Crustal Anisotropy from the Birefringence of P-to-S Converted Waves: Bias Associated with P-Wave Anisotropy

Abstract

Many researchers have used the birefringence of P‑to‑S converted waves from the Moho discontinuity to constrain the anisotropy of Earth’s crust. However, this practice ignores the substantial influence that anisotropy has on the initial amplitude of the converted wave, which adds to the splitting acquired during its propagation from Moho to the seismometer. We find that large variations in Ps birefringence estimates with back-azimuth occur theoretically in the presence of P‑wave anisotropy, which normally accompanies S‑wave anisotropy. The variations are largest for crustal anisotropy with a tilted axis of symmetry, a geometry that is often neglected in birefringence interpretations, but is commonly found in Earth’s crust. We simulated globally-distributed P‑coda datasets for 36 distinct 4‑layer crustal models with combinations of elliptical shear anisotropy or compressional anisotropy, and also incorporated the higher-order anisotropic Backus parameter C. We tested both horizontal and tilted symmetry-axis geometries and tested the birefringence tradeoff associated with Ps converted phases at the top and bottom of a thin high‑ or low‑velocity basal layer. We computed composite receiver functions (RFs) with harmonic regression over back azimuth, using multipletaper correlation with moveout corrections for the epicentral distances of 471 events, to simulate a realistic data set. We estimate Ps birefringence from the radial and transverse RFs, a strategy that is similar to previous studies. We find that Ps splitting can be a useful indicator of bulk crustal anisotropy only under restricted circumstance, either in media with no compressional anisotropy, or if the symmetry axis is horizontal throughout. In other, more-realistic cases, the inferred fast polarization of Ps birefringence estimated from synthetic RFs tends either to drift with back-azimuth, form weak penalty-function minima, or return splitting times that depend on the thickness of an anisotropic layer, rather than the birefringence accumulated within it.