Representing anisotropic subduction zones with isotropic velocity models: A characterization of the problem and some steps on a possible path forward

Despite the widely known fact that mantle flow in and around subduction zones produces the development of considerable seismic anisotropy, most P-wave tomography efforts still rely on the assumption of isotropy. In this study, we explore the potential effects of erroneous assumption on tomographic images and explore an alternative approach. We conduct a series of synthetic tomography tests based on a geodynamic simulation of subduction and rollback. The simulation results provide a self-consistent distribution of isotropic (thermal) anomalies and seismic anisotropy which we use to calculate synthetic delay times for a number of realistic and hypothetical event distributions. We find that anisotropy-induced artifacts are abundant and significant for teleseismic, local and mixed event distributions. The occurrence of artifacts is not reduced, and indeed can be exacerbated, by increasing richness in ray-path azimuths and incidence angles. The artifacts that we observe are, in all cases, important enough to significantly impact the interpretation of the images. We test an approach based on prescribing the anisotropy field as an a priori constraint and find that even coarse approximations to the true anisotropy field produce useful results. Using approximate anisotropy, fields can result in reduced RMS misfit to the travel time delays and reduced abundance and severity of imaging artifacts. We propose that the use of anisotropy fields derived from geodynamic modeling and constrained by seismic observables may constitute a viable alternative to isotropic tomography that does not require the inversion for anisotropy parameters in each node of the model.