Considering parameter uncertainty in a GIS-based sliding surface model for large areas

The GIS-based open source software r.slope.stability computes broad-scale spatial overviews of shallow and deepseated slope stability through physically-based modelling. We focus on the landslide-prone 90 km² Collazzone area, central Italy, exploiting a comprehensive set of lithological, geotechnical and landslide inventory data available for that area. Inevitably, the geotechnical and geometric parameters are uncertain, particularly for their three-dimensional variability. Considering the most unfavourable set of geotechnical parameters (worst case scenario, appropriate for engineering purposes) is less useful to obtain an overview of the spatial probability (susceptibility) of landslides over tens of square kilometres. Back-calculation of the parameters based on topographic and geotechnical considerations would be experimental. Instead, we estimate the slope failure probability by testing multiple combinations of the model parameters sampled deterministically. Our tests indicate that (i) the geotechnical parameterization used allows to reproduce the observed landslide distribution partly (a challenge consists in the appropriate treatment of the variation of the geotechnical parameters with depth); (ii) the evaluation outcome depends strongly on the level of geographical aggregation; and (iii) when applied to large study areas, the approach is computingintensive, and requires specific strategies of multi-core computing to keep the computational time at an acceptable level.