11:00 AM - 11:15 AM
[STT36-07] Estimation of shallow seismic velocity structure along National Route 47 using traffic noise in DAS records
Keywords:DAS, Surface wave exploration, shallow structure
Surface wave exploration employing existing telecommunication optical fiber cables and Distributed Acoustic Sensing (DAS) has gained significant attention in recent years due to its potential to drastically reduce costs and labor compared to traditional geophysical exploration methods. The attempt was first made by Ajo-Franklin et al. (2019) near Sacramento, California, and has been applied in other areas such as Ridgecrest (Yang et al., 2021) and Yunnan Province, China (Song et al., 2021), and has also been conducted within Japan in investigations along parts of national routes by Yabu et al. (2021, SSJ) and Hamanaka et al. (2023, SSJ). Leveraging DAS’s capacity to concurrently measure sections spanning several tens of kilometer, we utilized telecommunication cables installed along a major route to conduct a comprehensive surface wave exploration, thus enabling the imaging of subsurface structures ranging from small to large scale. Here, we present the results of our DAS observation in Japan.
In this study, we utilized a fiber-optic cable buried beneath National Route 47, which extends from Furukawa to Naruko Onsen in Osaki City, Miyagi Prefecture. The channel spacing was 6.4 m, the gauge length was 9.6 m, the total number of channels was 4855, covering a length of about 31 kilometers, and the sampling frequency was 200 Hz. The measurement period spanned from March 28 to April 21, 2022. In the following analyses, data during March 28 and April 4 were used.
Continuous, prominent traffic noise was observed in the DAS recordings. The dataset was organized into subsections, each comprising 50 channels, with a 25-channel overlap between consecutive subsections. Analysis was conducted on 130 of these subsections where the azimuths between all channels, viewed from the position of the smallest channel number within each subsection, was contained within 10 degrees. For each subsection, we constructed minute-by-minute dispersion spectra of surface waves using the Phase-shift method (Park et al., 1998), assuming that all channels are aligned in a straight line and that traffic signals propagate consistently parallel to this alignment. To minimize the influence of non-traffic noise sources, only recordings from late night hours (0:00-5:00) were utilized. Data from a week were compounded and averaged, yielding dispersion spectra with sufficient signal-to-noise ratios for each subsection. We inverted those spectra for seismic velocity structures within each subsection. For layers deeper than those with Vs = 1600 m/s, we employed the J-SHIS model, while the shallower layers were modeled using a nine-layer horizontal stratification, estimating the P-wave velocity, S-wave velocity, and thickness of each layer. Density was assumed to be a function of P-wave velocity (Gardner et al., 1974). Dispersion spectra were normalized for each frequency, and following Yuan et al. (2020), the total value of power spectral density that coincided with the theoretical dispersion curve was used as an objective function. The particle swarm optimization method was used to determine the best parameters. Due to the distinct modes and frequencies observed in each dispersion spectrum, the highest order of the modes and the frequencies for each mode to be used for inversion were determined visually for each subsection.
A pseudo two-dimensional seismic velocity structure spanning approximately 30 kilometers was obtained. A notable large-scale structure emerges: the depths of layers with higher seismic velocities become shallower from Furukawa towards Naruko Onsen. For instance, the layer with a Vs = 1 km/s is located about 500 meters deep near Furukawa, while at the midpoint, Iwadeyama, it is 300 meters deep, and at Kawatabi it is approximately 30 meters deep. In the section parallel to the Eai River, it was observed that layers with a Vs = 400 m/s or less exist up to about 20 meters deep. This is interpreted as soft ground deposited by the river's sedimentary actions, which is of significant importance for hazard assessment. We will compare these results with analyses using seismic interferometry and microtremor survey.
Acknowledgments:
We would like to thank the Ministry of Land, Infrastructure, Transport and Tourism for allowing us to use their optical fibers and the staff at the Sendai National Highway Office for their cooperation.
In this study, we utilized a fiber-optic cable buried beneath National Route 47, which extends from Furukawa to Naruko Onsen in Osaki City, Miyagi Prefecture. The channel spacing was 6.4 m, the gauge length was 9.6 m, the total number of channels was 4855, covering a length of about 31 kilometers, and the sampling frequency was 200 Hz. The measurement period spanned from March 28 to April 21, 2022. In the following analyses, data during March 28 and April 4 were used.
Continuous, prominent traffic noise was observed in the DAS recordings. The dataset was organized into subsections, each comprising 50 channels, with a 25-channel overlap between consecutive subsections. Analysis was conducted on 130 of these subsections where the azimuths between all channels, viewed from the position of the smallest channel number within each subsection, was contained within 10 degrees. For each subsection, we constructed minute-by-minute dispersion spectra of surface waves using the Phase-shift method (Park et al., 1998), assuming that all channels are aligned in a straight line and that traffic signals propagate consistently parallel to this alignment. To minimize the influence of non-traffic noise sources, only recordings from late night hours (0:00-5:00) were utilized. Data from a week were compounded and averaged, yielding dispersion spectra with sufficient signal-to-noise ratios for each subsection. We inverted those spectra for seismic velocity structures within each subsection. For layers deeper than those with Vs = 1600 m/s, we employed the J-SHIS model, while the shallower layers were modeled using a nine-layer horizontal stratification, estimating the P-wave velocity, S-wave velocity, and thickness of each layer. Density was assumed to be a function of P-wave velocity (Gardner et al., 1974). Dispersion spectra were normalized for each frequency, and following Yuan et al. (2020), the total value of power spectral density that coincided with the theoretical dispersion curve was used as an objective function. The particle swarm optimization method was used to determine the best parameters. Due to the distinct modes and frequencies observed in each dispersion spectrum, the highest order of the modes and the frequencies for each mode to be used for inversion were determined visually for each subsection.
A pseudo two-dimensional seismic velocity structure spanning approximately 30 kilometers was obtained. A notable large-scale structure emerges: the depths of layers with higher seismic velocities become shallower from Furukawa towards Naruko Onsen. For instance, the layer with a Vs = 1 km/s is located about 500 meters deep near Furukawa, while at the midpoint, Iwadeyama, it is 300 meters deep, and at Kawatabi it is approximately 30 meters deep. In the section parallel to the Eai River, it was observed that layers with a Vs = 400 m/s or less exist up to about 20 meters deep. This is interpreted as soft ground deposited by the river's sedimentary actions, which is of significant importance for hazard assessment. We will compare these results with analyses using seismic interferometry and microtremor survey.
Acknowledgments:
We would like to thank the Ministry of Land, Infrastructure, Transport and Tourism for allowing us to use their optical fibers and the staff at the Sendai National Highway Office for their cooperation.