Shear wave splitting intensity tomography: theory, application, and coupling with geodynamical models

The interpretation of shear wave splitting observations in complex regions where the geometry of anisotropy is highly heterogeneous is extremely difficult. This is particularly true for subduction zone regions, where the processes that control the three-dimensional distribution of anisotropy are not well understood. The inversion of splitting measurements for anisotropic stucture is not often attempted due to the limitations imposed by sparse data, but carefully constructed splitting data sets from dense broadband arrays can potentially illuminate regions of the upper mantle and facilitate a tomographic approach. We have developed a method for 2.5-D splitting intensity tomography that incorporates constraints from numerical models of geodynamical processes. We apply this method to study upper mantle anisotropy beneath southwestern Japan, using a previously published set of local and teleseismic shear wave splitting measurements. We use the splitting intensity metric introduced by Chevrot (2000) and calculate finite-frequency sensitivity kernels for the orientation and strength of anisotropy using the Born approximation. We calculate sensitivity kernels in a heterogeneous starting model that is taken from a numerical modeling study of anisotropy development in the mantle wedge due to a subducting lithospheric plate, and use these kernels in the inversion. The calculation of such kernels in realistic 2D or 3D heterogeneity is essential for true multi-scale, finite frequency splitting intensity tomography. Finally, we use the results of the inversion to refine our flow models with control parameters that best fit the splitting observations, and further improve the models by using perturbed flow models as new starting models for the inversion.