1:45 PM - 2:00 PM
[S03-1-02] Statistical characteristics of scattered waves in random media based on 3D finite difference simulations
Short-period seismograms show complex waveforms due to the scattering by the small-scale heterogeneities widely distributed in the solid Earth. The scattering causes the broadening of seismogram envelope, the attenuation of the peak amplitude with travel distance increasing, and the excitation of long lasting coda waves. In this study, we conduct the 3D finite difference (FD) simulation of the scalar wavelet propagation in the random heterogeneous media by using the Earth Simulator, a vector supercomputer managed by JAMSTEC. We analyze the characteristics of the scattered waves and compare them with the direct envelope synthesis methods, such as the radiative transfer equation with the Born approximation and the improved Markov approximation based on the parabolic approximation (Sato, 2016).
We confirmed that Born envelopes adequately fit with FD envelopes in the whole lapse time range, but improved Markov envelopes fail to model even at the peak of the envelopes for the case of a short correlation distance. On the other hand, improved Markov envelopes well model FD envelopes around the peak for the case of a large correlation distance. In this case, although the Born envelopes can't model the peaks of FD envelopes, late coda part of the envelopes are well reproduced. We also investigate the distribution of the squared amplitudes of FD traces. We find that the distribution varies as lapse time increases from the log normal distribution at onset to the exponential distribution at the coda. For the case of a short correlation distance, squared amplitudes obey the exponential distribution soon after the onset. This means that the random scattered waves are dominant from the early lapse time, since wide angle scattered waves are dominant. According to the FK analysis of the FD traces, squared amplitudes obey the exponential distribution before the energy fluxes of the scatted waves become isotropic.
We confirmed that Born envelopes adequately fit with FD envelopes in the whole lapse time range, but improved Markov envelopes fail to model even at the peak of the envelopes for the case of a short correlation distance. On the other hand, improved Markov envelopes well model FD envelopes around the peak for the case of a large correlation distance. In this case, although the Born envelopes can't model the peaks of FD envelopes, late coda part of the envelopes are well reproduced. We also investigate the distribution of the squared amplitudes of FD traces. We find that the distribution varies as lapse time increases from the log normal distribution at onset to the exponential distribution at the coda. For the case of a short correlation distance, squared amplitudes obey the exponential distribution soon after the onset. This means that the random scattered waves are dominant from the early lapse time, since wide angle scattered waves are dominant. According to the FK analysis of the FD traces, squared amplitudes obey the exponential distribution before the energy fluxes of the scatted waves become isotropic.