Anisotropy beneath California: shear wave splitting measurements using a dense broadband array

We have determined the shear wave splitting parameters for a dense network of broad-band stations in the western United States using high-quality SKS and SKKS waveforms, with particularly high spatial resolution in the southern California region covered by the TriNet seismic network. The alignment of most fast polarization directions can be explained by plate-tectonic, extensional and compressional events. We find that the overall pattern of fast directions agrees well with the P-n anisotropy model by Hearn that images the uppermost mantle. Furthermore, the measured fast directions are generally orthogonal to the maximum horizontal compressive stress directions as determined from shallow crustal stress indicators (World Stress Map). This suggests that the pattern of anisotropy in the western US is predominantly uniform throughout the crust and upper mantle and that a 100-150 km thick layer (as estimated from the SKS delay time, assuming 4 per cent anisotropy) of anisotropic material has experienced coherent strain conditions and has undergone a similar deformation history. A more detailed investigation reveals small-scale lateral variations in anisotropy that are manifested by minor differences in splitting parameters between closely located stations as well as between SKS and SKKS for the same station-event pairs. We also identify a contrast in splitting parameters between central (the greater Bay area) and southern California. In central California, our measurements show evidence for variation of splitting parameters with backazimuth, while in southern California the pattern of measurements can be fit adequately with a single-layer anisotropy model. This contrast dominates any consistent effect of the San Andreas Fault (SAF). We can fit the variation of splitting parameters as a function of polarization azimuth for some stations in the vicinity of the SAF better with a two-layer anisotropy model than a single layer model, with one thin layer having a fast direction parallel to the SAF. However, many alternative models, which could incorporate dipping axes of anisotropy, lateral variation of anisotropy or a more continuous variation of fast direction with depth, would be able to produce a similar fit.