Fluid injection, steam extraction, and reservoir stimulation in geothermal systems lead to induced seismicity. While in rare cases induced events may be large enough to pose a hazard, on the other hand the microseismicity provides information on the extent and the space-time varying properties of the reservoir. Therefore, microseismic monitoring is important, both for mitigation of unwanted effects of industrial operations and for continuous assessment of reservoir conditions. Here we analyze induced seismicity at The Geysers geothermal field in California, a vapor-dominated field with the top of the main steam reservoir some 1-3 km below the surface. Commercial exploitation began in the 1960s, and the seismicity increased with increasing field development.
We focus our analyses on induced seismicity recorded between August 2007 and October 2011. Our calibrated waveform database contains some 15000 events with magnitudes between 1.0 and 4.5 and recorded by the LBNL Geysers/Calpine surface seismic network. We associated all data with events from the NCEDC earthquake catalog and re-picked first arrival times. Using selected events with at least 20 high-quality P-wave picks, we determined a minimum 1-D velocity model using VELEST. A well-constrained P-velocity model shows a sharp velocity increase at 1-2 km depth (from 3 to 5 km/s) and then a gradient-like trend down to about 5 km depth, where velocities reach values of 6-7 km/s. The station corrections show coherent, relatively high, positive travel time delays in the NW zone, thus indicating a strong lateral variation of the P-wave velocities. We determined an average Vp-to-Vs ratio of 1.67, which is consistent with estimates from other authors for the same time period. The events have been relocated in the new model using a non-linear probabilistic methods. The seismicity appears spatially diffused in a 15x10 km2 area elongated in NW-SE direction, and earthquake depths range between 0 and 6 km. As in previous seismicity studies of this geothermal field, we find that events occurring in the NW sector are on average deeper than in the SE area.
To infer the present stress regime, we computed focal mechanisms of a large event data set with M > 2, using P-wave first-arrival polarities. The found fault-plane solutions show a dominant strike-slip and normal faulting mechanisms, with P and T axes coherently oriented with expected regional stress field for the area.
We also determined the main seismic source parameters from a multi-step, iterative inversion of P-wave displacement spectra, assuming a four-parameters spectral model and a constant-Q attenuation mechanism. In particular, we computed seismic moments, source radii and stress drops. We observe a self-similar scaling of source parameters in the whole investigated magnitude range, with a nearly constant stress-drop of 20 and 120 MPa depending on the use of Brune (1970) or Madariaga (1976)’s source model respectively.