Serdar Özalaybey and Martha K. Savage
Seismological Laboratory, Mackay School of Mines, University of Nevada, Reno, NV 89557-0141
We have examined shear wave splitting in teleseismic shear waves from 26 broadband stations in the western U.S. Fast polarization directions (phi) and delay times (dt) show spatial variations that are coherent within geologic provinces. Stations located near the San Andreas fault show clear evidence for fault-parallel anisotropy in the crust and upper mantle (115-125 km thickness). This can be explained by the finite strain associated with the relative plate motion between the North American and Pacific plates. The lateral extent of this strain field is probably narrow to the west, because stations 55 km west of the San Andreas fault do not show fault-parallel anisotropy in southern California. Station LAC located 80 km east of the San Andreas fault shows large fault-parallel anisotropy. This suggests that the Pacific-North American plate boundary in the mantle might be displaced to the east in southern California. A deeper E-W oriented fast direction of anisotropy underlies the fault-parallel anisotropic layer in the vicinity of the San Andreas fault. An E-W fast feature is also present beneath the western Basin and Range and the foothills of the Sierra-Nevada, although local variations are present. The magnitude of delay times suggests that this feature resides in the asthenosphere. We interpret this feature as the asthenospheric flow in the slabless window left behind the Farallon plate. The flow-induced anisotropy may partially be frozen-in at shallow depths. Station ORV is located near the southern edge of the Gorda slab where no anisotropy is detected. The absence of anisotropy at this location could therefore mark a boundary between Farallon associated flow and regions where E-W oriented asthenospheric flow did not occur. The lack of evidence for NE-SW fast orientation within the Walker Lane Shear Belt of western Nevada suggests that this crustal feature does not extend into the mantle or that is not as well developed as that beneath the San Andreas fault. Stations located over the young subducting Gorda plate mark a change in the fast direction to nearly NE-SW. This direction aligns well with the maximum compressive stress direction in the overlying North American plate and the NE-SW directed internal shearing of the Gorda plate. The anisotropic thicknesses calculated from delay times suggest roughly double that expected for purely lithospheric contributions. This implies that the anisotropic thickness may include some of the asthenosphere. Alternatively, using a higher anisotropy of 8% can bring thicknesses in line with other measures of lithospheric thicknesses. The correspondence between the fast directions and the present plate tectonic deformations suggest that mapping upper mantle deformation through seismic anisotropy is a viable method, and that asthenospheric flow may be a significant contributor to seismic anisotropy.
Shear wave splitting results obtained from teleseismic shear waves have become a powerful source for examining seismic anisotropy and the related deformation history of the strain field in the uppermost mantle (Kind et al., 1985; Silver and Chan, 1991; Vinnik et al., 1992; Savage and Silver, 1993). In most cases, splitting measurements have been explained by models requiring single or double homogeneous anisotropic layers with horizontal symmetry axes localized in the upper mantle (Silver and Chan, 1991; Farra et al., 1992; Kaneshima and Silver, 1992; Savage and Silver, 1993; Vinnik et al., 1994). Savage and Silver (1993) have established the first anisotropic mantle model deduced from shear wave splitting that is consistent with the tectonics of the western U. S. They have shown that splitting parameters, fast direction of anisotropy (f) and the delay time (dt) between the fast and slow directions, vary azimuthally for stations along the northern San Andreas Fault (SAF) (Berkeley network) as well as station LAC located in the Mojave Desert east of the southern SAF. They modeled these variations by a system consisting of two homogeneous anisotropic layers; a top layer with fast orientation locally parallel to the strike of SAF and a lower layer with E-W fast orientation beneath the Berkeley network and NE-SW fast orientation beneath LAC. These intriguing results have also been confirmed using a double-layer waveform method and a more complete data set (Ozalaybey and Savage, 1994).
Although the hypothesis that the strain field produced by the relative plate motion between the North American and Pacific plates should give rise to seismic anisotropy and therefore shear wave splitting with fast axis parallel to the SAF has been supported with local measurements (Savage and Silver, 1993), the uniformity of this strain field and its spatial distribution has not yet been determined. In this paper, we present the results of a study of shear wave splitting from 26 broadband stations located in the California and Nevada regions. We aim to characterize the nature of mantle anisotropy and its relation to tectonics along and away from the SAF and find if this double-layer anisotropic mantle structure is also present beneath southern and northern California, the Sierra-Nevada, and the Basin and Range Province. This study extends earlier work by Savage and Silver (1993), adding more stations and more data at stations characterized in that paper. We interpret the fast polarizations and delay times as reflecting past and present plate tectonic deformations that the western U.S. has experienced.
We have analyzed SKS and S waveforms for shear wave splitting from 26 stations in the TERRAscope, University of California, Berkeley, and University of Nevada, Reno (UNR) regional broadband networks. We have examined more than 250 events to search for well-isolated SKS and S waveforms with a high signal to noise ratio (SNR). 91 events were found useful for 262 reliable splitting measurements. Events used in this study are listed in Table A1. S waves are only used to complement the SKS measurements when there is not enough azimuthal coverage. Unlike SKS phases, shear wave splitting in the S phases can be due to anisotropy in both source and receiver regions. To avoid source side splitting, deep focus events are used (Kaneshima and Silver, 1992; Fischer and Yang, 1994, Russo and Silver, 1994). We used only 25 S wave measurements of which only 6 were from shallow events (Table A1). We examine all the S results to make sure they are consistent with SKS results at a given station. Further, we observe that the results vary from station to station for the same event. This indicates that the splitting in the S phase is dominated by receiver side anisotropy. On the other hand, the S phases are of special interest because their initial polarization is independent of earthquake station geometry (Savage and Silver, 1993) and therefore, provide sampling of various polarizations which may not be possible with SKS phases, whose polarization is parallel to the earthquake-station back-azimuth.
The six TERRAscope network broadband stations that we used (GSC, ISA, PAS, PFO, SBC, and SVD) are located in southern California within close proximity of the southern SAF. The network description can be found in Kanamori and Hauksson (1991). The data for this network were obtained from the IRIS Data Management Center. The Berkeley network presently consists of 11 broadband stations (YBH, ARC, WDC, MIN, ORV, CMB, BKS, STA, MHC, SAO, and PKD) and covers most of the northern and central California along the SAF and the Sierra-Nevada region (Romanowicz, et al., 1992). The UNR broadband network consists of 10 stations (DNY, BMN, WHR, WCN, KVN, MNA, DSP, TRC, WCK, NEL) (Peppin and Nicks, 1992) and provides coverage for the Basin and Range Province. In addition to these regional network stations, we have included results from two portable stations (AFD and ADW) operated by UNR (Ozalaybey et al., 1991) (Figure 1).
We use the method of Silver and Chan (1991) to retrieve shear wave splitting parameters, assuming that shear waves traverse a single homogeneous anisotropic layer. The method of finding phi and dt is based on a grid search over the possible splitting parameters to reverse the process of generating the split shear waves. This is done by minimizing the component that is orthogonal to the incoming polarization of phip of the shear wave before entering the anisotropic medium. For SKS phases phip is equal to earthquake back-azimuth, therefore, the energy on the observed transverse component is minimized. The error bounds for the estimated splitting parameters are obtained through F-test analysis as described in Silver and Chan (1991).
Silver and Savage (1994) have shown that in the presence of two anisotropic layers, splitting parameters measured under the assumption of a single anisotropic layer will be apparent parameters (phia , dta), and will display azimuthal variations. They have derived analytical equations for the variation of (phia , dta) as a function of phip, the dominant period of incoming waveform, and the splitting parameters of the lower (phi1 , dt1) and upper (phi2, dt2) layers. We use these equations to compute theoretical two-layer variation curves for a comparison with the observed data. To retrieve the anisotropic parameters of two layers, we use the double-layer waveform inversion method developed by Ozalaybey and Savage (1994). The inversion for double layers is based on the fact that when an S phase passes at near vertical incidence through two homogenous, anisotropic layers, the incoming waveform will be split twice. For teleseismic S, the time delay between the split waves are much smaller than the period of S and therefore the split waves will result in a composite waveform instead of four individual arrivals at the receiver (Figure 2). The retrieval of double layer anisotropic parameters can be accomplished by a grid search method (see Ozalaybey and Savage, 1994 ). The resolution and the accuracy of this method are controlled by the range of incoming polarizations sampled by the data and by the SNR. Care must be taken when using waveforms from polarizations that are very close to the fast or slow axis of the lower layer. These cases are null for the lower layer anisotropy and will result in a single layer result indicating the upper layer anisotropy. To avoid this problem a suite of low-noise waveforms representing different polarizations are inverted at a single station. The double-layer splitting parameters are then found by stacking transverse energy misfit spaces associated with different waveforms. The double layer anisotropic parameters obtained by waveform inversion are checked for consistency by comparing the two-layer curves with the variation of observed fa and dta. To statistically quantify this consistency check, we compute the variances of phia and dta from the estimated theoretical two-layer curves and from the single layer parameters (simple mean of phia and dta). We then compare the variances of each model by computing the F-distribution probability function, which represents the confidence level at which the two variances (associated with double and single layer models) are different (Press et al., 1986). The probability function is zero when the two variances are not different and is one when they are different at 100% confidence level. The variances achieved in this manner are not strictly related to standard deviations of the waveform inversion calculation, because the variances are calculated compared to a theoretical formulation that is valid at a single frequency (Silver and Savage, 1994), while the waveform inversion is calculated using all the frequencies received at a station.
Figure 3 shows an example of a splitting measurement at station GSC assuming a single anisotropic layer. The transverse component SKS arrival is clearly seen well above the noise level (Figure 3a). The theoretical arrival times of S phases are calculated from the "iaspei91" earth model (Kennett, 1991). Figure 3a presents the original uncorrected radial and transverse components together with the components corrected for the effect of anisotropy. Transverse component SKS energy is removed after the correction. Figure 3b displays the SKS waveforms in the fast and slow coordinate frame together with the particle motion plot. After the removal of the anisotropy effect, the reconstructed waveform is almost linearly polarized. Table 1 lists the best weighted average single layer splitting pairs found for each station. In the calculation of weighted averages we included measurements having standard deviations in phi of less than 22o because measurements with 2 sigma = 45o is interpreted as failure to detect splitting (Silver and Chan, 1991). The measurements with no detectable splitting are reported as null and are fairly consistent with the estimated fast and slow directions (within ~20o). This is expected because when a shear wave enters an anisotropic medium polarized parallel to either fast or slow direction, there is no corresponding orthogonal component to split. Individual single layer splitting measurements are listed in Table A2. In figure 4, we plot the estimated splitting parameters as a function of phip modulo pi/2 for stations that have more than one non-null measurement to look for any azimuthal dependence that may arise from heterogeneities, and/or multiple anisotropic layers. Double anisotropic layers cause the estimated splitting parameters to vary with pi/2 periodicity as a function of phip (Silver and Savage, 1994). Table 2 lists the best double layer splitting parameters obtained from waveform inversion. Table 3 lists the results of variance comparison of the double and single layer models.
Splitting measurements in this region show scatter and complexity. Measurements at GSC, PFO, PAS, SBC and SVD are in general consistent with each other indicating an average E-W fast direction and a delay time of 1 to 1.5 s (Table 1, Figure 4). However, the number of measurements at stations SBC, SVD, and PFO is not sufficient to exclude possible variations in shear wave splitting as might occur in the cases of inhomogeneous anisotropy or double-anisotropic layers. Our results for station ISA are most scattered and indicate inconsistent measurements for a single or double anisotropic layers. We have only three non-null measurements obtained from one SKS and two S waveforms at this station. As discussed earlier, the measurements obtained from S waveforms may be due to both source and/or receiver side anisotropies. Five SKS measurements resulted in null-directions, and the waveforms from this station contained a high level of noise. Therefore, we presently will not interpret these measurements. A recent study by Liu et al. (1994) presented shear wave splitting results localized in the southern California region including stations we analyze here. Their results agree well with ours and interestingly our results differ most for station ISA. We also point out that their results for modeling double-anisotropic layers do not agree with ours at station PFO.
Station LAC is located close to these stations and has been modeled by a two-layered system consisting of an upper layer with fast direction locally parallel to the strike of SAF and a lower layer with NE-SW fast direction (Savage and Silver, 1993; Ozalaybey and Savage, 1994). The theoretical two-layer curves for this system are plotted in Figure 4. These curves fit fairly well to the variation of fa and dta as a function of fp at LAC. The deviations of fa and dta from the theoretical curves may be due to scattering effects in the data and due to using only a single period (10 s) in calculating the curves. The fit to phia variation has a higher confidence level (0.83) for double-layers than does dta (0.08) (Table 3). Shear wave splitting beneath PFO has previously been analyzed by Helffrich et al. (1994), who suggested that anisotropy was consistent with the two-layer results at nearby LAC. Our data set consists of 6 more events. The double-layer inversion of high SNR waveforms at PFO shows that the variations of splitting parameters are fit by a two-layer model with fault-parallel upper layer fast direction but a small delay time (0.6 s) and a lower layer oriented at N70oE with a delay time of 1.2 s (Figure 4). This model is less well-constrained because we could only use 3 high SNR waveforms for the inversion. The variance comparison also shows that there is a low confidence level in distinguishing between the double and single layer models with the available number of data points (Table 3).
To determine the possible crustal contribution to SKS splitting measurements at station LAC where two anisotropic layers were inferred, we have analyzed Landers earthquake aftershocks. This station was knocked off center by the mainshock and was fixed months later so we found only one event (a depth of 6.47 km and a distance of 8.98 km from LAC) within the shear wave window (i.e. angle of incidence < 35o). Delay time measurement for this event yields 0.046 s, suggesting a splitting value of 0.0042 s/km if the splitting is evenly distributed. Thus, for a 30 km thick crust the delay time would be on the order of 0.13 s, which suggests that the dominant source of anisotropy is in the subcontinental mantle. Usually, anisotropy decreases with increasing depth in the crust due to closing of cracks (assuming stress-aligned crack-induced anisotropy, Crampin and Lovell, 1991), so this is an upper estimate. Recently, Parsons et al. (1994), have measured velocity in lower crustal xenoliths in the Mojave desert at Cima, California. They find little anisotropy (less than 4%) in lower crustal rocks.
The splitting parameters at PAS and GSC are better constrained and show nearly no variation in phi or dt with pi/2 periodicity as a function of phip . Null measurements at these stations are consistent with the single layer model. We have tested these stations against the double layer model found for PFO. The F-test shows that there is a very low probability that the variations of fa and dta could be fit by this model at either station (Table 3). Li et al. (1994) also studied crustal splitting in the Los Angeles Basin. Their measurements include a site (SCS) located approximately 20 km southwest of station PAS . The crustal anisotropy (upper 20 km) is dominated by N-S fast direction and a maximum delay time of 0.12 s. Aster et al. (1990) found splitting delay times due to crustal anisotropy beneath the Anza network in southern California to vary from 0.02 to 0.15 s.
These results suggest that fault-parallel anisotropy from the SAF disappears within short distances to the west beneath southern California (55 km between PAS and SAF). The magnitude of delay times observed for the crust indicates that the contribution of crustal anisotropy to the SKS splitting is very small in this region.
By including data unavailable to Savage and Silver (1993), we find that stations BKS, STA, MHC, and SAO yield splitting parameters that are strongly indicative of a double-anisotropic layer system. Figure 4 shows that phia and dta vary with pi/2 periodicity as a function of phip for these stations. Both single and double-layer inversion of splitting parameters (Silver and Savage, 1994; Ozalaybey and Savage, 1994) indicate an upper layer with fast direction locally parallel to the strike of SAF and a lower layer with E-W fast direction. Splitting pairs for SAO are slightly different for both layers (Table 2). We also plotted theoretical two-layer curves for the splitting pairs obtained from double-layer waveform inversion listed in Table 2 for these stations. These curves fit well to the variation of phia and dta as a function of phip (Figure 4). The variance comparison for these stations also shows that the double-layer model fits the data much better than the single-layer model (Table 3). Seismic anisotropy in the shallow crust of the Loma Prieta segment of the SAF was studied by Zhang and Schwartz (1994), using aftershocks of the 1989 Loma Prieta earthquake. They found fault-parallel anisotropy with an average delay time of only 0.035 s at station SAO. This also supports the conclusion that the splitting in the SKS is too large to be explained by the crust and must originate in the subcontinental mantle. Vinnik et al. (1989) examined SKS splitting at GEOSCOPE station SCZ, near SAO, along the northern SAF, using only three records. Their result yielded an average E-W fast direction and a delay time of 1.3 s. Double or single layer models, however, are probably not distinguishable with their few measurements. At PKD, we have two non-null measurements with near E-W fast direction and a large delay time. Large delay times are often observed when using the single-layer method when there are actually two anisotropic layers of similar magnitude. We have inverted waveforms of PKD for double-layer anisotropy. Although not as well constrained as the other stations, the results show that the same double-layer model could fit the waveform data. We expect that more measurements for station PKD will follow the two-layer variation curve.
For other stations, measurements indicate a single anisotropic layer. Stations located at the foothills of the Sierra-Nevada (CMB, AFD, and ADW) are characterized by nearly E-W fast direction and large delay times (up to 1.8 s). Note, however, that AFD and ADW are single measurements, thus double layers can not be ruled out. At station ORV, which is located 100 km north of station AFD, we have obtained 8 measurements with no detectable splitting. Further north, station MIN (only two measurements) also shows a small delay time (dt =0.78 s) with approximately E-W fast direction. Stations WDC and YBH yield approximately NE-SW fast directions with 1 to 1.5 s delay times, respectively. Station ARC also shows little splitting (dt = 0.55 s). Stations to the north of our study region have recently been analyzed for shear wave splitting by Bostock and Cassidy (1994). They found E-W to NE-SW fast directions over the Juan de Fuca ridge subduction system but no splitting immediately to the northeast of the ridge-fault-trench triple junction.
We must address the question of source of anisotropy before any interpretation. In the above sections, we have ruled out source side anisotropy when using S phases because the S measurements are consistent with the SKS and only a small fraction of our results is dependent on the S wave data. Observed delay times are too large to be explained by the crustal anisotropy in our study region. Further, the magnitude of crustal splitting ( 0.1 < dt < 0.2s), regardless of fast direction, is too small to contaminate SKS measurements. An anisotropic layer with a split time of less than 0.2 s is not within the resolution of long period SKS measurements (Silver and Chan, 1991). The anisotropy may also arise in the lower mantle path traversed by the SKS and S phases. The lower mantle as the source of anisotropy is also ruled out. Observed splitting parameters vary over distance ranges as small as 100 km. For example station ORV has no detectable splitting while stations located at the foothills of the Sierra-Nevada (AFD, ADW, and CMB) show large splitting with fast directions that are consistent with each other (Figure 1). As argued by Silver and Chan (1991), the anisotropy must lie well within the upper mantle, in order to observe such short wave length variations. Based on these results, we focus on the upper mantle as the source of anisotropy. We interpret our measurements as caused by single or double layers of homogeneously, transversely anisotropic material with horizontal symmetry axes due to strain-induced preferred orientation of olivine in the upper mantle beneath our stations in the following sections.
One of the most important results from this study is the disappearance of the fault-parallel fast direction at very short distances from the fault trace. Stations BKS, STA, MHC, SAO, LAC (and possibly PKD, PFO, and SVD) are located close to the SAF and show fault-parallel anisotropy. In northern California, station CMB located 200 km east of the fault shows no fault-parallel anisotropy. In southern California, PAS and SBC are on the Pacific plate about 55 km west of the fault and show E-W fast anisotropy, again not parallel to the SAF. Double layers of significant thickness are ruled out for these stations.
Station LAC located in the Mojave Desert just east of the SAF shows large fault-parallel upper layer anisotropy. In recent years, the importance of right-lateral motion within the Eastern California shear zone or the Mojave domain near LAC has been highlighted. Dokka and Travis (1990) suggest that as much as 29% of crustal deformation from the Pacific-North American motion within the last 6-10 M.y. has taken place in this region. The 1992 Landers, California earthquake served to confirm their point (Unruh et al., 1994). It is somewhat surprising that the strongest evidence we have for a fast orientation parallel to the Pacific-North American plate motion in southern California comes from LAC. This implies that the crustal deformation in the Mojave extends to the mantle. If the lack of fault-parallel anisotropy at stations PFO and SVD is confirmed by more evidence, it might even suggest that the plate boundary in the mantle is displaced eastward of the surface trace of the SAF, as first suggested by Hadley and Kanamori (1977) to explain the lack of displacement observed across the Transverse Ranges high-velocity anomaly.
The fault-parallel anisotropy is expected to result from the finite strain associated with the relative transform plate motion between the North American and Pacific plates. The magnitude of delay times suggests that the upper layer is in the lithosphere and the lower layer resides in the asthenosphere. If we assume 4% anisotropy and an average isotropic shear wave velocity of 4.6 km/s for the entire lithosphere, then the fault-parallel anisotropy extends from surface to 115 km near the northern SAF and to 125 km near the southern SAF. The actual depth extent of this layer is of course dependent on the true value for the percentage of anisotropy. The lateral extent of this layer appears to be less than 55 km on the west side of the surface trace of the SAF in southern California.
Other studies exist for the estimation of the width and depth extent of the shearing zone associated with the SAF based on thermo-mechanical modeling and seismic velocity inversion (Lachenbruch and Sass, 1973; Zandt, 1981; Furlong et al., 1989; Benz et al., 1992; Nicolas, 1993 ). Zoback et al. (1987) propose a model for the SAF in which the near fault-normal compression occurs within a 100 km zone on either side of the fault that yields extremely low shear stresses resolved on the fault plane over a depth equivalent to the thickness of the entire lithosphere. Nicolas (1993) gives estimates for the width of the SAF assuming an olivine dry or wet rheology in the mantle of 10 to 40 km (half width), respectively. Seismic imaging and lithospheric strength modeling of the SAF in northern California (at the latitude of San Francisco) suggest a thin lithosphere (40 to 60 km) and a wide shearing zone (100 km) (Zandt, 1981; Furlong et al., 1989). Seismic imaging through travel-time inversion has poor resolution in the vertical direction, leading to less accurate absolute velocity and thus depth information (Zandt, 1981). Converting delay times to layer thicknesses is also uncertain depending on the true degree of anisotropy and the path length travelled in the anisotropic layer. Thus, the measurements of the width and depth extent for the SAF obtained by different methods may be made to agree with each other within their resolutions.
The fast direction in the lower layer beneath the northern SAF is nearly E-W. An E-W fast layer is also found for most of the stations in southern California, Sierra-Nevada, and Basin and Range. Measurements also show that the E-W fast layer does not extend to the north of station ORV, where no anisotropy is detected (Savage and Silver,1993). We can not directly conclude that this E-W feature is caused by the same phenomenon in each region. N-S compression in southern California across the Transverse Ranges, possibly associated with descending lithosphere (Humphreys and Hager, 1990), could cause E-W fast olivine axes and therefore explain the southern California results. Similarly, E-W oriented Miocene extension in the Basin and Range could explain results at MNA and DNY (Savage et al., 1990). However, the simplest hypothesis is that this E-W fast feature has a common origin. Alternative mechanisms that may cause olivine to align in this direction include past and present regional scale processes; large scale E-W oriented mantle flow in the asthenosphere occurring in the slabless window left behind the Farallon plate beneath the western edge of the continent, the E-W directed shear induced on the asthenosphere beneath the continent by the subduction of the Farallon plate, and presumably E-W frozen in anisotropy in the remnants of the Farallon plate left under the North American plate. Below, we consider all of these mechanisms as the possible source of an E-W alignment and therefore seismic anisotropy.
Humphreys and Dueker (1994a) present evidence of relatively high velocity upper mantle beneath northwesternmost California and the offshore region, the Sierra-Nevada, and the Transverse Ranges between depths of 70 to 200 km. They interpret this high velocity upper mantle as subducted old oceanic lithosphere beneath a marginal domain (Humphreys and Dueker, 1994b). Stations in northern California and the foothills of the Sierra-Nevada, where E-W fast anisotropy was found, are included within the boundaries of this marginal subduction domain (see Figure 7 of Humphreys and Dueker, 1994b). This correlation indicates that the E-W fast direction might be explained by the subducted oceanic plate, presumably the remnants of the Farallon plate, assuming the frozen-in anisotropy is oriented E-W. However, this correlation does not explain the absence of anisotropy at ORV, which is also situated within the marginal domain.
Dickinson and Snyder (1979) suggest that with the onset of the San Andreas the Mendocino Triple Junction moved northward over time leaving a slabless window where the Farallon plate had been. The subduction of Farallon plate has been oriented nearly E-W until the last 6 M.y. (Severinghaus and Atwater, 1989). The slabless window described by Dickinson and Snyder (1979) extended through most of western North America (as far east as the central Basin and Range and as far south as the gulf of California) affecting the entire region of mantle.The slabless window mechanism was used by Savage and Silver (1993) to explain their splitting measurements in the western U. S. They proposed an E-W alignment mechanism generated by the differential motion between the slab to the north and asthenosphere beside it to the south. When first proposed, the differential motion was considered to be a shear deformation. However, at higher temperatures differential motion could be related to flow. Figure 5 shows a schematic of the slabless window. This window was filled by upwelling, hot asthenospheric material. This model predicts a hole beneath the North American lithosphere created by the slab removal. The upwelling asthenospheric material would then enter this hole resulting in a horizontal component of mantle flow directed E-W. This arises because the lithospheric slab to the north acts as a barrier to flow of the asthenosphere within the slab window. When a laminar flow occurs along a barrier, the flow must be parallel to the boundary. Thus if the asthenosphere is flowing in a laminar fashion, the horizontal component along the southern edge of the slab must be parallel to the E-W directed boundary. To the south, the asthenospheric material that has already filled the space vacated by the slab partially cools and accretes to the base of adjacent Pacific and North American lithospheres over time (Figure 5) (Furlong et al., 1989). Thus, flow-induced anisotropy may be partially frozen-in. The existence of frozen-in anisotropy must be controlled by the temperature gradient at depth. The depths of 100 to 150 km may be taken as the limiting temperatures (~900oC) for frozen-in anisotropy to exist in the upper mantle of the cratons (Vinnik et al. 1992). The frozen-in anisotropy could later be modified by the Pacific-North American plate motion strain field or remain consistent with the flow beneath. If the accreted material is affected by this strain field, this may explain the thickness of fault-parallel oriented fast layer near the SAF.
Station ORV is located near the estimated southern edge of the Gorda plate (Severinghaus and Atwater, 1989; Jachens and Griscom, 1983), which is part of the remnant of the Farallon plate. No preferred fast direction is found at ORV (Figure 4). The absence of anisotropy at this location could therefore mark the boundary where E-W asthenospheric flow has not been occurring due to the slab removal. This mechanism could explain the widespread range of E-W fast anisotropy presumably in the asthenosphere without a need for complicated models. However, this model is speculative and must be checked with 3D mantle flow modeling incorporating the known slab migration. Garfunkel et al. (1986) give examples of 2D flow modeling including slab migration and shows how migrating slabs can modify the flow in the mantle. It does not, however, consider the case of slab windows.
Three Basin and Range stations (WCN, KVN, and WCK) and the lower layer at LAC display NE-SW fast axes with average 1.0 s delay time. This orientation is parallel to the absolute plate motion (APM) direction for the North American plate (Gripp and Gordon, 1990). The contribution of APM to shear wave splitting in active regions of tectonics is found to be small in the western U.S., if this contribution is required to be constant throughout the region (Savage and Silver, 1993; Sandvol et al., 1992). Thus, we are reluctant to invoke APM. These stations lie close to the 600 km long, NW-SE trending, right lateral shear zone of Walker Lane Belt (Bell and Slemmons, 1979) and other shear zones (Hardyman and Oldow, 1991 ) (Figure 1) . The Walker Lane shear zone is considered to be a major strike-slip fault zone, along which approximately 48 km of cumulative right-lateral slip offset occurred during the last 22 M.y. (Hardyman et al., 1975). The NE-SW fast orientation is perpendicular to the trend of this zone. This lack of correlation suggests that this shear zone, unlike the SAF, is probably a shallow structure confined more or less to the crust, or else that strain in the mantle has not been sufficient to align the olivine. The fast axis in this region, therefore, may be due to some other local disturbance to a regional process.
The stations located over the Gorda slab mark a change in the fast direction to nearly NE-SW (Figure 4). The Gorda plate is a young slab with very slow subduction rate (2 cm/yr) (Riddihough, 1984). The active internal deformation of the Gorda plate was suggested to explain the change in the direction of basement lineations and magnetic anomalies found in the Gorda Basin ( e.g., Wilson , 1986, 1989; Stoddard, 1987). The state of stress and deformation modeling shows that the deformation of the Gorda plate away from its ridge is taking place along a series of NE-SW directed left-lateral shear zones (Wilson, 1989). Similar studies also conclude that the maximum compressive stress in the overlying North American plate is oriented NE-SW (Jachens and Griscom, 1983; Spence, 1989). The fast directions found at YBH and WDC align quite well with these directions. The correlation between fast directions and plate tectonic deformations, and delay times suggest that subduction in this region causes anisotropy, which must be induced within last 6 M.y. (Severinghaus and Atwater, 1989). The results of gravity modeling, combined with results from other geophysical studies gives a lithospheric thickness of 20 km for the North American plate and a maximum of 32 km for the Gorda plate in this region (Jachens and Griscom, 1983). This suggests a 52 km combined lithospheric thickness assuming both plates are contributing to the splitting. Delay times, on the other hand, imply a 115 to 172 km thick anisotropic layer, extending into the asthenosphere if our estimate of 4% anisotropy and their estimates of lithospheric thickness are true. The anisotropic thickness may therefore, include some of the asthenosphere in this region. Alternatively, the anisotropy may be greater. The shear wave anisotropy of ophiolite samples is about 8%, which is twice the value obtained for subcontinental kimberlite nodules (Mainprice and Silver, 1993). Using this value as more representative of percentage of anisotropy for the Gorda slab region reduces above thicknesses by a factor of two, which are more comparable to the lithospheric thickness.
The results presented in this study lead us the following conclusions. (1) The fast directions near the SAF show clear evidence for fault-parallel anisotropy in the upper 115-125 km. This can be explained by the finite strain associated with the relative plate motion between the North American and Pacific plates. Near the sothern SAF, fault-parallel anisotropy is ruled out 55 km west of the fault, and is neither required nor ruled out directly over the fault. Yet 80 km east of the fault, evidence is strong for SAF-parallel anisotropy beneath station LAC in the Mojave Desert. This suggests a mantle strain with either a narrow lateral extent to the west, or with an eastward displacement. (2) A wide-spread, deeper, E-W oriented fast direction of anisotropy underlies the fault-parallel anisotropic layer. This E-W fast feature is also present beneath the western Basin and Range and the foothills of the Sierra-Nevada. The E-W fast orientation is correlated with tomographically resolved high velocity upper mantle interpreted as fragments of subducted oceanic Farallon plate and may therefore be caused by pre-existing anisotropy in deep seated lithospheric structures. Another model that explains this feature is that of E-W directed asthenospheric flow in the slabless window left behind the Farallon plate. If this model is correct then it predicts that the E-W fast anisotropy is dominated by flow induced by the slab removal. The flow-induced anisotropy may be partially frozen-in. (3) To the north, station ORV shows no anisotropy. The absence of anisotropy at this location could therefore mark the northern boundary of E-W oriented asthenospheric flow. (4) Stations located over the Gorda plate mark a change in fast direction to nearly NE-SW. This direction aligns well with the maximum compressive stress direction in the overlying North American plate and the NE-SW directed internal shearing of the Gorda plate. This correspondence implies that anisotropy is induced by subduction in this region. (5) The lack of evidence for NE-SW fast orientation within the Walker Lane Belt of eastern California and western Nevada suggests that this crustal feature does not extend into the mantle, or that it is not as well developed as that beneath the SAF and beneath the Mojave desert. (6) Shear wave splitting results in our study region suggest anisotropic thicknesses of roughly double that expected for purely lithospheric contributions, if an assumption of 4% shear wave anisotropy is used. Using a higher anisotropy of 8% in this region can bring thicknesses in line with other measures of lithospheric thickness. Alternatively, the anisotropic thickness may include some of the asthenosphere.
This work was supported by National Science Foundation grants EAR-8917161 and EAR-9206473. S. Ozalaybey was also supported by a fellowship from the Turkish Ministry of Education. B. Romanowicz and R. Uhrhammer provided data from the Berkeley network. R. Titus provided data from the IRIS Data Management Center. Keith Nakanishi provided LAC data from Landers earthquake. A. Aburto helped to extract UNR events. J. N. Louie provided the western U. S. elevation map. Discussions with P. G. Silver, K. M. Fischer, and N. I. Christensen helped to clarify ideas. The Pascal Instrument Center and the Carnegie Institute of Washington provided instruments for the portable deployment of stations AFD and ADW. Comments by R. M. Russo and two anonymous reviewers helped improve our presentation of the manuscript.
Aster, R. C., P. M. Shearer, and J. Berger, Quantitative Measurements of Shear Wave Polarizations at the Anza Seismic Network, Southern California: Implications for Shear Wave Splitting and Earthquake Prediction , J. Geophys. Res., 95, 12,449-12,473, 1990.
Bell, E. J., and D. B. Slemmons, Recent Crustal Movements in the Central Sierra-Nevada Walker Lane Region of California-Nevada: Part II, The Pyramid Lake Right-Slip Fault Zone Segment of the Walker Lane, Tectonophysics, 52, 571-583, 1979.
Benz, H. M., G. Zandt, and D. H. Oppenheimer, Lithospheric Structure of Northern California From Teleseismic Images of the Upper Mantle, J. Geophys. Res., 97, 4791-4807, 1992.
Bostock, M. G., and J. F. Cassidy, Variations in shear wave splitting across the Canadian National Seismograph Network, Geophys. Res. Lett., submitted, 1994.
Crampin, S. and J. H. Lovell, A decade of shear-wave splitting in the Earth's crust: what does it mean? what use can we make of it? what should we do next, Geophys. J. Int., 107, 387-407, 1991.
Dickinson, W. R., and W. S. Snyder, Geometry of Subducted Slabs Related to San Andreas Transform, J. Geol., 87, 609-627, 1979.
Dokka, R. K., and C. J. Travis, Role of the Eastern California Shear Zone in Accommodating Pacific-North American Plate Motion, Geophys. Res. Lett, 17, 1323-1326, 1990.
Farra, V., L. P. Vinnik, B. Romanowicz, G. L. Kosarev, and R. Kind, Inversion of teleseismic S particle motion for azimuthal anisotropy in the upper mantle a feasibility study, Geophys. J. R. Astron. Soc., 106, 421-431, 1991.
Fischer, K. M., and X. Yang, Anisotropy in Kuril-Kamchatka subduction zone structure, Geophys. Res. Lett., 19, 10,048-10,052, 1992.
Furlong, K. P., W. D. Hugo, and G. Zandt, Geometry and Evolution of the San Andreas Fault Zone in Northern California, ,J. Geophys. Res., 94, 3100-3110, 1989.
Garfunkel, Z. , C. A. Anderson, and G. Schubert, Mantle Circulation and the Lateral Migration of Subducted Slabs, J. Geophys. Res., 91, 7205-7223, 1994a.
Gripp, A. E., and R. G. Gordon, Current Plate Velocities Relative to the Hotspots Incorporating the NUVEL-1 Global Plate Motion Model, Geophys. Res. Lett., 17, 1109-1112, 1990.
Hadley, D., and H. Kanamori, Seismic structure of the Transverse Ranges, California, Geol. Soc. Am. Bull., 88, 1469-1478, 1977.
Hardyman, R. F., E. B. Ekren, and F. M. Byers, Cenozoic strike-slip, normal and detachment faults in northern part of the Walker Lane, west-central Nevada, Geol. Soc. Am. Abstr., 7, 1100, 1975.
Hardyman, R. F., and J. S. Oldow, Tertiary Tectonic Framework and Cenozoic History of the Central Walker Lane, Nevada, Geology and Ore Deposits of the Great Basin, April 1-5, 1990, Symposium Proceedings, published by Geological Society of Nevada, Reno, 666-678, 1991.
Helffrich G., P. G. Silver, and H. Given, Rapid Shear Wave Splitting Variation in Continental Compressional Tectonic Regions, Geophys. J. Int., in press, 1994.
Humphreys, E. D., and B. H. Hager, A kinematic model for the Late Cenozoic development of southern California crust and upper mantle, J. Geophys. Res., 95, 19,747-19,762, 1990.
Humphreys, E. D., and K. G. Dueker, Western U. S. Upper Mantle Structure, J. Geophys. Res., 99, 9615-9634, 1994a.
Humphreys, E. D., and K. G. Dueker, Physical State of the Western U.S. Upper Mantle, J. Geophys. Res., 99, 9635-9650, 1994b.
Jachens, R. C., and A. Griscom, Three dimensional geometry of the Gorda plate beneath northern California, J. Geophys. Res., 88, 9375-9392, 1983.
Kanamori, H., and E. Hauksson, TERRAscope, IRIS Newsletter, X, 1-3, 1991.
Kaneshima, S., and P. G. Silver, A search for source side mantle anisotropy, Geophys. Res. Lett., 19, 10,048-10,052, 1992.
Kennett, B. L. N. (Editor), IASPEI 1991 Seismological Tables, Research School of Earth Sciences, Australian National University Canberra, Australia, 1991.
Kind, R., G. L. Kosarev, L. I. Makeyeva, L. P. Vinnik, Observations of laterally inhomogeneous anisotropy in the continental lithosphere, Nature, 318, 358-361, 1985.
Lachenbruch, A. H., J. H. Sass, Thermo-Mechanical Aspects of the San Andreas Fault System, Stanford University Press, Palo Alto, 192-205, 1973.
Li, Y. G., T. Teng, and T. L. Henyey, Shear-wave splitting observations in the northern Los Angeles Basin, southern California, Bull. Seismol. Soc. Am., 84, 307-323, 1994.
Liu, H., P. M. Davis, and S. Gao, SKS splitting in southern California, written communication.
Mainprice, D., and P. G. Silver, Interpretation of SKS-waves using samples from the subcontinental lithosphere, Phys. Earth Planet. Inter., 78, 257-280, 1993.
McNamara, D. E., and T. J. Owens, Azimuthal shear wave velocity anisotropy in the Basin and Range Province using Moho Ps converted phases, J. Geophys. Res., 98, 12,003-12,017, 1993.
Nicholas, A., Why Fast Polarization Directions of SKS Seismic Waves are Parallel to Mountain Belts, Phys. Earth Planet. Inter., 78, 377-342, 1993.
Ozalaybey, S., and M. K. Savage, Double-layer anisotropy resolved from S phases, Geophys. J. Int., 117, 653-664, 1994.
Ozalaybey, S., M. K. Savage, and P. G. Silver, Regional variation in teleseismic shear wave splitting in east-central California and southern Nevada, EOS (Trans. Am. Geophys. Union), 72, 328, 1991 (abstract).
Parsons, T., N. L. Christensen, and H. G. Wilshire, Velocities of southern Basin and Range xenoliths: Insights on the nature of lower crustal reflectivity and composition, Geology, in press.
Peppin, W. A., and W. F. Nicks, Real-Time Analog and Digital Data Acquisition through CUSP, Seismol. Res. Lett., 63, 181-189, 1992
Press, W. H., S. A. Teukolsky, W. T. Vetterling, and B. P. Flannnery, Numerical Recipes in Fortran: The Art of Scientific Computing, Cambridge University Press, New York, 1986.
Riddihough, R. P., Recent movements of the Juan de Fuca plate system, J. Geophys. Res., 89, 6980-6994, 1984.
Romanowicz, B., L. Gee, and R. Uhrhammer, Berkeley digital seismic network: A broadband network for northern and central California, IRIS Newsletter, XI, 1-5, 1992.
Russo, R. M., and P. G. Silver, Trench-parallel flow beneath the Nazca Plate from seismic anisotropy, Science., 263, 1105-1111, 1994.
Sandvol, E., J. Ni, S. Ozalaybey, and J. W. Schlue, Shear-wave splitting in the Rio Grande rift, Geophys. Res. Lett., 19, 2337-2340, 1992.
Savage, M. K., and P. G. Silver, and R. P. Meyer, Observations of teleseismic shear-wave splitting in the Basin and Range from portable and permanent stations, Geophys. Res. Lett., 17, 21-24, 1990.
Savage, M. K., and P. G. Silver, Mantle deformation and tectonics: Constraints from seismic anisotropy in western United States, Phys. Earth Planet. Inter., 78, 207-228, 1993.
Severinghaus, J., and T. Atwater, Cenozoic geometry and thermal condition of the subducting slabs beneath western North America. In: B. Wernicke (Editor), Basin and Range Extensional tectonics near the Latitude of Las Vegas, Nevada, Geol. Soc. Am. Mem., 176, 1-22, 1989.
Silver, P. G. and W. W. Chan, Shear-wave splitting and subcontinental mantle deformation, J. Geophys. Res., 96, 16,429-16,454, 1991.
Silver, P. G. and M. K. Savage, The interpretation of shear-wave splitting parameters in the presence of two anisotropic layers, Geophys. J. Int., in press, 1994.
Spence, W., Stress Origins and Earthquake Potentials in Cascadia, J. Geophys. Res., 94, 3076-3088, 1989.
Stoddard, P . R., A kinematic model for the evolution of the Gorda plate, J. Geophys. Res., 92, 11,524-11,532, 1987.
Unruh, J. R., W. R. Lettis, and J. M. Sowers, Kinematic Interpretation of the 1992 Landers Earthquake, Bull. Seismol. Soc. Am., 84, 537-546, 1994.
Vinnik, L. P., V. G. Krishna, R. Kind , P. Bormann, and K. Stammler, Shear wave splitting in the records of the German regional seismic network, Geophys. Res. Lett., 21, 457-460, 1994.
Vinnik, L. P., L. I. Makeyeva, A. Milev, and A. Yu. Usenko, Global patterns of azimuthal anisotropy and deformations in the continental mantle, Geophys. J. Int., 111, 433-447, 1992.
Vinnik, L. P., V. Farra, and B. Romanowicz, Azimuthal anisotropy in the earth from observations of SKS at GESOSCOPE and NARS broad band stations, Bull. Seismol. Soc. Am., 79, 1542-1558, 1989.
Wilson, D. S., Deformation of the so-called Gorda Plate, J. Geophys. Res., 94, 3065-3075, 1989.
Zandt, G., Seismic Images of the deep structure of the San Andreas fault system, central Coast Ranges, California, J. Geophys. Res., 86, 5039-5052, 1981.
Zhang, Z., and S. Y. Schwartz, Seismic anisotropy in the shallow crust of the Loma Prieta segment of the San Andreas fault system, J. Geophys. Res., 99, 9651-9661, 1994.
Zoback, M. D., M. L. Zoback, V. S. Mount, J. Suppe, J. P. Eaton, J. H. Healy, D. Oppenheimer, P. Reasenberg, L. Jones, D. B. Raleigh, I. G. Wong, O. Scotti, and C. Wentworth, New evidence on the state of stress of the San Andreas fault system, Science, 238, 1105-1111, 1987.
Table 1. Weighted mean single layer splitting parameters. Starred stations have measurements based on 3 or fewer records. Station locations are also indicated. Stations are listed in a north to south ordered fashion as they appear in Figure 1.
Table 2. Double layer splitting parameters obtained from waveform inversion. ( phi2, dt2) and (phi1, dt1) correspond to splitting pairs of the upper and lower layers, respectively. Starred stations have measurements based on 3 waveforms. Waveform inversion was computed at stations BKS, MHC, SAO, STA, LAC, PFO, and BMN. Results at stations PKD and SVD are from inversion of nearby stations that had measurements consistent with these stations.
Table 3. Variance comparison of single and double layer model fits. The variances are computed for fast direction, phi, and delay times, dt, separately. Subscripts d and s are used to indicate variances for double and single layer models, respectively. Probf and Probdt are the calculated confidence levels for distinguishing the double and single layer models. n is the degrees of freedom used in evaluating F-distribution probability function. Variances for double layer models were calculated compared to the theoretical fa and dta curves expected for 10 s waves with lower and upper layers determined from the waveform inversion results of Table 2. GSC and PAS were calculated compared to the two-layer model derived for station PFO.
Table A1. Events used to measure splitting parameters in this study.
Table A2. Individual single layer splitting measurements at each station. phi is measured clockwise from north. Error bars are one standard deviations. dt=0.00 indicates null-measurements. Back-azimuth of the event, phase type, and event numbers from Table A1 are indicated.
Figure 1. Shear wave splitting results (Tables 1 and 2) from 26 stations plotted on western U. S. elevation map. Bars by station names denote mean phi from the best measurements; length of bars is proportional to dt as in the lower left of the figure. Red bars indicate fast direction in the upper layer where resolved from the waveform data. Dashed bars are used to indicate unconstrained measurements based on 3 or fewer records. Note (Table 3) that two-later solutions for stations PFO and SVD above the souhern San Andreas fault are not strongly preffered over a single-layer. At station ORV, a cross sign is used to indicate absence of splitting. Approximate present boundary of the Gorda plate is shown from Jachens and Griscom (1983) and Severinghaus and Atwater (1989). Queries indicate regions where knowledge of the boundaries is poor. Central Walker Lane Belt (WLB) (Hardyman and Oldow, 1991), main trace of the San Andreas and Garlock Faults (GF) are also shown. Absolute plate motion (APM) direction for the North American Plate (Gripp and Gordon, 1990), Pacific and North American relative plate motion are indicated by thick black arrows.
Figure 2. Schematic diagram showing a doubly split SKS phase through the upper mantle. (phi1, dt1) and (phi2, dt2) correspond to the splitting parameters of the asthenosphere and lithosphere, respectively. The incoming waveform is split twice, once by each layer, resulting in a composite waveform of four individual arrivals at the receiver. Arrows denote fast direction in each layer.
Figure 3. Example of splitting measurement from an SKS phase recorded at station GSC, for an event occurring on day 337, 1991 at 10:33:39.9, with depth 561 km, mb=6.0, D = 86.50, and back-azimuth=234.10. (a) Seismograms in radial, transverse, and vertical directions are indicated before and after the correction for anisotropy. Predicted arrival times of S phases, calculated from the IASPEI 91 earth model. Vertical solid bars denote measurement time window. Note that in the corrected component transverse energy is nearly zero. (b) Top two traces are the superposition of fast (phi) (solid) and slow (phi + pi/2) (dotted) components, uncorrected for delay (left) and corrected (right). Bottom is the corresponding particle motion plots showing nearly linear particle motion after the removal of the anisotropy effect.
Figure 4. High-quality measurements of phi and dt as a function of phip modulo pi/2 for all the stations used in this study. Solid circles represent measurements from SKS or SKKS while open circles represent S phases. Measurements shown with no error bars indicate null-directions. Stations PFO, LAC, BKS, MHC, SAO, STA, and BMN have been modeled by double-anisotropic layers; theoretical two-layer variation curves are plotted for these stations. These curves agree well with the observed variation of measurements. Two-layer curve of PFO is also plotted for GSC and PAS. The fit is poor for these stations.
Figure 5. Schematic block diagram showing the hypothesized horizontal flow of the asthenospheric material within the slabless window. The solid thick arrows show the flow of material along the E-W directed Gorda slab boundary where the most recent opening has taken place. Thick dashed arrows are used to indicate the flow of asthenospheric material that already filled the space vacated by the Farallon plate. This material partially accretes to the base of Pacific and North American plates by cooling over time (gray shaded region). Relative motions between the plates are shown by gray thick arrows.