Artificial age-independent seismic anisotropy, slab thickening and shallowing due to limited resolving power of (an)isotropic tomography

Seismic anisotropy is key to constrain mantle flow, but it is challenging to image and interpret it. Existing large-scale tomography models of seismic anisotropy typically show large discrepancies, which can lead to completely distinct geodynamical interpretations. To better quantify the robustness of anisotropy tomography, we create a 2-D ridge-to-slab geodynamic model and compute the associated fabrics. Using the resulting 21 elastic constants, we compute seismic full waveforms, which are inverted for isotropic and radially anisotropic structure. We test the effects of different data coverage and levels of regularization on the resulting images and on their geodynamical interpretation. Within the context of our specific imposed conditions and source-receiver configuration, the retrieved isotropic images exhibit substantial artificial slab thickening and loss of the slab's high-velocity signature below similar to 100 km depth. Our results also show that the first-order features of radial anisotropy are well retrieved despite strong azimuthal anisotropy (up to 2.7 per cent) in the input model. On the other hand, regularization and data coverage strongly control the detailed characteristics of the retrieved anisotropy, notably the depth-age dependency of anisotropy, leading to an artificial flat depth-age trend shown in existing anisotropy tomography models. Greater data coverage and additional complementary data types are needed to improve the resolution of (an)isotropic tomography models.