Mapping stress and structurally controlled crustal shear velocity anisotropy in California Naomi L. Boness* ⎤ ⎥ Department of Geophysics, Stanford University, Stanford, California 94305, USA Mark D. Zoback ⎦ ABSTRACT earthquakes recorded on three-component We present shear velocity anisotropy data from crustal earthquakes in California and seismic stations; the procedure minimizes the demonstrate that it is often possible to discriminate structural anisotropy (polarization of scatter in the data and, with careful consider- the shear waves along the fabric of major active faults) from stress-induced anisotropy ation of ray paths and local geology, allows (polarization parallel to the maximum horizontal compressive stress). Stress directions interpretation of the observed anisotropy as from seismic stations located near (but not on) the San Andreas fault indicate that the stress induced or structural. For each station, maximum horizontal compressive stress is at a high angle to the strike of the fault. In we search the earthquake catalog for events contrast, seismic stations located directly on one of the major faults indicate that shear that are above 20 km depth and located within deformation has significantly altered the elastic properties of the crust, inducing shear- a cone under the station up to a maximum wave polarizations parallel to the fault plane. incidence angle of 40Њ within the S-wave win- dow (Nuttli, 1961; Booth and Crampin, 1985). Keywords: crustal stress, seismic anisotropy, San Andreas fault. We filter each event with a low limit of 1 Hz and a high limit between 10 and 25 Hz (cho- INTRODUCTION 2004), aligned macroscopic fractures associ- sen to be 5 Hz above the dominant S-wave It has been known for the past 25 yr that ated with regional tectonics (Mueller, 1991), frequency) to remove high-frequency noise carefully used earthquake, well-bore, and geo- sedimentary bedding planes (Alford, 1986), without degrading the waveforms. We only logic data can be used to map the direction and the alignment of minerals or grains (Say- use data with a signal-to-noise ratio Ͼ3:1 and relative magnitude of in situ horizontal ers, 1994). These mechanisms can be divided (amplitude of initial S-waves relative to pre– principal stresses in the Earth’s crust (Zoback into two major categories: stress-induced and S-wave coda) and visually inspect each seis- and Zoback, 1980, 1991; Zoback, 1992), and structural anisotropy (Fig. 1). Stress-induced mogram and the corresponding particle mo- tion to establish if the S-wave arrivals are a global database of more than 10,000 crustal shear anisotropy is the result of SHmax causing stress indicators, the World Stress Map,is microcracks to open and/or preexisting mac- impulsive enough to be picked with confi- now available (Zoback et al., 1989; Reinecker roscopic fractures to close, generating a fast dence. We use the two horizontal components to determine the anisotropy between the S- et al., 2005). direction parallel to SHmax. Stress-induced an- In this paper we present shear velocity an- isotropy is observed in the upper crust and the waves because we expect vertically propagat- isotropy data from local earthquake sources as effect decreases with depth as the confining ing waves within the S-wave window. an independent tool to analyze the state of pressure increases, closing fractures in all ori- We measure the S-wave splitting of each stress close to active faults on a regional scale entations. Borehole measurements indicate event that satisfies our quality criteria using a and in geographic regions where other types that anisotropy is largest in the near surface, technique that combines covariance matrix de- of stress measurements are lacking. We use with values of ϳ10% (e.g., Aster and Shearer, composition (Silver and Chan, 1991) with data from the Southern California Seismic 1991; Boness and Zoback, 2004; Liu et al., cross-correlation (Bowman and Ando, 1987). Network and Northern California Seismic 2004), and decreases with depth. However, ev- The covariance matrix of the horizontal S- Network, with an emphasis on Southern Cal- idence suggests anisotropy is still Ͼ3% in the wave particle motion is rotated over azimuths ifornia. Western California is a good place to granitic rocks in California at a depth of 3 km of Ϫ90Њ to ϩ90Њ in increments of 1Њ and over demonstrate this method because there are (Boness and Zoback, 2005). Structural aniso- delay times of 0–50 ms in increments of the many independent stress measurements, the tropy occurs when macroscopic features such sampling frequency. The fast direction and de- tectonic structures are well documented, and as fault-zone fabric, sedimentary bedding lay time that best linearize the particle motion are determined by minimizing the second ei- the direction of SHmax is, in general, at a high planes, or aligned minerals and/or grains po- angle to the faults, allowing us to differentiate larize the S-waves with a fast direction in the between the two mechanisms (e.g., Zinke and plane of the feature. With knowledge of the Zoback, 2000). structural elements in a region, including ma- Numerous examples of seismic anisotropy jor faults, it is possible to determine whether in the upper crust have been documented in anisotropy is caused by stress, structure, or a the literature over the past 25 yr since it was combination of both mechanisms (Zinke and first observed using microearthquakes (Cram- Zoback, 2000). In this paper we focus on the pin et al., 1980). The mechanisms that cause fast polarizations of the S-waves, because they shear waves (S-waves) to split into a fast and are mostly dependent on the last anisotropic slow component include dilatancy of micro- medium the wave passes through (e.g., Cram- Figure 1. Schematic illustrating stress- cracks due to stress (Crampin, 1991), prefer- pin, 1991) and more robust than the delay induced anisotropy in crust adjacent to fault ential closure of macroscopic fractures in an time. zone, where vertically propagating shear anisotropic stress field (Boness and Zoback, waves are polarized with fast direction par- allel to SHmax due to preferential closure of METHODOLOGY fractures (dashed lines), and structural an- *Present address: Chevron, 6001 Bollinger Can- We have devised a quality control proce- isotropy of shear waves inside fault zone yon Road, San Ramon, California 94583. dure to measure S-wave splitting using micro- with fast direction parallel to fault fabric. ᭧ 2006 Geological Society of America. For permission to copy, contact Copyright Permissions, GSA, or [email protected]. Geology; October 2006; v. 34; no. 10; p. 825–828; doi: 10.1130/G22309.1; 3 figures; Data Repository item 2006180. 825 genvalue, and we compute the degree of rec- tilinearity (Jurkevics, 1988) from the ratio of the two nonzero eigenvalues. We only include measurements with a degree of linearity Ͼ0.8 (1 being perfectly linear). We use the maxi- mum cross-correlation coefficient of the rotat- ed waveforms to confirm that the seismograms have been rotated into the fast and slow po- larization directions and discard any measure- ment with a cross-correlation coefficient Ͻ0.8. After the fast shear polarizations have been determined for all earthquakes at a given sta- tion, we compute the mean orientation of the fast S-waves using Fisher statistics (Fisher et al., 1987). If the azimuthal standard deviation is Ͻ20Њ, we believe that the fast direction is well constrained and probably contains valu- able information about either stress or struc- ture. In contrast, standard deviations Ͼ20Њ probably indicate that a mix of mechanisms (e.g., Peng and Ben-Zion, 2004), scattering (e.g., Aster et al., 1990), or complicated local geology (e.g., Aster and Shearer, 1992) is causing the anisotropy. RESULTS We apply this method to data from 86 three- component stations and achieve a well-defined mean fast direction with a standard deviation of Ͻ20Њ at 62 stations (Fig. 2). The 10 stations shown in red exhibit a mean fast direction that is within 20Њ of the local San Andreas fault (SAF) strike (and other nearby major subpar- Figure 2. Map of California showing mean fast shear polarizations at stations with stan- ؇ allel faults). The red rose diagrams in Figure dard deviation <20 . Red stations exhibit fast polarization that is subparallel to San Andreas fault and are typically located along main fault trace. Corresponding red rose diagrams 2 are the individual S-wave splitting measure- show measurements of fast polarization observed at each of these stations with adjacent ments used to compute the mean at each of numbers indicating number of good quality measurements out of total number of earth- those stations (see GSA Data Repository Ta- quakes analyzed. Blue stations exhibit fast shear polarizations parallel to direction of SHmax ble DR11). The numbers associated with each (in black) determined from focal mechanism inversions (sticks) and borehole breakouts bowties). Stations where fast azimuth standard deviation is >20؇ (yellow diamonds) indicate) rose diagram indicate the number of good mix of mechanisms or scattering. Data for three white stations showing (a) stress, (b) struc- measurements that complied with our quality tural, and (c) mixed anisotropy are displayed in Figure 3. control criteria out of the total number of earthquakes analyzed. Similarly, the blue sta- tions and rose diagrams indicate stations with sistent with the direction of SHmax obtained northwest of Parkfield that shows structural mean fast directions that are at an angle Ͼ20Њ from focal mechanism inversions and well- anisotropy (Fig. 3B), the earthquakes are al- to the SAF. bore breakouts (data shown are from Townend most directly beneath the station and most are Nearly all of the stations exhibiting a mean and Zoback, 2004). There are a few stations relatively shallow (5–8 km). One station that fast direction subparallel to the structural fab- located on major faults like the San Jacinto shows the mixed effects of structural and ric are located on the SAF.