11:00 〜 13:00
[SSS03-P02] Shallow tremor epicenter determination based on simulation results of high-frequency seismic wave propagation in a local three-dimensional velocity model
キーワード：スロー地震、微動、海底地震計、地震波伝播シミュレーション
Deployments of permanent ocean bottom seismometer networks (e.g., DONET and S-net) allow us to analyze seismic phenomena in offshore regions even for small signals of shallow slow earthquakes. By using these offshore observations, tectonic tremors and very low frequency earthquakes on the shallow plate boundary have been investigated (e.g., Araki et al., 2017; Nakano et al., 2018; Nishikawa et al., 2019). To locate tectonic tremors, the envelope cross-correlation method (ECM) has been widely used. However, shallow tremor locations by ECM include large uncertainties due to shallower heterogeneous structures in offshore regions. Takemura et al. (2020) investigated the effects of offshore heterogeneities on tremor waveforms using numerical simulations in a local 3D model southeast off the Kii Peninsula. They also demonstrated that shallow sedimentary structure dominantly controls the propagation characteristics of high-frequency seismic waves from shallow tremors to DONET stations. To achieve reliable monitoring of shallow slow earthquakes along the Nankai Trough, we proposed the matched filter technique based on simulated Green’s function envelopes in a local 3D model.
Green’s functions at virtual source grids every 0.01º on the plate boundary were calculated via reciprocal calculation of the OpenSWPC (Maeda et al., 2017). The 3D velocity model was the same in Takemura et al. (2020). In this model, a 3D sedimentary structure was constructed by interpolation/extrapolation of 1D structures beneath DONET (Tonegawa et al., 2017). We synthesized seismograms of low-angle thrust faulting mechanism. Then, we calculated the horizontal and vertical component envelopes for frequencies of 2-8 Hz. To detect and locate shallow tremor, we calculated cross-correlation coefficients (CCs) between synthetic and observed envelopes at DONET stations. Envelope CC (CCe) is evaluated by averaging CCs for each station and component. We also evaluate amplitude CC (CCa), which is cross-correlation coefficients between observed and simulated maximum amplitudes at used stations. We finally evaluated normalized CC (CCn) by using an equation 0.5*(CCe+CCa) for each time window. When the largest CCn in each 60-s time window exceeds 0.7, these events are selected as shallow tremor candidates.
We tested our method for the shallow slow earthquake episode that started from 6th December 2020. Spatiotemporal distributions of our shallow tremor candidates well corresponded with those of shallow VLFEs (Takemura et al., 2021; Yamamoto et al., 2021). During the episode, shallow tremor activity initiated in the eastern part of the activity area, where the paleo-Zenisu ridge is subducted (Park et al., 2004). Shallow tremors migrated in south-southwest directions, near the trench. The activity areas of shallow tremors also well corresponded to the regions with heterogeneous changes, which can be considered as a result of the propagation of fluid or rupture of slow slip events (Tonegawa et al., 2021).
More details of shallow tremor migrations in the episode from December 2020 will appear in our JpGU presentation.
Acknowledgments
We used NIED DONET data. Numerical simulations were conducted on the Fujitsu PRIMERGY CX600M1/CX1640M1 (Oakforest-PAC) and PRIMEPC FX1000 (Wisteria/BDEC-01) in the Information Technology Center, University of Tokyo. We also used the computational resources of the Earth Simulator in the Japan Agency for Marine-Earth Science and Technology. This study was supported by the ERI JURP 2020-S-04. This study was also supported by JSPS KAKENHI Grant-in-Aid for Scientific Research (C), grant number 21K03696, and a Grant-in-Aid for Scientific Research on Transformative Research Areas “Science of Slow-to-Fast earthquakes,” grant numbers 21H05203 and 21H05205.
Green’s functions at virtual source grids every 0.01º on the plate boundary were calculated via reciprocal calculation of the OpenSWPC (Maeda et al., 2017). The 3D velocity model was the same in Takemura et al. (2020). In this model, a 3D sedimentary structure was constructed by interpolation/extrapolation of 1D structures beneath DONET (Tonegawa et al., 2017). We synthesized seismograms of low-angle thrust faulting mechanism. Then, we calculated the horizontal and vertical component envelopes for frequencies of 2-8 Hz. To detect and locate shallow tremor, we calculated cross-correlation coefficients (CCs) between synthetic and observed envelopes at DONET stations. Envelope CC (CCe) is evaluated by averaging CCs for each station and component. We also evaluate amplitude CC (CCa), which is cross-correlation coefficients between observed and simulated maximum amplitudes at used stations. We finally evaluated normalized CC (CCn) by using an equation 0.5*(CCe+CCa) for each time window. When the largest CCn in each 60-s time window exceeds 0.7, these events are selected as shallow tremor candidates.
We tested our method for the shallow slow earthquake episode that started from 6th December 2020. Spatiotemporal distributions of our shallow tremor candidates well corresponded with those of shallow VLFEs (Takemura et al., 2021; Yamamoto et al., 2021). During the episode, shallow tremor activity initiated in the eastern part of the activity area, where the paleo-Zenisu ridge is subducted (Park et al., 2004). Shallow tremors migrated in south-southwest directions, near the trench. The activity areas of shallow tremors also well corresponded to the regions with heterogeneous changes, which can be considered as a result of the propagation of fluid or rupture of slow slip events (Tonegawa et al., 2021).
More details of shallow tremor migrations in the episode from December 2020 will appear in our JpGU presentation.
Acknowledgments
We used NIED DONET data. Numerical simulations were conducted on the Fujitsu PRIMERGY CX600M1/CX1640M1 (Oakforest-PAC) and PRIMEPC FX1000 (Wisteria/BDEC-01) in the Information Technology Center, University of Tokyo. We also used the computational resources of the Earth Simulator in the Japan Agency for Marine-Earth Science and Technology. This study was supported by the ERI JURP 2020-S-04. This study was also supported by JSPS KAKENHI Grant-in-Aid for Scientific Research (C), grant number 21K03696, and a Grant-in-Aid for Scientific Research on Transformative Research Areas “Science of Slow-to-Fast earthquakes,” grant numbers 21H05203 and 21H05205.