PROCEEDINGS, Twenty-Sixth Workshop on Geothermal Reservoir Engineering Stanford University, Stanford, California, January 29-31, 2001 SGP-TR-168 SHEAR WAVE SPLITTING AND FRACTURE PATTERNS AT THE GEYSERS (CALIFORNIA) GEOTHERMAL FIELD D. Erten, M. Elkibbi and J.A. Rial Wave Propagation Laboratory, Department of Geological Sciences, University of North Carolina at Chapel Hill , NC 27599 ABSTRACT along the ray joining the source and the re- ceiver. In addition, the source time history, me- Shear-wave splitting analyses from recorded dium heterogeneity, a weathered surface layer, microearthquakes at The Geysers (California) the presence or absence of fluids in the cracks, indicate that subsurface predominant fractures and the surface topography tend to obscure the in the field are oriented N10oE to N50oE under effect of crack-induced anisotropy. the NW Geysers seismic array and N40-50oE Ideally, the approaching S-wave is split and N40-50oW under the SE Geysers array. into a fast wave polarized along the predomi- These crack alignments seem consistent with a nant crack direction, and a slow wave, polarized local structure characterized by a pull-apart perpendicular to it and arriving a few tens of block structure limited by two right-lateral milliseconds later. Polarization diagrams are strike-slip faults oriented NW-SE and regionally used to detect the marked switch in polarity of controlled by the San Andreas fault. the two arrivals, which provide the clearest indi- cation of medium anisotropy (Figure 1). It is the INTRODUCTION assumption here that the splitting of shear waves Shear-wave splitting due to patterns of stress- in The Geysers is indeed induced by the pres- aligned cracks in the crust has been widely ob- ence of oriented fractures in an otherwise homo- served in a variety of earthquake settings and geneous medium. Typically, the orientation of controlled-source seismic data. It has also been the fast wave polarization can be measured with recognized that the polarization of the leading errors not exceeding 10%. split shear wave is usually parallel to the local The time delay between the fast and strike of crack systems (or normal to the direction slow shear waves is measured directly after the of the minimum horizontal stress). In addition, seismogram is rotated to orient the fast and slow the time delay between fast and slow shear waves arrivals along the instrument’s horizontal com- is directly related to the number of cracks per unit ponents. This operation cleanly separates the volume (crack density) in the medium (Crampin, two shear wave arrivals allowing direct and ac- 1987; Crampin and Lovell, 1991). Therefore, the curate measurement of the time delay. Occa- interpretation of shear-wave splitting is an im- sionally, rotation of all three components is portant diagnostic tool to determine the direction performed to align the vertical component with and evaluate the bulk density of subsurface frac- the approaching ray. Most delay times recorded tures in hydrocarbon and geothermal reservoirs. at The Geysers range from 20 to 70 ms (see Fig- ure 1). The smallest time interval that can be METHODOLOGY measured is 2.1 ms (480 samples/sec). The shear-wave splitting method has been de- At each receiver polarization data are scribed extensively (e.g., Crampin, 1981; Lou collected in equal area projection diagrams and and Rial, 1997; Lou et al., 1997). Although rose diagrams. Numerous measurements per straightforward, the analysis of shear-wave station (Figures 2, 3, 4, 5) allow for a statisti- splitting for the purpose of crack detection must cally robust determination of the preferred di- be based on a large set of shear-wave seismo- rections. gram data. This is in part because the behavior One important restriction to the shear- of shear waves in cracked media is usually as wave splitting analysis is that the arriving rays complex as the distribution of cracks may be need to be within the shear wave window de- 2 -1 fined by a critical angle ic = sin (b/a), where a · Automatic selection of event from data and b are the P-wave and S-wave surface ve- bases locities, respectively. For angles of incidence · Interactive display of 3-component greater than ic, shear-waves interact strongly seismograms and their power spectra with the free surface, thus distorting the incom- · Interactive filter selection and win- ing waveform (Crampin and Booth, 1985). dowing of the S-wave train · Interactive horizontal and vertical rota- tion of windowed components · Interactive display of rotated seismo- grams and determination of time delay · Two- and three-dimensional displays of particle motion · Automatic recording of all measured parameters. THE GEYSERS The Geysers geothermal field is located north- east of the San Andreas Fault in the northern Coast Range of California (110 km north of San Francisco). There is abundant microseismicity in the field, some of which appears induced by steam production. Two major right-lateral Figure 1. Shear wave window for a selected event. strike-slip faults related to the San Andreas fault Top row shows the two horizontal components and the system, the Mercuryville and Collayami fault horizontal particle motion of the S-wave train. Note that the angle between the two split shear waves is close to zones, appear to form the southwest and north- 90 degrees. The lower row shows how the rotated seis- east boundaries of the steam field (McLaughlin, mograms separate the arrivals into fast and slow. The 1981). (uncorrected) crack orientation is N65E. Time delay is There have been extensive seismic 64.5ms, measured directly from the rotated seismo- grams. studies at the Geysers geothermal field in recent years. Conventional seismological studies were The critical angle is 35o in a half-space with a first carried out to obtain P- and S-wave veloci- Poisson’s ratio of 0.25. In The Geysers it is ties, seismic Q values within the geothermal possible to increase the angle to 45o due to the reservoir etc (e.g., Zucca et al., 1994, Romero et presence of the near surface low velocity layers. al, 1995, Julian et al, 1993). More recently, Routine measurement of shear wave studies by our UNC-CH group (Erten and Rial, splitting is tedious because it requires direct 1998, Erten and Rial, 1999) and at Duke Uni- inspection of each seismogram by a trained op- versity (Shalev and Lou, 1995) have focused on erator. Our experience with The Geysers data the determination of the crack directions using suggests that automatic picking of polarization the shear-wave splitting method. directions and time delays without human inter- For simplicity we shall subdivide The vention is unreliable. This is because of the Geysers reservoir into two areas: the NW and great variability and diversity of wave patterns SE Geysers. The two areas are different in the that result from the interaction of shear waves number and quality of data available as well as with complex cracked media. Thus, to expedite in the type of instrumentation used. For in- and simplify the measuring process we devel- stance, seismographs in NW Geysers are down- oped computational graphic user interfaces hole instruments, while in SE Geysers all in- (GUIs) based on MatlabTM software (e.g., Figure struments are on the ground surface. 1). These automate the measurement process through a series of user-friendly, interactive SHEAR-WAVE SPLITTING FROM measurement devices and multidimensional NW GEYSERS displays that include the following: The NW Geysers area is a fairly active seismic zone with an average of 17 microearthquakes per day. Most of the earthquake activity is con- 3 centrated in the southeastern part of the Cold- polarization directions, large signal-to-noise water Creek Steam Field (CCSF), coinciding ratio, clear, robust shear-wave splitting and arri- with the production area. The data used for this val angle within the shear-wave window (e.g., study were recorded by a 16-station, digital, Figure 1). Since the seismometers at NW Gey- three component network operated by Lawrence sers are down-hole instruments, it is necessary Berkeley Laboratory during 1988 and part of to determine the true geographic orientations of 1994. All 16 geophones record at 400 sam- the horizontal components. To do this, we com- ples/sec and are buried about 30 meters below pared predicted and observed P-wave arrival the surface. angles for events surrounding the station. An After processing and analyzing over average of 20 events per station was used to 6000 events we have collected a large number of estimate the ‘text-book’ seismograms that show consistent actual horizontal component orientations. Figure 2. The rose diagrams at each station indicate the preferred polarization. Station names are indicated. An average of 70 events per station were used to generate the rose diagrams. Station S13 (NW) used the least number of events (11) and station S12 (SE) the largest (204). This resulted in standard deviations within 10o to N500E. Two exceptions are stations 9 and 14 for most station corrections. The geographic where the orientations are NW-SE. In general, orientations of the cracks are obtained after cor- the crack pattern as suggested by the fast shear recting for station component orientation. wave polarizations is consistent with the exis- tence of aligned fractures held open by a re- STRIKE OF LEADING SHEAR-WAVES: The gional northeast-southwest maximum horizontal preferred orientations for the leading shear- compressive stress in the Northwest Geysers waves within the steam field range from N100E (Romero et al., 1995). It is also consistent with a 4 study by Nielson et al. (1991) which measured fracture orientations in cores inside the CCSF and found predominant north-northeast strikes of fractures parallel to the maximum horizontal compressive Figure 3.