Intermediate-period Rayleigh and Love waves propagating across Tibet indicate marked radial anisotropy within the middle-to-lower crust, consistent with a thinning of the middle crust by about 30%. The anisotropy is largest in the western part of the plateau, where moment tensors of earthquakes indicate active crustal thinning. The preferred orientation of mica crystals resulting from the crustal thinning can account for the observed anisotropy. The middle-to-lower crust of Tibet appears to have thinned more than the upper crust, consistent with deformation of a mechanically weak layer that flows as if confined to a channel.

Teleseismic P waves are followed by a series of scattered waves, particularly P-to-S converted phases, that form a coda. The sequence of scattered waves on the horizontal components can be represented by the receiver function (RF) for the station and may vary with the approach angle and azimuth of the incoming P wave. We have developed a frequency-domain RF inversion algorithm using multiple-taper correlation (MTC) estimates, instead of spectral division, using the pre-event noise spectrum for frequency-dependent damping. The multitaper spectrum estimates are leakage resistant, so low-amplitude portions of the P-wave spectrum can contribute usefully to the RF estimate. The coherence between vertical and horizontal components can be used to obtain a frequency-dependent uncertainty for the RF. We compare the MTC method with two popular methods for RF estimation, timedomain deconvolution (TDD), and spectral division (SPD), both with damping to avoid numerical instabilities. Deconvolution operators are often biased toward the frequencies where signal is strongest. Spectral-division schemes with constant waterlevel damping can suffer from the same problem in the presence of strong signalgenerated noise. Estimates of uncertainty are scarce for TDD and SPD, which impedes developing a weighted average of RF estimates from multiple events. Multiple-taper correlation RFs are more resistant to signal-generated noise in test cases, though a "coherent" scattering effect, like a strong near-surface organ-pipe resonance in soft sediments, will overprint the Ps conversions from deeper interfaces. The MTC RF analysis confirms the broad features of an earlier RF study for the Urals foredeep by Levin and Park (1997a) using station ARU of the Global Seismographic Network (GSN), but adds considerable detail, resolving P-to-S converted energy up to f ‫ס‬ 4.0 Hz.

S U M M A R YP-SH conversion is commonly observed in teleseismic P waves, and is often attributed to dipping interfaces beneath the receiver. Our modelling suggests an alternative explanation in terms of flat-layered anisotropy. We use reflectivity techniques to compute three-component synthetic seismograms in a 1 -D anisotropic layered medium. For each layer of the medium, we prescribe values of seismic velocities and hexagonally symmetric anisotropy about a common symmetry axis of arbitrary orientation. A compressional wave in an anisotropic velocity structure suffers conversion to both SVand SH-polarized shear waves, unless the axis of symmetry is everywhere vertical or the wave travels parallel to all symmetry axes. The P-SV conversion forms the basis of the widely used 'receiver function' technique. The P-SH conversion occurs at interfaces where one or both layers are anisotropic. A tilted axis of symmetry and a dipping interface in isotropic media produce similar amplitudes of both direct ( P ) and converted (Ps) phases, leaving the backazimuth variation of the P-Ps delay as the main discriminant. Seismic anisotropy with a tilted symmetry axis leads to complex synthetic seismograms in velocity models composed of just a few flat homogeneous layers. It is possible therefore to model observations of P coda with prominent transverse components with relatively simple 1 -D velocity structures. Successful retrieval of salient model characteristics appears possible using multiple realizations of a geneticalgorithm (GA) inversion of P coda from several backazimuths. Using GA inversion, we determine that six P coda recorded at station ARU in central Russia are consistent with models that possess strong (> 10 per cent) anisotropy in the top 5 km and between 30 and 43 km depth. The symmetry axes are tilted, and appear aligned with the seismic anisotropy orientation in the mantle under ARU suggested by SKS splitting.

[1] Isotropic and anisotropic seismic structures across the Northern Apennines (Italy) subduction zone are imaged using a new method for the analysis of teleseismic receiver functions (RFs). More than 13,000 P-wave coda of teleseismic records from the [2003][2004][2005][2006][2007] Retreating-Trench, Extension, and Accretion Tectonics (RETREAT) experiment are used to provide new insights into a peculiar subduction zone between two continental plates that is considered a focal point of Mediterranean evolution. A new methodology for the analysis of receiver functions is developed, which combines both migration and harmonic decomposition of the receiver function data set. The migration technique follows a classical "Common Conversion Point" scheme and helps to focus on a crucial depth range (20-70 km) where the mantle wedge develops. Harmonic decomposition of a receiver function data set is a novel and less explored approach to the analysis of P-to-S converted phases. The separation of the back-azimuth harmonics is achieved through a numerical regression of the stacked radial and transverse receiver functions from which we obtain independent constraints on both isotropic and anisotropic seismic structures. The application of our method to the RETREAT data set succeeds both in confirming previous knowledge about seismic structure in the area and in highlighting new structures beneath the Northern Apennines chain, where previous studies failed to clearly retrieve the geometry of the subducted interfaces. We present our results in closely spaced profiles across and along the Northern Apennines chain to highlight the convergence of the Tyrrhenian and the Adriatic microplates which differ in their crustal structure where the Adriatic microplate subducts beneath Tuscany and the Tyrrhenian sea. A signature of the dipping Adriatic Moho is clearly observed beneath the Tyrrhenian Moho in a large portion of the forearc region. In the area where the two Mohos overlap, our new analysis reveals the presence of an anisotropic body above the subducted Moho. There is a strong Ps converted phase with anisotropic characteristics from the top of the Adriatic plate to a depth of at least 80 km. Because the Ps conversion occurs much deeper than similar Ps phases in Cascadia and Japan, dehydration of oceanic crust seems unlikely as a causative factor. Rather, the existence of this body trapped between the two interfaces supports the hypothesis of lower crustal delamination in a postsubduction tectonic setting.

Abstract.Observations of shear wave splitting in the northeastern U.S. Appalachians and in the foredeep of the Urals vary significantly with the back azimuth and incidence angle of the incoming phase. These variations suggest two or more layers within the upper mantle with different anisotropic properties. Synthetic seismograms for simple multilayered anisotropic structures show that shear wave splitting parameters tend to vary substantially with the direction of approach.