By continuing to browse this site you agree to us using cookies as described in About Cookies
Notice: Wiley Online Library will be unavailable on Saturday 7th Oct from 03.00 EDT / 08:00 BST / 12:30 IST / 15.00 SGT to 08.00 EDT / 13.00 BST / 17:30 IST / 20.00 SGT and Sunday 8th Oct from 03.00 EDT / 08:00 BST / 12:30 IST / 15.00 SGT to 06.00 EDT / 11.00 BST / 15:30 IST / 18.00 SGT for essential maintenance. Apologies for the inconvenience.
 Episodic slow slip events have been recognized by means of tilt changes in the western Shikoku area, southwest Japan. The crustal tilt deformation was observed repeatedly with a recurrence interval of approximately six months coincident with the occurrences of major non-volcanic deep tremor activities in this area. Observed tilt changes can be explained by slow slip events occurring around the source area of tremors. In each episode, the source of the slow slip event and tremor migrate simultaneously. The spatial and temporal coincidence of tremors and slow slip events indicates that they both may be coupling phenomena reflecting the stress accumulation process at the subducting plate.
 In southwest Japan, megathrust earthquakes of magnitude greater than 8 occur repeatedly in association with the subduction of the Philippine Sea plate. To better understand the great earthquake cycle, it is important to resolve the subduction process. Recently, Obara  has discovered non-volcanic deep tremors in the fore arc side of the subduction zone in southwest Japan by NIED Hi-net, which is a densely distributed high-sensitivity seismograph network operated by National Research Institute for Earth Science and Disaster Prevention (NIED). The tremors are located at the deeper part of the slip distribution of megathrust earthquakes [Sagiya and Thatcher, 1999], therefore they might reflect a part of the subduction process.
 Such tremor activity has been also revealed in the Cascadia subduction zone correlated temporally and spatially with the periodic slow slip event [Rogers and Dragert, 2003]. The coupled phenomena called ETS (Episodic Tremor and Slip) occur with a recurrence rate from 13 to 16 months as a chatter of plate motion. The source model of the slow slip event detected by GPS monitoring is estimated on the plate interface between depths of 25 and 45 km [Dragert et al., 2001]. Tremors in the Cascadia margin and in southwest Japan have some similarities in their source depth and the tectonic environment with the subducting young plate. Therefore, we expect a possibility of the slow slip along the tremor source area in southwest Japan.
 In this paper, the tremor activity and crustal deformation were investigated. As the result, we succeeded in detecting the slow slip event coincident with the major tremor activity by monitoring the crustal tilt change in the western part of Shikoku, southwest Japan.
2. Observation of Tremor and Tilt Change
 Tremors detected in southwest Japan subduction zone are characterized by long lasting wave-trains of low-frequency components between 1 and 10 Hz, which is lower than that of same sized normal earthquakes. Because of lack of distinct P or S phases in tremor signals, envelopes of filtered seismograms are used to locate by means of a correlation technique [Obara, 2002]. Unlike the frequently observed deep low-frequency micro-earthquakes beneath active volcanoes and faults in Japan [Ohmi and Obara, 2002; Katsumata and Kamaya, 2003], the subduction tremors are extremely large phenomena characterized by long time duration from hours to weeks and a wide source area, over 600 km in length. The source of tremors are not distributed homogeneously in a narrow belt along the strike of the slab but spatially clustered. The time sequence of the tremor activity in the western Shikoku area, where is the most active part on the belt-like distribution of the subduction tremor, is plotted in Figure 1. Minor activities of tremors were frequently detected during two years of 2001 and 2002; however major tremor activities with duration longer than a few days occurred four times with a recurrence interval of approximately 6 months.
 At every Hi-net observatory, a high-sensitivity accelerometer with a wide response range in frequency from 5 Hz to DC component is installed in a borehole sensor capsule accompanied by a set of velocity seismometers. Horizontal component of the sensor is used as the tiltmeter. The effect of the earth tide on the tilt data is removed using Baytap-G [Tamura et al., 1991]. The tilt movement observed at station HIYH, western Shikoku, is plotted together with the time sequence of the tremor activity in this region (Figure 1). In the time period from January 2001 to December 2002, four episodes of the step-like tilt change can be clearly identified because of the coincidence in time to the major tremor activities. The tilt step of up to 0.1 micro radian is too small to be recognized as a meaningful signal by using only the tilt data. The consistency of the tilt change and the tremor activity indicate that these are coupling geophysical phenomena.
Figure 2 shows the detailed pattern of step-like tilt changes and the epicentral distribution of tremors in four episodes. There are two significant features: the tilt steps are not instantaneous but gradually change for a time period of several days; these four tilting episodes can be classified into the winter (episodes of January 2001 and February 2002) and summer (episodes of August 2001 and August 2002) groups. For the winter group, the tilt change in N-S component started three days in advance of the change in E-W component. On the other hand, for the summer group, the tilt change started simultaneously in both components, and the N-S component keeps changing for 1–2 days after the tilt change in E-W component has stopped. To clarify the seasonal difference, tilt vectors for the first and second stage of the tilt step are plotted on bottom panels of Figure 2. It indicates south down at the beginning then changes to southwest down for the winter type. For the summer type, the tilt vector changes from southeast down to south down. The pattern of tilt change is closely related to the migration of the tremor. Actually, the tremor source propagates smoothly, but the migration of the tremor is simply represented by the difference in epicentral distribution between the first and second stage of each episode. The tremor migrates from the northeast to southwest in winter season, and from the southwest to northeast in summer season. Comparing tilt data observed at station HIYH with migrating tremors, the tilt vector indicates south-down and southeast-down movements when tremors are mainly located in northeast and southwest part, respectively. This indicates that the source of the crustal deformation propagates coincident with the migrating tremor temporally and spatially. As shown in Figure 2, the tremor activity usually starts in advance of the beginning of the tilt step.
 Similar step-like tilt changes have also been observed at other Hi-net stations in a wide area with a diameter of more than 200 km. Figure 3 shows the example of tilt changes at some stations for the episode of August 2002. There was no significant change in atmospheric pressure and rainfall during the time period of tilt steps. From August 16, tilt data were slightly contaminated by a major typhoon with great pressure depression and strong wind; however the disturbance is much smaller than tilt steps. This indicates that the tilt step accompanied by tremors is not noise. The tilt step is regarded as a static crustal deformation because the linear trend is parallel before and after the period of the step-like tilt change as shown in Figure 3. The spatial pattern of the tilt vector is distributed systematically. At stations in the northwest side of the tremor source area plotted on the upper half of Figure 3, the ground tilt is down toward the northwest direction but down to the southwest in the southeast stations plotted on the lower half. These crustal tilt deformations were not detected by GPS monitoring system, GEONET managed by Geographical Survey Institute Japan.
3. Source Model of Slow Slip Event
 The source area of the slow slip seems to migrate gradually with propagation of tremors. To simplify the time-dependent phenomena, the slow slip model parameters for the episode of August 2002 were estimated for two time stages from the corrected tilt data as shown in Figure 3; the first stage of four days from August 6 to 9 and the second stage of three days from August 10 to 12. Strike and dip slips were calculated by means of a least square method with Okada's  formula and the fault geometry was optimized by a genetic algorithm. In this process, parameters of the fault geometry for the first stage are estimated with taking into account the plate geometry, and the same parameters were used for the second stage. The observed data can be explained by slow slips with dislocations of 3 cm and 0.7 cm on two successive reverse faults. The horizontal dimension combining two faults is about 100 km (Figure 4) and the total corresponding moment magnitude is approximately Mw6.0. The faults are located just above the dipping seismic zone in the subducting slab. Considering the depth error, the fault plane might be roughly located around the plate interface. The updip limit of the slip corresponds to the source area of tremors and the deeper part of the rupture area of megathrust earthquakes [Sagiya and Thatcher, 1999] which is the coupling zone. The downdip limit of the slow slip event might correspond to the junction of the subducting plate interface and the continental Moho discontinuity according to the guided wave analysis [Ohkura, 2000]. It indicates that the slow slip is considered to take place on the interface between the oceanic crust and the continental lower crust. The displacement expected from the estimated slow slip model is less than 2 mm on the ground surface; therefore it is very difficult to detect the crustal deformation by GPS observation network. The small amount of the observed tilt step brings difficulty to obtain the unique solution; however the simultaneous occurrence of the slow slip event and the deep tremor is the striking observational fact.
4. Discussion and Conclusions
 In the Cascadia subduction zone, the slow slip events occur repeatedly with a period of 13–16 months [Dragert et al., 2001; Miller et al., 2002] and are always accompanied by deep seismic tremors [Rogers and Dragert, 2003]. Characteristics of Cascadia ETS are very similar to those detected in western Shikoku. The coupled phenomena of tremors and slips occur at the deepest portion of the transition zone on the subducting plate boundary in both subduction zones with young and warm plates. These observations indicate that the source area of ETS is usually locked, but sometimes slides with a certain recurrence interval. The difference in the recurrence time between Cascadia and southwest Japan might depend on the difference of conditions on the plate boundary. In the eastern Shikoku area, major tremor activity repeats with a time period of 2 or 3 months. At a station close to the tremor source area, minor tilt changes lasting for a few days are sometimes associated with tremor episodes. This association suggests that the boundary conditions may be different for each individual tremor cluster on the same subducting plate. Both slow slip events detected in Shikoku area and Cascadia margin accompanied by tremor activity usually last for several days to a few weeks. On the other hand, other kinds of slow slip event lasting for much longer durations have been detected in southwest Japan. In the Tokai area, which is located near the eastern edge of the belt-like tremor zone, the slow slip event has occurred in 2000 and still continues for a few years [Ozawa et al., 2002]. In Bungo channel, located near the western edge of the tremor zone, a slow slip event occurred from the middle of 1996 to 1997 detected by GPS monitoring [Hirose et al., 1999]. The source depth of slow slip events with durations of a year in Bungo channel is slightly shallower than that of short-term slow slip events accompanied by tremor activities as pointed in this paper. The coupling property at the plate interface changes from the locked zone to free slip according with increasing depth and temperature [Hyndman et al., 1997]. The duration of the slip event might reflect the difference in frictional properties of the plate interface.
 The tremor was thought to be a phenomenon at the Moho discontinuity related to fluid liberated from the subducting plate by dehydration process [Obara, 2002]. However, top of the oceanic plate and the continental Moho discontinuity are very close each other around the target area. Considering the similarity with ETS in Cascadia, the coupling phenomenon is seemed to occur along the plate boundary. The fluid might affect the friction properties at the slab interface, resulting in slow slips. These tremor and slow slip are thought to represent directly the condition of the plate boundary and stress accumulation of the locked zone. Continuous monitoring of the phenomenon may be very useful for modeling and monitoring the loading process in these subduction zones.
 We are grateful to anonymous two reviewers for helpful comments.