Forward modeling of the development of seismic anisotropy in the upper mantle

Development of seismic anisotropy in response to upper mantle flow is approached through an integrated numerical model. This model allows to predict the splitting parameters for a shear wave propagating across an upper mantle which deformed in response to a given geodynamic process. It consists of (1) thermo-mechanical modeling of the finite strain field, (2) modeling olivine lattice-preferred orientation (LPO) generated by this strain held, (3) calculation of the 3-D elastic properties associated with this LPO, and (4) estimation of the shear-wave splitting parameters: the time lag between the fast and slow split shear wave arrivals (delta t) and the polarization azimuth of the fast wave (phi). Modeled olivine LPO are constrained relative to LPO measured in naturally and experimentally deformed peridotites. Comparison of predicted shear-wave splitting parameters with seismological data allows us to quantify the possible contribution of the modeled upper mantle flow to the measured splitting values and, hence, to constrain the interpretation of shear-wave splitting data in terms of upper mantle flow. We use this forward model to investigate the seismic anisotropy generated in ocean basins by a velocity gradient between the plate and the deep mantle. Fast-shear wave polarizations calculated assuming a constant plate motion are in good agreement with both the SKS polarization and the fast propagation direction for P and Rayleigh waves observed in the Pacific and Indian oceans, suggesting that, away from mid-ocean ridges, seismic anisotropy in oceanic basins primarily results from asthenospheric deformation by resistive drag beneath the plate. Delay times are, however, overestimated. This may be ascribed to a stronger strain localization in nature or to partial erosion of the anisotropic layer by hotspots. Indeed, hotspot activity may explain the short length scale variations of delta t in the southern Pacific. Finally, two-layer models that simulate a change in Pacific plate motion as suggested by the bend in the Hawaii-Emperor chain fail to reproduce the observed shear-wave splitting. This is consistent with previous suggestions that the Emperor chain track may not faithfully record the Pacific plate absolute motion before 43 Ma. (C) 1998 Elsevier Science B.V. All rights reserved.