Rock slope failure preparation paced by total crack boundary length

This study investigates the factors controlling rock slope failure using seismic data from the Illgraben catchment in Switzerland. A hidden Markov machine learning model was applied to seismic data to reveal the temporal evolution of cracks before a major rockslide. The results show an S-shaped crack rate pattern in the day before the rockslide, explained by a mechanistic model indicating that total crack boundary length is the key factor controlling failure plane evolution. This suggests that cracks should be treated as 2-D objects, and that slope failure can be internally driven. The model offers a novel approach for early warning of slope failures.

Sophie Lagarde, Michael Dietze, Conny Hammer, Martin Zeckra, Anne Voigtländer, Luc Illien, Anne Schöpa, Jacob Hirschberg, Arnaud Burtin, Niels Hovius, Jens M. Turowski