[1] We analyze moment tensor solutions from deep subduction zone earthquakes to determine global slab deformation patterns. Inferred strain rates are compared to predicted deformation patterns from fluid models to help constrain the first-order radial and lateral viscosity structure of the Earth. While all slabs that reach the lower mantle are compressed at their tip, intermediate depth patterns are more complex. We compute 3-D spherical flow with various slab rheologies and compare the angular misfit between the compressive eigenvectors of the resultant stress field and global centroid moment tensor (gCMT) solutions. We find that upper mantle slab viscosities of ∼10-100 and lower mantle viscosities of ∼30-100 times the upper mantle produce the best match to gCMTs. A 0.1 viscosity reduction in the asthenosphere seems preferred. Slab geometry and lower mantle viscosity exert significant control on deformation. Inclusion of the phase changes at 410 km and 660 km increases extensional deformation at intermediate depth and compressional deformation at the lower mantle, improving the match to gCMTs for strong slabs. Our conclusions are fairly insensitive to surface boundary conditions. However, models which include net rotations of the surface with respect to the lower mantle produce compression at intermediate depths for west directed slabs and extension for east directed slabs. Without allowing for regional variations, these models yield the best match to gCMTs. While significant deviations between model and seismicity remain, our results show that seismicity provides an underutilized constraint for slab dynamics.

[1] Upper mantle heterogeneity beneath the Alboran Sea (western Mediterranean) as inferred from seismology has been associated with a range of subduction and lithospheric delamination scenarios. However, better constraints on the deep dynamics of the region are needed to determine the cause and consequence of complex surface tectonics. Here, we use an improved set of shear wave splitting observations and a suite of mantle flow models to test a range of suggested structures. We find that the observed seismic anisotropy is best reproduced by mantle flow models that include a continuous, deeply extending slab beneath the Alboran which elongates along the Iberian margin from Granada to Gibraltar and curves southward toward the High Atlas. Other models with detached slabs, slabs with spatial gaps, or drip-like features produce results inconsistent with the splitting observations. SW-directed shear flow, when combined with sublithospheric deflection in response to a dense sinker, generates NNW-splitting orientations most similar to the patterns observed along Gibraltar. Slab viscosities of 250 times that of the upper mantle are preferred because they provide a balance between the poloidal flow induced by any sinker and toroidal flow induced by stiff slabs. The best match to anisotropy across the Atlas is a model with a stiff continental keel in northwestern Africa which deflects flow northward. Our results show that quantitative predictions of seismic anisotropy are useful in distinguishing the spatial and depth extent of regional density structures which may otherwise be ambiguous.

.[1] We assess the co-seismic deformation inferred from earthquake moment tensor solutions for subducting slabs at depths greater than 50 km globally. We rotate each moment tensor into a local slab reference frame, then sum tensors within 50 km depth bins to approximate long term deformation characteristics. This builds upon previous analyses by using the up-to-date global Centroid Moment Tensor catalog, incorporating a more complete slab geometry, and focusing on the 3-D aspects of slab deformation. Results show a general consistency with Isacks and Molnar (1969), who found that most slabs can be divided into intermediate-extensional, intermediate-extensional-deep-compressional, and intermediate to deep-compressional categories. Exceptions to these three categories can be related to slab bending in the top 100 km, plate convergence that is oblique to the trench normal direction, and regions of higher focal mechanism heterogeneity. The regions of higher focal mechanism heterogeneity appear where there are along-strike changes in slab geometry and/or evidence of double-seismic zones. We find that the sense of deformation in the intermediate strain axis direction is opposite to that of the down-dip direction, in agreement with Kuge and Kawakatsu (1993). By quantitative comparison to numerical models of global mantle flow, we show that these observations are consistent with deformation of viscous slabs responding to their own negative buoyancy and an upper to lower mantle viscosity increase.