Contribution of the Surface and Down-Hole Seismic Networks to the Location of Earthquakes at the Soultz-sous-Forêts Geothermal Site (France)

Abstract

Seismicity induced in geo-reservoirs can be a valuable observation to image fractured reservoirs, to characterize hydrological properties, or to mitigate seismic hazard. However, this requires accurate location of the seismicity, which is nowadays an important seismological task in reservoir engineering. The earthquake location (determination of the hypocentres) depends on the model used to represent the medium in which the seismic waves propagate and on the seismic monitoring network. In this work, location uncertainties and location inaccuracies are modeled to investigate the impact of several parameters on the determination of the hypocentres: the picking uncertainty, the numerical precision of picked arrival times, a velocity perturbation and the seismic network configuration. The method is applied to the geothermal site of Soultz-sous-Forets, which is located in the Upper Rhine Graben (France) and which was subject to detailed scientific investigations. We focus on a massive water injection performed in the year 2000 to enhance the productivity of the well GPK2 in the granitic basement, at approximately 5 km depth, and which induced more than 7000 earthquakes recorded by down-hole and surface seismic networks. We compare the location errors obtained from the joint or the separate use of the down-hole and surface networks. Besides the quantification of location uncertainties caused by picking uncertainties, the impact of the numerical precision of the picked arrival times as provided in a reference catalogue is investigated. The velocity model is also modified to mimic possible effects of a massive water injection and to evaluate its impact on earthquake hypocentres. It is shown that the use of the down-hole network in addition to the surface network provides smaller location uncertainties but can also lead to larger inaccuracies. Hence, location uncertainties would not be well representative of the location errors and interpretation of the seismicity distribution possibly biased. This result also emphasizes that it is still necessary to properly describe the seismic propagation medium even though the addition of down-hole sensors increases the coverage of a surface network.