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

[SSS07-P02] High-resolution imaging of sedimentary structures by autocorrelation function from ground motion records

*Tomotsugu Watanabe¹, Hiroshi Takenaka¹ (1.Okayama University )

Keywords:seismic motion, strong motion, autocorrelation function

We develop a new method to estimate sedimentary structures with a high resolution by autocorrelation analysis based on seismic interferometry. Seismic interferometry is widely used for elucidating real structures of Earth’s interior. Recently, application of autocorrelation to ground motion records has been studied to detect the boundary between the bedrock and sedimentary layers. The trace of autocorrelation function may be equivalent to the zero-offset velocity profile in seismic exploration, where the events on the trace can be recognized as reflected waves from the interface of subsurface structures. Since seismic records include both of source and propagation effects, we need to apply both whitening and filtering to its autocorrelation function to emphasize the reflected waves. The former removes the source effect, and the latter extracts the components over the frequency range to be interested. The optimal parameters for the operations are selected empirically by trial and error. While the autocorrelation analysis is easy and tells us useful information on subsurface structures, it may not to be easy to estimate the complex shallow structures. For example, the signals reflected from shallow sedimentary layers are hidden because the autocorrelation function has a large and broad peak around zero seconds in the time domain. In this study, we propose a novel way to treat the signals of autocorrelation function for investigating the sedimentary structures more clearly. We exploit the logarithmic power spectra and its inverse Fourier transform which is called cepstrum. First, we vanish the lower order term of the cepstrum instead of applying spectral whitening. After Fourier transforming and getting rid of logarithm, we subtract the average from the derived power spectra. The estimated waveform then has reflected waves only without a peak at around zero seconds. Eventually, we can identify the signals reflected from shallow sedimentary layers with higher resolution than ones from the conventional method.