Complex deformation beneath Sulawesi from local and teleseismic shear-wave splitting observations

Located at the junction the Eurasian, Indo-Australian, and Philippine Sea Plates, the Sulawesi region is a complex, active subduction environment. Constraints on the mantle flow-field, mantle composition, and state of stress in the area are lacking, but these parameters are nevertheless expected to be reflected in measurable anisotropic fabrics. We present new shear-wave splitting results for both local and teleseismic data for northern Sulawesi, where two subduction zones converge at ca. 200 km depth, the Celebes Sea Plate dipping south and the Molucca Sea Plate dipping northwest beneath Sulawesi. Results for the northwest dipping Molucca Sea subduction show little variation in delay time with depth for local earthquakes, indicating that the anisotropy is confined to the lithosphere, whereas the mantle wedge is probably isotropic. Delay times of SKS splitting are significantly higher (an average of ~1.36 s for SKS splitting as opposed to a local-S average of ~0.56 s), suggesting that there is additional anisotropy in or around the slab not sampled by local events. Fast direction orientations vary considerably, especially for local-S splitting, exhibiting nearly the full 180° range of directions and thereby indicating that the anisotropic fabrics are complex. However, to a large extent, the local-S and SKS fast directions are approximately between 10° and 80°. The likely cause of the complexity is due to two subducting slabs converging at depth as well as the ongoing clockwise rotation of the north arm of Sulawesi. We forward model the shear-wave splitting results with a subduction zone model based on regional tomographic images in order to place constraints on the location and orientation of anisotropy: We calculate ray paths for each shear-wave used in our study for a standard 1-D velocity model. We then calculate the travel-time for each of these rays through the tomographic model and compare with that predicted from the 1-D model. The discrepancy between the two travel-times gives a proxy as to how much time each ray may have spent within the subducting slab, hence constraining possible locations of the anisotropic fabrics. Without exception, the local events arrive earlier than predicted by the reference Earth model, highlighting the fact that all rays are likely to have spent some time in the subducting slab. However, we see no apparent correlation of arrival time perturbation with the splitting results, indicating that, at least for local events, the ray path within the slab is unlikely to harbour the source of the measured anisotropy, making a shallow lithospheric fabric the most likely explanation. This implies a second, deeper (deep slab interior or subslab) region of anisotropy to explain the larger SKS results.