Along-strike structural changes controlled by dehydration-related fluids within the Philippine Sea plate around the segment boundary of a megathrust earthquake beneath the Kii peninsula, southwest Japan
Earthquake Research Institute, University of Tokyo, Tokyo, Japan
Corresponding author: E. Kurashimo, Earthquake Research Institute, University of Tokyo, 1-1-1, Yayoi, Bunkyo-ku, Tokyo, 113-0032, Japan. (firstname.lastname@example.org)
 Active and passive seismic experiments in the Kii Peninsula, southwest Japan, revealed prominent structural features around the segment boundary of a megathrust earthquake associated with the subduction of the Philippine Sea Plate (PHS). A distinct reflection band in the uppermost part of the PHS shows significant lateral variation along its strike. The thicker reflection band, which corresponds to the deeper extension of the fault area of the 1944 Tonankai Earthquake, is interpreted to be a zone of high pore fluid pressure, causing a state of conditionally stable slip on the plate boundary by the reduction in effective normal stress. A thinner reflective band corresponds to the deepest part of the rupture area of the 1946 Nankai earthquake, where plate coupling is stronger due to less effect of fluids. This along-strike structural variation controls the differences in frictional properties and the lateral limit of rupturing along the plate boundary.
 Subduction zones located along convergent margins are areas of high seismicity, where great earthquakes occur at regular temporal and spatial intervals [e.g., Thatcher, 1990]. The boundary between two neighboring earthquakes, referred to as a segment boundary, controls the extent of fault slip and the magnitude of an earthquake. In some cases, the rupture propagates beyond the segment boundary to become a single very large earthquake [e.g., Lay et al., 2005]. It is important to identify the structural factors that control the spatial extent of fault slip, as this should provide insights into the dynamics of earthquake rupture. Recent studies have identified several seismic structural factors that control the lateral limits of rupture propagation; for example, upper-plate transverse crustal faults [Collot et al., 2004]. In addition, various slip motions with a different time scale, including episodic tremors and very low-frequency earthquakes, have been identified at or near the updip and downdip limits of the seismogenic zone [e.g., Obara, 2002; Ito and Obara, 2006]. These fault slips may be related to fluids released by dehydration of subducted oceanic lithosphere [e.g., Kodaira et al., 2004]. Although previous studies indicate the importance of fluid behavior in understanding fault slip at plate interfaces, little is known about the structure around segment boundaries, particularly in their deeper parts, which are strongly affected by dehydration [e.g., Obara, 2002].
 The Philippine Sea Plate (PHS), which is subducting beneath southwest Japan in the Nankai Trough region, forms a well-defined interplate seismogenic zone (Figure 1). The rupture zones of the last two large earthquakes in this region, the 1944 Tonankai and 1946 Nankai earthquakes, have been well documented by tsunami and geodetic data [e.g., Baba and Cummins, 2005] (Figure 1). The southern part of the Kii Peninsula is an ideal area in which to study the structural factors causing rupture segmentation, because the rupture boundary between the 1944 and 1946 events extends northward beneath the peninsula (Figure 1). Previous research has indicated that zones of high pore fluid pressure are marked by high reflectivity and/or high Vp/Vs [e.g., Kodaira et al., 2004]. Such pore fluid pressure, which probably controls spatial variations in frictional properties upon the plate interface [e.g., Rice, 1992], is undoubtedly of importance to stress accumulation and rupture processes in and around the locked zone. Here we present the seismic structure around a segment boundary beneath the Kii Peninsula, based on the results of active- and passive-source seismic experiments, focusing on structural heterogeneity controlled by dehydration-related fluids.
2 Seismic Array Observations and Data
 To investigate the detailed structure of the PHS plate beneath the Kii Peninsula, we conducted an integrated analysis of both active-source seismic and natural earthquake data. A seismic experiment was conducted during November 2004 in the Kinki district of southwest Japan [Ito et al., 2006] as part of the Special Project for Earthquake Disaster Mitigation in Urban Areas (DAIDAITOKU). The N–S-oriented seismic profile extended between the Pacific and Japan Sea coasts (Figure 1), approximately parallel to the direction of subduction of the PHS. Therefore, the southern part of the profile, herein referred to as the N–S profile, was expected to reveal structural variation in the direction of plate subduction. We conducted another deep seismic reflection profile to complement the DAIDAITOKU seismic experiment to reveal the structural variations transverse to the direction of subduction of the PHS, that is, along the strike of the plate boundary (Figure 1). This involved the deployment of 279 seismic stations on a 60 km long line (E–W profile), and shooting of three explosives (SP-A, SP-B, and SP-C) as controlled seismic sources. An explosive (SP2) on the N–S profile was located at the intersection of the E–W and N–S profiles.
 Natural earthquake data were also used to determine S wave velocity variations and the deeper structure of the subducted oceanic lithosphere. These data were acquired by preprogramming the 279 seismic stations deployed on the E–W profile and the 1036 seismic stations on the N–S profile for continuous recording during active-source experiments on the nights of 17–19 or 17–20 November 2004. Data from an additional 13 three-component portable recording systems operated from January to February 2004 (Figure 1) and from 74 permanent stations around the profile lines (Figure 1b) were also incorporated into our analysis to obtain a high-resolution velocity model. Eventually, 9611 P wave and 4125 S wave arrival times from 105 local earthquakes and 11 explosive shots (SP1–SP8, SP-A, SP-B, and SP-C) were used in an inversion analysis. Details of the data specification are provided in the supporting information text S1.
3 Analysis and Results
3.1 Reflection Imaging
 The relatively high-amplitude reflections observed on both the E–W and N–S profiles were mapped into depth sections using a poststack migration technique after common midpoint stacking (Figures 2a and 2b). Normal moveout and depth-conversion velocities were based on a tomography-derived P wave velocity model, as described in section 3.2 (Figures 3a and 3c). The E–W profile shows a distinct reflection band that varies in thickness along strike (Figure 2a). This reflection band is about 6 km thick in the eastern part of the E–W profile but only about 3 km thick in the western part (Figure 2c). Along the N–S profile, this band dips to the northwest (Figure 2b). According to the previous refraction profile by Nishisaka et al.,  obtained beneath the Pacific coastline, the top of the PHS is located at a depth of about 18 km, which is almost identical to the top of the reflective zone (P1) in our N–S profile (Figure 2b).
3.2 Tomographic Imaging
 The simul2000 tomography code [Thurber and Eberhart-Phillips, 1999] was applied to P and S wave arrival-time data to determine the detailed seismic velocity structure, using an X–Y coordinate system, as shown in Figure 1b (see supporting information text S2). For the velocity model and station corrections obtained, we relocated a total of 8879 events occurring from 1 January 2002 to 31 December 2009 to investigate the relationship between crustal heterogeneity and seismicity (Figure S5). The P and S wave arrival-time data were provided by the Japan Meteorological Agency (JMA).
 Figures 3a and 3b shows a vertical cross section of Vp and Vp/Vs values along the X axis (E–W profile), together with the relocated hypocenters. The prominent feature in Figure 3 is a high Vp (6.0–7.0 km/s) zone located in a horizontal range of −10 to 10 km beneath SP2 (dashed white circle in Figure 3a). This high-velocity body is more clearly recognized in the Y axis cross section (dashed white circle in Figure 3c). Another structural feature is a distinct change in Vp/Vs at about X = −10 km and at depths greater than 35 km (Figure 3b). Beneath the western part of the E–W profile, a high intraslab seismicity is found in a zone of high Vp and low Vp/Vs (Figures 3a and 3b). Low-frequency earthquakes (LFEs) were relocated in a zone of low Vp and high Vp/Vs at a depth of 35–40 km (Figures 3c and 3d), probably at the top of the PHS, as inferred from the seismic reflection imaging.
4 Discussion and Conclusions
 Reflection images show a bright reflective band within the uppermost part of the PHS (Figures 2a and 2b). The top of the PHS beneath SP2 is located at ~27 km depth, about 10 km shallower than that previously reported from seismicity data [Miyoshi and Ishibashi, 2004]. The eastern part of the E–W profile (−2 < X < 23 km) shows a ~6 km thick distinct reflective zone at a depth of 27–33 km. As shown in Figure 4, the thicker reflection band is located downward of the rupture zone of the 1944 earthquake. In contrast, the thinner reflection band (about 3 km in thickness) at −32 < X < −10 km corresponds to the deepest part of the rupture area of the 1946 earthquake. This image is in marked contrast to a sharp reflection signature in the locked zone offshore of the Kii Peninsula [e.g., Park et al., 2002]. A similar reflection character was observed at the Cascadia subduction zone [Nedimović et al., 2003], where its locked part shows a thin, sharp reflection, whereas its deeper part, characterized by aseismic slip, exhibits a thick reflection band. Therefore, a strong relationship between the thickness of the reflection band and the degree of frictional strength is expected. In the present case, the seismic coupling in the western part of the E–W profile may be stronger than that in the eastern part. In the Cascadia subduction zone, several structural models have been proposed [e.g., Bostock, 2012]. We interpreted the top of the reflective band as the top of the PHS, on which fault slips represented by LFEs are occurring (Figure 3). Our observation is consistent with the model proposed by Bostock .
 A comparison of the migrated depth section (E–W profile) with the Vp/Vs structure indicates that the zone of relatively high Vp/Vs extends up to the thick reflection band (Figure 4). This relationship is similar to that observed in the deeper extension of the expected rupture area of the Tokai earthquake in central Japan [Kodaira et al., 2004], where a zone of high pore fluid pressure, formed by dehydration of the subducted oceanic crust, is responsible for the high reflectivity and high Vp/Vs ratio observed at the top of the oceanic crust. Such structural features cause a physical state of conditionally stable slip along the plate boundary. Similar zones of high pore fluid pressure have been documented in other subduction zones [e.g., Peacock et al., 2011]. A highly conductive (<10 Ωm) zone is also found at depths of 30–45 km beneath SP2 (below the top of the PHS) [Yamaguchi et al., 2009], indicating the existence of fluids within the relatively high Vp/Vs body located within and below the thick reflection band. Therefore, it is likely that the thick reflection band represents a zone of high pore fluid pressure (Figure 4). The high pore fluid pressure would reduce the effective normal stress on the plate interface, resulting in weakened plate coupling [e.g., Rice, 1992]. In Figures 1b and 4, it is also noted that the downward extension of the thick reflection band corresponds to the active tremor area. Conversely, along the downward extension of the thin reflection band, few tremors are recognized. The seismic coupling in this area may be stronger than that in the active tremor area.
 With respect to an effective mechanism for the accumulation of high-pressure fluid, Kato et al.  found an impermeable barrier with a high velocity just above the plate interface in the Tokai region. In our case, a high Vp body is also present above the PHS (dashed white circle in Figures 3a, 3c, and 4), which shows high-resistivity (>1000 Ωm) [Umeda et al., 2006]. The southeastern part of the Kii Peninsula underwent significant magmatic activity in the middle Miocene that formed the Kumano acidic intrusion and the Omine granitic intrusion (Figure 1b) [e.g., Miura and Wada, 2007]. The high Vp and high-resistivity body may be the expression of these intrusions, probably acting as an impermeable barrier to confine high-pressure pore fluids.
 To date, several seismic structural factors that control segmentation have been proposed [e.g., Collot et al., 2004]. Such knowledge, however, is limited with regard to the locked seismogenic zone. The seismic images presented here show along-strike structural variations present in the brittle–ductile transition zone along a plate boundary. These variations are most likely controlled by the behavior of high-pressure fluids confined by the impermeable barrier of the Kumano acidic intrusion and Omine granitic intrusion. Consequently, in the area downward of the 1944 earthquake fault, fluids are more abundant compared with the neighboring area downward of the 1946 earthquake fault. This difference gives rise to the contrast in frictional properties between the two areas, which determine the lateral limits of dynamic rupturing during megathrust earthquake events.
 We thank two anonymous reviewers for their constructive comments. This study was supported by the Special Project for Earthquake Disaster Mitigation in Urban Areas from the Ministry of Education, Culture, Sports, Science, and Technology of Japan. We also thank the National Research Institute for Earth Science and Disaster Prevention, the Japan Meteorological Agency, and the University of Tokyo for allowing the use of their waveform data.
 The Editor thanks Eli Silver and an anonymous reviewer for their assistance in evaluating this paper.