There is a regional variation of seismicity in the Japan trench region characterized by cluster distribution of seismically active zones which persist for at least a few decades. We conducted seismic refraction-reflection experiments there to clarify the relationship between seismicity and crustal structure. A velocity structure model was obtained by travel-time inversion. Although there does not seem to be distinct relationship between seismicity and the bulk velocity structure, we found a good relationship between seismicity and variation of reflective amplitude. Large amplitude reflected waves generated at the plate boundary were observed at low seismicity region and vice versa.
 Many destructive earthquakes as well as microearthquakes have occurred along the Japan Trench. Figure 1 shows epicentral distribution in Japan trench area, determined by the microearthquake observation network. There is an aseismic region south of latitude 39°N, that is, the seismic activity is not uniform along the trench axis. It is suggested that the seismic coupling of the Pacific Plate and the Continental Plate is not uniform along the trench axis.
 Revealing the physical state of the plate boundary is important to understand the mechanism of these earthquakes. The physical state can be controlled by roughness, pressure, temperature and porosity on the plate boundary, and by the presence of fluid. The objective of the present survey is to investigate the physical state on the plate interface in relation to the velocity structure and its relation to earthquake generation (or earthquake occurrence).
 A number of seismic reflection and refraction experiments using artificial sources and Ocean Bottom Seismometers (OBSs) have been conducted in this region (e.g. [Suyehiro and Nishizawa, 1994]). In addition, observation of seismic activities using OBSs have been carried out for several times [Kasahara et al., 1982; Nagumo et al., 1984; Hirata et al., 1985; Nishizawa et al., 1992; Hino et al., 1996]. These experiments and observations revealed the characteristics of crustal structure and seismic activities in this region. However, the relationship between the regional variation of seismicity and of the plate coupling along the trench axis had not become clear. In this study, we intend to present a relationship that we found between seismic activities and the physical state on the plate interface at the Japan Trench subduction area.
 In 1996, we conducted a seismic refraction-reflection experiment at the Japan Trench region. We deployed 39 pop-up type OBSs, which were aligned along 4 lines (Figure 2). We call the longest north-south line NS-line, which is almost parallel to the trench axis and 140 km long. In this study, we focus only on the NS-line to discuss the physical state on the plate interface parallel to the trench axis.
 On the NS-line, fourteen OBSs were deployed, and 71 explosives and 1900 airgun shots were fired. The charge size of each explosive was 20 kg and shot spacing was 2 km. We used 4 airgun chambers totally 71 liters. Airgun shot interval varied with place and approximately 50 m and 100 m. Ten OBSs had digital recording system, while the others had analog recording system. The digital OBSs consist of two types, 24 bit A/D converter type and 16 bit type [Kasahara et al., 1995, 1997; Shiobara et al., 1996]. During airgun operation, a 24-channel hydrophone streamer was towed behind the research vessel. The hydrophone spacing was 50 m.
3. Structure Analysis Method
 The analysis procedure to model the crustal P-wave velocity structure was as follows. First of all, we constructed the first velocity structure model using results of τ − p analysis ([Stoffa et al., 1981]) and MCS records. Next, we refined it by trial and error approach using traveltimes of both first-breaks and reflections; however, we did not give horizontal heterogeneity to deeper structure than 7 km because of lack of deep penetration of refracted waves. Using this preliminary structure as initial model, we obtained a result model by iterative non-linear travel-time inversion after 10 times iterations. During inversion, we computed traveltimes and raypaths using a regular grid method based on the Fermat's principle [Fujie et al., 2000], because it is robust and accurate.
 We used reflected travel-times as well as refracted travel-time to determine velocity values and interface depths. We observed three prominent reflected phases from the depth deeper than 10 km, and we interpreted these phases come from the plate boundary, the interface just beneath the plate boundary and the Moho discontinuity, respectively.
4.1. Velocity structure
 The final model of P-wave velocity structure under the NS-line and its resolution are shown in Figure 3 and Figure 4, respectively. Adequate resolution with regard to both velocity values and interface depths was obtained except for one interface, which lies at about 7.5 km depth (Figure 4). Since we set this interface to simplify the deeper part of the initial model to 1-D structure, the velocity change across this interface is small. Therefore, there is little influence on deeper structure although resolutions of this interface depth are low. Referring to the results of the former studies (e.g. [Suyehiro and Nishizawa, 1994]), we interpret the 6-th interface as the top of the subducting Pacific plate.
4.2. Observed Reflected Waves
 Although we only used traveltime information during structural analysis, the observed amplitudes of reflected waves depended largely upon reflection points (Figure 5). Because it is not straightforward to compare the absolute amplitudes among OBSs due to each instrumental characteristics, it has little meaning to compare directly the amplitude of different traces. Therefore, we normalized each trace by the maximum amplitude of the first arrival to minimize discrepancies of both sources and receivers.
Figure 6 shows the distribution of distinct reflectors at the plate boundary inserted in the present structural result. The intensity of reflection at the plate boundary is evaluated by the amplitude ratio of the reflected wave to the first arrival, and it varies greatly from place to place.
 In our inverse analysis of P-wave velocity structure, we adopted a simple initial model, and could determine the deeper structure objectively. Although the plate boundary and the Moho discontinuity become shallow toward north, we could not obtain any distinct relationships between the velocity structure and the regional variation of seismic activities.
 Paying attention to the amplitude information, we found very interesting relationship between seismic reflection intensity and seismic activity (Figure 6); large amplitude reflected waves were observed from low seismic region and vice versa. According to hypocenter determinations using OBSs (e.g. [Hino et al., 1996]), most of local microearthquakes was found to occur around the plate boundary. Since the amplitude of reflected waves can give knowledge on the characteristics of the reflected boundary, this relationship suggests that seismic activity has certain relations with the characteristics of the reflective interface, that is, the plate boundary.
 In Figure 6, we evaluated the amplitude ratio by comparing it with the synthetic amplitude ratio, which were computed by using ray theory [Zelt and Smith, 1992]. We assumed that the density of the Layer-A (Figure 3) is constant during our evaluation. At the north of NS-line, where only weak reflections were observed, the observed amplitudes of reflected waves from the Layer-A are consistent with the result by synthetics. Therefore, the velocity in Layer-A is about 6 km/s at the north, as in the result model. However, at the low seismic region, where large amplitude reflected waves were observed, the observed amplitude is not consistent with the synthetics at all. According to the synthetic amplitude computations, the velocity at the top of Layer-A must be approximately 3 ∼ 4 km/s or 8 km/s in the low seismic region. Since it is not realistic that the velocity becomes as large as 8 km/s at the Layer-A, it is more natural to consider the velocity is 3 ∼ 4 km/s. Hence, the velocity at the top of Layer-A varies horizontally from 3 km/s to 6 km/s. However, it is unlikely that the layer has such a large velocity change laterally.
 Other possibility for the generation of large amplitude reflected waves is a very thin layer with low velocity at plate boundary, and we will call it Layer-X. The Layer-X must be thin, for example, several hundred meters, enough to generate reflective waves and not to affect the traveltime analysis.
 It is difficult to specify the materials constituting the Layer-X from only the relationship between seismicity and amplitude of reflection. We, however, can infer some influences of water as one of the possibilities. In the case of Barbados, fluid layer exists in the decollement zone [Moore et al., 1998]. Although the Layer-X is too deep to suppose the same situation as Barbados, water as fluid phase or hydrated minerals can be present at the plate boundary. The subducting oceanic crust, which passes through blueschist facies stage (at 10–50 km depth), undergoes dehydration and produces aqueous fluids [Iwamori, 1998]. Thus, aqueous fluids can exist along the plate boundary. The existence of fluids generates large amplitude reflected waves and weaken the coupling of the both plate.
 As another possibility, aqueous fluids can generate hydrated minerals around the plate boundary. For instance, peridotite in the mantle will be serpentinized if water can reach the mantle wedge, and serpentine will ascend along the plate boundary because the density of it is comparatively small. Since hydrated rocks as serpentine are apt to undergo plastic deformation [Raleigh and Paterson, 1965] and their P-wave velocity is comparatively small, it is easy to release the strain between both plate where the strong reflected waves were observed. In the Izu Bonin Trench, the presence of serpentinized mantle wedge has been proposed [Suyehiro et al., 1996; Kamimura et al., 2000]. According to the previous study in this region [Suyehiro and Nishizawa, 1994], mantle wedge may be located not so far from the NS-line. Therefore, we infer serpentine is one of the candidates which constitute the Layer-X.
 We studied the relationship between seismic activities and crustal structures at the Japan Trench region. According to the result of refraction-reflection analysis, we found the relationship between seismic activities and reflection intensities.
 We proposed a hypothesis that thin layer with slow velocity exists along the plate boundary in low seismic region, and it explains the relationship between regional variation of reflection intensities and seismic activities quite well. We infer the thin layer is affected by aqueous fluid and/or hydrated rocks, such as serpentine.