On the trade-off between seismic anisotropy and heterogeneity: Numerical simulations and application to Northeast Japan

P wave tomography has been recently used to study 3-D azimuthal and radial anisotropy of subduction zones and continental regions. However, the fundamental issue about the trade-off between the isotropic and anisotropic structures is still unclear. In this study, we investigate this issue systematically with comprehensive synthetic tests. Our results indicate that good ray coverage in the azimuth (for azimuthal anisotropy) and incidence (for radial anisotropy) is required for determining reliable anisotropic models. The isotropic and anisotropic structures are strongly coupled, and smearing effects are significant when the rays used in the inversion are limited in a small range of azimuth or incidence. We therefore plot ray azimuth and ray incidence ellipses at every grid nodes and propose to use the normalized length of the short axis (i.e., the ratio of the short-axis and long-axis lengths) for estimating the ray coverage quantitatively. Applying our novel approach to a large number of high-quality arrival time data of local shallow- and intermediate-depth earthquakes, we obtained new tomographic images of 3-D P wave azimuthal and radial anisotropy in Northeast Japan. Both the azimuthal and radial anisotropy results are determined reliably for the shallow parts of the study region, whereas the smearing effects are significant in the deeper part of the mantle wedge and the subducting slab. Our results show dominant trench-normal and vertical-fast anisotropy in the mantle wedge while trench-parallel and horizontal-fast anisotropy in the subducting slab, which indicates different dynamics in different domains of the subduction zone.