Seismic data as an additional layer of information for large rockslide modelling through back analysis

When studying large landslides any data matters. In fact, in contrast to other scientific branches, data scarcity is a pressing issue especially for high altitude landslides. When large rockslides occur in steep valleys, long runouts can threaten buildings and infrastructures, even though the source area is well above the exposed elements. In such cases, usually no ground-truth monitoring data is available on site, which is needed to understand the causes and processes of the collapse.

Seismic data from a widespread network of seismographs can help to partially fill gaps in the characterization of the above mentioned processes. The most straightforward information that can be inferred from ground motion recordings is the duration of the event – or at least the duration of the most intense and violent phase of the runout. By coupling the velocity of the event with some topographic data leading to the estimation of the detached volume and the deposit distribution, the user may gather sufficient information to produce a satisfactory numerical model through back analysis. In case the topographic data are characterized by high uncertainty and/or poor resolution, seismic records are particularly useful for ground-truthing because they represent an independent source of data.

This study describes the modelling approach used to understand the dynamic of a 365,000 m³ rockslide in the Dolomites (UNESCO World Heritage, North-East Italy). The landslide detached from a steep slope located between 3100 and 2800 m a.s.l. and almost free fell for 600 m. Then it crashed and fragmented in a small rocky hanging valley of glacial origin, subsequently reaching 1400 m a.s.l. with a runout of approximately 2 km. For this area, pre-event and post-event DEMs were available but with different resolution, alignment and coverage. The comparison of pre- and post-event topography allowed the identification and quantification of erosional and depositional areas, the estimation of landslide volume and of the potential errors associated with this type of analysis.

A DAN3D numerical model of the landslide was calibrated using both DEM of Difference (DoD) maps and seismic data. The ground motion records proved to be remarkably useful, as they ensured the reliability of the model notwithstanding the DoD maps intrinsic uncertainties. The seismic data provide a new layer of information in a swiss cheese model of reliability focused on reducing model equifinality and on increasing the overall robustness of the analysis. This finding is fundamental as the results of the back analysis may be used to model and test future scenarios, which can be used to support risk assessment and mitigation.