A workflow for building an automatic earthquake catalog from near-real time DAS data recorded on offshore telecommunications cable in central Chile.

International audience ; Distributed Acoustic Sensing (DAS) is of critical value for the offshore expansion of seismological networks. The work presented here is part of the 5-years ERC ABYSS project, which aims at building a permanent seafloor seismic observatory leveraging offshore telecommunication cables along the central coast of Chile. In preparation for this project, a first experiment named POST was conducted from October to December 2021 on a submarine fiber-optic cable connecting the city of Concón to La Serena. DAS data were recorded continuously for 38 days over a distance of 150 km, constituting more than 37,500 virtual sensors sampled at 125 Hz. We develop a workflow to detect more than 3500 local, regional and teleseismic events with local magnitudes down to ML = 0.5, automatically processing over 72 TB of data. We show that applying those methods to DAS data combined with data from the national onland seismic network greatly increases the accuracy of the earthquake hypocenter localizations. As a first step, we perform automatic seismic phase arrival picking using PhaseNet pretrained on conventional seismological stations, followed by phase association with GaMMA. We then apply a correction of the phase picks to account for shallow sedimentary layers and invert for the event hypocenter with VELEST. Finally, we estimate a local magnitude based on peak ground displacements. The ABYSS project near-real time data collection started the 30th of September 2023 using three DAS units to sense two offshore telecommunications cables connecting the cities of Concón to La Serena and La Serena to Caldera. The DAS data covers over 500 km of cable, comprising 30,000 virtual sensors sampled at 62.5 Hz. These data are synchronized once a day with a storage server located in France, the volume of which is anticipated to reach an estimated 608 TB by the end of the project. By applying our workflow, tested and validated on the POST experiment, to our daily data, we are able to process data in near-real time to ...