<p>Seismic anisotropy across the Hikurangi subduction zone measured from shear-wave splitting exhibits strong lateral changes over distances of about 250 km. Teleseismic S-phases show trench-parallel fast polarisations with increasing delay times across the forearc and arc region. In the arc region, delay times reach up to 4.5 s, one of the largest delay times measured in the world. Such large delay times suggest strong anisotropy or long travel paths through the anisotropic regions. Delay times decrease systematically in the backarc region. In contrast, local S-phases exhibit a distinct change from trench-parallel fast orientations in the forearc to rench-perpendicular in the backarc, with average delay times of 0.35 s. In the far backarc, no apparent anisotropy is observed for teleseismic S-phases. The three different anisotropic regions across the subduction zone are interpreted by distinct anisotropic domains at depth: 1) In the forearc region, the observed "average" anisotropy (about 4%) is attributed to trench-parallel mantle flow below the slab with possible contributions fromanisotropy in the slab. 2) In the arc region, high (up to 10%) frequency dependent anisotropy in the mantle wedge, ascribed to melt, together with the sub-slab anisotropy add up to cause the observed high delay times. 3) In the far backarc region, the mantle wedge dynamic ends. The apparent isotropy must be caused by different dynamics, e.g. vertical mantle flow or small-scale convection, possibly induced by convective removal of thickened lithosphere. The proposed hypothesis is tested using anisotropicwave propagation in two-dimensional finite difference models. Large-scale models of the subduction zone (hundreds of kilometres) incorporating the proposed anisotropic domains of the initial interpretation result in synthetic shear-wave splittingmeasurements that closely resemble all large-scale features of real data observations across the central North Island. The preferred model constrains the high (10%) anisotropy to the mantle wedge down to about 100 kmunder the CVR, bound to the west by an isotropic region under the western North Island; the slab is isotropic and the subslab region has average (3.5%) anisotropy, down to 300 km. This model succeeds in reproducing the constant splitting parameters in the forearc region, the strong lateral changes across the CVR and the apparent isotropy in the far backarc region, as well as the backazimuthal variations. The influence of melt on seismic anisotropy is examined with different small-scale (tens of kilometres) analytical modelling approaches calculating anisotropy due to melt occurring in inclusions, cracks or bands. Conclusions are kept conservative with the intention not to over-interpret the data due to model complexities. The models show that seismic anisotropy strongly depends on the scale of inclusions and wavelengths. Frequency dependent anisotropy for local and teleseismic shear-waves, e.g. for frequency ranges of 0.01-1Hz can be observed for aligned inclusions on the order of tens of meters. To test the proposed frequency dependence in the recorded data, two different approaches are introduced. Delay times exhibit a general trend of -3 s/Hz. A more detailed analysis is difficult due to the restricted frequency content of the data. Future studies with intermediate frequency waves (such as regional S-phases) are needed to further investigate the cause of the discrepancy between local and teleseismic shear-wave splitting. An additional preliminary study of travel time residuals identifies a characteristic pattern across central North Island. Interpretation highlights the method as a valuable extension of the shear-wave splitting study and suggests a more detailed examination to be conducted in future.</p>
Seismic anisotropy and heterogeneity in the Alaska subduction zone
