Active monitoring of upper crust using ACROSS-seismic array system

Temporal variations of S- and surface-wave travel times were continuously monitored using ACROSS source and seismic array. We made an experiment lasting 5 months at a site near the Nojima fault which ruptured during the 1995 Kobe earthquake (M7.2). Elastic waves generated by ACROSS vibrators are received by two seismic arrays. One is located at about 300m northwest and the other is about 300m southwest of the vibrators. Each array has an aperture size of about 50 m and consists of ten seismometers that are three component velocity sensors with natural frequency of 4.5Hz. In this experiment, we used solar-battery systems to enable the long-term experiment, and we succeeded in continuous data recording without any troubles. To obtain the signal in time domain, in which P, S and some later phases were included, we executed the following procedure in the frequency domain. We extracted the ACROSS signals from the every stacked data. The extracted signal was divided by the force which was generated by the source. In this study, we used the spectrum of the theoretical force calculated from the frequency-modulated rotation. We regarded the result as a transfer function (or band-limited impulse response) between the source and the receivers. Applying appropriate window function and inverse Fourier transformation, we could obtain S wave and big surface wave. To emphasize later part of ACROSS signal, we stacked the data of all N-array sensors for every one hour and transformed its envelope using Hilbert transformation. We may detect some phase around 8, 13, 16 -seconds in the envelope. There were a few candidates for a cause of the phases, random noise or coherent noise, or reflected signals from deeper portion of the crust. We examined these possibilities one by one. The phases were found all through the experiment period. Therefore they must not be due to random noises. Next, we synthesized transfer function between the vibrator and the seismic array to examine the effect of coherent noises theoretically. As a result, we figured out that the later phases should be excited due to coherent noise which should be distributed uniformly over all frequency band. A strong candidate of such noise is uncertainty of the source function which was used in calculation of the transfer function. We estimated relation between the main phases and later phases excited by the uncertainty of the source function quantitatively. In this case, we cannot insist that the later phases were correlated with reflected phase. If they are due to uncertainty of the source function, the amplitude of the later phases, which was three orders of magnitude smaller than the maximum amplitude, suggests that the uncertainty of the source function used in this study should be about 1 percent.