Modelling surface waves in anisotropic structures II: Examples

Surface waves in several anisotropic geological settings are studied by numerical simulation. Asymptotic point-source synthetic seismograms are obtained using formulas from a companion paper, the group velocities and mode-excitation factors being evaluated via variational integrals. Some alternative waveform approximations and ray-path effects are explored and the possible significances of peculiarities of anisotropy, such as wavefront folding, are at least partially assessed. Model one is a 5-layer oceanic environment with +/-5.6% aligned vertical-crack S-wave anisotropy in the upper crust, The 2-3 s waveforms often have highest group velocities at 45 degrees to the cracks and there are very strong transverse components from a vertical source. Maslov and ray/WKBJ waveforms agree well, even though for some modes and frequencies azimuthal group-velocity variations imply the latter is suspect. The Maslov-integral end-point signals mean that it too requires care and the so-called quasi-WKBJ approximation is a viable alternative. Periods approaching 1 s and/or high crack densities lead to very complicated group-velocity geometry indeed. The second model has +/-2% upper-mantle mineral-alignment anisotropy, with its isotropic component constrained to agree with model ak135. The 20-30 s waveforms from a transversely-isotropic source force have fundamental 'Rayleigh' and 'Love' packets showing clear azimuthal phase anomalies but surprisingly standard polarizations and amplitudes, Only higher modes show such variable amplitudes and phases that they might be identified as likely due to anisotropy. Around 10-s period the higher-mode family has an interesting banded group-velocity structure and as the period is lowered further, two modes separate out and the band tightens in what may be a characteristic feature. Two candidate models of mantle-plume anisotropy show distinct azimuthal dependencies of mode velocities. Point-source surface-wave rays tracked in one plume model display shadows and focusing and show how initially folded wavefronts can unfold in the lateral variations. The upper-mantle models are numerically taxing. R/T coefficients and mode slownesses can be found using double-precision arithmetic, but depth profiles and variational-integrals sometimes require greater accuracy. We find that while displacements near the stack top and bottom may be obtained quite easily, as frequency increases it becomes increasingly difficult to compute those at mid-depth. We find no reason to believe that the Kennett invariant-embedding scheme has lesser accuracy than others. (C) 1997 Elsevier Science B.V.