Delay and Migration of the 2008 Iwate-Miyagi Early Aftershocks, Observed Using High-Resolution Waveform Data

The aftershock activity immediately following a large earthquake is usually not recorded completely in seismic catalogs (e.g., Kagan, 2004). Continuous high-resolution waveforms recorded at stations situated closely to the aftershock distributions were used in several recent studies (e.g., Peng et al., 2006; Enescu et al., 2007; Peng and Zhao, 2009) to help quantify early aftershocks. Recovering these missing early events is important for understanding the physical mechanisms of earthquake triggering, and tracking post-seismic deformation around the mainshock rupture zone. We have applied the matched-filter technique to identify early aftershocks occurred in the first day after the 2008 Iwate-Miyagi earthquake (Mw6.9), northeast Japan. The method (Peng and Zhao, 2009) uses waveforms of existing events as templates to search for similar patterns in the continuous waveform recordings. We selected 5556 template events occurred in the first month after the mainshock, which were recorded by at least six out of ten NIED, Hi-net seismic stations, located relatively close the aftershock distribution. The 3-component template seismograms, as well as the continuous waveform data recorded by the 10 stations, were 2 to 8 Hz band-pass filtered. A 4-s time window centered on the S-wave arrivals is used as waveform template window. We shift the 4-s window through the 1-day continuous waveforms and compute the cross-correlation (CC) coefficient, which we assign to its origin time by subtracting the S-wave arrival time. We stack and average the CC values for all stations and components, and compute the median absolute deviation (MAD) of the mean CC trace for each template event and use 9 times of MAD as the detection threshold (Peng and Zhao, 2009; Shelly et al., 2007). Similar to Peng and Zhao (2009), we have assigned for each detected event the same location as the corresponding template and determine its magnitude based on amplitude calibration. We have detected in this way 9727 events (M > -1.4), which is more than 5 times the number of aftershocks listed in the JMA earthquake catalog in the first day after the mainshock. The spatio-temporal patterns of early aftershocks show the following remarkable features: a) There is a ~100-s delay in the onset of aftershocks, relative to the mainshock occurrence time. To check about this delay, we have scrutinized the continuous waveforms in several high-frequency bands: 2-8, 10-20, 20-30, 30-40 and > 40 Hz and found no clear seismic events within 100 s after the mainshock; b) The aftershocks start occurring slightly north from the mainshock and expand in the next ~20 min. to the whole aftershock area; c) Most of the very early events occur outside the mainshock slip area, which suggests some kind of anti-correlation between aftershocks and mainshock slip.