5:15 PM - 6:30 PM
[MIS34-P14] Change in b-values following the three M6 or 7-class earthquakes in the 2016 Kumamoto earthquake sequence
Keywords:b-value, stress, detectability of earthquake, Bayesian estimation, aftershock sequence, Gutenberg-Richter law
This study investigates whether or not b-values of the Gutenberg-Richter law change following the three events (M6.5 on 14 April, M6.4 on 15 April, and M7.3 on 16 April) in the 2016 Kumamoto earthquake sequence. To examine the statistical significance of the change, a Bayesian approach developed and used in Iwata [2008, 2013, GJI; 2016, Pageoph] was applied to the sequence. This approach introduces a statistical model of a magnitude-frequency distribution covering the entire magnitude range proposed by Ogata & Katsura [1993, GJI]. The proposed distribution contains a parameter representing the quality of detection capability of earthquakes. With the aforementioned Bayesian approach, we can estimate a b-value with the consideration of the temporal change in the quality of detectability; all observed events are used in the estimation while we discard events of which magnitude is less than a cut-off magnitude in an conventional method to find a b-value. This is the advantage of the Bayesian approach to examine the significance of the change in b-values.
Data analyzed in this study was taken from the JMA catalog (last accessed on 11 May 2016). A study area was chosen to cover the aftershock area just before the M7.3 event and the epicenter of the M7.3 event, and four periods were considered: from 1 August 2015 to the occurrence of the M6.5 event (Period 0), between the occurrences of the M6.5 and M6.4 events (Period 1), between the occurrences of the M6.4 and M7.3 events (Period 2), and after the M7.3 event (Period 3). For each of the four periods, the b-value and temporal change in detection capability (i.e, the parameter of the magnitude-frequency distribution proposed by Ogata & Katsura [1993]) were estimated simultaneously. Cases with the constraint of a common b-value over successive periods (e.g., b-values of Periods 1, 2, and 3 are the same while the one in Period 0 is different.) were also considered. In total, there were eight cases (models), depending on which periods have a common b-value, and the best model was selected from the eight with Akaike's Bayesian Information Criterion (ABIC).
The best model is the case where Periods 0, 1, and 2 have a common b-value of 0.736 and only Period 3 has a different value of 0.941; the b-value varies at the timing of the M7.3 event whereas it does not change with the M6.5 and M6.4 events. The difference of ABIC values between this model and the one where the b-values over the four periods are the same is 7.9, which suggests the statistical significance of the best model. Remarkable increase in the b-value following a major/megathrust earthquake has been found [e.g., Tormann et al., 2015, Nature Geoscience]. This increase will correspond to decrease of stress due to a large slip, because it has been suggested that the b-value is correlated inversely with stress [e.g., Scholz, 2015, GRL]. From this viewpoint, the result of this study suggests that the high stress in the focal area of the Kumamoto sequence was not decreased by the M6.5 and M6.4 events but the M7.3 event has released it.
Data analyzed in this study was taken from the JMA catalog (last accessed on 11 May 2016). A study area was chosen to cover the aftershock area just before the M7.3 event and the epicenter of the M7.3 event, and four periods were considered: from 1 August 2015 to the occurrence of the M6.5 event (Period 0), between the occurrences of the M6.5 and M6.4 events (Period 1), between the occurrences of the M6.4 and M7.3 events (Period 2), and after the M7.3 event (Period 3). For each of the four periods, the b-value and temporal change in detection capability (i.e, the parameter of the magnitude-frequency distribution proposed by Ogata & Katsura [1993]) were estimated simultaneously. Cases with the constraint of a common b-value over successive periods (e.g., b-values of Periods 1, 2, and 3 are the same while the one in Period 0 is different.) were also considered. In total, there were eight cases (models), depending on which periods have a common b-value, and the best model was selected from the eight with Akaike's Bayesian Information Criterion (ABIC).
The best model is the case where Periods 0, 1, and 2 have a common b-value of 0.736 and only Period 3 has a different value of 0.941; the b-value varies at the timing of the M7.3 event whereas it does not change with the M6.5 and M6.4 events. The difference of ABIC values between this model and the one where the b-values over the four periods are the same is 7.9, which suggests the statistical significance of the best model. Remarkable increase in the b-value following a major/megathrust earthquake has been found [e.g., Tormann et al., 2015, Nature Geoscience]. This increase will correspond to decrease of stress due to a large slip, because it has been suggested that the b-value is correlated inversely with stress [e.g., Scholz, 2015, GRL]. From this viewpoint, the result of this study suggests that the high stress in the focal area of the Kumamoto sequence was not decreased by the M6.5 and M6.4 events but the M7.3 event has released it.