16:00 〜 16:15
[S09-29] 低周波地震活動統計モデルの比較ー低周波地震活動の発生メカニズムへの示唆ー
Slow earthquakes are a general term for various slow fault-slip events. In subduction zones, the relationship between slow earthquakes and megathrust earthquakes has been actively studied (e.g., Obara & Kato, 2016). Quantifying characteristics of slow earthquake activity and monitoring them are important because they may change before a megathrust earthquake (e.g., Matsuzawa et al., 2010; Luo & Liu, 2019).
Statistical seismicity models are useful for quantifying seismicity characteristics. For the activity of ordinary fast earthquakes, the epidemic-type aftershock-sequence (ETAS) model is widely used (e.g., Ogata, 1988). However, there is still no standard statistical model for slow earthquake activity. Statistical modeling of low-frequency earthquakes (LFEs), a type of slow earthquake, has begun recently. Lengliné et al. (2017) and Tan & Marsan (2020) proposed statistical models similar to the ETAS model (L model), in which the LFE occurrence rate is expressed as the summation of a stationary background rate and the effect of each LFE inducing future LFEs. Ide & Nomura (2022) proposed a statistical model (IN model) for tectonic tremors, a swarm of LFEs. Their model was based on an approach different from the L model; they described the probability distribution of interevent times using log-normal and Brownian passage time distributions. Furthermore, their model does not explicitly include the effect of event-to-event triggering.
Identifying the best statistical model for LFE activity is important for elucidating the mechanisms governing LFE activity. However, the existing statistical models have never been compared. Here, I applied the existing models to M 0.6 or larger LFEs in the Nankai Trough (Kato & Nakagawa, 2020) and examined their performance using Akaike’s information criterion (Akaike 1974). As a result, I found the IN model often outperforms the L and ETAS models. This result indicates that the effect of LFE-to-LFE triggering is not significant. In other words, the cascade model (e.g., Ellsworth & Bulut, 2018), in which the rupture of one asperity leads to the rupture of the next asperity, cannot explain LFE activity. This is also evident from the fact that intense LFE bursts do not decay gradually but rather stop abruptly. My result places constraints on future physical modeling of LFE activity.
Statistical seismicity models are useful for quantifying seismicity characteristics. For the activity of ordinary fast earthquakes, the epidemic-type aftershock-sequence (ETAS) model is widely used (e.g., Ogata, 1988). However, there is still no standard statistical model for slow earthquake activity. Statistical modeling of low-frequency earthquakes (LFEs), a type of slow earthquake, has begun recently. Lengliné et al. (2017) and Tan & Marsan (2020) proposed statistical models similar to the ETAS model (L model), in which the LFE occurrence rate is expressed as the summation of a stationary background rate and the effect of each LFE inducing future LFEs. Ide & Nomura (2022) proposed a statistical model (IN model) for tectonic tremors, a swarm of LFEs. Their model was based on an approach different from the L model; they described the probability distribution of interevent times using log-normal and Brownian passage time distributions. Furthermore, their model does not explicitly include the effect of event-to-event triggering.
Identifying the best statistical model for LFE activity is important for elucidating the mechanisms governing LFE activity. However, the existing statistical models have never been compared. Here, I applied the existing models to M 0.6 or larger LFEs in the Nankai Trough (Kato & Nakagawa, 2020) and examined their performance using Akaike’s information criterion (Akaike 1974). As a result, I found the IN model often outperforms the L and ETAS models. This result indicates that the effect of LFE-to-LFE triggering is not significant. In other words, the cascade model (e.g., Ellsworth & Bulut, 2018), in which the rupture of one asperity leads to the rupture of the next asperity, cannot explain LFE activity. This is also evident from the fact that intense LFE bursts do not decay gradually but rather stop abruptly. My result places constraints on future physical modeling of LFE activity.