Using a seismic network for automatic detection, localization and characterization of mass movements in the Mont-Blanc massif

In the Mont-Blanc massif (western European Alps), seismic stations record numerous signals originating from surface mass movements, such as rockslides, rockfalls, and serac avalanches. The large number of recorded signals makes the automation of the processing workflow essential for practical application. These seismic waveforms differ significantly from those generated by earthquakes, making standard algorithms unsuitable for their analysis. The signals typically exhibit an emergent onset, making it challenging to precisely determine their start time. Moreover, the arrival times of P and S waves, routinely used for earthquake localization, cannot be easily identified. The seismic records also vary in length, reflecting the differing durations of the associated phenomena.

To analyze such data using a seismic network, we adapted selected algorithms to address these challenges. For detection, we chose the STA/LTA algorithm, and for localization, we used amplitude decay algorithm and BackTrackBB software, which exploits wave field coherence. To test these algorithms, we created a reference dataset consisting of large, well-documented mass movements. The dataset was developed using regional mass movement databases, webcam image analysis, direct observations made by a network of observers, and seismic data from the Sismalp network. This reference dataset enabled us to fine-tune the algorithms and automate the processing of waveforms related to mass movements.