SH-SV Polarization Anisotropy:Interpretation of Experimentally Measured Love and Rayleigh Wave Phase Velocities

Abstract:

It is sometimes not possible to find a single isotropic structure whose computed phase velocities fit both the experimental, fundamental-mode Love and Rayleigh wave data, for earth models that are perfectly elastic and are composed of thick, low contrast layers. Usually, velocity anisotropy of the body waves is applied to the earth models to fit the data. A few early studies used thin, high contrast layers in perfectly-elastic isotropic models to obtain approximate fit to the experimental data; here, we improve and expand this successful isotropic modelling by generalizing to realistic, anelastic layers, and by also requiring a fit to the fundamental-mode Love and Rayleigh wave amplitude-attenuation data.

We treat the Love and Rayleigh wave data from the central United States, where this Love-Rayleigh "discrepancy" was discovered by McEvilly. Using only the experimental phase-velocity data, with the insertion of a thin, high contrast LVZ in each of the granitic, basaltic-grabbroic, and olivine regions, we find a continuum of isotropic models that give successful fits to the experimental data. Then by adding experimental amplitude-attenuation to the data, we attempt to reduce this huge volume of isotropic solutions: with the three thin LVZs, we successfully restricted the solutions by simultaneously fitting the experimental data for both Love and Rayleigh wave, phase-velocity and amplitude-attenuation dispersions.

However, in the solution the body-wave velocities and Q values of these thin layers are improbably low, and these single-layer LVZs can only be considered effective representations; the true, physical situation requires the replacement of any one of these single-layer LVZs by a vertical distribution of N layers, each having the same thickness as the original thin layer. A simple scaling of the single-layer, seismic velocities and Qs then provides completely reasonable values for these parameters in the N-layer representation (which yields the same successful fit to the experimental data).