Seismic wave propagation in complex crust‐upper mantle media using 2‐D finite‐difference synthetics

Abstract

A description is given of the numerical FD scheme used to solve the elasti wave equation, including a few remarks on the source functions used. Our FD method has been used for computing synthetic seismograms for 2‐D crust/upper mantle models of size 150 × 400 km2, with options for free‐surface topography. The strategy was to introduce successively more complex lithosphere models for generating the synthetics; the reference model was laterally homogeneous lithosphere. The interface scattering was visualized through displays of free‐surface synthetic waveforms and snapshots for models with a corrugated Moho only and free surface topography only. Near the free surface the latter seems to dominate, in the form of P‐to‐Rg and S‐to‐Rg conversions. Lithosphere randomizations were introduced through von Kármán functions of order n= 0.3, with rms velocity fluctuations of 3–4 per cent and correlation distances (horizontal and vertical) at 2.5 or 10 km. In case of a medium with only sub‐Moho heterogeneities, those with horizontal anisotropy (ax= 10 km; az= 2.5 km) produced relatively strong Pn and Sn phases. The respective codas were dominated as in most of our experiments by P‐to‐S and S‐to‐S scattering wavelets excluding Rg scattering at a free surface with topography. For a medium with crustal heterogeneities, the distortions of the P and S wave trains with distance were clearly demonstrated. For full‐scale heterogeneous lithosphere models, characteristic features of the synthetics were quantitatively similar to observational records of local events. Dominant attributes were a pronounced P coda consisting mainly of P‐ and Rg‐scattered wavelets, and a relatively strong S coda consisting mainly of P‐to‐S and S‐to‐S scattered wavelets. The P and S waveforms are severely distorted pointing at the futility of reliably picking many secondary arrivals in local event recordings. Most of the scattering wavelets are confined to the crustal waveguide and to surface waves, since coda excitations for sensors at a depth of 100 km were weak and, moreover, consisted mainly of S wavelets. This implies that a strong teleseismic P coda does not reflect scattering within the crust in the source region but, rather, a complex source. Observational results from analysis of NORESS and ARCESS local event recordings are also presented. Clearly the lithosphere is not isotropically inhomogeneous. The essence of our 2‐D FD synthetic seismogram experiments is that a simple lithosphere model, being moderately heterogeneous, gives rise to complex seismograms which are grossly similar to the observational recordings. In contrast, complex models derived from profiling surveys (but lacking the fine‐scale random variations) give simple,‘ray tracing’like synthetics, not necessarily similar to the observed records. Copyright © 1994, Wiley Blackwell. All rights reserved