Probabilistic mapping of susceptibility and hazard of rainfall-triggered landslides in pyroclastic slopes of Campania (Italy).

Frequently occurring rainfall-induced landslides in pyroclastic soil deposits of Campania (southern Italy) are very threatening for the local community and the infrastructure. The major factor behind the triggering of such landslides is assuredly the rainfall. However, there are several other hydrological processes which are responsible for predisposing the slopes to failure. The study area is Partenio massif of Campania, where slopes are covered with coarse grained loose pyroclastic soils deposited in alternate layers of ashes and pumices laying on densely fractured limestone bedrock. The assessment of landslide triggering considers both static and dynamic factors. The former account for landslide susceptibility assessment, while the latter are responsible for the assessment of time-dependent landslide hazard. Several studies reported in literature explored various methodologies for the assessment of landslide susceptibility. However, landslide hazard assessment still needs attention especially in terms of reliably predicting triggering location and its probability under dynamically varying conditions.

Landslide susceptibility is evaluated with a probabilistic approach based on available historical precipitation records, considering only slope inclination and soil thickness as geomorphological controlling factors. In fact, owing to the characteristics of the considered area, the rest of the features influencing the landslide susceptibility like geo-lithology, geomorphology, vegetation and soil characteristics were assumed to be homogenous. The slopes of the area have been thus grouped in eight classes according to their inclination and soil thickness. The response of the slopes to precipitation is assessed by applying an 1D model of unsaturated flow and slope stability to the eight slope classes, considering the hourly rainfall recorded in 22 years (2002-23) by ten rain gauges around the study area. Slope susceptibility is evaluated as the historical (static) probability of landslide occurrence, based on the number of predicted slope failures from model simulations. Susceptibility mapping is carried out based on slope units, which are assigned to a slope class according to their inclination and thickness and are associated to the nearest rain gauge.

Landslide hazard is also assessed with a probabilistic approach, based on Bayes’ theorem, by integrating susceptibility with dynamic controls, i.e., triggering rainfall and antecedent rootzone soil moisture. Landslide triggering hazard is evaluated as the dynamic conditional probability, i.e. based on the number of failures for each slope class and for given event rain depth and antecedent soil moisture conditions. Hazard mapping is finally carried out based on slope unit susceptibility, and dynamic controls derived from the simulations with the nearest rain gauge data.

The obtained maps were tested by comparing them with actual reported landslides. Specifically, the susceptibility map well agrees with the locations of landslides recorded between 1999 and 2022. The operational applicability of the proposed hazard mapping was carried out by replacing the modelled antecedent conditions with those obtained from ERA5-Land. The dynamic triggering probability maps well identify the dates and the zones where landslides have been reported.