The role of vegetation on susceptibility modelling of landslides in pyroclastic slopes: a case study in Campania, Italy.

Many sloping areas of Campania in Southern Italy are widely covered by pyroclastic deposits derived from the activity of the Vesuvius and Campi Flegrei volcanic complexes. These areas are highly susceptible to rapid flow-like landslides triggered by intense precipitation, often after antecedent wet periods. Vegetation can play a relevant role in increasing the stability of such slopes. Indeed, the influence of plants on the mechanical and hydraulic behaviour of slopes is significant within the first few meters of the subsoil, where root systems may provide substantial soil reinforcement and alter the flow dynamics in the unsaturated zone, affecting the stress state of the soil.

This research explores how cover, type, and seasonal dynamics of vegetation may influence the susceptibility of pyroclastic deposits to shallow landslides, providing valuable insights into how vegetation may mitigate slope instability in such environments. To achieve these objectives, a tool based on Machine Learning (ML) algorithms which estimates the spatial-temporal probability of landslide triggering at regional scale is being developed. The use of ML facilitates the identification of potential interactions between different types of vegetation and geomorphological, geomechanical, and atmospheric factors. To ensure the operational functionality of the model, both static and dynamic datasets have been utilized.

The study area is the "Camp3" zone of the regional landslide early warning operational in Campania. Data about geomorphology, lithology, soil cover thickness, land use and land cover are made available mainly from thematic maps developed by river basin authorities at regional scale. The model employs a Digital Terrain Model of the study area with a resolution of 1x1 m, obtained from LiDAR data. The analysis also includes data on the hydrographic network, roads, and railways. Landslide events are derived from two landslide catalogs: ITALICA and FraneItalia. Atmospheric data is taken from the ERA5-Land reanalysis data provided by the C3S service, which allows for the reconstruction of precipitation patterns, temperature, and soil moisture at three levels up to 1 m depth.  ERA5-Land data are also used to consider the role of vegetation, specifically considering information on vegetation types, the percentage cover per vegetation class and subclass, the Leaf Area Index (LAI) and their seasonal variations. The study also integrates the Corine Land Cover map as provided in its 2018 version.

Finally, to validate the outcomes of the ML model, a physically based 1D mathematical model is adopted to simulate unsaturated flow and assess slope stability. 1D modelling is deemed suitable for the involved slopes based on geological, geomorphological, and geotechnical information. The model also considers the mechanical reinforcement due to the roots through an additional cohesive term. The apparent cohesion associated with the moisture content, in turn is calculated accounting for the coupled hydro-thermal behaviour of the involved soil. Modelling investigates the importance of different vegetation properties (e.g., LAI, root depth, root density, vegetation height, and plant moisture limit) on the stability conditions for typical slope scenarios in the study area.