Anisotropy and XKS-splitting from geodynamic models of double subduction: Testing the limits of interpretation

We utilize three-dimensional geodynamic models to predict XKS-splitting in double subduction scenarios characterized by two outward-dipping slabs. These models are highly applicable in various realistic settings, such as the central Mediterranean. Our primary focus is on the analysis of XKS-splitting, a key geophysical observable used for inferring seismic anisotropy and mantle flow patterns.Our models simulate the concurrent subduction of two identical oceanic plates separated by a continental plate. The variation in the strength of the separating plate causes a transition from a retreating to a stationary trench. The models offer detailed insights into the temporal evolution of mantle flow patterns, particularly the amount of trench-parallel flow induced by this specific type of subduction.In the subsequent step, we employ the well-known D-Rex model to estimate Crystallographic Preferred Orientation (CPO) development in response to plastic deformation resulting from mantle flow. Based on the D-Rex model results, which incorporate the full elastic tensor of a deformed multiphase polycrystalline mantle aggregate, we derive synthetic apparent splitting parameters and splitting intensities at virtual receivers placed at the surface using multiple-layer anisotropic waveform modeling. To identify regions with pronounced depth-dependent variations of anisotropic properties, particularly the fast polarization directions, we define a complex anisotropy factor dependent on the apparent splitting parameters and splitting intensities.Finally, using the apparent splitting parameters, we conduct two-layer model inversions at selected locations characterized by a large complex anisotropy factor. The two-layer model provides apparent splitting parameters as a result of analytical waveform modeling for two anisotropic layers. We observe that while several models can effectively explain the apparent splitting parameters, only a subset can accurately reproduce the depth-dependent anisotropic properties. Our findings unequivocally demonstrate that a classical XKS-splitting analysis can effectively identify areas characterized by complex anisotropy and provide accurate approximations of the depth-dependent variations of anisotropic properties within these regions. However, caution is warranted when interpreting results obtained through inversion based on a two-layer analysis.