On the relationship between azimuthal anisotropy from shear wave splitting and surface wave tomography

Seismic anisotropy provides essential constraints on mantle dynamics and continental evolution. One particular question concerns the depth distribution and coherence of azimuthal anisotropy, which is key for understanding force transmission between the lithosphere and asthenosphere. Here, we reevaluate the degree of coherence between the predicted shear wave splitting derived from tomographic models of azimuthal anisotropy and that from actual observations of splitting. Significant differences between the two types of models have been reported, and such discrepancies may be due to differences in averaging properties or due to approximations used in previous comparisons. We find that elaborate, full waveform methods to estimate splitting from tomography yield generally similar results to the more common, simplified approaches. This validates previous comparisons and structural inversions. However, full waveform methods may be required for regional studies, and they allow exploiting the back‐azimuthal variations in splitting that are expected for depth‐variable anisotropy. Applying our analysis to a global set of SKS splitting measurements and two recent surface wave models of upper‐mantle azimuthal anisotropy, we show that the measures of anisotropy inferred from the two types of data are in substantial agreement. Provided that the splitting data is spatially averaged (so as to bring it to the scale of long‐wavelength tomographic models and reduce spatial aliasing), observed and tomography‐predicted delay times are significantly correlated, and global angular misfits between predicted and actual splits are relatively low. Regional anisotropy complexity notwithstanding, our findings imply that splitting and tomography yield a consistent signal that can be used for geodynamic interpretation.