The southern limit of the 2011 Tohoku earthquake is considered to be located around off Ibaraki. However, it is not well constrained how far south the large slip extended. To give better constraints, we investigated seismicity including small earthquakes before and after the Tohoku earthquake around off Ibaraki using dense ocean bottom seismic array data. We automatically identified epicenters by backprojecting semblance values resulting in a considerable increase in the number of detected events compared with those listed in the catalog based on onshore observation. The results revealed a couple of seismicity-activated region. The largest aftershock also occurred ~30 min after the main shock. Our detailed results suggest that this highly activated seismicity was initiated by the largest aftershock instead of the main shock. It, then, suggests that the large coseismic slip zone of the Tohoku earthquake may not have extended off Ibaraki.
- There are some coincidences between seismicity and tectonic features
- Tectonic features are likely major factors to control rupture propagation
- Marine data in the study area enable discussions with fine resolution
A number of studies on the rupture process of the 2011 Mw 9.0 Tohoku earthquake have been conducted by applying inversion analysis [e.g., Fujii et al., 2011; Ide et al., 2011; Yokota et al., 2011; Iinuma et al., 2012] and by applying backprojection analysis [e.g., Honda et al., 2011; Ishii, 2011] to various geophysical data. The source area extended about 500 km and 200 km in along-strike and along-dip directions, respectively. Its southern limit is considered to be located around off Ibaraki [e.g., Yagi and Fukahata, 2011; Kato and Igarashi, 2012] (Figure 1). However, the southern limit of the slip zone has not been well constrained by both onshore seismic and geodetic observations due to a large distance offshore from the network. Furthermore, due to an intensively high-occurrence rate of earthquakes, the detection level of aftershock magnitudes remained high (M ~ 2.5) over months after the main shock, and the JMA (Japan Meteorological Agency) hypocenter catalog was considered highly incomplete. Understanding seismicity after the main shock including small earthquakes around the source area provides essential information to better define the limit of the rupture area.
Mochizuki et al.  reported a tectonic feature that appears coincident with the southern limit of the main shock slip zone. They revealed a subducting seamount of about 3 km height at 10 km depth on the Pacific Plate and a trace of multiple subducted seamounts by seismic survey (Figure 1). Kubo et al.  and Honda et al.  proposed that the subducting seamount played an important role in rupture termination of the Mw 7.8 largest aftershock of the Tohoku earthquake by estimating its rupture process from an inversion analysis and from a backprojection analysis, respectively. Although theoretical [e.g., Cloos, 1992; Scholz and Small, 1997; Wang and Bilek, 2011] and computational [e.g., Yang et al., 2013] studies investigated the effect of subduction of seamounts on plate coupling and earthquake rupture propagation, it still remains controversial.
On the basis of these backgrounds, we investigated spatiotemporal variation of seismicity before and after the Tohoku earthquake around off Ibaraki using data from ocean bottom seismometers (OBSs). Furthermore, we discuss the southern limit of the Tohoku earthquake based on the resulted seismicity.
2 OBS Array Data
We use velocity waveforms of 1 Hz long-term OBSs which were densely deployed over the frontal region of the subducting seamount [Mochizuki et al., 2008] at station intervals of about 6 km (Figure 1). The data sampling rate is 200 Hz, and the total observation period was about a year including the occurrence of the 2011 Tohoku earthquake. Twenty-four OBSs were first deployed in October 2010, 10 OBSs were added to the array in February 2011, and 31 of them were successfully recovered by September 2011. This OBS array has a couple of advantages. The array is closest to the study area so that we can determine hypocenters with good accuracy. The other advantage is that the array is very dense with 31 stations in a region of the dimensions of about 20 km in width and 30 km in length which are comparable to the station intervals of the other marine aftershock observations covering the rupture area of the main shock [Shinohara et al., 2011, 2012]. This density contributes to reduce effects from the large ambient noise after a series of large earthquakes by stacking observed waveforms.
Conventional event identification by manually picking P wave and/or S wave onsets is not manageable well in the target period due to intensively high activity of aftershocks of the Tohoku earthquake (Figure S1 in the supporting information). Additionally, such manual on-set picking inherently bears problems in repeatability and reproducibility. We applied a semblance analysis [e.g., Neidell and Taner, 1971] which estimates source of energy release by stacking OBS waveform data. In this method, visual identification of events and manual picking of P and S wave onsets are not required. Even though OBS waveforms are often affected by the large ambient noise, it is possible to overcome this problem by stacking waveforms observed by a highly dense array.
- We detected events by applying a Short-Term Average to Long-Term Average ratio (STA/LTA) algorithm [e.g., Allen, 1978; Urabe and Hirata, 1984] to 1–3 Hz band-pass filtered vertical-component OBS seismograms and extracted their waveforms.
- We converted the original event waveforms to envelope waveforms so that the P wave polarity varying about different focal mechanisms is not considered in equation 1.
- We calculated semblance values at each small area. The time window is for 2 s with a 1 s overlap.
By visually comparing time series of waveforms and the corresponding semblance values, we found the noise level of semblance value is 0.2 or smaller. Therefore, we automatically identified events whose semblance values are larger than 0.3, which is 0.1 larger than the above observed noise level, in order to obtain reliable epicenters from backprojected semblance values. Detailed procedures of event identification are described in the supporting information (Text S1 and Figure S2). Here we assume that all earthquakes occurred on the plate interface. In order to evaluate this assumption, we referred to the depth distribution of the earthquakes listed in the JMA catalog whose hypocenters were precisely relocated by the marine aftershock observation [Shinohara et al., 2011, 2012]. The depth distribution shows that about 70% of the earthquakes occurred within a 7 km band on both sides of the plate interface. Therefore, we consider this assumption reasonable.
4.1 Accuracy Evaluation
Before applying the above method to the observed data, we conducted synthetic tests by using a 2 Hz sine wave from a point source with several different sets of signal-to-noise (S/N) ratio and focal depths in order to evaluate the accuracy of resulted epicenters. Thirty epicenters with three different focal depths relative to the plate interface of −5 (in the hanging wall), 0 (on the plate interface), and 5 km (in the slab) are selected as given hypocenters. We backprojected the synthetic waveforms from the selected hypocenters. As the results, points of the maximum semblance value coincide with the original point sources within less than 10 km error when the given epicenters are within a range of about 40 km from the OBS array, regardless of S/N ratio and focal depths (nonmasked region in Figure 2). These results ensure that the method is noise robust due to stacking waveforms and that we can discuss the epicenters with good accuracy for those earthquakes within the 40 km range from the OBS array. However, the resulted epicenters rapidly shifted from their true locations at points beyond 40 km from the array, especially when the hypocenters are off the plate interface (masked region in gray in Figure 2). In consequence of these results, we will specifically discuss the epicenters only within the 40 km range from the array (nonmasked region in Figure 2).
4.2 Changes in Seismicity Before and After the Tohoku Earthquake
For the post-Tohoku period, we successfully determined epicenters of a much large number of events than those listed in the catalog based on onshore observation. The detected number of events was more than twice as many as that in the JMA catalog, while the completeness of the JMA catalog became low in this period.
Figure 3 shows the resulted seismicity in the form of rate of earthquake occurrence. The rate was calculated for a 1.5 km by 1.5 km grid as the number of events per month. It indicates drastic change before and after the occurrence of the Tohoku earthquake. Before the Tohoku earthquake, slight seismic activity is located around the western edge of the subducting seamount (a blue dashed region in Figure 3a). The lineation of this seismicity is in good agreement with the subduction direction of the seamount. By contrast, after the Tohoku earthquake, there are two characteristic distributions with a considerable number of earthquakes. One is outlined as region A in Figure 3b. This small cluster is located between the largest slip zone of the Mw 7.8 aftershock [Kubo et al., 2013] and the western front of the subducting seamount. The other distribution is depicted as region B in Figure 3b. The highly activated seismicity is distributed in the vicinity of the frontal region of the subducting seamount, especially along the northern edge. The highest rate of earthquake occurrence in region B is more than five events per grid (1.5 km × 1.5 km) per month. The region remained seismically active during the post-Tohoku period, whereas only a small number of events occurred in this region before the Tohoku earthquake. It may be the result of changes in the stress field caused either by the main shock of the Tohoku earthquake or by the largest Mw 7.8 earthquake that occurred 30 min after the main shock. In order to verify this hypothesis, we investigated the detailed spatiotemporal transition of the seismicity and qualitatively evaluate the change of the stress field. Hence, we applied the semblance analysis to continuous OBS waveform data without applying the STA/LTA algorithm over two periods (25 min long each); between the end of the main shock rupture and the initiation of the largest aftershock, and after the largest aftershock rupture. Time series of the OBS waveforms and corresponding semblance values are shown in Figure S3. We identified all events during each period from semblance values and conducted visual confirmation through the waveforms. Figure 4 shows the epicenter distribution for both periods. In the former period, the seismic events (blue circles on Figure 4) are sparsely distributed and only five events occurred in region B. In the latter period, on the other hand, about 3 times as many earthquakes (red circles on Figure 4) occurred intensively in region B as those in the previous period. The activity in the north disappeared in this latter period. These results indicate that the highly activated seismicity mainly started after the largest aftershock.
From the results of the seismicity before and after the Tohoku earthquake, we can observe spatial characteristics. These characteristics are not found from the result of the existing earthquake catalog (Figure S4). One characteristic is related to the slip zone of the Mw 7.8 largest aftershock. In Figure 3b, there is a highly activated cluster between the largest slip zone and the western front of the subducting seamount as introduced in the previous section (region A in Figure 3b). In addition, the seismicity within the source area of the largest aftershock appears relatively low. This contrast of the seismicity may reflect the high stress drop due to the occurrence of the largest aftershock and the postseismic stress concentration [e.g., Das and Henry, 2003].
The other most distinguished feature is the highly activated earthquake cluster in the northern edge of the subducting seamount. In the pre-Tohoku period, the seismicity is active in the frontal region of the seamount subduction (Figure 3a). This activity may be the result of stress concentration in front of the subducting seamount. In the post-Tohoku period, on the other hand, the most highly activated region is located along the northern edge of the seamount (region B in Figure 3b). The spatiotemporal transition of seismicity after the main shock shows that this highly activated seismicity was caused by the largest aftershock rather than the main shock (Figure 4). The previous seismic observation conducted in 2005 revealed similar activity [Mochizuki et al., 2008]. These results suggest that a region around the northern edge of the subducting seamount is prone to seismic activity that can be initiated by large earthquakes and that the largest aftershock of the Tohoku earthquake may have triggered such activity. The above results may provide important information on better constraining the southern limit of the main shock rupture area.
On the basis of our results on time series of hypocenter distribution and discussions on change of the state of stress, we propose a scenario on what happened during and after the Tohoku earthquake around its southern limit. First, the main shock rupture was initiated ~250 km to the north from the off Ibaraki region, propagated down to the south, and stopped before the off Ibaraki region. After the main shock rupture, the seismic energy was released sparsely in the off Ibaraki region. Then, the largest aftershock occurred 30 min after the main shock. And finally, after the largest aftershock rupture, the stress field may have been changed and initiated high seismicity around the subducting seamount.
We successfully determined seismic events including small events before and after the 2011 Tohoku earthquake from dense OBS array data. One of the main objectives of this study is to understand the spatiotemporal variation of the seismicity around the southern limit of the Tohoku earthquake. We found the highly activated seismicity in the vicinity of the frontal region of the subducting seamount after the largest aftershock instead of the main shock (Figures 3 and 4). This relationship can be evidence of the rupture termination of the main shock. The results of the seismicity suggest the southern limit of the large coseismic slip zone of the main shock may not have extended off Ibaraki.
We are grateful to Editor A. Newman and three anonymous reviewers for their valuable comments. We thank the Japan Meteorological Agency for allowing us to use the earthquake catalog and the Institute for Research on Earth Evolution of Japan Agency for Marine-Earth Science and Technology for providing us with the crustal structural data. This study was supported by the Ministry of Education, Culture, Sports, Science, and Technology (MEXT) of Japan, under its Observation and Research Program for Prediction of Earthquakes and Volcanic Eruptions. OBS data are available for collaborative studies on request. Generic Mapping Tools [Wessel and Smith, 1998] was used to draw figures.
The Editor thanks three anonymous reviewers for their assistance in evaluating this paper