10:00 AM - 10:15 AM
[SSS09-20] Detection of the Velocity Boundary between the Sedimentary Layer and the Bedrock in Kumamoto Prefecture Using S-wave Autocorrelation Functions of Strong Motion Records
Keywords:autocorrelation function, strong motion record, S-wave velocity structure model, Kumamoto Prefecture
The 2016 Kumamoto earthquake ruptured a part of the Futagawa and Hinagu fault zones. The other parts of them did not rupture at that time, and there are other active faults in and around Kumamoto Prefecture. Therefore, it is important to improve strong ground motion prediction in Kumamoto Prefecture. The strong ground motion prediction requires a source model and a velocity structure model. The 3-D velocity structure model of Kumamoto Prefecture was constructed in the previous project (Asano et al., 2019). This model is based on the exploration data from the seismic reflection survey, the microtremor array survey and the hot spring well logs. The accuracy of the model differs from point to point, so it is necessary to test the model in order to improve it.
As one way to extract a reflection wave from an underground structure boundary, autocorrelation functions calculated from seismic observation records are used (e.g., Pham and Tkalcic, 2017). Chimoto and Yamanaka (2019, 2020) calculated the autocorrelation functions from strong motion records, detected the reflected waves from the layer boundaries of the Kanto Basin, and tried to adjust the S-wave velocity structure model of the sedimentary layers. Fukutome et al. (2021) also calculated S-wave autocorrelation functions from the strong motion records in the Osaka Basin, validated the existing velocity structure model, and discussed about a criteria of data selection associated with incident angles. In this study, we calculated autocorrelation functions of the S-wave parts of the strong motion records in Kumamoto and detected the reflected waves from the velocity boundary between the sedimentary layer and the bedrock.
First, we computed the S-wave autocorrelation functions for the strong motion records of the K-NET station KMM016 (Hitoyoshi) between March 22, 2005, when KMM016 was relocated, and December 31, 2017, on a trial basis. KMM016 is located in the southwestern part of the Hitoyoshi Basin. We read the S-wave arrival time manually, and did not use the records whose S-wave arrival was unclear. In the end, 114 records were used (the JMA magnitude: 2.8-6.4, the hypocentral depth: 2.2-89.1 km, the epicentral distance: 11.3-136.2 km). We set a time window of 11 s starting 1 s ahead the S-wave arrival, applied a bandpass filter, rotated to the transverse component, applied the spectral whitening, and computed autocorrelation functions. We stacked all the autocorrelation functions by using the Phase Weighted Stack (Schimmel and Paulssen, 1997). Among several frequency ranges of the bandpass filter tested, the one with a clear negative peak was selected, which was considered to be the reflected wave from the velocity boundary on the top of the bedrock. The most suitable filter ranging was 0.3-5 Hz, and the phase peak showed a lag time of 1.18 s. In addition, the two-way time of S-wave vertically propagating between the top of the bedrock and the surface is 1.05 s in the calculation using the S-wave velocity structure model (Asano et al., 2019). The contrast of the impedance on the boundary is strong, so we assumed the vertical propagation in the sedimentary layers.
Manually reading S-wave arrival time is hard work when the amount of data is large. Then, we tried to calculate the theoretical arrival time of S-wave using the JMA2001 velocity structure model (Ueno et al., 2002) and used this arrival time for setting the time window for the autocorrelation analysis. The peak time associated with the reflected S-wave from the boundary was 1.20 s.
The lag time of the reflected S-wave from the velocity boundary, calculated from the autocorrelation functions of the S-wave portions of the records at KMM016, was slightly later than the two-way time of the S-wave between the surface and the bedrock from the existing velocity structure model. Around KMM016, the boundary may be deeper than modeled, or the S-wave velocity may be slower, or both. The S-wave arrival time was faster than the reading in most of the records by 1-2 s. The reason why the reflected S-wave responses indicated almost the same value is considered that the used time window started 1 s before the S-wave arrival time and was set long enough to contain the reflected S-wave on the boundary. Thus, a theoretical S-wave arrival time can be used to set a time window. After that, we will calculate the lag time of S-wave responses at the other stations and validate the present S-wave velocity structure model (Asano et al., 2019).
Acknowledgments: The JMA unified earthquake catalog and the strong motion records from K-NET of the National Research Institute for Earth Science and Disaster Resilience (NIED) were used. We thank them for their cooperation.
As one way to extract a reflection wave from an underground structure boundary, autocorrelation functions calculated from seismic observation records are used (e.g., Pham and Tkalcic, 2017). Chimoto and Yamanaka (2019, 2020) calculated the autocorrelation functions from strong motion records, detected the reflected waves from the layer boundaries of the Kanto Basin, and tried to adjust the S-wave velocity structure model of the sedimentary layers. Fukutome et al. (2021) also calculated S-wave autocorrelation functions from the strong motion records in the Osaka Basin, validated the existing velocity structure model, and discussed about a criteria of data selection associated with incident angles. In this study, we calculated autocorrelation functions of the S-wave parts of the strong motion records in Kumamoto and detected the reflected waves from the velocity boundary between the sedimentary layer and the bedrock.
First, we computed the S-wave autocorrelation functions for the strong motion records of the K-NET station KMM016 (Hitoyoshi) between March 22, 2005, when KMM016 was relocated, and December 31, 2017, on a trial basis. KMM016 is located in the southwestern part of the Hitoyoshi Basin. We read the S-wave arrival time manually, and did not use the records whose S-wave arrival was unclear. In the end, 114 records were used (the JMA magnitude: 2.8-6.4, the hypocentral depth: 2.2-89.1 km, the epicentral distance: 11.3-136.2 km). We set a time window of 11 s starting 1 s ahead the S-wave arrival, applied a bandpass filter, rotated to the transverse component, applied the spectral whitening, and computed autocorrelation functions. We stacked all the autocorrelation functions by using the Phase Weighted Stack (Schimmel and Paulssen, 1997). Among several frequency ranges of the bandpass filter tested, the one with a clear negative peak was selected, which was considered to be the reflected wave from the velocity boundary on the top of the bedrock. The most suitable filter ranging was 0.3-5 Hz, and the phase peak showed a lag time of 1.18 s. In addition, the two-way time of S-wave vertically propagating between the top of the bedrock and the surface is 1.05 s in the calculation using the S-wave velocity structure model (Asano et al., 2019). The contrast of the impedance on the boundary is strong, so we assumed the vertical propagation in the sedimentary layers.
Manually reading S-wave arrival time is hard work when the amount of data is large. Then, we tried to calculate the theoretical arrival time of S-wave using the JMA2001 velocity structure model (Ueno et al., 2002) and used this arrival time for setting the time window for the autocorrelation analysis. The peak time associated with the reflected S-wave from the boundary was 1.20 s.
The lag time of the reflected S-wave from the velocity boundary, calculated from the autocorrelation functions of the S-wave portions of the records at KMM016, was slightly later than the two-way time of the S-wave between the surface and the bedrock from the existing velocity structure model. Around KMM016, the boundary may be deeper than modeled, or the S-wave velocity may be slower, or both. The S-wave arrival time was faster than the reading in most of the records by 1-2 s. The reason why the reflected S-wave responses indicated almost the same value is considered that the used time window started 1 s before the S-wave arrival time and was set long enough to contain the reflected S-wave on the boundary. Thus, a theoretical S-wave arrival time can be used to set a time window. After that, we will calculate the lag time of S-wave responses at the other stations and validate the present S-wave velocity structure model (Asano et al., 2019).
Acknowledgments: The JMA unified earthquake catalog and the strong motion records from K-NET of the National Research Institute for Earth Science and Disaster Resilience (NIED) were used. We thank them for their cooperation.