Evidence for Anisotropic Crust Within the Tibetan Plateau From Teleseismic Receiver Functions

A close examination of teleseismic receiver functions can reveal lateral heterogeneity or anisotropy in the earth's crust beneath a seismic recording station. The 1991-1992 Sino-American PASSCAL experiment spanning the length of the Tibetan Plateau, from Lhasa to Golmud, yielded a large, high-quality data set of broadband teleseisms. Receiver functions calculated from this data consist of radial and transverse component traces that vary depending upon the back azimuth between a recording station and an earthquake. Transverse component receiver functions have amplitudes comparable to those of the radial component for many back azimuths, and several arrivals show distinctive polarity reversal patterns that are a function of back azimuth. The presence of both significant transverse energy and radial component arrivals that show systematic variation as a function of back azimuth suggests that seismic energy has been rotated out of the P-SV plane during P to S conversion. This P to SH conversion can occur as a result of lateral heterogeneity, such as dipping isotropic layers, or the presence of anisotropy in one or more subsurface layers. Unlike shear wave splitting studies, the examination of P to S conversions from within the crust unambiguously constrains the causal structure to be above the Moho. Forward modeling shows that Tibetan Plateau receiver functions for several stations can be well matched with synthetics produced for several flat layers over a halfspace, with alternating anisotropic and isotropic layers and anisotropy in as much as one third to one half of the crust. Anisotropy is assumed to have hexagonal symmetry, with a single fast axis direction and slow axis directions in the plane perpendicular to the fast axis (sometimes referred to as "melon" anisotropy). The orientation of the symmetry axis can be varied for each anisotropic layer, but the simplest case of a constant axis strike and tilt in all anisotropic layers for a given model yields synthetics that provide a good match to observed data in many cases. The amount of anisotropy is expressed as a percent deviation of the fast axis velocity from the slow axis velocity; observed transverse component amplitudes can be mimicked with models having values of 10-15 percent, which are reasonable based on experimental studies of anisotropic rocks and common rock-forming minerals. Though the forward modeling discussed here is non-unique, even the simplistic nature of the anisotropy in these models fits more aspects of the data than any isotropic model. In addition, dipping layer models do not seem to adequately explain the observed high amplitudes of transverse receiver functions without unrealistically large velocity contrasts between steeply dipping layers. While reasonable models for many stations share the aforementioned anisotropy characteristics, the stations are too far apart to correlate individual layers, and velocities and thicknesses of layers in models for different stations vary slightly. In addition, stations in the southern portion of the Plateau seem to require more complex models, with more layers, than stations in the north. Despite some structural variations among stations, these basic observations provide strong evidence for anisotropic crust in the Tibetan Plateau. This anisotropy could be a result of past or present tectonic processes involved in the development of the Plateau, and further constraints on its orientation and spatial distribution could provide insight into deformational mechanisms currently or formerly active within the Plateau.