10:45 AM - 12:15 PM
[SCG45-P30] Hypocenter locating of shallow tremors in the northern Japan Trench with S arrival times detected by seismogram polarization
Keywords:tremor, hypocenter determination
Objective
Slow earthquakes are often observed around seismogenic zones and are considered to interact with regular earthquakes. Detailed characterization of the interaction will provide useful information on the occurrence of large earthquakes. However, tectonic tremors, a kind of slow earthquakes, have small signals and their onset on seismograms is not clear, making it difficult to identify absolute arrival times. Therefore, the differential arrival times measured by the envelope correlation method (ECM) are generally used to determine the hypocenters.
In the shallow plate boundary in the northern Japan Trench, Takahashi et al. (2021) applied ECM to the seismograms obtained by Pop-up Ocean Bottom Seismometers (OBS) and determined hypocenters of tremors. Matsumoto et al. (2022) extracted regular earthquake events from the event catalog by Takahashi et al. (2021) as those with a duration of 20 s or less and determined their hypocenters by arrival times of P- and S-wave. They concluded that regular earthquakes and tremors occur close to each other both in time and space; however, a detailed discussion of their relationship is not easy due to the difficulty of evaluating the possible biases caused by differences in hypocenter location methods.
In the Nankai subduction zone, Hendriyana & Tsuji. (2021) pointed out that the similarity of envelope seismograms among the stations might be deteriorated and proposed a method to detect onsets of tremor signals based on the polarization analysis. They measured the differential travel times of detected onsets to locate the tremor hypocenters. In this study, we applied the method to the OBS data used by Takahashi et al. (2021) and Matsumoto et al. (2022) to relocate the tremor hypocenters in the northern Japan Trench.
Method
Among 49 OBSs installed in the northern Japan Trench from October 2007 to June 2008, we used the velocity waveforms of two horizontal components velocity seismograms from 12 OBSs surrounding a tremor swarm.
The covariance matrix S(t) of horizontal velocities is calculated by sliding the time window of 1 s length by 1 sample (50 ms), and the maximum eigenvalue λ of the matrix is obtained. Since λ is expected to increase when the S-wave comes in, we computed recursive STA/LTA of λ to identify the S-wave arrival. The sensitivity to the arrival of S-waves is further increased by using the characteristic function CF(t), which is defined by taking the positive part of the time derivative of STA/LTA.
The CF is normalized and cross-correlated between stations with half-overlapping 20 s time windows to identify the S-wave onsets. A significant event is defined when the number of station pairs for which the cross-correlation (CC) is greater than or equal to 0.85 exceeds 10. The lag time with the maximum correlation was measured as the arrival time difference of the S-wave between the stations.
Results
When the hypocenters of regular earthquakes are determined using S-wave alone, most of the RMS travel time residuals were less than 1.6 seconds. So, among the tremor sources determined by the above method, those with RMS runtime residuals of 1.6 seconds or larger were excluded as containing inaccurate values. As a result, the relocated hypocenters of the tremors were distributed near the plate boundary, like the results of Takahashi et al. (2021) by ECM.
Since ECM analyzes the envelope waveforms of wavetrains over a relatively long time, it correlates a wavetrain to a single tremor event, but the wavetrain can be a superposition of multiple low-frequency earthquakes (LFEs). In contrast, the present method is expected to be sensitive to the rise of individual LFEs, it might be possible to detect and determine the hypocenter of the individual LFEs that comprise the tremor event. So far, we have been able to detect multiple S-wave arrivals in a single tremor but have not yet been able to determine their hypocenter accurately enough. There may be a problem with the event association.
Slow earthquakes are often observed around seismogenic zones and are considered to interact with regular earthquakes. Detailed characterization of the interaction will provide useful information on the occurrence of large earthquakes. However, tectonic tremors, a kind of slow earthquakes, have small signals and their onset on seismograms is not clear, making it difficult to identify absolute arrival times. Therefore, the differential arrival times measured by the envelope correlation method (ECM) are generally used to determine the hypocenters.
In the shallow plate boundary in the northern Japan Trench, Takahashi et al. (2021) applied ECM to the seismograms obtained by Pop-up Ocean Bottom Seismometers (OBS) and determined hypocenters of tremors. Matsumoto et al. (2022) extracted regular earthquake events from the event catalog by Takahashi et al. (2021) as those with a duration of 20 s or less and determined their hypocenters by arrival times of P- and S-wave. They concluded that regular earthquakes and tremors occur close to each other both in time and space; however, a detailed discussion of their relationship is not easy due to the difficulty of evaluating the possible biases caused by differences in hypocenter location methods.
In the Nankai subduction zone, Hendriyana & Tsuji. (2021) pointed out that the similarity of envelope seismograms among the stations might be deteriorated and proposed a method to detect onsets of tremor signals based on the polarization analysis. They measured the differential travel times of detected onsets to locate the tremor hypocenters. In this study, we applied the method to the OBS data used by Takahashi et al. (2021) and Matsumoto et al. (2022) to relocate the tremor hypocenters in the northern Japan Trench.
Method
Among 49 OBSs installed in the northern Japan Trench from October 2007 to June 2008, we used the velocity waveforms of two horizontal components velocity seismograms from 12 OBSs surrounding a tremor swarm.
The covariance matrix S(t) of horizontal velocities is calculated by sliding the time window of 1 s length by 1 sample (50 ms), and the maximum eigenvalue λ of the matrix is obtained. Since λ is expected to increase when the S-wave comes in, we computed recursive STA/LTA of λ to identify the S-wave arrival. The sensitivity to the arrival of S-waves is further increased by using the characteristic function CF(t), which is defined by taking the positive part of the time derivative of STA/LTA.
The CF is normalized and cross-correlated between stations with half-overlapping 20 s time windows to identify the S-wave onsets. A significant event is defined when the number of station pairs for which the cross-correlation (CC) is greater than or equal to 0.85 exceeds 10. The lag time with the maximum correlation was measured as the arrival time difference of the S-wave between the stations.
Results
When the hypocenters of regular earthquakes are determined using S-wave alone, most of the RMS travel time residuals were less than 1.6 seconds. So, among the tremor sources determined by the above method, those with RMS runtime residuals of 1.6 seconds or larger were excluded as containing inaccurate values. As a result, the relocated hypocenters of the tremors were distributed near the plate boundary, like the results of Takahashi et al. (2021) by ECM.
Since ECM analyzes the envelope waveforms of wavetrains over a relatively long time, it correlates a wavetrain to a single tremor event, but the wavetrain can be a superposition of multiple low-frequency earthquakes (LFEs). In contrast, the present method is expected to be sensitive to the rise of individual LFEs, it might be possible to detect and determine the hypocenter of the individual LFEs that comprise the tremor event. So far, we have been able to detect multiple S-wave arrivals in a single tremor but have not yet been able to determine their hypocenter accurately enough. There may be a problem with the event association.