Detecting seismo-volcanic events based on inter-channel coherency of a DAS cable

Abstract

Distributed Acoustic Sensing (DAS) has been successfully employed to monitor volcanic seismicity and to infer volcanic subsurface structures. Here, we analyse data recorded in September 2018 at Mount Etna by the 9N seismic network. The multi-instrument network includes a 1.3 km long fibre-optic cable that was buried 2-2.5 km away from the main craters. Additionally, 15 geophones were installed along the trajectory of the DAS cable, allowing for a comparison of strain-rate and ground velocity data.During the acquisition period, tiny seismic events, likely caused by fluid movement and degassing, are visible with inter-event times in the range of 1 min. Volcanic explosions and volcano-tectonic earthquakes also occur frequently. We detect events over all frequency ranges by calculating the coherence matrix for very short time windows (stacking 15 windows of 5 seconds length). An eigendecomposition of the coherence matrices allows to extract the first eigenvectors, corresponding to the dominant source in the time window. The principal eigenvectors can be clustered to find groups of events with similar source properties. We also use the principal eigenvector of already known events as a matched filter to scan the whole dataset. The results of the DAS cable are compared to the observations of the geophone array. While largely obtaining similar findings, the DAS cable seems to better capture high-frequency features of certain events. We also explore the effects of stacking and downsampling of the DAS data prior to detection, which influences both resolution and computational efficiency of the algorithm.