Mass wasting events are preceded by the appearance and the propagation of cracks, during the preparation phase. Identifying these precursor signals in a robust and consistent way is crucial for hazard anticipation. In many cases, the signal-to-noise ratio (SNR) of crack seismic signal is too low to be effectively retrieved using seismic data, unless the sensors are located directly at the failing site. We adapt a state-of-the-art machine learning technique (Hidden Markov model), to overcome this problem and monitor the spatial and temporal crack evolution prior to a rockslide that happened in the Illgraben catchment on the 01/01/2013. The llgraben catchment is located in the Swiss Alps, in the canton of Valais. The steep flanks (30 to 80 degrees) of the catchment are notable source areas for rockslides. The site was instrumented with a network of seismometers from November 2012 to September 2013. We then use the spatial and temporal crack evolution prior to the rockslide to propose a physical explanation for the destabilization driving mechanism at the slope scale.
- Lagarde, S., Dietze, M., Gimbert, F., Laronne, J. B., Turowski, J., Halfi, E. (2021): Grain‐size distribution and propagation effects on seismic signals generated by bedload transport. - Water Resources Research, 57, 4, e2020WR028700. https://doi.org/10.1029/2020WR028700