A 7.2-magnitude earthquake occurred on the plate boundary off Miyagi Prefecture, northeastern Japan, on 16 August 2005. From seafloor geodetic observations using the GPS/acoustic combination technique in this area, we have detected notable seafloor movements associated with, and subsequent to this event. The time series of horizontal coordinates obtained at the seafloor reference point MYGW, located about 10 km east of the epicenter, indicates that the strain accumulated for the interseismic period was partly released by the event and a constant strain accumulation restarted after 1–2 years of the post-seismic period with an erratic movement. The horizontal velocity of MYGW after the restart above is estimated to be about 5.7 cm/year toward WNW relative to the Eurasian plate. Our result is consistent with those derived from terrestrial GPS measurements and implies that the interplate locking was restored in the rupture area of the 2005 event around 2007.
 On 16 August 2005, a large earthquake (M7.2) occurred on the plate boundary off Miyagi Prefecture, northeastern Japan, where large interplate earthquakes with magnitude of about 7.5 have repeatedly occurred about every 37 years (Headquarters for Earthquake Research Promotion (HERP), Long-term evaluation of the Miyagi-Oki earthquakes (in Japanese), 2003, available at http://www.jishin.go.jp/main/index.html). This event re-ruptured the southern/southeastern part of the asperities which caused the previous 1978 Miyagi-ken-oki earthquake (M7.4) [e.g., Okada et al., 2005; Miura et al., 2006]. Miura et al.  estimated co- and post-seismic slip distributions of the event based on inversion of terrestrial GPS data.
 In order to monitor crustal movements around offshore plate boundary regions, the Hydrographic and Oceanographic Department of Japan (JHOD) has been carrying out seafloor geodetic observation using the GPS/acoustic combination technique on the landward side of the major trenches around Japan, such as the Japan Trench and the Nankai Trough, under technical cooperation with the Institute of Industrial Science, the University of Tokyo [Asada and Yabuki, 2001; Mochizuki et al., 2003; Fujita et al., 2006].
 In the region off Miyagi Prefecture, we have two seafloor reference points labeled “MYGI” and “MYGW”, which were installed in 2001 and 2004, respectively (Figure 1). At MYGI, located about 100 km landward from the axis of the Japan Trench, the intraplate movement toward WNW was found through our first couple of years observations [Fujita et al., 2006]. In association with the 2005 Off-Miyagi Prefecture earthquake, an eastward co-seismic movement of about 10 cm was detected at MYGW, located about 10 km east of the epicenter, while no prominent movement was found at MYGI, about 50 km further east of MYGW [Matsumoto et al., 2006].
 In this paper, we report and discuss a notable behavior appearing in the time series of horizontal coordinates at MYGW after the 2005 event, which has been obtained by our observations through 2010.
2. Measurement System and Analysis
 A schematic picture of the seafloor geodetic observation system that we have developed is shown in Figure 2. This system consists of a seafloor unit with three or four acoustic mirror-type transponders and an on-board unit with a GPS antenna/receiver, an undersea acoustic transducer and a dynamic motion sensor.
 The system measures ranges from the on-board transducer to the seafloor acoustic transponders through round-trip acoustic travel times, while simultaneously gathering kinematic GPS data. The attitude data of the survey vessel are also acquired on board by the dynamic motion sensor, which are used to determine the coordinates of the on-board transducer relative to those of the GPS antenna. The acoustic velocity profiles in the seawater, which are necessary to transform travel time into range, are obtained from temperature and salinity profilers every several hours.
 The data analysis consists of three procedures: (1) acoustic wave analysis to obtain the round-trip travel time by the correlation processing method, (2) kinematic GPS analysis to determine the position of the GPS antenna, and (3) a combination of results from (1) and (2) to determine the positions of the seafloor transponders by a linear inversion method based on least squares formulation. The positions of grouped transponders are finally averaged to be a virtual position of the seafloor reference point. Fujita et al.  described the methodology of our observation technique in detail.
3. Observation Results at Seafloor Reference Point MYGW
 The locations of our seafloor reference points offshore Miyagi Prefecture are shown in Figure 1. In this study, we focus on the seafloor reference point labeled as MYGW, located only about 10 km apart from the epicenter of the 2005 event, where the clear co-seismic movement was detected [Matsumoto et al., 2006]. A set of four acoustic transponders has been installed on the seafloor, at a depth of about 1,100 m. We carried out campaign observations 14 times for the period from April 2005 just before the event to March 2010.
Figure 3 shows the time series of estimated horizontal coordinates of MYGW. The position reference is the Shimosato site in Wakayama Prefecture, central Japan, which is the Satellite Laser Ranging (SLR) station of the Japan Coast Guard. The intraplate velocity of about 3 cm/year toward WNW within the Eurasian plate, due to the pressure of the Philippine Sea plate subduction, was estimated from the SLR observations [e.g., Sengoku, 1998].
 The whole time series in Figure 3 can be specified as three phases of different features. The first phase is the co-seismic phase right at the event, showing an eastward co-seismic movement of about 10 cm, already reported by Matsumoto et al. . A closer look at the time series after the event leads us to observe that they can be further divided into two phases at around the end of 2006 in view of the difference in the trend. The early period exhibits a slight eastward trend as a whole but with some erratic motion especially in the NS component, while the latter shows a westward linear trend. It should be mentioned that, during this period, there were no large earthquakes which may have significantly affected the position of MYGW other than the 2005 event.
 A linear fit to the latter time series, after December 2006, gives us a rate of 2.6 ± 0.2 cm/year westward and 0.2 ± 0.4 cm/year southward. The root mean squares around the fitted lines are 0.7 cm in the EW component and 1.2 cm in the NS component. Adding the intraplate velocity of Shimosato (3.2 cm/year, 291° [Sengoku, 1998]) to the above rate, we have obtained the velocity vector of 5.7 cm/year with an azimuth of 280°, relative to the stable part of the Eurasian plate, which is exhibited with an arrow in Figure 4.
 In the region around MYGW, offshore Miyagi Prefecture, it is known that interplate earthquakes with magnitude of about 7.5 have repeatedly occurred in the past. It is understood from the mechanism of such earthquakes that this region is under the state of strain accumulation due to the pressure of the Pacific plate subduction during an interseismic period.
 At MYGI also in this region, located about 50 km east of MYGW, Fujita et al.  found the westward intraplate movement with a constant velocity from the seafloor geodetic observations conducted for about three years before the 2005 event. This movement is interpreted to reflect the strain accumulation process mentioned above. The westward trend at MYGI has been basically continuing after the event, though it is not shown and discussed in detail in this paper because of our present focus on the MYGW behavior. The similar features, i.e., westward movements, have been observed also at terrestrial GPS stations on the coastal area around Miyagi Prefecture [e.g., Miura et al., 2004]. Thus, the westward movement is also expected at MYGW before the 2005 event, though we do not have observation data for this period. This is confirmed by the back slip model by Suwa et al. , which was estimated by inverting three-dimensional velocity data of terrestrial GPS stations during an interseismic period before the 2005 event. The velocity of MYGW relative to the Eurasian plate calculated from this model is 5.6 cm/year with an azimuth of 282°.
 Thus, it is reasonable to interpret that the westward movement appearing in the last phase of the observed time series at MYGW reflects a strain accumulation process, which is expected to have had continued until the 2005 event.
 On the other hand, the second phase with an erratic movement subsequent to the event likely corresponds to the post-seismic deformation often reported to be caused by post-seismic slip accompanying large interplate earthquakes. Miura et al.  estimated co- and post-seismic slip distribution associated with the 2005 event by geodetic inversion methods using terrestrial GPS data. The post-seismic slip appears to have had continued at least until the end of the period of their study, i.e., July 2006, which is consistent with our results.
 Summarizing all these discussions including the co-seismic phase discussed by Matsumoto et al. , the three phases in the time series at MYGW indicate that the strain accumulated on the plate boundary was partly released by the 2005 event and started to accumulate again after the period of 1-2 years affected by the post-seismic slip on the rupture area.
 For comparison, Figure 5 shows the time series of the horizontal coordinates of the terrestrial GPS station closest to MYGW, ‘Oshika’ on the coast of Miyagi Prefecture (Figure 2), which is operated by the Geospatial Information Authority of Japan (GSI) [Sagiya et al., 2000]. The coordinate values plotted in Figure 5 are the GSI solution [Nakagawa et al., 2009] represented in ITRF2005 [Altamimi et al., 2007]. As shown in Figure 5, an eastward co-seismic movement of about 5 cm associated with the 2005 event clearly appeared. The time series indicates that the linear trend characterized in the pre-seismic period at Oshika was restored in the first half of 2007. Therefore, our result that the re-accumulation of the strain started at around the end of 2006 is consistent with that of the terrestrial GPS measurements, in spite of the number of our campaign observations limiting the estimation of the exact epoch of the restoration.
 The obtained velocity vector of MYGW in the last phase, 5.7 cm/year WNW, is between those observed at terrestrial GPS stations (3–4 cm/year [e.g., Miura et al., 2004]) and that of the Pacific plate (8-9 cm/year [e.g., DeMets et al., 1994]) in amount and consistent in direction. In addition, it should be noted that it is in good agreement with the synthetic vector, 5.6 cm/year WNW, already shown above from the back slip model by Suwa et al.  based only on the observations before the 2005 event, though, in the offshore area, the resolution of the model, consequently the reliability of the calculated velocity, would be quite limited. We should therefore conclude that our result implies that the interplate locking was restored in the rupture area of the 2005 event around 2007.
 From seafloor geodetic observations off Miyagi Prefecture, we have succeeded in detecting seafloor movements associated with and subsequent to the 2005 Off-Miyagi Prefecture Earthquake. The time series obtained at the seafloor reference point MYGW close to the epicenter can be specified as three phases of different features: (1) a westward co-seismic movement, (2) an erratic behavior for 1-2 years and (3) a west-northwestward linear trend equivalent to the horizontal velocity of 5.7 cm/year relative to the Eurasian plate. These features indicate that the strain accumulated for the interseismic period was partly released by the 2005 earthquake and a constant strain accumulation started again after 1-2 years of the post-seismic period with an erratic movement. It is the first successful observation of co- and post-seismic movement on the seafloor until the state of the constant strain accumulation is restored. The restoration epoch obtained in this study shows good agreement with those derived from terrestrial GPS measurements. Our result suggests that the interplate locking was restored in the rupture area of the 2005 event around 2007.
 We thank the Geospatial Information Authority of Japan for providing us with the GPS data and releasing daily coordinates of Oshika on the GSI's website. We also thank Oscar L.Colombo of the NASA Goddard Space Flight Center for providing us with the kinematic GPS software ‘IT’ (for Interferometric Translocation) [Colombo et al., 2000]. We are indebted to Motoyuki Kido and Yukihito Osada of Tohoku University for their support about calculating the synthetic velocity from the back slip model by Suwa et al., . A lot of staff members of Hydrographic and Oceanographic Department of Japan Coast Guard, including the crew of S/Vs Meiyo and Kaiyo, have been supporting us in observations and data analyses. Comments from the anonymous reviewers have improved the paper substantially. Some figures were produced with the GMT software [Wessel and Smith, 1991].