Shear-Wave Splitting and Seismic Anisotropy in the Wellington Region, New Zealand

Abstract

The phenomenon of shear-wave splitting is investigated using the shear-waves from local earthquakes recorded on the Wellington Peninsula, New Zealand. Three separate deployments of three-component digital seismographs resulted in the recording of perhaps the best data set for the study of shear-wave splitting currently available. Clear evidence of shear-wave splitting is demonstrated, and the most likely cause of the phenomenon is shown to be seismic anisotropy in the Earth's crust. The results of modelling the observed polarizations indicate that the Wellington Peninsula has a complex anisotropic structure. Both manual and automated means are used to derive shear-wave splitting parameters from the recorded arrivals. The shear-wave splitting parameters measured are the polarization of the first arriving shear-wave, and the delay between the two shear-wave arrivals. No delays are measured unless shear-wave splitting is positively identified: two arrivals with similar pulse shapes being required. A comparison between the shear-wave polarizations measured using the manual and automated techniques shows that the automated technique gives 30% less scatter than the manual technique at a similar level of estimated measurement error. For this reason the results derived using the automated technique are used for all subsequent analysis. A significant number (= 37%) of the earthquakes recorded within the shear-wave window show clear evidence of shear-wave splitting: identifiable fast and slow shear-wave arrivals with similar pulse shapes. Consistent polarization alignments on individual stations are also observed, even when poor signal-to-noise or scattering means that no slow shear-wave arrival can be identified. Correcting for the observed shear-wave splitting improves the fit between the measured shear-wave polarizations, and those calculated assuming a double-couple focal mechanism. The cause of the observed shear-wave splitting is therefore most likely to be seismic anisotropy. However, the measured shear-wave delays show a large degree of scatter, and comparisons between stations only 900 m apart give inconsistent results. Thus, either the method used to assign errors to the measured delays is inadequate, or the natural processes involved produce a large amount of scatter. The shear-wave splitting is used to map the complex anisotropic structure of the Wellington Peninsula. Three anisotropic regions are identified on the peninsula, two of which show orthogonal horizontal symmetry axes. The north-westem part of the Wellington Peninsula (Region 1), west of the Ohariu fault and north of a line through the Karori valley is seismically anisotropic with a symmetry axis perpendicular to the local geological structure (NE-SW). The southern region (Region 2) between the Wellington and Ohariu faults has the symmetry axis rotated by 90˚, and the region east of the Wellington fault (Region 3) has a symmetry axis intermediate between the other two regions. Arrivals from some azimuths on two stations are complicated by the closeness of the fracture zone of the Wellington Fault. All three regions exhibit the properties of hexagonal symmetry, as judged by the observed shear-wave polarization alignments. The existence of the two regions with orthogonal symmetry axes allows an estimate of the depth extent of the anisotropy to be made; this is not possible for arbitrary differences in the symmetry axes because then the measured polarizations rely only on the symmetry of the material underlying the recording station. Modelling using synthetic seismograms and a block structure for the Wellington Peninsula, confirms the interpretation that the observed shear-wave splitting is caused by seismic anisotropy and allows the depth extent of the two regions with orthogonal symmetry axes to be estimated. The polarization alignments in Region 1 (= 50˚) are coincident with the NE-SW strike of the active faults in the Wellington region, while the adjoining region (Region 2) has a shear-wave polarization alignment normal to this (= 140˚). These regions show approximately 6% velocity anisotropy extending to a depth of approximately 4 km. The polarization alignment in Region 3 (= 90˚) is close to the maximum horizontal compressive stress direction in the Wellington Region. Converted phases are used to confine the volume causing the observed shear-wave splitting to above the plate interface, and there is weak evidence for perhaps 2% shear-wave velocity throughout the overlying Australian plate from the measured delays. The most likely cause of the observed seismic anisotropy is aligned cracks and microcracks in the Earth's crust. Both crack-induced anisotropy and periodic thin layer anisotropy (PTL) can be used to model the observed shear-wave splitting. Both mechanisms produce an effectively anisotropic medium with hexagonal symmetry, and a similar pattern of shear-wave delays. However, the lack of correlation between the measured bedding azimuths and the observed shear-wave polarization alignments suggests that it is unlikely that the observed anisotropy is caused by PTL. On the other hand, the large changes in the observed polarization alignments do not favour the preferred mechanism of crack alignment which suggested that cracks and microcracks can be aligned by the current regional stress field; a process which produces extensive-dilatancy anisotropy (EDA). One prediction of EDA is that the shear-wave splitting parameters should vary as the stress field varies. However, comparisons of the shear-waves of similar earthquakes recorded in this study with time separations of up to 14 months show no sign of temporal change. There were no large earthquakes near the study area so this does not preclude such changes. The preferred explanation of the observed seismic anisotropy relies on the existence of orthogonal joint sets on the Wellington Peninsula. In some regions one joint set dominates, while in others the orthogonal joint set is most important. As the measured polarizations and delays will depend on an average over the ray path between the source and the seismograph station, this could also explain the large degree of scatter, particularly in the measured delays.