11:15 AM - 11:30 AM
[STT36-08] Estimation of temporal changes in seismic velocities at Sakurajima based on seismic interferometry analysis of DAS data
Keywords:DAS, Sakurajima, seismic velocity change
Seismic interferometry is a technique to extract seismic waveforms propagating between two seismic stations by calculating the cross-correlation function (CCF) of seismic ambient noise at the two stations [Shapiro & Campillo, 2004]. Taking advantage of this technique, which does not require natural earthquakes or active shot records, estimates of seismic velocity changes associated with large earthquakes and volcanic activities have been reported in various regions. Distributed acoustic sensing (DAS) is a technique for measuring strain/strain velocity in the direction of fiber-optic cables at intervals of several meters. Distributed acoustic sensing (DAS) is a technique for measuring strain/strain velocity in the direction of fiber-optic cables at intervals of several meters. Recently, seismic interferometry has been applied to DAS data to estimate temporal changes in seismic velocity, and it is expected that DAS data will enable us to estimate subsurface structural changes with higher spatial resolution than seismic interferometry analysis using conventional seismometers. In this study, we applied seismic interferometry to DAS data at Sakurajima to estimate seismic velocity changes before and after the M4.1 earthquake that occurred in Kagoshima Bay at around 22:30 on December 8, 2022, during the DAS observation period.
We used DAS continuous records (channels 100-2495) from 10:00 on December 7, 2022 to 13:00 on December 9, 2022. The seismic ambient noise CCFs were calculated for 87 channel pairs set up in the following manner. First, one channel number was set at 25 channel intervals (100, 125, ..., 2250) in the range 100-2250. The other channel number was set so that the linear distance between the channels was 1,000 m. In the calculation of CCFs, we spatially stacked the channel in which the seismic ambient noise CCF is calculated with the continuous DAS recordings of five channels before and after to reduce the effect of incoherent noise [Maass et al. 2023]. Then, a 1-4 Hz bandpass filter and one-bit normalization were applied to the continuous records after spatial stacking. The CCFs calculated every 10 minutes were then stacked for 6 hours (SCCF). This calculation was repeated with a 75% overlap of time windows to obtain SCCFs every 5,400 seconds. In the measurement of relative velocity change (dv/v), a reference CCF (RCCF) was calculated by stacking SCCFs over the entire period. The wavelet transform stretching method [Yuan et al., 2021] was used to measure dv/v between RCCF and an SCCF for each frequency.
The dv/v was measured every 5400 seconds, and a velocity decrease was estimated in some channel pairs before and after the earthquake at 22:30 on December 8. The maximum amount of velocity reduction was about 1 % at 2-3 Hz. At frequencies below 2 Hz, measured dv/v varied by several percent within a day. Therefore, it was difficult to detect changes in seismic velocity before and after the earthquake. At frequencies above 3 Hz, the waveform correlation between RCCF and SCCF was less than 0.6 in many cases, suggesting that dv/v measurements are not reliable.
The spatial distribution of seismic velocity changes was examined by plotting dv/v at 2-3 Hz for each channel pair on a map. Specifically, dv/v were assigned to two-channel locations, assuming a higher sensitivity to seismic velocity changes in the vicinity of the channels. The velocity decrease before and after the earthquake was particularly noticeable (up to about 1 %) in the channel 100-600 interval. We compared this result with the AVS30 map provided on J-SHIS (https://www.j-shis.bosai.go.jp). The AVS30 around channels 100-600 was about 300 m/s, while the AVS30 at other channels was about 500 m/s. By using the JMA short-period seismograms at Sakurajima, the PGV of the earthquake was estimated to be about 0.2 cm/s. Our results suggest that the seismic velocity reduction due to the earthquake ground motion might have been more pronounced in channels 100-600, where the ground is softer than in other sections.
Acknowledgments: We used fiber optic cables from the Ministry of Land, Infrastructure, Transport and Tourism. We would like to thank the Osumi River National Highway Office for helping us with the DAS observation. We thank the Japan Meteorological Agency for providing continuous seismograms.
We used DAS continuous records (channels 100-2495) from 10:00 on December 7, 2022 to 13:00 on December 9, 2022. The seismic ambient noise CCFs were calculated for 87 channel pairs set up in the following manner. First, one channel number was set at 25 channel intervals (100, 125, ..., 2250) in the range 100-2250. The other channel number was set so that the linear distance between the channels was 1,000 m. In the calculation of CCFs, we spatially stacked the channel in which the seismic ambient noise CCF is calculated with the continuous DAS recordings of five channels before and after to reduce the effect of incoherent noise [Maass et al. 2023]. Then, a 1-4 Hz bandpass filter and one-bit normalization were applied to the continuous records after spatial stacking. The CCFs calculated every 10 minutes were then stacked for 6 hours (SCCF). This calculation was repeated with a 75% overlap of time windows to obtain SCCFs every 5,400 seconds. In the measurement of relative velocity change (dv/v), a reference CCF (RCCF) was calculated by stacking SCCFs over the entire period. The wavelet transform stretching method [Yuan et al., 2021] was used to measure dv/v between RCCF and an SCCF for each frequency.
The dv/v was measured every 5400 seconds, and a velocity decrease was estimated in some channel pairs before and after the earthquake at 22:30 on December 8. The maximum amount of velocity reduction was about 1 % at 2-3 Hz. At frequencies below 2 Hz, measured dv/v varied by several percent within a day. Therefore, it was difficult to detect changes in seismic velocity before and after the earthquake. At frequencies above 3 Hz, the waveform correlation between RCCF and SCCF was less than 0.6 in many cases, suggesting that dv/v measurements are not reliable.
The spatial distribution of seismic velocity changes was examined by plotting dv/v at 2-3 Hz for each channel pair on a map. Specifically, dv/v were assigned to two-channel locations, assuming a higher sensitivity to seismic velocity changes in the vicinity of the channels. The velocity decrease before and after the earthquake was particularly noticeable (up to about 1 %) in the channel 100-600 interval. We compared this result with the AVS30 map provided on J-SHIS (https://www.j-shis.bosai.go.jp). The AVS30 around channels 100-600 was about 300 m/s, while the AVS30 at other channels was about 500 m/s. By using the JMA short-period seismograms at Sakurajima, the PGV of the earthquake was estimated to be about 0.2 cm/s. Our results suggest that the seismic velocity reduction due to the earthquake ground motion might have been more pronounced in channels 100-600, where the ground is softer than in other sections.
Acknowledgments: We used fiber optic cables from the Ministry of Land, Infrastructure, Transport and Tourism. We would like to thank the Osumi River National Highway Office for helping us with the DAS observation. We thank the Japan Meteorological Agency for providing continuous seismograms.