Abstract. Although shallow landslides are a very widespread phenomenon, large area (e.g. thousands of square kilometres) early warning systems are commonly based on statistical rainfall thresholds, while physically based models are more commonly applied to smaller areas. This work provides a contribution towards the filling of this gap: a forecasting chain is designed assembling a numerical weather prediction model, a statistical rainfall downscaling tool and a geotechnical model for the distributed calculation of the factor of safety on a pixel-by-pixel basis. The forecasting chain can be used to forecast the triggering of shallow landslides with a 48 h lead time and was tested on a 3200 km2 wide area.

Abstract:To properly evaluate weather variables regulating the occurrence of geo-hydrological hazards, the current constraints of climate models imply the need of adopting statistical approaches in cascade to GCM/RCM for the assessment of the potential variations associated to climate changes. Since, in the last years, several approaches, often freely available, have been proposed and applied to investigate various hazards in different geographical areas and geomorphological contexts, a deeper understanding about their performances and constraints is crucial; in the work, it is carried out focusing the attention on two kind of approaches widely adopted in impact studies: bias correction methods (in particular, quantile mapping tools) and weather generators. Both methodology have been applied to outputs of an high resolution RCM simulation carried out on Italian territory for analyzing two very localized (and then challenging) landslide case studies. Beyond an assessment about relative performances in reproducing weather variables on the areas, the goal concerns an increasing awareness about how these approaches could affect the climate signal, physically detected by RCM, not only in outputs weather variables but also in derived components of soil surface budgets strictly governing the occurrence of landslide phenomena.

No abstract

No abstract