2:30 PM - 2:45 PM
[SSS02-04] Ambient noise tomography in the offshore forearc region along the Japan trench using S-net data
★Invited Papers
Keywords:S-net, Ambient noise tomography, Japan Trench, Subduction zone
The innovative seafloor seismic network (S-net) is providing us new findings of megathrust zone seismicity and subduction zone dynamics. Although shear wave velocity structure can be a key to better understand them, our knowledge about shear velocity structure in the shallow subduction zone is still limited. Ambient noise surface wave tomography is now a standard tool for high-resolution imaging of shear velocity structure. We apply ambient noise tomography to S-net data to infer shear velocity structure in the shallow subduction zone, analyzing three-year records of three-component accelerometers.
However, the application of ambient noise tomography to S-net was not trivial because of technical and essential issues. One technical issue is unknown sensor orientation. Because the S-net sensors are fixed on the cables, even Z component is not in the vertical direction. Takagi et al. (2019) estimated the sensor orientations of all 150 stations by using the gravitational acceleration recorded on direct-current components of three-component accelerometers and the polarization of long-period teleseismic Rayleigh waves. Another technical issue is the high instrument noise level. In our target frequency range below 0.1 Hz that is sensitive to the seismogenic zone, the ambient seismic noise is below the instrument noise level of S-net. Takagi et al. (2021) found that coherent components of the instrument noise strongly contaminate the cross-correlation function of S-net records and developed solutions to correct the effect of coherent instrument noise.
A more essential issue is complex surface wave propagations in the ocean area: strong dispersion, multimode propagation, and their large lateral variations. We developed a robust method of multimode phase velocity measurement by combining the spatial auto-correlation method (Aki, 1957; Nishida et al. 2008) and the nonlinear multimode waveform fitting (Yoshizawa & Kennett, 2002). In this method, we use a one-dimensional velocity structure as unknown model parameters to synthesize ambient noise cross spectra. By fitting the synthetic cross spectra to observed ones with simulated annealing, we measured phase velocities of multimode surface waves. We also utilized full-component cross spectra (vertical-vertical, vertical-radial, radial-radial, and transverse-transverse) to maximize available observation information. With this method, we succeeded in measuring the path-averaged phase velocities of the fundamental mode Rayleigh and Love waves in 0.04-0.09 Hz and those of the first overtone Rayleigh wave in 0.06-0.09 Hz.
Surface wave tomography reveals that offshore extension of known low-velocity structure captured by land-based observations and along-strike variations in the phase velocity near the trench. We found low-velocity anomalies off the west of the Hidaka mountain range and off the Boso Peninsula, which may be associated with the Kuril arc collision (Iwasaki et al., 2004) and with the overlaid subduction of the Philippine Sea Plate (Uchida et al., 2009), respectively. In the forearc region near the trench, Love wave phase velocity shows a clear contrast at 38.5ºN: low velocity at the north and high velocity at the south, which is consistent with active source survey (Tsuru et al., 2002). The velocity boundary is located at the northern limit of the large coseismic slip of the 2011 Tohoku earthquake, and the large slip area is within the high-velocity anomaly. In addition, another high-velocity anomaly can be found in the low-seismicity region off Nemuro where the occurrence of a large earthquake is anticipated. The spatial correlation suggests that the structural control of the interplate coupling and interplate slip behavior. This study also suggests that even though its application to ocean bottom networks is not trivial, ambient noise tomography can make a significant contribution to understating the shallow subduction zone and to ocean bottom seismology.
However, the application of ambient noise tomography to S-net was not trivial because of technical and essential issues. One technical issue is unknown sensor orientation. Because the S-net sensors are fixed on the cables, even Z component is not in the vertical direction. Takagi et al. (2019) estimated the sensor orientations of all 150 stations by using the gravitational acceleration recorded on direct-current components of three-component accelerometers and the polarization of long-period teleseismic Rayleigh waves. Another technical issue is the high instrument noise level. In our target frequency range below 0.1 Hz that is sensitive to the seismogenic zone, the ambient seismic noise is below the instrument noise level of S-net. Takagi et al. (2021) found that coherent components of the instrument noise strongly contaminate the cross-correlation function of S-net records and developed solutions to correct the effect of coherent instrument noise.
A more essential issue is complex surface wave propagations in the ocean area: strong dispersion, multimode propagation, and their large lateral variations. We developed a robust method of multimode phase velocity measurement by combining the spatial auto-correlation method (Aki, 1957; Nishida et al. 2008) and the nonlinear multimode waveform fitting (Yoshizawa & Kennett, 2002). In this method, we use a one-dimensional velocity structure as unknown model parameters to synthesize ambient noise cross spectra. By fitting the synthetic cross spectra to observed ones with simulated annealing, we measured phase velocities of multimode surface waves. We also utilized full-component cross spectra (vertical-vertical, vertical-radial, radial-radial, and transverse-transverse) to maximize available observation information. With this method, we succeeded in measuring the path-averaged phase velocities of the fundamental mode Rayleigh and Love waves in 0.04-0.09 Hz and those of the first overtone Rayleigh wave in 0.06-0.09 Hz.
Surface wave tomography reveals that offshore extension of known low-velocity structure captured by land-based observations and along-strike variations in the phase velocity near the trench. We found low-velocity anomalies off the west of the Hidaka mountain range and off the Boso Peninsula, which may be associated with the Kuril arc collision (Iwasaki et al., 2004) and with the overlaid subduction of the Philippine Sea Plate (Uchida et al., 2009), respectively. In the forearc region near the trench, Love wave phase velocity shows a clear contrast at 38.5ºN: low velocity at the north and high velocity at the south, which is consistent with active source survey (Tsuru et al., 2002). The velocity boundary is located at the northern limit of the large coseismic slip of the 2011 Tohoku earthquake, and the large slip area is within the high-velocity anomaly. In addition, another high-velocity anomaly can be found in the low-seismicity region off Nemuro where the occurrence of a large earthquake is anticipated. The spatial correlation suggests that the structural control of the interplate coupling and interplate slip behavior. This study also suggests that even though its application to ocean bottom networks is not trivial, ambient noise tomography can make a significant contribution to understating the shallow subduction zone and to ocean bottom seismology.