The modification of the stress field induced by fluid injection into the ground can generate seismic motions. Their monitoring is a key point to limit the occurrence of impacting events. Generally, this is performed using seismic surface networks, which can be limited by a significant ambient noise level especially in urban contexts. An alternative consists in the installation of stations in the depth of wells to increase the distance with surface ambient noise sources. However, few are the industrial projects fitted with such technologies, because of their cost and complexity of installation. Another possibility is to operate with dense seismic networks (seismic arrays), combined with appropriate data processing, to limit the impact of anthropogenic noise by distinguishing it from earthquakes. Here we investigate the case of the “Strasbourg induced earthquake sequence”, occurring since mid-2018 around the Geoven deep geothermal doublet operated by the Fonroche company in Vendenheim (France). So far, the BCSF-Rénass (national observatory service in charge of the french seismicity monitoring) has recorded 567 induced earthquakes using traditional local and regional seismic networks. Their catalogue has an estimated magnitude of completeness of Mc=0.6 at best, containing event with a local magnitude (Mlv) up to 3.9, including 22 with Mlv>2 and 4 with Mlv>3. These events are organized into two distinct swarms: a first cluster in the vicinity of the Geoven wells and a second one 4-5km South from it. Although the project has been forced to stop because of the felt induced seismicity, the Northern cluster is still very active, with the largest event occurring the 26th of June 2021. To improve our knowledge of this seismic crisis, we deployed 3 mini seismic arrays of 21 SmartSolo nodes each around the active cluster, recording at a sampling rate of 1000Hz for 4 months starting a few days after the Mlv=3.6 event of 4th of December 2020. The aperture of each array is around 70m, allowing good wave number resolution in the frequency range of interest for local seismic events. Beamforming and match field processing techniques allow us to characterize the local ambient noise, which consists mostly in surface waves with slow apparent velocities. As the arrays are located roughly on the top of induced seismic events hypocenters, the front waves illuminate the arrays with a significantly higher apparent velocity. Therefore, stacking brutally the waveforms increases drastically the SNR. We improve it even more by considering the signal instantaneous phase as a coherency parameter during the stacking process, what is called phase-weighted stacking. This allows us to detect events down to magnitude -0.5, which leaves us with 4 to 5 times more events than the BCSF-Rénass catalogue. In parallel, we also investigate how much these arrays can improve event location as a complement to traditional networks.
Rémi Fiori, Jérôme Vergne, Jean Schmittbuhl, Dimitri Zigone, Sophie Lambotte
EOST/ITES, University of Strasbourg/CNRS, 5 rue René Descartes, 67000, Strasbourg