11:45 AM - 12:00 PM
[SCG45-23] Global investigation of foreshock acceleration prior to large earthquakes
Keywords:Foreshock, ETAS model
Investigating precursory phenomena of large earthquakes is crucial for evaluating the earthquake predictability. Foreshock activity before large earthquakes has been actively studied. Bouchon et al. (2013) claimed that many large earthquakes at circum-Pacific plate boundaries were preceded by accelerating foreshock activities. Nishikawa & Ide (2018) also reported an accelerating foreshock swarm immediately before a M 6.9 interplate earthquake in the Japan Trench subduction zone in 2008. Similar foreshock acceleration has been reported by numerous laboratory experiments and numerical simulations (e.g., Marty et al., 2023; Ito & Kaneko, 2023). However, Bouchon et al. (2013) have been criticized for inadequately accounting for the effects of earthquake clustering (Felzer et al., 2015). Furthermore, Nishikawa & Ide (2018) analyzed only a few large earthquakes, leaving the general prevalence of accelerating foreshock swarms unclear.
Here, we investigated foreshock activity before large earthquakes globally using the epidemic-type aftershock-sequence (ETAS) model (Ogata, 1988), a standard statistical model of seismicity. Based on previous numerical simulations and rock experiments (e.g., McLaskey, 2019), we incorporated a new term representing the foreshock swarm acceleration into the ETAS model. This term, similar to the inverse Omori law, describes a power-law acceleration of the seismicity rate leading to a mainshock (= L/(Teq - t + d)q), where t is time, Teq is the mainshock origin time, and L, d, and q are new model parameters. We applied the model to seismicity (M 4.5 or larger and within 100 km of a large earthquake) preceding 358 M 6.5 or larger earthquakes from 2000 to 2024 in the ANSS earthquake catalog.
Our results show that approximately 2% of the large earthquakes exhibited significant foreshock acceleration. For example, the M 6.9 earthquake analyzed by Nishikawa & Ide (2018) and a M 7.3 interplate earthquake in the Vanuatu subduction zone in 2008 showed notable foreshock acceleration. The M 6.9 earthquake in Japan was preceded by 15 foreshocks within three days (L = 0.42, d = 0.042 days, and q = 2.1), and the M 7.3 earthquake in Vanuatu was preceded by 10 foreshocks within three days (L = 0.73, d = 0.061 days, and q = 1.7).
In addition to the above analysis, we conducted the same analysis on seismic activity preceding randomly selected earthquakes to examine whether the acceleration of seismic activity is a phenomenon specific to large earthquakes. Furthermore, to evaluate the impact of the earthquake cascading process on our results, we applied the same analysis to synthetic earthquake catalogs generated by the ETAS model. As a result, we found that even for randomly selected earthquakes, a significant acceleration phenomenon was observed in approximately 2% of cases. This indicates that there is no significant difference between seismic activity preceding randomly selected earthquakes and that preceding large earthquakes. In contrast, in the ETAS synthetic catalogs, the proportion of cases exhibiting accelerated seismic activity was significantly lower than 2% (~0.3%). We also performed a similar analysis using the Japan Meteorological Agency catalog, a more complete local catalog in Japan, and obtained similar results.
The analysis of randomly selected earthquakes suggests that seismicity acceleration is not unique to large earthquakes, while the analysis of the ETAS synthetic catalogs indicates that the apparent acceleration is unlikely to result from the earthquake cascading process. Based on these findings, we propose that an aseismic process, such as slow slip, may induce seismicity acceleration. However, this phenomenon is not exclusive to the period preceding large earthquakes. Our study demonstrates that foreshock acceleration, which has been considered one of the most plausible precursors to large earthquakes based on observational studies, experiments, and simulations, has only limited predictive power.
Here, we investigated foreshock activity before large earthquakes globally using the epidemic-type aftershock-sequence (ETAS) model (Ogata, 1988), a standard statistical model of seismicity. Based on previous numerical simulations and rock experiments (e.g., McLaskey, 2019), we incorporated a new term representing the foreshock swarm acceleration into the ETAS model. This term, similar to the inverse Omori law, describes a power-law acceleration of the seismicity rate leading to a mainshock (= L/(Teq - t + d)q), where t is time, Teq is the mainshock origin time, and L, d, and q are new model parameters. We applied the model to seismicity (M 4.5 or larger and within 100 km of a large earthquake) preceding 358 M 6.5 or larger earthquakes from 2000 to 2024 in the ANSS earthquake catalog.
Our results show that approximately 2% of the large earthquakes exhibited significant foreshock acceleration. For example, the M 6.9 earthquake analyzed by Nishikawa & Ide (2018) and a M 7.3 interplate earthquake in the Vanuatu subduction zone in 2008 showed notable foreshock acceleration. The M 6.9 earthquake in Japan was preceded by 15 foreshocks within three days (L = 0.42, d = 0.042 days, and q = 2.1), and the M 7.3 earthquake in Vanuatu was preceded by 10 foreshocks within three days (L = 0.73, d = 0.061 days, and q = 1.7).
In addition to the above analysis, we conducted the same analysis on seismic activity preceding randomly selected earthquakes to examine whether the acceleration of seismic activity is a phenomenon specific to large earthquakes. Furthermore, to evaluate the impact of the earthquake cascading process on our results, we applied the same analysis to synthetic earthquake catalogs generated by the ETAS model. As a result, we found that even for randomly selected earthquakes, a significant acceleration phenomenon was observed in approximately 2% of cases. This indicates that there is no significant difference between seismic activity preceding randomly selected earthquakes and that preceding large earthquakes. In contrast, in the ETAS synthetic catalogs, the proportion of cases exhibiting accelerated seismic activity was significantly lower than 2% (~0.3%). We also performed a similar analysis using the Japan Meteorological Agency catalog, a more complete local catalog in Japan, and obtained similar results.
The analysis of randomly selected earthquakes suggests that seismicity acceleration is not unique to large earthquakes, while the analysis of the ETAS synthetic catalogs indicates that the apparent acceleration is unlikely to result from the earthquake cascading process. Based on these findings, we propose that an aseismic process, such as slow slip, may induce seismicity acceleration. However, this phenomenon is not exclusive to the period preceding large earthquakes. Our study demonstrates that foreshock acceleration, which has been considered one of the most plausible precursors to large earthquakes based on observational studies, experiments, and simulations, has only limited predictive power.