Probabilistic inversion of shear-wave splitting for D'' anisotropy

Abstract:

The lowermost mantle - D'' - remains an enigmatic region of the Earth. From seismic tomography we know that D'' is heterogenous and rich in structures, such as the LLSVPs, ULVZs, and fast anomalies attributed to subducted slab material. Our current understanding of the origin, composition and evolution of these features is limited. Resolving azimuthal seismic anisotropy offers new insights into features across D''. Seismic anisotropy, however, can result from a range of plausible mineralogical mechanisms and dynamic scenarios. Discerning between these requires constraint of more general styles of anisotropy.

To properly image such azimuthal anisotropy requires combining observations from multiple shear-wave phases at difference azimuths and incidence angles. This is difficult to treat robustly in conventional shear-wave splitting studies and forward modelling exercises. An additional challenge in studying D'' is solving a two-layer seismic anisotropy problem imposed by the significant seismic anisotropy in the upper mantle, whilst preserving directional information. Current approaches, such as differential shear-wave splitting or splitting intensity tomography, offer partial solutions, but cannot fully resolve D'' anisotropy.

Here, we present a method for the direct probabilistic inversion of shear-wave data for seismic anisotropy. We divide D'' into a set of geographical domains and search for the most likely orientation and strength of an elastic tensor representing each domain using a Metropolis-Hastings Markov Chain Monte Carlo sampler. This allows us to easily combine data from multiple shear-wave phases and incorporate results from mineral physics to invert for, for example, flow geometry.

We demonstrate our approach on ScS, SKS and SKKS waveform data for a region of D'' beneath the Eastern Pacific. Our results show clear evidence for azimuthal seismic anisotropy which we evaluate with different candidate mechanisms for D'' anisotropy including the lattice-preferred orientation of post-perovskite and the shape-preferred orientation of melt inclusions.