5:15 PM - 6:45 PM
[SSS07-P04] Epicenter determination by the correlation method for shallow very low-frequency earthquakes
Keywords:slow earthquake, very low-frequency earthquake, Hyuga-nada, correlation method, cycle skipping
Shallow very low-frequency earthquakes (SVLFEs) are suitable for real-time monitoring of source-area expansion of shallow slow earthquakes because their signals can be clearly detected even at distant seismic stations. Array analysis (Asano et al. 2008) and grid search (e.g. Asano et al. 2015) have been mainly used to determine the epicenter of SVLFE in the Nankai Trough area by analyzing broadband seismic records from F-net and DONET. The envelope correlation method has also been used to analyze OBS records (Tonegawa et al. 2020). However, it appears that the correlation method has not yet been applied to bandpass waveforms to determine the epicenter of SVLFE. In this study, we confirmed that the SVLFE epicenter can be determined with sufficient accuracy by applying the correlation method to F-net bandpass waveforms.
Generally, the correlation method treats the time difference that gives the maximum value of the cross-correlation coefficient calculated from the waveforms of two stations as the travel-time difference data. However, in narrow-band waveforms, the cycle skipping often occurs, in which the time difference deviates by an integer multiple of the dominant period from the true travel-time difference. In this study, we solved this problem by treating time differences that give maxima of the cross-correlation coefficient and by setting conditions other than the cross-correlation coefficient.
The above method was applied to determine the epicenters of three large-scale SVLFE activities that occurred in the Hyuga-nada area: January-March 2010, May-July 2015, and April-May 2023, when the active area extended eastward to the south of Cape Ashizuri. The data used were the vertical component of broadband seismometer records from 27 F-net stations. They were band-pass filtered at periods between 20 and 50 sec, and then resampled to 1 sec. The analysis was performed continuously on the waveforms reduced by the theoretical travel-time for a certain reference point in the Hyuga-nada area, shifting the analysis window of 300 seconds in length by 150 seconds. The apparent wave velocity was assumed to be 3.8 km/s. The obtained epicenters were adopted as the results if the bootstrap standard error was less than 7 km.
The distribution of epicenters shows four distinct clusters in all three activities, which surround the subducting seamounts inferred from the geomagnetic anomalies. This feature is almost the same as for shallow tremors (Arai et al. 2023). The present results are consistent with those of the grid search using template waveforms, and more detailed epicenter distributions are obtained. This indicates that the SVLFE epicenters can be determined with sufficient accuracy without considering the three-dimensional seismic velocity structure as far as the epicenter determination is concerned.
Acknowledgments: F-net records were used in this study. We thank all parties involved at the National Research Institute for Earth Science and Disaster Resilience.
Generally, the correlation method treats the time difference that gives the maximum value of the cross-correlation coefficient calculated from the waveforms of two stations as the travel-time difference data. However, in narrow-band waveforms, the cycle skipping often occurs, in which the time difference deviates by an integer multiple of the dominant period from the true travel-time difference. In this study, we solved this problem by treating time differences that give maxima of the cross-correlation coefficient and by setting conditions other than the cross-correlation coefficient.
The above method was applied to determine the epicenters of three large-scale SVLFE activities that occurred in the Hyuga-nada area: January-March 2010, May-July 2015, and April-May 2023, when the active area extended eastward to the south of Cape Ashizuri. The data used were the vertical component of broadband seismometer records from 27 F-net stations. They were band-pass filtered at periods between 20 and 50 sec, and then resampled to 1 sec. The analysis was performed continuously on the waveforms reduced by the theoretical travel-time for a certain reference point in the Hyuga-nada area, shifting the analysis window of 300 seconds in length by 150 seconds. The apparent wave velocity was assumed to be 3.8 km/s. The obtained epicenters were adopted as the results if the bootstrap standard error was less than 7 km.
The distribution of epicenters shows four distinct clusters in all three activities, which surround the subducting seamounts inferred from the geomagnetic anomalies. This feature is almost the same as for shallow tremors (Arai et al. 2023). The present results are consistent with those of the grid search using template waveforms, and more detailed epicenter distributions are obtained. This indicates that the SVLFE epicenters can be determined with sufficient accuracy without considering the three-dimensional seismic velocity structure as far as the epicenter determination is concerned.
Acknowledgments: F-net records were used in this study. We thank all parties involved at the National Research Institute for Earth Science and Disaster Resilience.