulative rainfall over different time intervals and varying earthquake magnitudes), can generate predictive displacement scenarios in a multi-hazard perspective, enabling the quantitative assessment of the role of preparatory and/or triggering factors.