14:45 〜 15:00
[STT40-04] Classes of low-frequency earthquakes based on inter-time distribution reveal a precursor event for the 2011 Great Tohoku Earthquake
キーワード：低周波地震、クラスター分析、東北地方太平洋沖地震
Detecting and identifying precursors to large earthquakes is essential in order to mitigate the devastating damage of earthquakes. In relation to the 11 March 2011 Great Tohoku Earthquake (Tohoku-oki EQ, hereafter) Mw 9.0, several precursors were reported retrospectively. With a long-term perspective, seismic quiescence was observed near the epicenter 23 years prior to Tohoku-oki EQ. Also, strong correlations were observed between tidally induced stresses and earthquake occurrence times near the epicenter ten years before it occurred. In the short term, it was observed that earthquake activities moved toward the epicenter one month before. Regarding non-seismic phenomena, positive anomalies of ionospheric total electron content (TEC) were detected around the focal region 40 minutes before. All these precursors are important not only for prediction of large earthquakes, but also for a better understanding of their underlying mechanisms.
In the present study, we examine the relevance of slow earthquakes to Tohoku-oki EQ, focusing in particular on low-frequency earthquakes (LFE) that occurred in northern Japan, several hundred kilometers from the epicenter of the Tohoku-oki EQ. As is well-known, it can be observed that LFEs occurred in specific areas in a cluster-like manner. The underlying physical mechanism of LFE in this region is not fully understood, but it is inferred that LFEs in this region are related to aqueous fluids supplied by the subducted slab. It is speculated that a sudden movement of such fluids near the Moho discontinuity may cause LFEs. With regard to the Tohoku-oki EQ, it is reported that in a spatial analysis, LFEs became generally less active after the Tohoku-oki EQ, which contrasts with activation of conventional EQs in this region. However, such a spatial analysis does not reveal by itself whether such a change in seismicity occurred before the Tohoku-oki EQ.
To address this point, we employed a different approach to analyze LFEs, focusing on the inter-time distribution, the time between consecutive events, rather than on their spatial distribution. For conventional EQs, the distribution of inter-time has recently attracted attention because the inter-time with a cutoff magnitude (from 5 to 6.5) seems to entail a universal law in the form of a generalized gamma distribution. With respect to the Tohoku-oki EQ, the parameters of such a generalized gamma distribution in the Tohoku region changed. In the context of LFEs, it is not obvious what distribution the inter-time follows. Nonetheless, it is of great interest to examine whether the distribution of inter-time changed with respect to the Tohoku-oki EQ. If we can identify exactly when the change of distribution occurred, this provides useful information on a possible precursor for the Tohoku-oki EQ. To shed light on this, we focused on the inter-time of LFEs with an epicenter in close proximity (less than 10km) to the preceding LFE, which captures spatio-temporal correlations of consecutive LFEs. To identify homogeneous distributions, we performed cluster analyses on logarithms of LFE inter-time in a data-driven manner, which identified four homogeneous classes. Remarkably, examination of the relevance of these classes to the Tohoku-oki EQ suggests that the activity of LFEs in the shortest inter-time class (median 24 seconds) diminished significantly at least three months before the Tohoku-oki EQ, and that complete quiescence occurred in this class 30 days before the Tohoku-oki EQ (p-value=0.00014). In contrast with LFEs, conventional earthquakes did not become inactive during the same period. In conclusion, it is implied that this class together with a similar class of volcanic tremor may have served as a precursor of the Tohoku-oki EQ.
In the present study, we examine the relevance of slow earthquakes to Tohoku-oki EQ, focusing in particular on low-frequency earthquakes (LFE) that occurred in northern Japan, several hundred kilometers from the epicenter of the Tohoku-oki EQ. As is well-known, it can be observed that LFEs occurred in specific areas in a cluster-like manner. The underlying physical mechanism of LFE in this region is not fully understood, but it is inferred that LFEs in this region are related to aqueous fluids supplied by the subducted slab. It is speculated that a sudden movement of such fluids near the Moho discontinuity may cause LFEs. With regard to the Tohoku-oki EQ, it is reported that in a spatial analysis, LFEs became generally less active after the Tohoku-oki EQ, which contrasts with activation of conventional EQs in this region. However, such a spatial analysis does not reveal by itself whether such a change in seismicity occurred before the Tohoku-oki EQ.
To address this point, we employed a different approach to analyze LFEs, focusing on the inter-time distribution, the time between consecutive events, rather than on their spatial distribution. For conventional EQs, the distribution of inter-time has recently attracted attention because the inter-time with a cutoff magnitude (from 5 to 6.5) seems to entail a universal law in the form of a generalized gamma distribution. With respect to the Tohoku-oki EQ, the parameters of such a generalized gamma distribution in the Tohoku region changed. In the context of LFEs, it is not obvious what distribution the inter-time follows. Nonetheless, it is of great interest to examine whether the distribution of inter-time changed with respect to the Tohoku-oki EQ. If we can identify exactly when the change of distribution occurred, this provides useful information on a possible precursor for the Tohoku-oki EQ. To shed light on this, we focused on the inter-time of LFEs with an epicenter in close proximity (less than 10km) to the preceding LFE, which captures spatio-temporal correlations of consecutive LFEs. To identify homogeneous distributions, we performed cluster analyses on logarithms of LFE inter-time in a data-driven manner, which identified four homogeneous classes. Remarkably, examination of the relevance of these classes to the Tohoku-oki EQ suggests that the activity of LFEs in the shortest inter-time class (median 24 seconds) diminished significantly at least three months before the Tohoku-oki EQ, and that complete quiescence occurred in this class 30 days before the Tohoku-oki EQ (p-value=0.00014). In contrast with LFEs, conventional earthquakes did not become inactive during the same period. In conclusion, it is implied that this class together with a similar class of volcanic tremor may have served as a precursor of the Tohoku-oki EQ.