Aftershocks of the 18 October 1989 Loma Prieta, California, earthquake are located using S-P arrival-time measurements from stations of the PASSCAL aftershock deployment. We demonstrate the effectiveness of using S-P arrival-time data in locating earthquakes recorded by a sparse three-component network. Events are located using the program QUAKE3D (Nelson and Vidale, 1990) with both 2-D and 3-D velocity models that have been developed independently for this region. The dense coverage of the area around the Loma Prieta rupture zone by instruments of the California Network (CALNET) has allowed the U.S. Geological Survey (USGS) to find P-wave earthquake locations for both velocity models, which we compare with our solutions. We also perform synthetic calculations to estimate realistic location errors resulting from uncertainties in both the 3-D velocity structure and the timing of arrivals. These calculations provide a comparison of location accuracies obtained using S-P arrival times, S and P arrival times, and P times alone. We estimate average absolute errors in epicentral location and in depth for the Loma Prieta aftershocks to be just under 2 km and 1 km, respectively, using S-P phase data and the sparse PASSCAL instrument coverage. The synthetic tests show that these errors are much smaller than those predicted using P-wave data alone and are nearly the same as those predicted using S- and P-phase data separately. This suggests that future aftershock recording deployments with sparse networks of three-component data can retrieve accurate event locations even if absolute timing is problematic.
We find moderate differences between our locations and those determined by the USGS from a larger network of stations; however, common characteristics in both seismicity patterns are apparent. Neither set of locations yields earthquake patterns that can be easily interpreted in terms of simple faulting geometries. The absence of a simple pattern in both sets of earthquake locations indicates that this complexity is not the result of earthquake mislocation but is a genuine feature of the seismicity. A deep southwesterly dipping plane and a near-vertical fault extending from the surface to at least 7-km depth beneath the surface trace of the San Andreas Fault are imaged by both sets of earthquake locations. Although earthquake locations indicate the existence of many more fault segments, the complexity of this region requires that a definitive picture of the faulting geometry will have to await improvement in our knowledge of the P- and S-wave velocity structures.