Azimuthal seismic anisotropy in the Earth’s upper mantle and the thickness of tectonic plates

S U M M A R Y Azimuthal seismic anisotropy, the dependence of seismicwave speeds on propagation azimuth, is largely due to fabrics within the Earth’s crust and mantle, produced by deformation. It thus provides constraints on the distribution and evolution of deformation within the upper mantle. Here, we present a new global, azimuthally anisotropic model of the crust, upper mantle and transition zone. Two versions of this new model are computed: the rough SL2016svAr and the smooth SL2016svA. Both are constrained by a very large data set of waveform fits (∼750 000 vertical component seismogram fits). Automated, multimode waveform inversion was used to extract structural information from surface and S wave forms in broad period ranges (dominantly from 11 to 450 s, with the best global sampling in the 20–350 s range), yielding resolving power from the crust down to the transition zone. In our global tomographic inversion, regularization of anisotropy is implemented to more uniformly recover the amplitude and orientation of anisotropy, including near the poles. Our massive waveform data set, with complementary large global networks and high-density regional array data, produces improved resolution of global azimuthal anisotropy patterns. We show that regional scale variations, related to regional lithospheric deformation and mantle flow, can now be resolved by the global models, in particular in densely sampled regions. For oceanic regions, we compare quantitatively the directions of past and present plate motions and the fast-propagation orientations of anisotropy. By doing so, we infer the depth of the boundary between the rigid, high-viscosity lithosphere (preserving ancient, frozen fabric) and the rheologically weak asthenosphere (characterized by fabric developed recently). The average depth of thus inferred rheological lithosphere-asthenosphere boundary (LAB) beneath the world’s oceans is ∼115 km. The LAB depth displays a clear dependence on the age of the oceanic lithosphere, closely matching the 1200 ◦C half-space cooling isotherm for all oceanic ages. In continental regions, azimuthal anisotropy is characterized by smaller-scale 3-D variations. Quantitative comparisons of the tomographic models with global SKS splitting measurements confirm the basic agreement of the two types of anisotropy analysis; they also offer a new insight into the average rheological thickness of continental lithosphere. In spite of significant recent improvements in the resolution of upper-mantle anisotropic structure, correlations between the anisotropic components of current global tomographic models remain much lower than between the isotropic ones. Our comparisons of the current models show which features are resolved consistently by different models, and therefore provide a means to estimate the robustness of anisotropic patterns and amplitudes. Significantly lower correlations are observed at depths greater than ∼300 km, compared to those shallower, which suggests that global azimuthal anisotropy models are yet to reach consensus on the nature of anisotropy in the transition zone.