PSI - Issue 62

Salvatore Misiano et al. / Procedia Structural Integrity 62 (2024) 576–584 Author name / Structural Integrity Procedia 00 (2019) 000 – 000

577

2

1. Introduction Shallow landslides (soil slips) are natural phenomena which involve the topsoil, that is generally debris layers of less than two meters in depth. Such events are triggered by intense and prolonged rainfall, which penetrates the topsoil and consequently causes instability; once triggered, they can evolve into debris flows or muds, characterized by high speeds and destructive power, when placed in urban contexts amid different facilities. A recent case, for instance, is “Madonna del Monte” , that is an overpass located in the Italian highway A6 (Torino-Savona), collapsed in November 2019 due to the interference with a large soil slip. These situations are not rare or isolated but, since they occur more easily in mountainous areas, they often involve small-size road infrastructure and, consequently, they get less attention from people. Regardless, in recent years, this type of phenomena is increasing due to climate change (Gariano et al. 2016, Jakobet al. 2021) and with that, also the need of methods for their prediction and prevention. Specifically, a crucial part of this process is the prediction of the possible path that the landslide, once triggered, might follow. In such a manner, it becomes easier to determine which infrastructures a soil slip might interfere with, and to act accordingly. Regarding the prediction, namely the process by which the possible triggering areas are detected, different approaches could be used, classifiable into three main categories: Machine Learning (ML); statistical; and Physically Based (PB) approaches. While the first two are focused on pure data, trying to extract patterns or thresholds and without investigating the principles of cause and effect, the PB approaches are aimed at creating a model that, albeit with simplifications, represents the true nature of the problem. According to the degree of simplification and to the focus on the susceptibility factors selected, different models exist (Borga 1998, Selvatici 2018). In this study, SLIP model is used, due to its good prediction quality showed in previous applications (Montrasio 2000), but also due to the short analysis time required, compared to others. Concerning the prediction of the landslide path and the volume involved, a simplified approach, based on the gradient descent theory and on the uniformly accelerated motion, is presented. Different similar studies adopted these simplified techniques over the last years (Guthrie et al. 2021, Komu et al. 2023, Lie et al. 2022) but most of them did not consider equations related to the evolution of the mass in motion. For this purpose, a modified dynamic FS is evaluated in different points of the pre-determined path, and according to it, the amount of volume eventually incorporated is assessed starting from the triggering point to the deposit point. The case study of this work dates back to 2014 in Enna, Sicily (southern Italy), where the 2 nd of February a shallow landslide turned into a debris flow and interfered with Provincial Road SP2 and the State Road SS117bis, causing significant economic damage and traffic viability problems. The phenomenon is studied entirely in a MATLAB apposite integrated platform: X-SLIP.

Nomenclature SLIP

Shallow Landslides Instability Predictor

ML PB

Machine Learning Pysically-Based

FS Factor of Safety DEM Digital Elevation Model

2. Methods 2.1. SLIP

SLIP is a simplified physically-based model that uses the limit equilibrium theory and the shear resistance criteria for partially-saturated soils. The main concept consists in homogenizing the mechanical properties and simplifying the process of imbibition and drainage of slopes. Based on this, the ratio between stabilizing and destabilizing forces is the FS. The latter is a function of various parameters, such as: geometrical (inclination of slope δ ; depth of