Mapping rockfall hazard and detecting precursory damage in rock slopes with passive seismic: Lessons from the La Praz case-study
Repeated passive seismic surveys were conducted over more than a decade on the unstable La Praz rock slope in Savoie, France, which involved an estimated rock volume of approximately 13,000--15,000 m${}^3$. The final survey took place two months before the complete collapse of the slope in August 2023. The amplitude of the Horizontal-to-Vertical Spectral Ratio at the peak frequency (*aHVSR(fp)*) was used to distinguish stable and unstable areas and identify fractures associated with significant mechanical decoupling. Rock compartments were also classified according to their rockfall hazard. The compartments most susceptible to failure exhibited the highest *aHVSR(fp)* values, ranging from 26 to 157, compared with values of 7.5--19 in intermediate-hazard areas and 2.7 in the adjacent stable rock mass. The main HVSR peak frequency (*fp*) exhibited distinct decreases of 15%, 14%, and 11% in three unstable rock compartments before the collapse. These reductions substantially exceeded the range of environmentally induced resonance-frequency fluctuations observed at the site and commonly reported in the literature as a limitation of this method. The frequency decreases were interpreted as indicators of progressive damage extending into the rock slope. Notably, these changes were detected exclusively in the three compartments identified as the most susceptible to failure, which also exhibited the highest *aHVSR(fp)* values. This study demonstrates, for the first time, the potential of passive seismic monitoring to reveal progressive in-depth damage within a complex, heterogeneous, and highly fractured rock slope. The findings highlight the value of combining HVSR amplitude and peak-frequency variations to characterize unstable rock compartments and identify potential precursors to catastrophic slope failure.