11:00 AM - 1:00 PM
[SSS03-P08] Attempt of seismic velocity change detection by seismic interferometry using OBSs in the Japan-Kuril trenches junction
Keywords:seismic interferometry, ocean bottom seismometer
The temporal variation in seismic velocity associated with the occasion of earthquakes is often detected by using Seismic interferometry (SI); for example, the 2011 Tohoku-oki earthquake and preceding slow slip event (Uenumra et al., 2018), and the 2018 Hokkaido earthquake (Ikeda&Takagi, 2019). The Japan-Kuril trenches junction where we study is known as high slow earthquake activity regions, such as very low frequency earthquakes (Baba et al., 2020) and tectonic tremors (Nishikawa et al., 2019; Kawakubo, 2021). Tectonic tremors often show epicenter migration, and consistency of migration speed with the fluid diffusion model implies the relationship between slow earthquake activity and fluid migration (e.g., Tanaka et al., 2019). Such fluid migration may also appear as a temporal change in subsurface structure. We examine to detect a temporal change in the subsurface structure associated with earthquakes by applying SI to the long-term OBS data in 2006-2007, including tremor activities, detected by Kawakubo (2021), in addition to two large earthquakes off the Kuril Islands (Lay et al., 2009) and the near-filed earthquake (Mj 6.2).
The OBS observation was conducted from October 25, 2006, to June 5, 2007. The network is composed of 42 long-term OBS. The northern half area of the network with 21 OBSs was installed on November 24, 2007. The OBSs had a 1-Hz three-component velocity sensor, and observed waveforms were recorded with a 200-Hz sampling rate.
We calculated an auto-correlation function (ACF) of ambient noise records at each station. We resampled vertical-component waveform data down to 10 Hz and applied band-pass filter at 0.25-2 Hz and a one-bit normalization. ACFs were calculated for each 120-s window with a 60-s overlap, and they are averaged over 1439 windows as the daily ACF. Then, we took an average of the daily ACFs in December 2007 as the reference ACF. We computed cross-correlation function (CCF) between the reference ACF and 15-day-averaged ACFs to investigate temporal changes of the ACFs. The CCFs are calculated within moving lag-time windows. The phase shift of a 15-day ACF against the reference (dt) corresponds to the amount of change in two-way-time to the scattering source. The seismic velocity changes are calculated from, where is the lag-time of the ACF used to measure dt.
The obtained ACFs are featured as follows; clear signals at lag times earlier than ~ 20 s perturbing through the observation period, commonly shown at most OBS, and smaller signals randomly found at lag times from ~ 20 to ~ 90 s at a part of OBSs. The CCFs between the 15-days-averaged and reference ACFs indicate a high CC coefficient value over 0.5 at the lag time earlier than ~ 90 s in contrast to later lags. However, at most OBSs, the CC coefficient around the lag time of 10 s is ~0.6 which is quite lower than ~0.8-0.9 in the later lag time window. Finally, the estimated varies ~ 2% in maximum around the lag time of 10 s but is almost stable in the later lag time.
Our result presents that the variation in is visible especially at lag times near ~ 10 s. However, the low CC coefficient value at that lag time implies less significance of present variation so careful discussion is necessary. Uemura et al. (2018) has observed similar ACF changes during the tremor period activated off Miyagi and interpreted as a disturbance of the surrounding wavefield resulting from the high tremor activity. We will investigate the significance of estimated variation carefully and then would like to interpret those meanings.
The OBS observation was conducted from October 25, 2006, to June 5, 2007. The network is composed of 42 long-term OBS. The northern half area of the network with 21 OBSs was installed on November 24, 2007. The OBSs had a 1-Hz three-component velocity sensor, and observed waveforms were recorded with a 200-Hz sampling rate.
We calculated an auto-correlation function (ACF) of ambient noise records at each station. We resampled vertical-component waveform data down to 10 Hz and applied band-pass filter at 0.25-2 Hz and a one-bit normalization. ACFs were calculated for each 120-s window with a 60-s overlap, and they are averaged over 1439 windows as the daily ACF. Then, we took an average of the daily ACFs in December 2007 as the reference ACF. We computed cross-correlation function (CCF) between the reference ACF and 15-day-averaged ACFs to investigate temporal changes of the ACFs. The CCFs are calculated within moving lag-time windows. The phase shift of a 15-day ACF against the reference (dt) corresponds to the amount of change in two-way-time to the scattering source. The seismic velocity changes are calculated from, where is the lag-time of the ACF used to measure dt.
The obtained ACFs are featured as follows; clear signals at lag times earlier than ~ 20 s perturbing through the observation period, commonly shown at most OBS, and smaller signals randomly found at lag times from ~ 20 to ~ 90 s at a part of OBSs. The CCFs between the 15-days-averaged and reference ACFs indicate a high CC coefficient value over 0.5 at the lag time earlier than ~ 90 s in contrast to later lags. However, at most OBSs, the CC coefficient around the lag time of 10 s is ~0.6 which is quite lower than ~0.8-0.9 in the later lag time window. Finally, the estimated varies ~ 2% in maximum around the lag time of 10 s but is almost stable in the later lag time.
Our result presents that the variation in is visible especially at lag times near ~ 10 s. However, the low CC coefficient value at that lag time implies less significance of present variation so careful discussion is necessary. Uemura et al. (2018) has observed similar ACF changes during the tremor period activated off Miyagi and interpreted as a disturbance of the surrounding wavefield resulting from the high tremor activity. We will investigate the significance of estimated variation carefully and then would like to interpret those meanings.