Abstract
A statistical method of analyzing vertical array seismic motion data is proposed. This method separates observed seismic motions into coherent wave components ("signal components") and incoherent wave components ("noise components") in the frequency domain, where the signal components are defined as perfectly correlated components between array observation points, whereas the noise components have no correlation. The method is applied to the MIYA vertical array data in order to reveal amplitudes of the signal components and the noise components at each observation point. The results indicate that the signal components are in general more dominant than the noise components at all observation points. However, the noise components become dominant in the frequencies where cancellations of up-going and down-going waves occur at down-hole observation points and in the higher frequencies at the surface point. This indicates that the noise components are not negligible when interpreting wave propagation phenomena in the ground using vertical array seismic motion data. In fact, ground transfer functions evaluated from only the signal components have higher peaks and deeper troughs compared with the usual ones calculated directly from observed data. In order to clarify differences of ground parameters inferred from these ground transfer functions, optimization analyses of ground parameters are conducted using the transfer functions of the signal components and the usual ones, respectively. The results indicate that damping factors inferred from the ground transfer functions of the signal components are smaller than those from the usual ones especially in the low frequency range. In conclusion, the noise components in vertical array seismic motion data suppress peaks of the ground transfer functions between array points and consequently cause overestimation of ground damping factors at low frequencies.
