[S-SS07] Seismic wave propagation: Theory and Application

[SSS07-12] Vector wavelet propagation through random elastic media: Hybrid Monte Carlo simulation using the spectrum division (Simulation)

*Kentaro Emoto¹, Haruo Sato² (1.Kyushu University, 2.Tohoku University)

Keywords:seismic waves, body waves, random media, scattering, finite difference, Monte Carlo simulation

We investigate seismic wave propagation in random heterogeneous media. The narrow-angle scattering due to the large-scale component of the heterogeneity leads to the envelope broadening of a seismic wavelet. On the other hand, wide angle scattering by the small-scale component generates coda waves. We proposed the method to simultaneously model both effects by dividing the power spectrum of the random heterogeneity. In the previous talk, we present the theoretical basis of the spectrum division method. In this talk, we explain the practical implementation of the method based on the Monte Carlo (MC) simulation and validate the proposed method by comparing it with the envelope derived from the 3D finite difference (FD) simulation of the seismic wave propagation. The conventional Born approximation is not appropriate for the cases when the wavelength is smaller than the characteristic scale of the heterogeneity, or the fluctuation is strong. The MC simulation is impractical in these cases due to the strong scattering coefficient. In the spectrum division method, the MC simulation becomes stable by reducing the strong forward scattering power and compensating it with the narrow-angle scattering based on the Eikonal approximation. Compared with the FD simulation, the envelopes derived by MC simulation with the spectrum division method well model the whole envelopes from the onset until the coda, even when the conventional Born approximation is not theoretically appropriate. In the spectrum division method, the travel time fluctuation due to the large-scale heterogeneity predicted by the Eikonal approximation is also stochastically implemented in the same framework of the MC simulation. There is a slight discrepancy between the MC simulation with thespectrum division method and the FD simulation. We may need to include the near and intermediate terms of the source radiation in the spectrum division method. It is worth noting that the correlation distance, at which the auto-correaltion function falls off to 1/e of the value at r=0, is an important scale to decide to use the spectrum division method in the von Karman type random heterogeneous medium.