We study seismic anisotropy beneath the Santa Cruz Mountains using digital recordings of S waves from 60 aftershocks of the 1989 Loma Prieta earthquake (Mω=7.0) recorded at 18 local stations. Three-component particle motion covariance matrix decomposition combined with a cross-correlation method is used to estimate the polarization direction of the fast shear wave arrival and the delay time between the split shear waves. Results from this procedure are compared with those obtained through visual inspection of horizontal particle motions. Inclusion of the vertical component in the particle motion analysis allows the recognition and elimination of S to P converted phases from our study. Uncertainties in the polarization of the fast shear wave and in the delay time between arrival of the fast and slow waves are estimated to be about 20¿ and 10 ms, respectively. About half of the S wave records examined show clear evidence of shear wave splitting. Two persistent polarization directions of fast shear wave arrivals are observed regardless of the focal mechanism and hypocenter location of the individual events; 52% of the polarizations are parallel to the strike of the San Andreas fault (NW-SE), and 24% are parallel to the probable direction of local maximum horizontal compressive stress (NE-SW).

The fault-parallel fast polarization direction dominates at 10 stations and may result from mineral or fracture alignment caused by shearing along the plate boundary. The fast polarization direction perpendicular to the local axes of minimum horizontal compressive stress dominates at one station farthest from the San Andreas fault and is consistent with the hypothesis of extensive dilatancy anisotropy (EDA), where parallel alignment of fluid-filled fractures produces the anisotropy. The predominance of fault-parallel fast shear wave polarizations indicates that the use of shear wave splitting results to deduce the orientation of local axes of principal stress may not always yield accurate results. The splitting time between fast and slow shear waves varies from 10 to 125 ms and shows no systematic relationship with either hypocentral distance or focal depth. This suggests that the seismic anisotropy is no deeper than 2 km, the depth of the shallowest earthquakes analyzed. For an anisotropic zone 2 km thick, the average velocity anisotropy is about 5% in our study region. ¿ American Geophysical Union 1994