14:00 〜 14:15
[SCG40-32] Tectonic Tremors and Tremor Clusters from 0.3 Seconds to 10 Days Detected by Event Feature- and Seismicity-based Clustering: Application to Western Japan
キーワード：スロー地震、テクトニック微動、クラスタリング
Slow earthquakes are often categorized into tectonic tremors and low-frequency earthquakes (LFEs), very low-frequency earthquakes (VLFEs), and slow slip events (SSEs). These phenomena are considered different manifestations of a unified broadband process observed at different frequency bands (Ide et al., 2007; Ide & Beroza, 2023). Episodic migration events of tectonic tremors, which we term tremor clusters, are also considered as slow earthquakes between VLFEs and SSEs (Wech et al., 2010; Bletery et al., 2017; Gombert & Hawthorne, 2023). Compiling a high-quality catalog of tectonic tremors and tremor clusters is crucial for understanding the mechanisms of slow earthquakes and the stress state of plate interfaces.
In this study, we develop a method for distinguishing tectonic tremors from earthquakes and anthropogenic events without subjective selection criteria, using a clustering method based on event features. Combining this method with an envelope cross-correlation method to detect seismic events (Mizuno & Ide, 2019), we can construct a more complete tremor catalog. We evaluate five event features: depth, duration, amplitudes at high and low frequencies, and the amplitude ratio. The duration is estimated as the minimum time period that accounts for 50 % of the seismic wave energy at each channel.
Applying this method to southwest Japan, we successfully detected more than 1.5 times the number of tectonic tremors compared to the previous study. Most of the newly detected tremors have a duration shorter than several seconds. The duration of the detected tremors ranges from 0.3 s to approximately 100 s, which bridges the gap between LFEs and VLFEs. The relationship between seismic moment and duration appears to be consistent with the scaling law of slow earthquakes.
Furthermore, extending the clustering method of Zaliapin and Ben-Zion (2013), which is based on the spatiotemporal distance of seismic events, to tectonic tremors, we detect tectonic tremor clusters. Some tremor clusters overlap spatiotemporally with SSEs determined by Okada et al. (2022), while some tremor clusters correspond to only the initial phase of SSEs. The durations and estimated seismic moments of the established clusters are 5 seconds to 10 days and 1013 to 1018 Nm, respectively, bridging the gap between VLFEs and SSEs. The ratio of durations and estimated seismic moments is approximately one order of magnitude smaller than the value of Aguiar et al. (2009), which is based on the observation of ETS in Cascadia. However, it is noteworthy that our study solely relies on the seismic energy in the 2-8 Hz frequency band for seismic moment estimation.
Most of the tremor clusters in Tokai, Kii Peninsula, and the western part of Shikoku exhibit spatiotemporally continuous distribution and occasional migration. Conversely, in the central to eastern parts of Shikoku, repeating tremor clusters which are isolated from other tremor clusters in space and time are observed. Such regional variations could offer crucial insights into understanding the heterogeneity of plate interfaces.
In this study, we develop a method for distinguishing tectonic tremors from earthquakes and anthropogenic events without subjective selection criteria, using a clustering method based on event features. Combining this method with an envelope cross-correlation method to detect seismic events (Mizuno & Ide, 2019), we can construct a more complete tremor catalog. We evaluate five event features: depth, duration, amplitudes at high and low frequencies, and the amplitude ratio. The duration is estimated as the minimum time period that accounts for 50 % of the seismic wave energy at each channel.
Applying this method to southwest Japan, we successfully detected more than 1.5 times the number of tectonic tremors compared to the previous study. Most of the newly detected tremors have a duration shorter than several seconds. The duration of the detected tremors ranges from 0.3 s to approximately 100 s, which bridges the gap between LFEs and VLFEs. The relationship between seismic moment and duration appears to be consistent with the scaling law of slow earthquakes.
Furthermore, extending the clustering method of Zaliapin and Ben-Zion (2013), which is based on the spatiotemporal distance of seismic events, to tectonic tremors, we detect tectonic tremor clusters. Some tremor clusters overlap spatiotemporally with SSEs determined by Okada et al. (2022), while some tremor clusters correspond to only the initial phase of SSEs. The durations and estimated seismic moments of the established clusters are 5 seconds to 10 days and 1013 to 1018 Nm, respectively, bridging the gap between VLFEs and SSEs. The ratio of durations and estimated seismic moments is approximately one order of magnitude smaller than the value of Aguiar et al. (2009), which is based on the observation of ETS in Cascadia. However, it is noteworthy that our study solely relies on the seismic energy in the 2-8 Hz frequency band for seismic moment estimation.
Most of the tremor clusters in Tokai, Kii Peninsula, and the western part of Shikoku exhibit spatiotemporally continuous distribution and occasional migration. Conversely, in the central to eastern parts of Shikoku, repeating tremor clusters which are isolated from other tremor clusters in space and time are observed. Such regional variations could offer crucial insights into understanding the heterogeneity of plate interfaces.