UPPER MANTLE ANISOTROPY FROM LONG-PERIOD P POLARIZATION

Abstract

We introduce a method to infer upper mantle azimuthal anisotropy from the polarization, i.e., the direction of particle motion, of teleseismic long-period P onsets. The horizontal polarization of the initial P particle motion can deviate by >10° from the great circle azimuth from station to source despite a high degree of linearity of motion. Recent global isotropic three-dimensional mantle models predict effects that are an order of magnitude smaller than our observations. Stations within regional distances of each other show consistent azimuthal deviation patterns, while the deviations seem to be independent of source depth and near-source structure. We demonstrate that despite this receiver-side spatial coherence, our polarization data cannot be fit by a large-scale joint inversion for whole mantle structure. However, they can be reproduced by azimuthal anisotropy in the upper mantle and crust. Modeling with an anisotropic reflectivity code provides bounds on the magnitude and depth range of the anisotropy manifested in our data. Our method senses anisotropy within one wavelength (250 km) under the receiver. We compare our inferred fast directions of anisotropy to those obtained from Pn travel times and SKS splitting. The results of the comparison are consistent with azimuthal anisotropy situated in the uppermost mantle, with SKS results deviating from Pn and Ppol in some regions with probable additional deeper anisotropy. Generally, our fast directions are consistent with anisotropic alignment due to lithospheric deformation in tectonically active regions and to absolute plate motion in shield areas. Our data provide valuable additional constraints in regions where discrepancies between results from different methods exist since the effect we observe is local rather than cumulative as in the case of travel time anisotropy and shear wave splitting. Additionally, our measurements allow us to identify stations with incorrectly oriented horizontal components.