Differences in rupture process between the 2001 and 2008 repeating earthquakes off Kamaishi

The cluster of repeating earthquakes of M4.8 +/- 0.1, which have occurred on the plate boundary off Kamaishi, NE Japan, is an example supporting the hypothesis of persistent asperities. Matsuzawa et al. (1999) found that the earthquakes had occurred repeatedly in the cluster from 1957 to 1995 with the mean recurrence interval of about 5.5 years. After the 1995 event, the 2001 and 2008 events occurred within the time range of 99% probability predicted by Matsuzawa et al. (1999, 2002). Comparing the rupture area of the 2001 event with that of the 1995 event inferred from waveform inversions, Okada et al. (2003) showed that rupture areas of the two events were almost completely overlapped. In this study, we compared the rupture areas of the 2001 and 2008 events. Firstly, by using the same method as Okada et al. (2003), we estimated the rupture area of the 2008 event and re-estimated those of the 1995 and 2001 events in the frequency range of 1.0-5.0Hz. The result shows that the rupture areas of the three events were almost completely overlapped. Thus, the 2008 event was thought to be caused by the rupture of the same asperity as the 1995 and 2001 events. Next, in order to investigate the slip distributions of repeating earthquakes off Kamaishi in more detail, we estimated those of the 2001 and 2008 events in the frequency range of 1.0-10.0Hz. For the analysis in higher frequency range, we picked up the first arrivals of P- and S-waves more accurately. Since the rupture times of the repeating earthquakes off Kamaishi were estimated to be about 0.3 seconds, we had to read first arrivals with errors less than 0.03 seconds for inferring rupture processes. In order to achieve such accuracy, we picked up onsets very carefully taking into consideration that a double difference (DD) of S-wave arrival times should be a product of DD of respective P-wave arrivals and VP/VS ratio. We estimated the slip distributions of only the 2001 and 2008 events because the data of the 1995 event was so poor qualitatively and quantitatively that we could not apply the technique mentioned above to the analysis of the 1995 event. Roughly speaking, these two slip distributions almost completely overlapped with each other. However, slip amount of the 2001 event was greater than that of the 2008 event to the north of the epicenter while the 2008 event shows larger slip to the east and west of the epicenter. Uchida et al. (2008) estimated the centroids, rupture-area radii, and stress drops of the 2001 and 2008 events and smaller repeating earthquakes around the two events from spectral analyses. Comparison of our result with the result of Uchida et al. (2008) shows that the areas with greater slip in the 2008 event seem to correspond to the locations of the smaller repeating earthquakes. Moreover, these smaller repeating earthquakes were active just before the 2001 event, but they were not active just before the 2008 event. Thus, it is indicated that slip distribution of an earthquake may be controlled by the activity of nearby smaller events just before the occurrence.