Induced seismicity and environmental change at The Geysers geothermal area in California
Intensive exploitation of the reservoir at The Geysers geothermal area, California, induces myriads of small-magnitude earthquakes that are monitored by a dense, permanent seismometer network that covers most of the reservoir. However, majority of the seismic stations, which belong to the UNOCAL network are poorly calibrated. Station polarities, and sensor orientations for the 8 three-component stations of this network were determined by using accurate focal mechanism solutions from a temporary network and using a simple method of observing the waveform from known earthquake locations. Using data from the UNOCAL network, tomographic inversions were performed for the three-dimensional Vp and Vp/Vs-ratio structure of the reservoir for February 1993, October 1996 and August 1998, adding to the inversions for April 1991 and December 1994 that had already been performed by other investigators. The extensive low -Vp/Vs anomaly known to characterise the reservoir grew progressively in strength from a maximum of 9% to a maximum of 12.4% at sea level during the seven-year study period. The anomaly growth is attributed to the depletion of pore liquid water in the reservoir and its replacement with steam. This causes Vp to decrease by increasing compressibility, and Vs to increase because of the reduction in pore pressure and the drying of argillaceous minerals, e.g., illite, which increases the shear modulus. All these effects serendipitously combine to lower the Vp/Vs ratio, resulting in an exceptionally strong overall effect that provides a convenient tool for monitoring reservoir depletion in the seismogenic zone. Variations in the separate Vp and Vs fields indicate that water depletion was the most important process in the central part of the exploited reservoir, and that pressure reduction and mineral drying were the dominant effects more northwesterly and southeasterly. Relative relocation of micro earthquakes was also performed using the same network. Four regions were studied. Although most multiplets relocated into tighter clusters and the reduction in the RMS of the relative relocations was good, further work is needed to substantiate these initial findings.