| Abstract |
Hypocenter location of earthquakes is one of the most important sources of information to understand the physical processes at the origin of earthquakes, to describe the subsurface and to quantify earthquake seismic hazard. This characteristic is necessary to compute several other attributes of the seismic source, e.g. origin time, seismic moment, focal mechanism, which will complete the earthquake catalogue. However, location errors exist and need to be properly quantified because with the earthquake hypocenter they determine the meaningful scale of investigation. We analyze for the Rittershoffen geothermal field (Upper Rhine Graben, France) the effect of velocity model errors on the determination of earthquake absolute locations. To do so, we first generate synthetic earthquakes in the geothermal reservoir and calculate in a 3D fault model the associated travel times on several seismic networks. Then, we relocate the events using a non-linear absolute location procedure, however in a reference 1D velocity model, as it is typically assumed for initial data processing. Thereby, we introduce velocity model errors between the synthetic and the relocation phases. The synthetic events are distributed in a volume where seismicity was induced during stimulations of the geothermal well GRT1. They range approximately between 1 and 4 km depth within a radius of 1.5 km around the well-head. The lay-out of the seismic network monitoring these operations is used as well as two other ones that are representative of the network densification over time. The results show that the reference 1D velocity model, despite built from well log data, is not a good representative of a more realistic 3D model including a fault and its associated block shift. The seismic network coverage and the velocity model control the amplitude and orientation of the induced relocation uncertainties and inaccuracies in the zone of interest, which are neither constant nor aleatoric. Although a denser network with better coverage clearly decreases location uncertainties, location inaccuracies can still increase and be much larger than the uncertainties. This emphasizes the very different behavior and physical meaning between both quantities, which should not be confused. The induced location inaccuracies may be such that the positioning and orientation of features delineated by seismicity are strongly distorted and difficult to correctly interpret, even in the case of a very dense seismic network. However, we show that a calibration shot can remove most of these effects. |