A data-driven approach to assess the role of the groundwater conditions in triggering shallow landslides initiating with frictional failure

In the Campania region (southern Italy), sloping pyroclastic surfaces in partially saturated conditions are frequently affected by flow-like landslides and shallow slip. In this context, safeguarding human life is entrusted to early warning systems (EWSs), typically based on empirical precipitation intensity-duration thresholds that may suffer from missed and false alarms. Indeed, the consequences of heavy rainfall depend on the hydraulic conditions in the soil before the rainstorm. If this aspect is neglected, precipitation intensity-duration thresholds can fail. However, although the current hydraulic condition in the soil is a discriminating factor, it is rarely measured and taken into account. In this regard, the present study focuses on the assessment of the role of the groundwater conditions for the initiation of shallow landslides, and it may be the first step towards the implementation of an innovative early warning system. At this aim, a numerical approach to reproduce the current hydraulic condition in the slope is presented. This study is based on the back-analysis of two debris flows that occurred in the Campania region by means of uncoupled hydromechanical numerical modelling. The stratigraphy and hydromechanical characteristics of the soil layers are known due to surveys and investigations carried out at an experimental test site set up a few kilometres from the two debris flows that occurred. The results show that in the slopes examined, the hydraulic state in the soil before landslide triggering was correctly reproduced through an infiltration analysis starting a few months before, in which rainfall recorded on site was imposed as an entering flow at ground level. Moreover, it was shown that the water storage in the entire loose soil cover was an effective indicator of the average hydraulic state of the slope; more than local variables, such as matric suction and water content.