A growing body of evidence indicates that many earthquakes are triggered by static and dynamic stress changes following large earthquakes. However, it has been difficult to differentiate the roles of static and dynamic stress transfer on earthquake triggering. Here we focus on two adjacent seismic clusters aligned E-W beneath the Hida Mountain Range, central Japan, where spatially uniform static and dynamic stress changes resulted from the Tohoku-Oki earthquake. The rate of seismicity in the western cluster, detected by the matched filter technique, showed a slight increase after the mainshock, which can be explained by static stress transfer. In contrast, seismicity in the eastern cluster showed a marked increase due to dynamic stress change. The difference in the number of triggered events between these clusters indicates that dynamic stress change is more important than the static one in terms of remote triggering, provided that large-sized potential earthquake-nucleation patches are close to failure.
 The large extensional stress perturbations associated with the 11 March 2011 Tohoku-Oki earthquake (magnitude (M) 9.0) have resulted in a widespread increase in seismicity across NE and central Japan [e.g., Kato et al., 2011; Toda et al., 2011]. Many examples of induced seismicity within the rupture length of the Tohoku-Oki earthquake can be explained by static stress transfer, induced by the mainshock rupture, on the nodal planes of aftershock focal mechanisms [e.g., Ishibe et al., 2011]. On the other hand, seismicity triggered far beyond the rupture length of a large earthquake (i.e., at remote sites located far from the large earthquake) may be governed by dynamic stress change associated with the propagation of surface waves [e.g., Gomberg et al., 2004]. For example, Miyazawa  reported that a front of dynamically triggered seismicity propagated over Japan toward the southwest following the Tohoku-Oki earthquake, with a traveling speed equal to that of surface waves. However, it is difficult to differentiate the influences of static and dynamic stress transfer on earthquake triggering [e.g., Stein, 1999; Felzer and Brodsky, 2005].
 Here we focus on seismicity at intermediate distances from the Tohoku-Oki earthquake rupture, where both static and dynamic stress transfers are effective in triggering seismicity. Two adjacent seismic clusters, separated by about 8 km, occurred in the Hida Mountain Range, central Japan and were recorded before and after the Tohoku-Oki earthquake (Figure 1). This region is a site of active volcanoes and is located ~500 km from the Tohoku-Oki rupture. The JMA (Japan Meteorological Agency) catalogue shows that a western cluster broke out 12 days prior to the Tohoku-Oki earthquake (on 27 February 2011) accompanied by several moderate events (M3.6–5.5; red circles in Figure 1a). Although the rate of seismicity in this cluster has decayed gradually with time, a period of quiescence occurred immediately after the Tohoku-Oki earthquake (gray area in Figure 1c). Subsequently, the rate of seismicity showed a gradual increase (within a few days) to its previous level. In contrast, on the eastern side of the studied region (blue circles in Figure 1b), an intensive seismic sequence started shortly after the Tohoku-Oki earthquake.
 Given that spatially uniform static and dynamic stress changes were induced in the studied area by the Tohoku-Oki earthquake, it is interesting to note the marked differences between the two clusters in terms of temporal changes in seismicity rate. Therefore, a detailed analysis of the seismicity would provide us with a unique opportunity to differentiate the effects of static and dynamic stress transfer on earthquake triggering. To improve on the precision of seismicity in the JMA catalogue, we applied a matched filter technique to continuous seismograms recorded by a dense and highly sensitive seismic network. We evaluated the completeness magnitude threshold (Mc) of the newly detected events, and we discuss the temporal changes of seismicity rate in the Hida Mountain Range before and after the Tohoku-Oki earthquake, as well as assess the relative contributions of static and dynamic stress transfers to earthquake triggering.
2 Data and Methods
 Many seismic events in existing catalogues are missing after large earthquakes because small-magnitude earthquakes are masked by the overlapping arrivals of large-amplitude waves from aftershocks [e.g., Felzer and Brodsky, 2005; Peng and Zhao, 2009]. In the present case, to recover missing events between 27 February and 17 March 2011 (19 days), we applied a matched filter technique to the continuous waveforms for this period [e.g., Peng and Zhao, 2009; Aso et al., 2011; Kato et al., 2012]. We analyzed the continuous three-component velocity seismograms recorded at 12 seismic stations in and around the target region (Figure 2), operated by the National Research Institute for Earth Science and Disaster Prevention (NIED) of Japan and the JMA. We used all 1558 available earthquakes listed in the JMA catalogue between 27 February and 17 March (Figure 1) as template events and searched the continuous records for events that strongly resembled these template events.
 For each template event at each station, we used 4 s of the waveform, starting from 1.5 s before the S wave arrival time, as computed using a one-dimensional velocity model proposed by the JMA. After applying a two-way 4–8 Hz Butterworth filter to the data, we downsampled the data from a sampling frequency of 100 Hz to 50 Hz. The use of this relatively high-frequency range helped to reduce the noise excited by long-period waves associated with large aftershocks, enabling the detection of local seismic events. We calculated the correlation coefficient between a template event waveform and a target waveform as a function of time, shifting the window by increments of 0.02 s through the continuous waveforms. At each point, the correlation coefficient was calculated on every component at every station and was averaged throughout. The same procedure was repeated for each template event.
 We used the mean correlation coefficient as the detection statistic. To define a detection threshold, we used the median absolute deviation (MAD) of the mean correlation coefficients for each event and each day. We set a threshold for a positive detection at 9.0 times MAD. This level was chosen to suppress spurious detections while retaining as many legitimate detections as possible by visual inspection of detected events. When the mean correlation peak exceeded this threshold level, it was labeled as a statistically significant positive detection. If the detected events have similar waveforms at multiple stations as compared with the template event, their hypocenter locations must be similar or identical. We assigned the location of the detected event to that of the template event. For multiple detections in each ±4 s window, we used the location of the template event with the highest mean correlation coefficient. After removing multiple detections, the matched filter technique identified 15,932 events, approximately 10 times the number in the JMA catalogue (1558) for the same period (Figure S1 in the supporting information).
 We assigned the time associated with the highest mean correlation coefficient to the origin time of the detected event by subtracting the computed S wave arrival time. Finally, we computed the magnitude of the detected event based on the median value of the maximum amplitude ratios for all channels between the template and detected events, assuming that a tenfold increase in amplitude corresponds to a one-unit increase in magnitude [Peng and Zhao, 2009].
 Figure 2 shows an example of detections for two small-magnitude events on 13 March 2011, over a period of 13:17:20 JST (UT + 9 h). One of the detected events (M0.2) was matched with template event #592 located in the western cluster (red lines in Figure 2), and the other event (M–0.2) was matched with template event #1466 in the eastern cluster (blue lines). The relative arrival times of S waves at each station differ between the two detected events. Therefore, the matched filter technique enables us to reliably assign events to either the western or the eastern cluster. Of note, both of these detected events are not listed in the JMA catalogue.
 Figure 3a shows a space-time diagram of the newly detected events. In contrast to the JMA catalogue (Figure 1c), the seismicity in the western cluster continued after the Tohoku-Oki earthquake, with no period of quiescence. Figure 3b shows an example of a newly detected event for template event #1194 (western cluster) on 11 March 2011, at 20:14:00 JST. The propagation of S waves from this template to the seismic network is clearly recognized. Including this event, we identified a total of around 35 events (magnitudes from −0.2 to 1.2) in the western cluster during a time period when the JMA catalogue showed no seismic events (gray area in Figure 1c). We also visually checked these events to ensure their reliability.
 In the eastern cluster, the intensity of the seismicity triggered immediately after the Tohoku-Oki earthquake was greater than that in the JMA catalogue (Figures 1c and 3a). We found that the initial detected event in the eastern cluster (M4.0) following the Tohoku-Oki earthquake was triggered during the passage of surface waves (Figures 4d and S2) [Miyazawa, 2011], of which long-period signals are recovered by a time domain recursive filter [Maeda et al., 2011].
 We calculated temporal changes in the Mc of the newly constructed catalogue using the ZMAP code [Wiemer and Wyss, 2000] (Figure 3c). The value of Mc remained around −0.3 before the Tohoku-Oki earthquake but jumped to around 0.5 immediately following this event and then gradually decreased to the previous level. Although we recovered a large number of missing events by applying the matched filter technique, this was insufficient to keep Mc values at around −0.3 during the period immediately following the Tohoku-Oki earthquake.
 Considering a maximum value in the sequence of Mc = 0.5, we plotted the temporal changes in the cumulative number of earthquakes with a magnitude greater than 0.5 (Figures 4a and 4b). In the western cluster, the cumulative number showed no abrupt change coincident with the Tohoku-Oki earthquake, but it showed a slight increase over the 3 days following the event. In contrast, the cumulative number of earthquakes in the eastern cluster showed a marked increase coincident with the Tohoku-Oki earthquake and thereafter a gradual slowing in the rate of increase over the 3 days following the event. Of note, the number of extra events in the eastern cluster (following the Tohoku-Oki earthquake) is approximately 30 times greater than that in the western cluster, despite the fact that spatially uniform static and dynamic stress changes are likely to have been imposed on the studied area by the Tohoku-Oki earthquake.
4 Discussion and Conclusions
 We applied the matched filter technique to two seismic clusters in the Hida Mountain Range and recovered a large number of events that are missing from the JMA catalogue. The rate of seismicity in the western cluster showed a slight increase in the 3 days following the Tohoku-Oki earthquake (compared with a period of quiescence in the JMA catalogue). Therefore, we conclude that the quiescence apparent in the JMA catalogue (Figure 1c) is an artificial feature that reflects the drastic increase in Mc in the catalogue. This abrupt reduction in seismicity detection may have occurred due to overlapping arrivals of large-amplitude waves from numerous aftershocks induced by the Tohoku-Oki earthquake. It is therefore necessary to apply caution when evaluating sequences that show a reduction in seismicity rate immediately following a large earthquake [e.g., Felzer and Brodsky, 2005]. Reductions in seismicity rate have been used to identify stress shadows where seismicity is suppressed by a decrease in static stress following a mainshock [e.g., Ma et al., 2005]. Toda et al.  examined the possible development of stress shadows during thrust faulting in inland Tohoku, identifying several areas with reductions in seismicity rate. However, it is necessary to reevaluate periods of reduced seismicity rate by carefully assessing temporal variations in Mc or by reconstructing the earthquake catalogue by applying the matched filter technique, as in the present study, if the number of earthquakes is insufficient to reevaluate the rate change.
 The representative focal mechanisms of the studied area indicate thrust and strike-slip faulting with the P axis aligned NW-SE (Figures 1a and 1b), which is typical of the stress field in this region [e.g., Terakawa and Matsu'ura, 2010]. The focal mechanisms in the eastern cluster were obtained using P wave first-motion polarities and the amplitude ratio of S and P waves [Hardebeck and Shearer, 2003]. We calculated the static Coulomb stress changes on both nodal planes of the thrust and strike-slip faulting earthquakes caused by rupture of the Tohoku-Oki earthquake, assuming an effective friction coefficient of 0.4 (Figure 4c). We assumed an elastic half-space model with a shear modulus of 30 GPa and Poisson's ratio of 0.25, giving the coseismic slip distribution along the plate boundary inverted from geodetic data (Figure S3 and Text S1). The static stress changes ΔCFF show positive values ranging from 0.01 to 0.03 MPa, which slightly exceed a threshold (~0.01 MPa) for the effective triggering of seismic events [e.g., Stein, 1999]. Based on the laboratory-based rate/state constitutive law, the seismicity rate immediately increases by a factor of exp(ΔCFF/Aσ) due to a stress step with ΔCFF, where Aσ is the constitutive parameter multiplied by the total normal stress acting on the fault [e.g., Dieterich, 1994; Toda and Stein, 2003]. Previous studies have suggested that Aσ ranges from 0.01 to 0.04 MPa in some source regions of shallow inland earthquakes and in seismic swarms associated with dyke intrusion in Japan [e.g., Toda et al., 2002; Toda and Stein, 2003]. The ΔCFF of ~0.02 MPa induced by the Tohoku-Oki earthquake could raise the seismicity rate by a factor of 1.6–7.4. This relatively small increase in the seismicity rate is roughly consistent with the slightly increased seismicity observed in the western cluster. Thus, the static stress transfer to the Hida Mountain Range is a plausible triggering mechanism in the western cluster.
 In contrast, the rate of triggered events in the eastern cluster is remarkably higher than that in the western cluster (Figure 4). The seismicity rate in the eastern cluster increased more than 104-fold, relative to the measured background rate (~0.03 day−1). This marked change in the seismicity rate is difficult to explain solely in terms of the small static stress step of ~0.02 MPa and the conventional Aσ values mentioned above. Of note, the initial detected event in the eastern cluster following the Tohoku-Oki earthquake was triggered during the passage of surface waves (Figures 4d and S2). Based on a calculation of the maximum magnitude of eigenvalues of the dynamic/static stress tensor in the Hida Mountain Range, the amplitude of dynamic stress change associated with surface wave propagation is approximately 10 times as large as that of static change (distance range ~500 km) [Miyazawa, 2011]. We therefore consider that the dynamic stress is a plausible triggering mechanism in the eastern cluster.
 The small increase in seismicity rate in the western cluster (Figure 4a) may indicate that the intensive activity prior to the Tohoku-Oki earthquake had already expended the seismic energy at the sites of potential earthquake-nucleation patches (Figure 3a). In such a case, the dynamic stress change would have been unlikely to trigger abundant events at remote distances from the source rupture. In contrast, the eastern cluster may have been the site of many large-sized potential earthquake-nucleation patches close to a state of failure because there had been no significant activity in the area of the eastern cluster during the 7–8 years prior to the Tohoku-Oki earthquake. Giving the occurrence of dynamic triggering in an area close to active volcanoes, it is likely that elevated fluid pressures resulting from dynamic shaking would have reduced the frictional strength of potential earthquake-nucleation patches and promoted intensive failures [e.g., Gomberg et al., 2004]. Preparedness for failure in earthquake patches is an important prerequisite for remote triggering by surface waves.
 We thank NIED and JMA for permitting the use of waveform data and the JMA catalogue. We acknowledge the Geospatial Information Authority of Japan for providing GPS data.
 The Editor thanks two anonymous reviewers for their assistance in evaluating this paper.