Seismically induced rockfall hazard from a physically based model and ground motion scenarios in Italy

The majority of landslide susceptibility and hazard zonations are implemented with statistical methods, especially on large scales: mostly because the data needed for physical simulations are only available in small areas. Physically-based simulations for slope stability are conceptually different from widely used statistical approaches. Both methods have specific advantages, depending on available data, their type and resolution, and the aim of the study. Here, we perform a hazard zonation based on the physical model STONE for the simulation of rockfalls, at 10 m resolution consistently all over Italy, and aggregating results at the slope unit level. The novelties, here, are: (i) the introduction of a seismic trigger for rockfalls, which adds a temporal component to an intrinsically static model and allows to obtain an estimate of seismically induced rockfall hazard, (ii) high-resolution application of the model at national scale, and (iii) implementation of the results in a WebGIS. Peak ground acceleration maps with different return times including seismic amplification represent the earthquake trigger. A data-driven map of possible rockfall sources all over Italy, mapped by experts in sample representative locations, allowed statistical generalization to unsurveyed areas, at national scale. Eventually, application of a simple linear transformation, to map values of peak ground acceleration into activation probability of sources, links "static" rockfall simulations with "dynamic", time-dependent triggering. Results are maps of rockfall susceptibility with different return times, i.e., a step forward to the full assessment of rockfall hazard. Maps of hazard values and corresponding uncertainties, aggregated at slope unit level and categorized, are readily available for download, and for visualization in the new WebGIS. The new model for seismic triggering of rockfalls can be applied at the local and regional scale, calibrated with specific earthquake events instead of the return time scenarios considered here. On the temporal scale, this approach in principle is suited for application in near-real time.