Subduction dynamics and mantle anisotropy: modeling and clustering of olivine textures

The mantle near Earth's subduction zones experiences significant deformation, forming anisotropic rock textures. These textures can be detected using seismic methods and simulated in geodynamic models. This study employs time-series clustering to examine tracers in subduction models, identifying regions with similar deformation histories, olivine crystallographic-preferred orientation (CPO) development, and CPO-induced anisotropic viscosity. We compare the evolution of olivine textures predicted by various numerical methods (e.g. D-Rex, MDM, and MDM+AV) for both retreating and stationary trench subduction settings.

Our modeling shows notable variations in olivine texture around the slab and as a function of subduction dynamics. These variations, which are illuminated by the clustering analysis, show that texture, seismic, and viscous anisotropy can vary greatly within the mantle wedge, sub-slab, and subducting plate regions of the upper mantle. In the retreating-trench model, the strongest textures are observed in the mid-depth mantle wedge region and beneath the slab at the 660 km transition zone. Trench-normal olivine a-axis orientations are predominant in the center of subduction zones, while toroidal flow around slab edges produces a mix of trench-normal, trench-parallel, and oblique fast seismic directions. On the other hand, in the stationary-trench model, the trench-normal signal in front of the slab is weaker while there are stronger trench-normal signals behind the slab at shallow depths between 100 and 300 km. At the edge of the slab, weak toroidal flow produces trench-oblique orientations while trench-parallel and trench-normal orientations are missing. In general, the retreating trench model exhibits stronger textures and anisotropy due to increased deformation from trench motion.

These results provide valuable insights into seismic anisotropy in subduction zones and underscore the importance of considering texture heterogeneity when interpreting geodynamic models and seismic data. The use of time-series clustering algorithms highlights the intricate pattern of evolution and the relationship between deformation history, CPO, and CPO-induced viscous anisotropy occurring within subduction zones.