Modelling shear wave splitting observations from Wellington, New Zealand

Abstract

SUMMARY Frequency-dependent anisotropy was previously observed at the permanent broad-band station SNZO, South Karori, Wellington, New Zealand. This has important implications for the interpretation of measurements in other subduction zones and hence for our understanding of mantle flow. This motivated us to make further splitting measurements using events recorded since the previous study and to develop a new modelling technique. Thus, in this study we have made 67 high-quality shear wave splitting measurements using events recorded at the SNZO station spanning a 10-yr period. This station is the only one operating in New Zealand for longer than 2 yr. Using a combination of teleseismic SKS and S phases and regional ScS phases provides good azimuthal coverage, allowing us to undertake detailed modelling. The splitting measurements indicate that in addition to the frequency dependence observed previously at this station, there are also variations with propagation and initial polarization directions. The fast polarization directions range between 2 ◦ and 103 ◦ , and the delay times range between 0.75 s and 3.05 s. These ranges are much larger than observed previously at SNZO or elsewhere in New Zealand. Because of the observed frequency dependence we measure the dominant frequency of the phase used to make the splitting measurement, and take this into account in the modelling. We fit the fast polarization directions fairly well with a two-layer anisotropic model with horizontal axes of symmetry. However, such a model does not fit the delay times or explain the frequency dependence. We have developed a new inversion method which allows for an inclined axis of symmetry in each of the two layers. However, applying this method to SNZO does not significantly improve the fit over a two-layer model with horizontal symmetry axes. We are therefore unable to explain the frequency dependence or large variation in delay time values with multiple horizontal layers of anisotropy, even allowing for dipping axes of symmetry. For the upper layer of anisotropy, the modelled fast polarization directions lie in the range 30 ◦ to 110 ◦ , subparallel to the major geological features as well as the relative direction of plate motion. We suggest that the upper layer of anisotropy consists of crust and subducting slab, and that the lower layer consists of subslab mantle. For the lower layer of anisotropy, the modelled fast polarization directions lie in the range −50 ◦ to 30 ◦ , which are intermediate