Search

Tsunami travels across the open sea very fast. It slows down as it approaches the coast, begins to increase in height and the waves becomes steeper and higher. Thus, the destructive tsunamis are believed to be generated in the harbors or near the shore. Moreover, most of the documented destructions related to tsunami are usually those near and along the shore areas. However, fishermen onboard fishing vessels located about 0.5–20 km from the west coast of northern Sumatra when the 2004 Sumatra–Andaman tsunami happened recounted some observed destructions. Almost all interviewed fishermen felt the earthquake shocks while others even had difficulty controlling their vessels during the shaking. Notably, shortly after the shaking, 10–20 m high waves struck their vessels. These waves were more or less isolated high waves and hit their vessels repeatedly. A numerical simulation using a dispersive nonlinear long wave equation suggested that solitary waves were generated on the steeply uprising waves of long-period tsunamis, which are similar to the waves that struck the vessels. Based from the interviews and simulation, it is ascertained that tsunami even several kilometers offshore can pose danger when the water depths are shallower than 40 m.

In order to clarify people's behavior after the March 11, 2011, Tohoku-Oki earthquake ( M w 9.0), we interviewed 164 tsunami survivors in 10 cities mainly in Iwate and Miyagi prefectures and also city officials in charge of the disaster-prevention measures in Miyako and Kamaishi cities, Iwate prefecture, Tohoku region. During the 150 years prior to this event, three large tsunamis have struck the Tohoku region. Ever since these tsunamis, disaster mitigation efforts have been undertaken in the region, for example, constructions of the huge breakwaters, practicing of earthquake early warning schemes, holding periodical tsunami evacuation drills, and publicizing geohazard maps. Despite these efforts, 18,554 deaths and missing persons were reported by the National Police Headquarters on June 10, 2013. Our interviews revealed that only around 30 percent of interviewed people escaped immediately after the shaking of seismic ground motion subsided. Therefore, we searched through the interviews as to why so many people were slow or reluctant to evacuate. We presupposed possible reasons for their slow reaction as follows: (1) the magnitude of the earthquakes expected in the offshore Tohoku region publicized by the governmental committee was much smaller than that of the 2011 Tohoku-oki earthquake; (2) the earthquake magnitude and tsunami height of the first warning issued by the Japan Meteorological Agency were much smaller than those of the actual event; (3) people living there made wrong conclusions or biased guesses of the magnitude of the tsunami height based on their experience of past tsunamis or on what they heard from their parents or grandparents; (4) many people had too much confidence in the breakwaters, which were known to be the sturdiest constructions in the world. From our interview, we found that the reasons of (1) and (2) above are not real causes of their slow action. Reasons (3) and (4) together may have been the real causes of people's slow action. On the other hand, we found that very few schoolchildren (∼1.8 percent of deaths and missing persons, although they constitute 8.5 percent of the population) lost their lives at schools in Iwate, Miyagi, and Fukushima prefectures. This indicates that the role of elementary schools should be reassessed in their role as emergency shelters, as well as the function of education and education about the earthquakes and tsunami hazards and natural hazards in general in the local communities.

In this paper we examine the simulation results of a fully nonlinear, three–dimensional dynamo and obtain inferences useful in the study of present and past geomagnetic field. This approach has importance because of the limitation in the available data of the real magnetic field: the present field is known with a high accuracy, but the time covered is only a small fraction of the time constant inherent to the geodynamo, while palaeomagnetic data provide data for a long time–span, but they are of poor quality and are distributed quite irregularly both in space and time. Thus, we compare what we see from the external field of the dynamo model with the features established or conjectured for the present or palaeomagnetic fields. We show that some of the statistical properties of the magnetic field generated by dynamo models compare well with those of the real magnetic field: dominance of the axial dipole, similar power in each degree of the harmonic at the core surface, nearly normal distribution of Gauss coefficients, and the lack of correlation among their variations. Differences were observed in other features, such as drift in the azimuthal direction and concentration of magnetic flux in small patches at the core surface. They can be attributed to either the shortness of the observational period, or to the difference in the resolution of the models, which suppresses small–scale features far below the level of the observation.

In the field of fracture mechanics, it is of great importance to obtain the stress intensity factor of a crack when the cracked material is under a load. In this paper, we report the results of a case of a block specimen with a crack under transient heating loads from both sides of the block. We experimentally inventigated the stress intensity factor for a block that is heated abruptly under a certain constant temperature. For the purpose of comparison, both caustic and photoelastic methods are employed and the effects on the stress intensity factor of block dimensions are considered and discussed.

The boundary between the Philippine Sea (PHS) and Eurasian (EUR) plates, which is presently seismically quiescent, is delineated on the basis of seismological data from the Kanto-Tokai (K-T) district during the period from January 1980 to April 1993. Most of the earthquakes in the K-T district occur either in the overriding plate (the EUR plate) or within the subducting slab (the PHS plate) at depths of less than around 30 km. The plate boundary between the PHS and EUR plates is characterized by a very low level of seismic activity; i.e., the plate boundary itself is presently seismically quiescent, and the seismicity consists of intra-plate events above or below the plate boundaries. In the Tokai district, earthquakes in the overriding (EUR) plate are mostly characterized by strike-slip faulting with an E-W-directed compressional axis, while those within the subducting (PHS) slab are characterized by strike-slip faulting with a N-S-directed compressional axis. In the Kanto district, the earthquakes are mostly characterized by faults with NW-SE-directed compressional axes. The above difference between the Tokai and Kanto districts is thought to be due to the difference in the slab configuration and/or regional stress conditions. Considering the results of the present paper together with geodetic data, the seismically quiescent plate boundary at depths from 10 to 30 km suggests strong coupling between the PHS and EUR plates, and is consistent with strain accumulation during an inter-seismic period.

Body wave data recorded at a small network of broadband seismograph stations are analyzed to investigate local events with focal depths deeper than about 50 km. For these events the initial portion of P-wave displacement represents well the source time function with a scaler correction for the seismic moment. The magnitudes of the analyzed earthquakes range from MW = 3.1 to 6.5. It is shown that the seismic moment M0 and the pulse width τ are well correlated as M0/τ3 = constant, indicating that the stress drop is largely constant. This dynamic similarity seems to be valid for a vast range of earthquake sizes: MW = 1 ∼ 8. It is also shown that source complexity such as a multiple shock nature is not a characteristic of only large earthquakes but is often observed even for small earthquakes.

The 1990 Odawara earthquake (M5.1) occurred within an area where a large earthquake of M7 class is expected to occur in the future. This area (the “Odawara area”) has been seismically quiescent since 1933. Analysis of near-field broadband seismograms and the extent of fore- and after-shocks suggests that the 1990 event had a high stress drop and that the Odawara area remains in a state of high stress which could potentially cause a large earthquake.

Meiginger's equation for a jet fan cannot be applied to the case in which jet fans operate against air flow to reduce its velocity in a tunnel. This operating mode is used during fire accident in the tunnel. The equation for the mode is shown in this paper and is examined through the model experiments. The efficiency and diffusion of jets equipped with a deflector flap for the normal tunnel ventilation mode are also examined through the model experiments including flow visualization. Application of the flap is recommended.

Recently, an idea was proposed that the eastern Japan Sea deformation zone represents a nascent convergent zone between the North American and Eurasian plates. This supposed convergent plate boundary extends southwards through the Fossa Magna, a colliding sector, to the Suruga Bay forming a triple junction near Mt. Fuji or the Kofu Basin where the three convergent boundaries among Eurasian, North American and Philippine Sea plates meet. In order to evaluate this idea, the seismic activity including focal mechanisms of large and small earthquakes in the South Fossa Magna and the northern margin of Philippine Sea plate are examined.The subducting Philippine Sea plate, descending northwestwards along the Suruga and Sagami troughs was well defined by using the hypocenters and focal mechanisms of high resolution and accuracy. However, the plate boundary passing through the inland region, the South Fossa Magna, between the Sagami and the Suruga troughs was not defined due to poor seismic activity. On the basis of available seismic data, the newly proposed convergent plate boundary passing through the South Fossa Magna could not be found.

[1] Strong scattering sources were detected in the upper mantle beneath central Japan from seismic data recorded by the high-sensitivity seismograph network (Hi-net). Anomalous seismograms, consisting of many strong later phases whose amplitudes did not decay significantly with time, were observed from an intermediate-depth earthquake. A comparison of those anomalous seismograms with others generated by other deep earthquakes shows that major scattering sources may exist predominately in the upper mantle, not in the crust. Three possible candidates have been considered as the sources of scattering in the upper mantle including magma conduits, the previous collision zone and slab melting. Although seismic waves may be scattered strongly from each potential candidate, a region associate with slab melting might be more suitable for generating the anomalous seismogram due to the existence of abundant melting spots within the subducted Philippine Sea slab that is extremely young and hot beneath central Japan. INDEX TERMS: 0669 Electromagnetics: Scattering and diffraction; 7218 Seismology: Lithosphere and upper mantle; 7230 Seismology:Seismicityandseismotectonics.Citation: Lin, C.-H., M. Ando, N. Fujii, K. Yamaoka, K. Tadokoro, A.-S. Jin, K. Obara, and M. Ishida, Anomalous seismograms generated by an intermediate-depth earthquake: Unusual scattering sources in the upper mantle of central Japan, Geophys. Res. Lett., 30(11), 1586, doi:10.1029/2002GL016837, 2003.

The 11 March 2011 Tohoku earthquake and tsunami devastated northeastern Japan, causing thousands of deaths and widespread destruction. Many people in the zones affected by the tsunami evacuated to safety, but others were not so fortunate. Why did some people evacuate immediately and others wait until it was too late? In interviews with survivors, we sought to find answers to that question. These interviews revealed that many residents did not get accurate warnings, and many chose to stay in dangerous locations in part because they misunderstood the risks. Better communication of warnings and improved education about the science behind earthquakes and tsunamis could help save lives in the future.

In the early morning of 27 May 2006, local time, central Java was jolted by strong seismic ground motion. In spite of the moderate size of the earthquake (Mw = 6.4), it caused severe damage nearYogyakarta city. According to a report from the Social Affairs Ministry of Indonesia, more than 5700 people were killed, 38,000 injured, and 423,000 evacuated. As a result of the shaking, more than 126,000 buildings collapsed, and more than 392,000—including those of the famous Prambanan temple complex, located about 17 kilometers east of Yogyakarta—were severely damaged (Figure 1).

In order to improve the accuracy of the hypocenter locations in the Kanto district, Japan, station corrections are estimated from the data of artificial explosions. About 500 earthquakes which occurred during the period from January to August 1975 have been relocated by using the station corrections. The area covered in this study is bounded by latitudes 34.25°N and 37.25°N and longitudes 138.6°E and 141.1°E. The absolute error in the spatial coordinates of the relocated epicenter is estimated to be about 1.5km. Depth of the local explosions is estimated to be within 2km of the actual location. The good agreement between the calculated and the actual depths suggests that the accuracy of the depth is significantly improved from that determined when the station corrections were not included. An aseismic zone in the middle of the Kanto district which was suggested by the old seismicity map is not substantiated by the present study.

Investigating the foreshock-main shock-aftershock sequence of the Shizuoka-Seibu earthquake (M = 2.8, 16 January 1981), central Japan, we found evidence that suggests high stress conditions for the foreshock period. The source dimension of the main shock was roughly estimated to be about 0.5 × 0.2 km2. The m value of the Ishimoto-lida's formula was 1.87 for foreshocks and 2.37 for aftershocks.Waveforms of the foreshocks, main shock, and aftershocks were similar to each other. However, the amplitude ratio of the low (0.9 to 0.8 Hz) to high (3.5 to 10 Hz) frequency band was different between the foreshocks and aftershocks. The ratio for the foreshocks was lower than that for the aftershocks, indicating that the foreshocks contain more high-frequency energy than the aftershocks. The Shizuoka-Seibu sequence included a larger number of foreshocks (70) compared to aftershocks (43). This characteristic, in contrast with large earthquakes, was found common to small-sized earthquakes occurring in the neighborhood areas. These observations can be interpreted based on the asperity model.

Recent studies on the velocity structure beneath the Kanto-Tokai District, Honshu, Japan, pointed out the large lateral heterogeneity of velocity structure. This heterogeneity is large enough to suspect that hypocenters determined by a laterally homogeneous velocity model are systematically deviated from true locations. Computation of travel times for a laterally inhomogeneous velocity model by three-dimensional (3-D) ray tracing revealed that travel time residuals of -1.0 to 1.2 s are produced if the lateral inhomogeneity is neglected. Residual distribution varied according to hypocenter locations. This indicates that a laterally homogeneous velocity model with station corrections is not sufficient to obtain accurate travel times in this district. According to a numerical experiment, the resultant deviation in hypocenter location is up to +-5 km and +-8 km for epicenter and focal depth, respectively. For obtaining reliable hypocenter distribution, about 10, 000 hypocenters of earthquakes which were observed by the stations of the National Research Center for Disaster Prevention for the period from 1983 to 1985 were relocated by using the 3-D velocity model. Shape of earthquake clusters became clearer after the relocation. For the computation of hypocenters, we used an approximate method of 3-D ray tracing, which gives a sufficient approximation of the exact ray tracing.

The spatial-temporal variation of seismicity of the 1980 earthquake swarm off the east coast of the Izu Peninsula, Japan, was investigated. Hypocentral distribution, focal mechanism, wave forms, and spectra of seismic waves were studied. The hypocenters were relocated by using the master event method. The forerunning earthquakes which started about one week before the largest shock (main shock), the 1980 Izu-Hanto-Toho-Oki earthquake (M = 6.7), occurred within the quiescent area of the earthquake activity for the preceding one year. The swarm area migrated toward the south with time and triggered the main shock in June 1980. The fault dimension and geometry were estimated from the aftershock area: the fault length and width are 14 km and 8 km; the strike and dip angles of the fault are N15°W and 65° to N75°E. Locations of the events in an earlier earthquake swarm (1978) were also examined by using difference in the S-P time at five selected stations distributed around the epicentral area. The 1978 swarm events were found to have clustered within a very small area of 8 ×1 km2 located about 2 km to the west of the 1980 swarm area. The earthquakes which occurred after the main shock of the 1980 swarm were classified into two groups, aftershocks and swarm events, according to the location of epicenters, wave forms, and spectra of S waves. The peak frequencies of spectra were distributed around 5 to 8 Hz for the aftershocks and around 10 to 15 Hz for the swarm events. Most of the aftershocks, characterized by low-frequency content, occurred to the south of the main shock within 2 weeks after the main shock. The number of aftershocks decayed following the modified Omori's formula with p = 1.5 ± 0.3. The swarm activity, on the other hand, continued intermittently for about 1 month after the main shock. The 1980 seismic activity is interpreted as a complex of a foreshock-main shock-aftershock sequence and swarm activity. The direction of the longer axis of the swarm area coincided with the direction of the maximum pressure axis of the main shock. The trend of the aftershock zone coincided with the strike of the fault planes of the main shock and aftershocks. This feature strongly suggests that tension cracks trending in the maximum stress direction opened prior to the occurrence of the main shock. The opening of cracks may be accounted for by increasing of interstitial pore pressure associated with increase in regional stress.

We examined the spatial distribution of earthquake hypocenters and the three-dimensional velocity structure of P-wave for the Kanto-Tokai District, central Japan, in order to investigate their relation. About 5, 000 earthquakes which occurred during the period from January 1980 to December 1981 were relocated. The double-planed structure of the intermediate-depth seismic zone is observed clearly beneath the Kanto District. The seismic zone underthrusting from Suruga trough toward the west and from Sagami trough toward the east were clearly traced throughout the troughs. Their dip angles are found to be steeper at the northern end compared with the southern part. These inclined seismic zones, however, could not be traced through the inland region to the north of the Izu Peninsula.The three-dimensional P-wave velocity structure was investigated by applying an inversion method for local earthquake data. The velocity struc-ture between 0 and 145 km in depth was modeled by six layers. Each layer was divided into 10 (latitude)×6 (longitude) blocks, the size of which is 28×26 km2. A high velocity zone appears in and around the Izu Peninsula (northernmost tip of the Philippine Sea plate) in the first layer (0-20 km) and to the east of the Izu Peninsula in the second layer (20-45 km). Low velocity zones appear beneath Tokyo Bay and in the western region in the first layer and to the northwest of the Izu Peninsula in the second layer. Coincidence of the high velocity zone with the inclined seismic zone is found under Sagami trough but such a correlation is not found under Suruga trough owing presumably to poor resolution.

A new method is proposed for determining the two-dimensional depth-distributions to the Conrad and the Moho together with station corrections and source parameters from first arrival time data. In the present study, the velocity structure is modelled by a hyperbolic function (tanh) and the configuration of each interface is expressed in a series of Chebyshev functions with coefficients which are unknown parameters. A non-linear inversion method by Tarantola and Valette is applied for inversion. An approximate ray tracing method with sufficient accuracy for a complex region is adopted for computation of seismic ray paths and travel times. The method is applied to P-wave first arrival time data from local earthquakes and explosions gathered in the seismological network by the National Research Center for Disaster Prevention to reveal the crustal structure beneath the Kanto-Tokai District, Japan. The results obtained in this study are very interesting and show the usefulness of the present method for the study of the crustal structure. The distribution of station corrections is reasonable where there is information on the shallow structure. Low-velocity zones appear around Tokyo Bay and the Cape of Omae and high-velocity zones appear in the mountainous regions. The depth to the Conrad is about 12–16 km and its lateral change is gentle. While the lateral change in the depth to the Moho is comparatively large and there is a narrow region where the depth to the Moho exceeds 40 km. Except for this region, the central-mountainous region, the agreement of the crustal structure obtained in the present study with those by the previous refraction studies is fairly good. In comparing our results to fractional velocity perturbation in Layer 1 (0–32 km) derived with Aki and Lee's method by Ishida and Hasemi, our work provides good information on the features causing the crustal velocity anomalies which have been unable to be constrained in Ishida and Hasemi's work.

All of the earthquakes which occurred in the epicentral area of the 1971 San Fernando earthquake during the period from 1960 to 1970 were relocated by using the master-event method. Five events from 1969 to 1970 are located within a small area around the main shock epicenter. This cluster of activity is clearly separated spatially from the activity in the surrounding area, so these five events are considered foreshocks. The wave forms of these foreshocks recorded at Pasadena are, without exception, very complex, yet they are remarkably similar from event to event. The events which occurred in the same area prior to 1969 have less complex wave forms with a greater variation among them. The complexity is most likely the effect of the propagation path. A well located aftershock which occurred in the immediate vicinity of the main shock of the San Fernando earthquake has a wave form similar to that of the foreshocks, which suggests that the foreshocks are also located very close to the main shock. This complexity is probably caused by a structural heterogeneity in the fault zone near the hypocenter. The seismic rays from the foreshocks in the inferred heterogeneous zone are interpreted as multiple-reflected near the source region which yielded the complex wave form. The mechanisms of the five foreshocks are similar to each other but different from either the main shock or the aftershocks, suggesting that the foreshocks originated from a small area of stress concentration where the stress field is locally distorted from the regional field. The number of small events with S-P times between 3.8 to 6 sec recorded at Mt. Wilson each month suggests only a slight increase in activity of small earthquakes near the epicentral area during the 2-month period immediately before the main shock. However, because of our inability to locate these events, the evidence is not definitive. Since the change in the wave forms is definite the present result suggests that detailed analyses of wave forms, spectra, and mechanism can provide a powerful diagnostic method for identifying a foreshock sequence.

The occurrence of large earthquakes in the Kanto-Tokai (K-T) district has been interpreted by the interaction among the Philippine Sea (PHS), the Eurasian (EUR) and the Pacific (PAC) Plates the PHS and PAC plates underthrust beneath the EUR plate in this region. In order to understand the tectonic process taking place beneath the K-T region, the seismic activity, velocity structure and source mechanisms were studied by using the data from recently developed seismological observation networks. The double-planed structure of the intermediate-depth seismic zone was observed clearly beneath the Kanto districts, and the PHS plate underthrusting from the Sagami and Suruga troughs were clearly delineated by the high velocity and high seismic region. The distribution of pressure axes corroborated the results derived from the hypocentral distribution.

世界の地震観測網が整備され,地球上に起こるマグニチュード5以上の地震は,ほとんど観測されるようになった今日でも,南極大陸およびその周辺に震源の決まった地震はない.しかし,南極各基地の地震観測の報告をみると,南極の基地だけにしか観測されないような,小さな地震が起こっているらしい.そこで,USCGSから発行されている"The Antarctic Seismological Bulletin"の資料に基づいて,震源の決定を試みた.少なくとも4点以上の基地で観測されているものにつき,その深さを33kmと仮定し,震央と発震時を求めた.震源決定を試みた地震72個のうち,24個の震央のみを,誤差±100km,O-C residual3秒以内で決めることができた.24個の地震のほとんどの震央は,大陸周辺に存在する地震帯の上に決まった.しかし,1個だけではあるが,南極大陸内に震源が決まった地震がある.この震源精度を確かめるために,多くの震源を仮定し,O-C residualの自乗平均値を調べた.その結果,この地震は大陸内20°W, 80°S,深さ1kmに起こり,マグニチュード4.3であることが分かった.現在の地震観測網では,震源を決められない小さな地震が,南極大陸に起こっていることは確実である.

According to the theory of plate tectonics proposed in the late 1960’s, the Earth’s surface is covered by about ten rigid plates, which are generated at mid-oceanic ridges, move to oceanic trenches, and there subduct into the mantle. A theory of plate tectonics explains most geologic phenomena such as seismicity, volcanism, and magnetic striping of the ocean-floor. However, the theory of plate tectonics explains only the orogenic phenomena occurring in the uppermost several hundred of kilometers of the Earth’s interior. It is, therefore, a theory covering less than 1/10 of the Earth’s diameter (left in Fig. 1). During the 1980’s studies of seismic tomography provided a new view of the Earth’s interior, and hence our understanding of the Earth’s dynamics has considerably improved. Beneath the southern Pacific a large-scale low-velocity region is present (e.g., Inoue et al., 1990), whereas beneath Asia a higher velocity region is present. The former region is characterized by a topographic high exceeding 1000 m (e.g., Crough, 1983) and also by several hot spots, suggesting the presence of large-scale mantle upwelling. The latter region corresponds to the area in which the largest amount of oceanic lithosphere has subducted during the last 300 Ma, suggesting the presence of mantle downwelling in central Asia (middle in Fig. 1). Maruyama (1994) and Maruyama et al. (1994) proposed “Plume tectonics” to explain not only the superficial layer but also dynamics in the whole Earth: Subducted slabs in the western Pacific are stagnant at the 670 km discontinuity because of the endothermic phase transition and eventually collapse to form a cold mantle downwelling to the core mantle boundary (CMB). Two large-scale mantle upwellings are present in the lower mantle under the south Pacific and Africa (hot superplume). The whole-mantle scale convection controlled by the large-scale flow is called plume tectonics and should play a major role in the Earth’s dynamics. The material circulation of the superplumes should carry not only heat but also incompatible elements in each layer, which should be identifiable in the chemical signature of hotspot magma (right in Fig. 1). The Superplume Project, funded by Japan’s Science and Technology Agency, started in 1996 to examine the above working hypothesis and establish a new view of whole Earth dynamics. The project is interdisciplinary and involves a broad range of Earth sciences: seismology, geology, ultra-high-pressure experiments, geodesy, and numerical simulation of mantle convection as described in detail below.

The spatio-temporal variation of seismicity prior to the 1971 San Fernando, California, earthquake is studied for the area within 35 km of the epicenter. During the period from 1932 to 1961, the seismicity in this area was relatively low and random. A remarkable NE-SW trending alignment of activity occurred during the period from 1961 to 1964, the period corresponding to the inferred onset of the Palmdale uplift. During the period from 1965 to 1968, the seismicity around the epicentral area became extremely low; no event was located within 13 km from the epicenter. During the period from 1969 to the occurrence of the San Fernando earthquake, activity around the epicentral area increased. This activity may be considered to be foreshock activity in a broad sense.

We determine the three-dimensional P wave velocity structure beneath the Kanto-Tokai district, central Japan, and simultaneously relocated by inverting travel time data for 91 events that were observed at 61 stations operated by the National Research Center for Disaster Prevention. The variance of the travel time data is reduced by 75%, as compared to the variance for the routinely used laterally homogeneous model. In general, our velocity model is consistent with tectonic models of the Kanto-Tokai district. Our velocity model is also in good agreement with seismic explosion data and Bouguer gravity anomalies. The uppermost layer (from 0 to 16 km depth) is characterized by lateral velocity differences of more than 20%. We find low-velocity zones beneath Tokyo Bay, the middle of the Boso Peninsula, and Omae-zaki, where there are thick sedimentary layers. A marked 30-km-thick high-velocity zone extends to the northeast and the west of the Izu Peninsula. This zone dips gradually from the surface to a depth of 70 km in the eastern region and to a depth of 40 km in the western region and is interpreted as the Philippine Sea (PHS) plate, which is being subducted at the Sagami and the Suruga troughs. The data unquestionably require a high-velocity zone beneath Mount Tsukuba that cannot be explained in terms of present tectonic models. We use our three-dimensional velocity model to relocate about 23,000 earthquakes that took place from 1983 through 1985. The accuracy of the hypocentral parameters improved considerably. As a result, the double seismic zone within the subducting Pacific plate is more clearly delineated than in previous studies. The inclined seismic zone for the subducting PHS plate coincides fairly well with the high-velocity zone of our three-dimensional velocity model.