Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Where and why do large shallow slab earthquakes occur? Tetsuzo Seno and Masaki Yoshida Earthquake Research Institute, University of Tokyo, Tokyo, Japan (submitted to J. Geophys. Res., on Jan. 18, 2001) Abstract. It has been believed that large earthquakes seldom occur within the shallow portion (20-60 km depth) of a subducting slab, because the differential stress there is generally expected to be low between bending at trench-outer rise area and unbending at intermediate-depth. However, there are several regions where large earthquakes (M≧7.0), including three events early in 2001, have occurred in this portion. Searching such events from published individual studies and the Harvard University centroid moment tensor catalogue, we find twenty events in E. Hokkaido, Kyushu-SW. Japan, S. Mariana, Manila, Sumatra, Vanuatu, N. Chile, C. Peru, El Salvador, Mexico, N. Cascadia and Alaska. Slab stresses revealed from the mechanism solutions of those large events and nearby smaller events are down-dip tensional. However, young ages of the subducting oceanic plates degrade a possibility of slab pull forces being a major cause for those large events. Except for E. Hokkaido, Manila, Sumatra and S. Vanuatu, there are pieces of evidence for a horizontal stress gradient within the upper plate in the across-arc direction. We infer that mantle drag forces operating beneath the upper plate produce the stress gradient, and drive the upper plate seaward. The upper plate driven by this drag would suck a young oceanic plate, resulting in extra-tension in the shallow portion of the slab, which might be a cause for the large shallow slab earthquakes. 1. Introduction It has been believed that large earthquakes seldom occur in the shallow portion of a subducting slab. However, early in 2001, fairly large shallow slab earthquakes succeeded on Jan. 13 (the El Salvador earthquake, Mw 7.6), on Feb. 28 (the Nisqually earthquake, Mw 6.8) in Washington, U.S. and on March 24 (the Geiyo earthquake, Mw 6.8) in Southwest Japan. Even though they were events within slabs, they caused severe damage on land areas because their hypocenters were relatively shallow. We are forced to take into account these shallow slab events for mitigation of disasters, hazard assessments of constructions, and then it would be important to understand where and why such large shallow slab earthquakes occur. We define here "shallow" portion of a slab by the depth range of 20-60 km. Events shallower than 20 km are excluded because they are located very close to a trench, which makes it difficult to discriminate them from trench - outer rise events. We also exclude events deeper than 60 km, since they are belonging to 1 so-called intermediate-depth earthquakes. Trench - outer-rise areas are generally characterized by the occurrence of earthquakes with horizontal T(P)-axes in the shallow (deep) portion (Seno and Yamanaka, 1998), consistent with the stresses produced by bending of an oceanic plate prior to subduction. Double seismic zones at intermediate-depth show usually down-dip compression (tension) in the upper (lower) zone (Hasegawa et al., 1978), consistent with unbending stresses (Engdahl and Scholz, 1977; Kawakatsu, 1986). Between bending at the trench - outer rise and unbending at the intermediate depth, a slab is usually expected to be in a low stress, which explains the scarcity of large events there. Although they are scarce, historically some large earthquakes are known to have occurred in the shallow portion of slabs. The 1931 Oaxaca earthquake (Mw 7.7) in Mexico (Singh et al., 1985), the 1949 Olympia earthquake (M 7.1) in Cascadia (Baker and Langston, 1987), and the 1970 Peru earthquake (Mw 7.9) (Abe, 1972) are such examples. Along with the recent large shallow slab events in 2001 mentioned earlier, occurrence of such events might not be so seldom as generally thought. In order to clarify how seldom they occur and in what tectonic regimes they tend to occur, we search for large shallow slab earthquakes systematically in this study. We particularly focus on stress state and age of the subducting slab and also stress state of the upper plate where those large shallow events occurred. Seno and Yamanaka (1998) showed that there is a relation between a stress in the shallow portion of a slab and a back-arc stress such that a tensional (compressional) slab tends to have a compressional (tensional) back-arc. Some arcs have tensional slabs with tensional back-arcs, violating this rule. We will show that these are the regions where large slab earthquakes tend to occur. They are also characterized by an across-arc horizontal stress gradient in the upper plate. Nakamura and Uyeda (1980) first noted this kind of a stress gradient in several arcs. Wang (2001) recently examined a stress state from a fore-arc to a back-arc in detail and concluded that the inner part of a fore-arc often has a low stress and there would be no significant change in stress state between fore- and back-arcs. In this study, we depict trajectories of σHmax (maximum horizontal stress) in the upper plate where large shallow slab events have occurred, in order to examine whether a stress gradient exists or not. Changes of σHmax directions are observed in most of the regions, implying existence of a stress gradient there. We finally discuss implications of the stress gradient for a role played by convection beneath the upper plate on the cause for large shallow slab earthquakes. 2. Large shallow slab earthquakes 2 We search for large shallow slab events from the Harvard centroid moment tensor (HCMT) catalogue for the period during 1977-2001 and from individual studies for the period prior to 1977. We restrict events with magnitudes larger than or equal to 7.0 and centroid or focal depths between 20 and 60 km. Moment magnitudes are used when available; if not, surface-wave magnitudes or their equivalents are used. We judge whether they occurred within a slab or not mainly based on their focal mechanisms, making references to relative plate motions and focal mechanisms of nearby smaller events; details of the procedure for each event are described in the following sub-sections. Table 1 lists the large shallow slab earthquakes thus selected. We include in Table 1 two recent events with moment magnitudes smaller than 7; they are the Feb. 28, 2001 Nisqually earthquake (Mw 6.8) and the March 24, 2001 Geiyo earthquake (Mw 6.8). This is because they occurred very close to historical events in the same region and are useful to clarify the faulting associated with those former events. Out of the twenty earthquakes listed in Table 1, fourteen events occurred during 1977-2001. This rate of occurrence of large shallow slab events (ca. 0.5 event per year) suggests that many slab events might have been omitted during the period prior to 1977, for which no routine determination of focal mechanisms like HCMT was available. Figure 1 shows the focal mechanisms (lower hemisphere) plotted at the epicenter of those events with event numbers (Table 1); for the three historical events near Japan (Events 2, 4 and 5), only their epicenters are plotted because reliable mechanisms are not available. Sources of the mechanism solutions are described in the caption of Table 1. The mechanism solutions in Figure 1 show both normal and reverse fault types. In the following sub-sections, we cite evidence for regarding those events as fracture within the shallow portion of a slab. We also show in more detail the stress state of the subducting slab for each region using published focal mechanism solutions of nearby smaller events. P- and T-axes of nearby smaller events (Mw>=5.8) using HCMT are plotted in cross-sections (Appendix, Figure A1). We also delineate σHmax directions in the upper plate for each region. When σHmax trajectories are available, we use them directly; these are for Kyushu-SW. Japan and Alaska. In other cases, we draw trajectories using published stress indicators, such as focal mechanisms, Quaternary active faults, dikes, monogenic volcanic centers, and in-situ stress measurements. Data sources are described in each sub-section. We call a stress state compression (tension) if a horizontal stress is larger (smaller) than a vertical stress. 3. Slab and upper plate stresses 3.1. Eastern Hokkaido 3 In and around Hokkaido, the Pacific plate is subducting beneath the Okhotsk plate with a rate of 79 mm/yr in the WNW direction (Seno et al., 1996). Off the east coast of Hokkaido, a large shallow earthquake occurred on Oct. 4, 1994 (the Hokkaido-toho-oki earthquake, Mw 8.3), having a reverse fault type mechanism solution different from typical underthrust types in this region (Event 1 in Figure 1a, HCMT; see also Kikuchi and Kanamori, 1995; Tanioka et al., 1995). The T-axis of the mechanism solution is dipping to the NE, subparallel to the slab dip. The hypocentral depth of this event was determined to be 33 km, along with the aftershock depths extending between 0-80 km, by the microearthquake network of Hokkaido University (Katsumata et al., 1995). Kikuchi and Kanamori (1995) and Tanioka et al. (1995) determined centroid depths of this event as 56 km and 50 km, respectively, from the body-wave analysis, which are shallower than the 68 km depth of HCMT. The mechanism solution along with the nearly vertical aftershock distribution leads to the interpretation that this event was fracture within the shallow portion of the slab. The age of the subducting plate near the epicenter is 123 Ma (Nakanishi et al., 1992). T-axes of nearby smaller events show that the slab is dominantly down-dip tension (Figure A1), consistent with the results of previous studies (Suzuki et al., 1983; Umino et al., 1984; Kosuga et al., 1996). Figure 2 shows the σHmax stress trajectories in Hokkaido. The small bars indicate P-axes of the focal mechanisms of shallow earthquakes occurring in the crust (Moriya, 1986; Kosuga, 1999), which were used to draw the trajectories. The stress regime is mostly of strike-slip fault type with E-W σHmax , except for the Hidaka area and its north where reverse fault types are dominant and Oshima Peninsula where σHmax is directed in the NW. The reverse faulting in the Hidaka area is probably representing the collision of the Kuril fore-arc sliver to western Hokkaido (Kimura, 1981; Seno, 1985; Tsumura et al., 1999). The NW σHmax in Oshima Peninsula may be due to a stress disturbance at the arc-arc junction. Except for these areas, the σHmax direction is oblique to the arc. 3.2. Kyushu-Southwest Japan The Philippine Sea plate is subducting along the Nankai Trough beneath SW. Japan and Kyushu with rates of 40-50 mm/yr in the WNW direction (Seno et al., 1993). The age of the subducting plate is 15-30 Ma along the Nankai Trough (Okino et al., 1994) and 50 Ma at the junction with the Ryukyu Trench near Kyushu (Seno, 1988). To select historical slab events in this region, we use the Catalogue of Major Slab Earthquakes near Japan (Association for the Development of Earthquake Prediction, 2000), in addition to individual studies. We restrict events since 1890 around when the Japan Meteorological Agency started systematic instrumental observations over Japan. 4 Three large (M>=7) earthquakes are listed; they are the 1899 Kii-Yamato earthquake (Event 2, M 7.0), the 1905 Geiyo earthquake (Event 4, M 7.2) and the 1931 Hyuganada earthquake (Event 5, M 7.1). The Kii-Yamato earthquake (Event 2) occurred beneath the southeast coast of Kii Peninsula (Figure 3). Nakamura (1997) inferred that this event occurred within the Philippine Sea slab at the 40-50 km depth, based on the fact that severe damage was localized at the epicenter, but strong shaking was felt in a wide area, which is a characteristic feature to the slab events in this area. The 1905 Geiyo earthquake (Event 4) occurred northwest of Shikoku. Close to the epicenter of this earthquake, a large earthquake occurred on March 24, 2001 (the 2001 Geiyo earthquake, Event 3, Mw 6.8) at the depth of 47 km, having a normal fault type mechanism with a W dipping T-axis (Figure 1a, HCMT). This event is regarded as fracture within the Philippine Sea slab because its focal mechanism is similar to those of smaller slab events in this region as mentioned later. We suppose that the 1905 Geiyo earthquake was a similar slab event to the 2001 event, since the iso-intensity map and the damage pattern are similar for both events, and estimate its depth around 50 km. The 1931 Hyuganada earthquake (Event 5) occurred under the sea between Kyushu and Shikoku at the depth of 40 km. Ichikawa (1971) suggested a normal fault type mechanism for this event. Near this event, two normal fault type earthquakes with W dipping T-axes occurred on August 6, 1984 (Mw 6.9, depth = 29 km) and on March 18, 1987 (Mw 6.6, depth = 38 km) (HCMT). We infer that the 1931 event had a similar focal mechanism to these smaller events and was fracture within the slab. Beneath SW. Japan, microearthquake activities are dipping to the NNW with a low angle to the depth of 60-80 km (Ukawa, 1982; Nakamura et al., 1997; Matsumura, 1997), and beneath Kyushu, they are dipping to the W with a high angle to the depth of 180 km (Shimizu et al., 2000). Beneath SW. Japan, the T-axes of small events within the slab are sub-horizontal and subparallel to the E-W strike of the slab (Ukawa, 1982; Matsumura, 1997). Moderate-size slab events beneath Kyushu have W-dipping T-axes along the slab (Seno, 1999). The stress state of the slab from Kyushu to SW. Japan is thus regarded as tension in the E-W direction along the slab. The T-axes plotted in Figure A1 also support this. Figure 3 shows the σHmax trajectories of the upper plate from Kyushu to SW. Japan (Seno, 1999). In S. Kyushu, the back-arc has a stress state of strike-slip fault type with NE σHmax . The fore-arc has a stress state of reverse fault type with NW σHmax . However, the latter σHmax direction was derived from the dike-volcano data in Nakamura and Uyeda (1980) and the fault type is not well constrained. CMT solutions from recent small and moderate-size events in the Ryukyu fore-arc show normal faulting with along-arc T-axes (Kubo and Fukuyama, 2001). If this stress regime extends to SE. Kyushu, the stress state of the fore-arc might be of normal fault type. In N. Kyushu and SW. Japan, σHmax is directed in the E-W. In SW. Japan and the eastern part of N. 5 Kyushu, strike-slip faulting is dominant, and in the westernmost part of N. Kyushu, normal faulting is dominant (Seno, 1999). 3.3. Southern Mariana The Pacific plate is subducting beneath the Philippine Sea plate along the Mariana Trench. However, due to the opening of the Mariana Trough, convergence direction of the Pacific plate beneath the Mariana fore-arc differs from that beneath the Philippine Sea plate (Seno et al., 1993). With the velocity between the Mariana fore-arc and Eurasia determined by GPS (Kotake, 2000) and the NUVEL1 Eurasia-Pacific velocity (DeMets et al., 1990), the convergence of the Pacific plate is calculated to be in the WNW direction with a rate of 76 mm/yr near Guam (Figure 4). The age of the subducting plate near Guam is 164 Ma (Nakanishi et al., 1992). On August 8, 1993, a large earthquake (Event 6, Mw 7.7) occurred south of Guam, having a reverse fault type mechanism with a nodal plane shallowly dipping to the NNW and an auxiliary plane dipping steeply to the SSE (Figure 1a). The HCMT centroid depth is 59 km and Tanioka et al. (1995) determined the source depth ranging 40-50 km from a body wave analysis. Because the NNW-SSE slip vector of this event is much different from the plate convergence direction described above, we regard this event as fracture within the subducting Pacific slab. The seismicity in the southernmost part of the Mariana arc where the 1993 event occurred is shallower than 200 km (Eguchi, 1984). Mechanism solutions of nearby smaller events are similar to that of the 1993 event (Figure A1; see also Eguchi, 1984), indicating that the short slab beneath the southern Mariana arc is down-dip tensional. Although data for stress state of the upper plate are scarce because the area is mostly submarine, the back-arc spreading in the Mariana Trough manifests that the back-arc is tensional with along-arc σHmax (Figure 4). In the fore-arc, Martines et al. (2000) depicted normal faults striking perpendicular to the arc, using side-scan sonar images. This indicates that σHmax is in the across-arc direction in the fore-arc (Figure 4). 3.4. Manila The South China Sea is subducting beneath Luzon with a rate of ca. 50 mm/yr in the ESE direction (Yu et al., 1999). On Dec. 11, 1999, a large earthquake (Event 7, Mw 7.2) with a normal fault type mechanism occurred east of the Manila Trench northwest of Manila at the depth of 35 km (HCMT), with an E dipping T-axis. The age of the subducting plate is 22 Ma near the event (Taylor and Hayes, 1980). The seismicity associated with the slab is shallower than 250 km along the Manila Trench (Seno and Kurita, 1978; 6 Cardwell et al., 1980). Mechanism solutions of nearby smaller events show T-axes more or less similar to that of the large event (Figure A1; see also Cardwell et al., 1980), indicating that the slab is down-dip tensional and the 1999 event was fracture within the slab. The stress state of the upper plate is characterized by across-arc compression as evidenced by the NW striking left-lateral strike-slip Philippine fault. However, the notion of fore- and back-arcs in the Philippines is obscured due to the subduction from both sides of the archipelago, and σHmax trajectories are not drawn in this region. 3.5. Sumatra The Australian plate is subducting beneath southern Sumatra with a rate of 70 mm/yr in the NNE direction (DeMets et al., 1990). The age of the subducting plate is 66 Ma (Liu et al., 1983). On June 4, 2000, a large earthquake (Event 8, Mw 7.8) occurred off southern Sumatra at the depth of 44 km (HCMT), having a strike-slip fault type mechanism with a NE dipping T-axis. The seismicity associated with the slab is shallower than 200 km in this region (Slancova et al., 2000; Newcomb and McCann, 1987). Mechanism solutions of nearby smaller events show similar T-axes to this event (Figure A1; see also Slancova et al., 2000), indicating that the slab is down-dip tensional and the 2000 event was fracture within the slab. The stress state of the upper plate is characterized by N-S σHmax as evidenced by the NW striking right-lateral strike-slip Semanko fault; the σHmax direction is thus oblique to the arc trend. We do not draw σHmax trajectories in this region due to the paucity of data. 3.6. Vanuatu Beneath the New Hebrides fore-arc, the Australian plate is subducting with rates of 100-150 mm/yr in the NEE direction (Louat and Pelletier, 1989). The arc changes its strike to the nearly E-W direction at the southern end, where the plate boundary becomes a transform fault in the east as the Hunter Fracture zone. Two large earthquakes with focal mechanisms different from the underthrust-type occurred in this region: one on July 13, 1994 (Event 9, Mw 7.1) in the central part of the arc at the depth of 25 km (HCMT), and the other on July 6, 1981 (Event 10, Mw 7.5) near the southern corner at the depth of 58 km (HCMT). The ages of the subducting plate are 52 Ma at Event 9 (Weissel et al., 1982) and 35 Ma at Event 10 (Malahoff et al., 1982). The seismicity associated with the slab is steeply dipping to the east and shallower than 300 km, which becomes shallower than 150 km to the south (Coudert et al., 1981). T-axes of the mechanism 7 solutions of the large events are oblique to the slab dip. Mechanism solution of nearby deeper smaller events are down-dip tensional (Figure A1; see also Coudert et al., 1981). Although the T-axes of the large events are not similar to those of the smaller events, we regard the large events as fracture within the slab, because their mechanism solutions are much different from the underthrust type. West of the epicentral location of Event 9, the d'Entrecasteaux Ridge is colliding with the New Hebrides fore-arc. The back-arc in this place is in compression as revealed by the thrust type HCMT solutions (Figure 5; Charvis and Pelletier, 1989). Further to the east, the North Fiji Basin is currently opening with the N-S striking ridge axis, as evidenced by the magnetic anomaly lineations, diffuse seismic activities, and hydrothermal features (Malahoff et al., 1982; Hamburger and Isacks, 1988; Auzende et al., 1988; Tanahashi et al., 1994). Figure 5 shows the σHmax trajectories in the New Hebrides - N. Fiji Basin region, drawn from the spreading axis and the focal mechanisms in Charvis and Pelletier (1989). The σHmax direction changes from the N-S in the center of the North Fiji Basin to the E-W in the New Hebrides arc. The place where Event 10 is located does not seem to be a typical subduction zone and the stress state of the upper plate above this event is difficult to depict. 3.7. Northern Chile The Nazca plate is subducting beneath the S. American plate in N. Chile with a rate of 84 mm/yr in the NEE direction (DeMets et al., 1990). Here on Feb. 23, 1965, a large earthquake (Event 11, Mw 7.0) having a normal fault type mechanism with an E dipping T-axis occurred at the depth of 60 km (Malgrange and Madariaga, 1983). The age of the subducting plate at Event 11 is 48 Ma (Herron, 1972). The seismicity associated with the slab beneath S. America is split into the intermediate-depth and deeper ones (Barazangi and Isacks, 1976). In N. Chile, intermediate depth earthquakes are distributed down to the 300 km depth with a dip angle of 20-25° (Barazangi and Isacks, 1976). Smaller events mostly have E dipping T-axes similar to the large event (Figure A1; see also Malgrange and Madariaga, 1983; Comte and Suarez, 1995), and focal mechanism solutions of intermediate-depth earthquakes also show that the slab is down-dip tensional (Stauder, 1973; Comte and Suarez, 1995). We then regard the large event as fracture within the slab. However, there are a few smaller events with down-dip P-axes (Figure A1), and seismic activities revealed by local microearthquake networks show more complexities such as double seismic zones (Comte and Suarez, 1994; Comte et al., 1999). Beneath N. Chile, though the slab stress shows a tendency of down-dip tension, its magnitude might not be large. The stress state of the upper plate will be described in the next sub-section along with Central Peru. 8 3.8. Central Peru On May 31, 1970, a large earthquake (Event 12, Mw 7.9) occurred off the coast of C. Peru at the depth of 43 km, having a normal fault type mechanism with an E dipping T-axis (Abe, 1972). The age of the subducting plate at Event 12 is 44 Ma (Herron, 1972). In north - central Peru, the slab seismicity shows flattening at the 100-150 km depth range (Hasegawa and Sacks, 1981). Mechanism solutions of nearby smaller events in the shallow and intermediate-depths show generally down-dip T-axes similar to that of the large event (Figure A1; see also Stauder, 1975; Hasegawa and Sacks, 1981), and we regard the large event as fracture within the slab. However, some earthquakes with down-dip P-axes have also been found (Stauder, 1975; Isacks and Barazangi, 1977). This indicates that the tensional stress of the slab might not be large, as in N. Chile. Figure 6 shows the σHmax trajectories in S. America drawn from the World Stress Map data (Zoback, 1992; Assumpcao, 1992). The sub-Andes is characterized by reverse faulting with E-W σHmax (See also Stauder, 1975; Suarez et al., 1983). The high Andes, Altiplano, Cordillera Blanca and Pacific lowland are characterized by normal faulting mostly with N-S T-axes (see also Sebrier et al., 1985; Suarez et al., 1983). The normal faulting can be explained by the high topography and thicker crust in the high Andes-Altiplano and Cordillera Blanca (Dalmayrac and Molnar, 1981; Froidevaux and Isacks, 1984; Coblentz and Richardson, 1996; Meijer et al., 1997; Liu and Yang, 2000), and the whole Andes area, if it had a normal crustal thickness, must turn to be in compression, similarly to the sub-Andes. The Quaternary normal faults in the Pacific coast (Sebrier et al., 1985) remain as an enigma because the crustal thickness there is not large. Because three reverse fault type earthquakes with E-W P-axes occurred along the coast of Peru and N. Chile (Figure 6), we guess that the present stress state along the coast is E-W compression. To the east across the continent, several reverse and strike-slip faults are seen in the Proterozoic-Archean shield area, and at the east coast, a few strike-slip and normal faults with N-S or E-W σHmax appear (Figure 6, Assumpcao, 1992; Zoback, 1992). The tension at the east coast along with the E-W compression in the sub-Andes indicates that there is a horizontal stress gradient across the S. American continent. 3.9. El Salvador The Cocos plate is subducting beneath the Caribbean plate near El Salvador with a rate of 79 mm/yr in the NNE direction (DeMets et al., 1990). On June 19, 1982 and January 13, 2001, two large earthquakes 9 (Event 13, Mw 7.3 and Event 14, Mw 7.7) occurred off the coast of El Salvador, at the depths of 52 and 56 km, respectively (HCMT). Both events had normal fault type focal mechanisms with NE dipping T-axes (HCMT). The age of the subducting plate off El Salvador is unknown, but would be older than 37 Ma, judging from the magnetic anomalies further offshore (Herron, 1972). The seismicity associated with the slab is shallower than 300 km, and has a steep dip angle of 60-70° below 50 km (Burbach et al., 1984). Nearby smaller events have similar focal mechanisms to the large events (Figure A1, see also Burbach et al., 1984), and we regard the large events as fracture within the slab. The stress state of the upper plate between the coast and the volcanoes is of strike-slip fault type with NW σHmax (World Stress Map, Zoback, 1992). Since the arc is elongated in the NW direction, the fore-arc is in across-arc tension in this region. We do not draw stress trajectories in this region due to the paucity of data. 3.10. Mexico The subduction zone off Mexico is divided into the Jalisco, Michoacan, Guerrero, and Oaxaca blocks from north to south (Figure 7, inset). The Rivera plate is subducting beneath the Jalisco block with rates of 20-50 mm/yr in the NE direction (Kostoglodov and Bandy, 1995), and the Cocos plate beneath the other three blocks with rates of 50-70 mm/yr in the NE direction (DeMets et al., 1990). On Jan. 11, 1997, a large earthquake (Event 17, Mw 7.1) occurred beneath the Michoacan block at the depth of 40 km (HCMT), and on Sept. 30, 1999, another large earthquake (Event 15, Mw 7.4) occurred beneath the Oaxaca block at the depth of 47 km (HCMT). Both events had normal fault type mechanisms with NNE dipping T-axes. Historically similar large earthquakes occurred beneath the Michoacan block in 1858 (Singh et al., 1996), and beneath the Oaxaca block in 1931 (Event 16, Singh et al., 1985). The 1931 event had a normal fault type mechanism similar to the recent large events (Singh et al., 1985). The age of the subducting slab at these events are 10-17 Ma (Klitgord and Mammerickx, 1982). The seismicity associated with the slab is shallower than 100 km, and has a steep dip angle (~50°) beneath the Jalisco block and a gentle dip angle (20-30°) beneath the Michoacan and Oaxaca blocks. Beneath the Guerrero block, the slab is flattened at 40-70 km depth range. Smaller earthquakes show T-axes more or less dipping to the NE or horizontal along the slab, similar to the large events (Figure A1; see also Suarez et al., 1990; Pardo and Suarez, 1995). Although the stress axes of the smaller events in Figure A1 are somewhat disturbed and a few earthquakes with down-dip P-axes have also been found within the slab beneath central S. Mexico (S. K. Singh, personal comm., 2000), we regard the large events as fracture within the slab. At depths larger than 60 km, several large earthquakes have occurred within the slab (Mikumo, 2000). They 10 have normal fault type focal mechanisms with down-dip or horizontal T-axes. The 1980 Huajuapan de Leon earthquake (Mw 7.0) at the depth of 65 km exceptionally had a S dipping T-axis (Yamamoto et al., 1984). Figure 7 shows the σHmax trajectories in the Jalisco-Michoacan blocks (Seno and Singh, 2001). The stress state of the upper plate is mostly of normal fault type. Along the Trans Mexican Volcanic Belt (TMVB), the σHmax direction is subparallel to the belt as seen in the strikes of active normal faults (Sutter, 1991; 1992) and focal mechanism solutions (J. F. Pacheco, personal communication, 2000). Borehole breakout data show that the σHmax direction changes to the NS in the northeast of TMVB (Sutter, 1991). To the south of TMVB, alignment of monogenic volcanic centers (Hasenaka, 1994) and strikes of normal faults in the Colima rift and the other two rifts (Allan, 1986; Sutter, 1991) indicate that the σHmax direction changes to the across-arc one (Figure 7). A few normal fault type small earthquakes within the overriding plate south of TMVB (Singh and Pardo, 1993; J. F. Pacheco, personal communication, 2000) are consistent with the across-arc σHmax in the fore-arc. The stress trajectories of reverse fault type near the coast in Figure 7 are inferred from the coastal mountains such as Sierra Madre del Sur. In some places, a normal fault type stress state may reach the coast as seen in the Colima Rift. 3.11. Northern Cascadia The Juan de Fuca plate is subducting beneath the North American plate in Cascadia with rates of 35-45 mm/yr in the NE direction (Riddihough, 1984; Wilson, 1993). On April 13, 1949, a large earthquake (Event 19, M 7.1) occurred in the Puge Sound area, western Washington, at the depth of 54 km. This event had a normal fault type mechanism with a SE dipping T-axis (Baker and Langston, 1987). Recently on Feb. 28, 2001, the Nisqually earthquake (Event 18, Mw 6.8) occurred 20 km northeast of the 1949 event at the depth of 47 km (HCMT), having a similar focal mechanism to the 1949 event, but had an E dipping T-axis (HCMT). The age of the Juan de Fuca plate off Washington is 10 Ma (Wilson, 1993). The seismicity associated with the slab is shallower than 100 km (Crosson and Owens, 1987; Ma et al., 1996). For smaller events, one solution available from HCMT shows a similar focal mechanism to the larger events (Figure A1). One moderate-size event on April 29, 1965 (M 6.5) at the depth of 59 km also had a similar focal mechanism (Algermissen and Harding, 1965). Microearthquakes occurring within the slab beneath western Washington show a slight tendency of E dipping T-axes (Ma et al., 1996). These all indicate that the large events were fracture within the slab and the slab is down-dip tensional beneath the northern part of Cascadia. Small earthquakes within the upper plate in Puge Sound west of the volcanic front have N-S horizontal 11 P-axes with T-axes distributed on a great circle in a vertical plane with an E-W strike (Ma et al., 1996). This indicates that the stress state of the inner fore-arc is nearly neutral in the across-arc direction with the along-arc σHmax (see Wang et al., 1995 for details). P-axes parallel to the arc have also been obtained for small earthquakes in the upper plate in the Vancouver island (Wang et al., 1995). We do not draw σHmax trajectories in this region, because the back-arc stress is complex (Zoback, 1992). 3.12. Alaska The Pacific plate is subducting beneath Alaska with a rate of ca. 60 mm/yr in the NNW direction (DeMets et al., 1990). The age of the subducting plate is around 55 Ma (Pitman et al., 1974). On December 6, 1999, a large earthquake (Event 20, Mw 7.0), having a strike-slip fault type mechanism solution with a NW dipping T-axis, occurred in the Kodiak Island at the depth of 54 km (HCMT). Hansen and Ratchkovski (2001) relocated this event at the 36 km depth, along with many aftershocks, using a joint hypocentral determination method. The focal mechanism solution along with the nearly vertical distribution of the relocated aftershocks indicates that this event was fracture within the slab. Nearby small events show similar T-axes to the large event (Figure A1). Figure 8 shows the σHmax trajectories in western Alaska (Nakamura et al., 1980). From the fore-arc to the rear side of the volcanic front, σHmax is in the across-arc direction, with a reverse fault type stress state in the fore-arc, and with a strike-slip type in the rear side of the volcanic front. Further to the NW, σHmax is in the along-arc direction with a normal fault type stress state. 4. Discussion 4.1. Slab stresses As we have shown in the previous sub-sections, T-axes of the large shallow slab events are more or less in the down-dip direction of the slab. Nearby smaller events have generally similar T-axes. This indicates that the slab is down-dip tensional in these regions. However, this does not necessarily mean that a slab pull force is dominant there. For N. Cascadia, SW. Japan, and Mexico, ages of the subducting slab are younger than 20 Ma. Since it is known that until 20 Ma, negative buoyancy of the slab does not overcome the positive buoyancy of the oceanic crust-harzburgite layer (Davies, 1992), slab pull forces in these areas would be negligible. In other regions, slabs are older than 20 Ma. However, they are not so old except for E. 12 Hokkaido and S. Mariana where the old Pacific plate is subducting (Table 1). We therefore believe that the slab pull force is not a major cause for the occurrence of these large shallow slab events. 4.2. Horizontal stress gradient in the upper plate We have seen changes in σHmax direction in many of the regions where large shallow slab events have occurred. Nakamura and Uyeda (1980) compiled several areas which have similar σHmax direction changes using Quaternary active faults, earthquake mechanisms, dikes and volcanic center alignments, and suggested existence of a horizontal stress gradient in the upper plate. Recently Wang (2001) showed that stress state of the fore-arc is in tension in many arcs due to a low coupling at the thrust zone and a gravitational collapse of the fore-arc topography, and there is not a significant stress gradient between the fore-arc and the back-arc. We here re-examine this issue, i.e., whether there is a stress gradient in the upper plate or not, by considering possible causes for the observed σHmax direction change. There are two possible sources for stress gradient; one is a crust-plate structural variation, and the other is shear traction at the base of the upper plate. The former is sometimes referred to as a gravitational potential energy force in recent literature (e.g., Sandiford and Coblentz, 1994; Jones et al., 1996). In this case, a differential force (a horizontal stress minus a vertical stress integrated over the plate thickness) at one point with respect to a reference point can be calculated by an integral of ∆ρ(z)z over the plate thickness, where ∆ρ(z) is density difference between these two points and z is the depth (Artyushkov, 1973; Parsons and Richer, 1980; Fleitout and Froidevaux, 1982). For a continental plate, crustal thickness change is the most important factor to affect the differential force, but change in the mantle structure also contributes to this. This kind of the differential force is the origin for the extension in High Himalayas-S. Tibet and Altiplano-Cordillera Blanca which have a very thick crust. In arcs, it produces more tension, in general to the landward, in the thicker crust - hotter plate area (Froidevaux et al., 1988; Seno, 1999). Figures 9a and b illustrate the differential stress change across the arc for this case. The level of σxx - σzz at the aseismic front, i.e., the landward edge of the thrust zone (Yoshii, 1975), is determined by the coupling at the thrust zone and the crustal thickness variation across the outer fore-arc (Seno and Yamanaka, 1998; Seno, 1999; Wang, 2001). We assume the differential stresses, σxx - σzz and σyy - σzz , are negative in this figure. From the aseismic front toward the back-arc, they change gradually due to a crust-plate structural change by an amount of less than a few hundred bars (Froidevaux et al., 1988; Seno, 1999). Because the plate thickness does not change much, the magnitude of the horizontal stresses itself neither changes much (Figure 9a). Thus the horizontal stress gradient is small for both x- and 13 y-directions, producing no exchange in σHmax direction. In contrast to this, σHmax changes its direction in many arcs as we have seen in S. Kyushu, S. Mariana, Vanuatu-N. Fiji Basin, Mexico, and Alaska. In these regions, the back-arc side has along-arc σHmax and the fore-arc side has across-arc σHmax generally. However, the boundary between these two stress regimes varies from a region to another region; for example in Mexico the boundary is located in the fore side of the volcanic front, and in Alaska it is in the rear side. Wang (2001) and Kubo and Fukuyama (2001) argued that the arc-parallel tension in the fore-arc may result from its seaward motion due to the back-arc spreading behind, producing the apparent rotation of the σHmax direction. However, this explanation might not be applicable in general since back-arc spreading does not occur in most of the arcs cited above and, even if a trench retreat occurs, the entire arc would suffer from the same strain due to the seaward migration, not only the fore-arc. Also note that arc's curvatures are small for Kyushu, Mexico, and Alaska, which makes the effect of the seaward migration to be small. We therefore believe that the σHmax direction change implies a horizontal stress gradient in the across-arc direction as shown in Figure 9c. Since such a stress gradient cannot be produced by interaction between plates or crust-plate structural change, drag forces operating at the base of the upper plate would be a most likely cause (Figure 9d), as Seno (1999) suggested for S. Kyushu. The drag forces would increase σxx seaward, and produce the desired σHmax direction change. Even in some regions without a σHmax direction change, we can recognize a horizontal stress gradient. In S. America, the sub-Andes is more compressional than the east coast as we have seen in Figure 6. Further to the east, there is a spreading center in the Atlantic. We infer that mantle drag forces are similarly operating at the base of the S. American continent, in a much larger scale than shown in Figure 9d. In N. Kyushu, σHmax is directed in the E-W, and the eastern part has a strike-slip fault type stress and the western part has a normal fault type (Figure 3). Since the crustal thickness does not change between these two areas, a plausible cause for the change in fault type would be a horizontal stress gradient, i.e., more compression in the east than in the west, consistent with the σHmax direction change in S. Kyushu (Seno, 1999). The tectonic situation where large shallow slab events tend to occur is thus summarized in Figures 10a and b. The drag forces at the upper plate base cause tension in the rear side and compression in the outer fore-arc. This compression should be balanced with tension within the shallow portion of the slab (see Seno and Yamanaka, 1998 for the force balance). The supposed drag forces would be due to flows within the mantle wedge beneath the arc or continent. The expected flows might not be a simple secondary flow associated with subduction (Hsui and Toksoz, 1981), which are common to all subduction zones (Nakamura and Uyeda, 1980). We advocate convection currents associated with thermal or chemical density anomalies, although their detailed nature remains unknown at present. In the case of S. America, the mantle convection 14 current originated from the accumulated heat beneath Gondowana would have caused the continent to drift toward the west, overriding the Nazca plate (Figure 10b). For Kyushu-SW. Japan and Cascadia where kinematics of the fore-arc is better known, we can depict more realistic cartoons of the tectonic situation (Figures 11a and b). A mantle upwelling has been expected in the South China Sea west of Kyushu from conductivity anomaly, seismic tomography, and OIB type volcanism (see Seno, 1999 for details). Recent GPS data show that the fore-arc of Kyushu and SW. Japan are migrating to the SE and NEE, respectively, with rates of more than 10 mm/yr relative to Eurasia (Ozawa et al., 1999), probably dragged by the lateral flow from the upwelling (Figure 11a, Seno, 1999). SW. Japan is then colliding with C. Honshu producing the E-W σHmax (Figure 3). Cascadia is likely to be a mirror image of Kyushu-SW. Japan. There might be a mantle upwelling in the Basin and Range (Gough, 1984) and paleomagnetic and GPS data show that the W. Oregon block has been migrating to the NW with respect to N. America, resulting in the collision with the Canadian Coastal Mountains (Figure 11b, Wells et al., 1998). The N-S σHmax in the Puge Sound area may be caused by this collision, which is similar to C. Japan. Both regions have down-dip tensional stresses in the slab, which are indicated by the diverging solid arrows in Figure 11, and thus belong to the tectonic regime shown in Figure 10a. 4.3. A cause for large shallow slab earthquakes The fact that most large shallow slab earthquakes have occurred in the tectonic situation illustrated in Figure 10 suggests a possible cause for these events. We have already degraded the possibility of a slab pull force for their cause. The numerical simulations of mantle convection involving continental plates, which are stable to convection, have been conducted (e.g., Lowman and Jarvis, 1993; Nakakuki and Honda, 2001). After the breakup of a supercontinent, a continent drifts over an oceanic plate (Figure 12a). The streamlines corresponding to an oceanic plate are deflected under the advancing continent, and the contour interval at the shallow part of the subducting slab, marked by the arrow in Figure 12a, becomes narrower. We call this phenomena "suction" by a continental plate. In this case, a continental plate sucks an oceanic plate, in contrast to the opposite case proposed by Elsasser (1971). The numerical simulation by Nakakuki and Honda (2001) involving a continent migrating seaward shows that the shallow portion of the slab is down-dip tensional (Figure 12b). Taking into account the fact that large shallow slab events tend to occur in the tectonic regime shown in Figure 10, we suppose that excess tension produced by the suction of an oceanic plate by a continental plate is a cause for large shallow slab events. 15 4.4 Exceptional cases In Manila, the back-arc is in the across-arc compression. In addition to the subduction of the South China Sea beneath Luzon from the west, the Philippine Sea is subducting along the Philippine Trench from the east. Thus the upper plate is sandwiched between the subducted slabs from both sides. In this case, an extra slab pull force, if it exits, can be supported by a large compressional stress within the upper plate. We cannot, however, find at present any reason for existence of such an extra slab pull force because the subducting South China Sea is young. In Sumatra and E. Hokkaido, horizontal stress axes of the upper plate are oblique to the arc trend, and it is difficult to discuss them in a simple two dimensional vertical cross-section. Including S. Vanuatu, causes for the large shallow slab events in these areas remain unresolved for future elaborate studies. 4.5 Relation to dehydration embrittlement Foci of the large shallow slab events are generally greater than 30 km (Table 1). This implies that the shear strength expected from the Coulomb-Navier fracture criterion is too large compared with the magnitude of tectonic stresses (a few kbars at most). A weakening mechanism is necessary to produce these slab events, similarly to produce intermediate-depth and deep earthquakes (See Kirby, 1995 for review). Dehydration embrittlement (Raleigh and Paterson, 1965) is a viable mechanism for intermediate-depth events (Kirby et al., 1996; Seno and Yamanaka, 1996; Peacock and Wang, 1999; Peacock, 2001; Seno et al., 2001). Then this might also be applicable to shallow slab events. If we accept the dehydration embrittlement hypothesis, there should be two types of origin for shallow slab events; one is due to dehydration of metamorphosed oceanic crust, and the other is dehydration of serpentinized slab mantle. Fault dimensions of large shallow slab events are on the order of tens km which exceed generally the oceanic crust thickness, and must involve fracture of the slab mantle. This implies that the slab where such large events occur should be serpentinized in the mantle portion. In SW. Japan, large slab events have occurred in Kii Peninsula and near Kyushu (Figure 3). These are the areas where Seno et al. (2001) inferred serpentinization of the Philippine Sea slab on the basis of seismic evidence. For major aftershocks of the 2001 Geiyo earthquake, Seno (2001) preliminarily examined existence of later phases traveling within the oceanic crust (see Ohkura, 2000 for a detailed study in this region), and found no such later phases, which suggests that this event might have ruptured the slab mantle. For the 2001 Nisqually earthquake, Creager et al. (2001) examined the aftershock distribution and favored a fault 16 plane lying horizontally within the oceanic crust. Because these two recent events are rather small (Mw 6.8), there remains an ambiguity whether these events have ruptured the mantle part of the slab or not. For many large events listed in Table 1, however, there should be little doubt that their rupture has extended into the mantle part, and thus serpentinization of the slab mantle must be a necessary condition for the occurrence of such large shallow slab earthquakes. 5. Conclusions We list (M>=7.0) earthquakes which occurred in the shallow portion (20-60 km depth) of slabs by searching published individual studies and the Harvard University CMT catalogue. The regions where such events are found are E. Hokkaido, Kyushu-SW. Japan, S. Mariana, Manila, Sumatra, Vanuatu, N. Chile, C. Peru, El Salvador, Mexico, N. Cascadia, and Alaska. These large events had focal mechanisms with down-dip T-axes, consistent with those of nearby smaller events. However, most of the slabs are young and apparently have no large slab pull force. The upper plates with these large shallow slab events are generally characterized by a horizontal stress gradient in the across-arc direction. This stress gradient is best manifested by the σHmax direction change from the along-arc to the across-arc from the back-arc side to the trench side. We infer that mantle drag forces beneath the upper plate are operating in these regions, driving the upper plate oceanward. These drag forces beneath the upper plate might have originated from small-scale or large-scale mantle convection currents. We propose that the upper plate driven by the convection currents overrides an oceanic plate, and suck the slab to be in extra tension, which may result in large shallow slab earthquakes. The place where large shallow slab earthquakes are expected in the future would thus be a region which has an across-arc stress gradient in the upper plate. Furthermore, if we accept the dehydration embrittlement hypothesis for large shallow slab events, the slab mantle in this region must be hydrated to form serpentine. Appendix P- and T-axes of HCMT solutions (Mw>=5.8) are plotted in a cross-section along a great circle passing through one of the large shallow events (Table 1) in the across-arc direction; events within a total width of two hundred km are plotted. Only the width for section "b" is extended to 500 km, in order to cover the more than one large slab events. The section is shown by the dotted line in Figure 1 with an alphabet corresponding to each cross-section. The P- and T-axes are projected on the vertical plane; thus a shorter 17 length indicates more obliquity to the plane. The P- and T-axes with small dots are those of the large shallow slab events in each section. Acknowledgments. We thank Ruth Ludwin and Natalia Ratchkovski for providing us useful information on the slab events in N. America, and Takeshi Mikumo, Krishna Singh, and Javier Pacheko for the slab events and stress data in Mexico, and Masahiro Kosuga, and Kazuhiko Goto for the stress data in Japan, and Masayuki Kikuchi for providing one of the authors (T.S.) a chance to visit Mexico, and Atsuki Kubo, Katsuhiko Ishibashi, and Yoji Kobayashi for discussion. References Abe, K., Mechanisms and tectonic implications of the 1966 and 1970 Peru earthquakes, Phys. Earth Planet. Interiors, 5, 367-379, 1972. Algermissen, S. T., and S. T. Harding, The Puget Sound, Washington earthquake of April 29, 1965, Preliminary Seism. Rep., 1-26, 1965. Allan, J. F., Geology of the northern Colima and Zacoalco Grabens, southwest Mexico: Late cenozoic rifting in the Mexican volcanic belt, Geol. soc. Am. Bull., 97, 473-485, 1986. Artyushkov, E. V., Stresses in the lithosphere caused by crustal thickness inhomogeneities, J. Geophys. Res., 78, 7675-7708, 1973. Association for the Development of Earthquake Prediction, Catalogue of major slab earthquakes near Japan, Seismotectonics Research Group, Association for the Development of Earthquake Prediction, 2000. Assumpcao, M., The regional intraplate stress field in south America, J. Geophys. Res., 97, 11889-11903, 1992. Auzende, J.-M., J.-P. Eissen, Y. Lafoy, P. Gente, and J. L. Charlou, Seafloor spreading in the North Fiji Basin (southwest Pacific), Tectonophysics, 146, 317-352, 1988. Baker, G. E., and G. A. Langston, Sourse parameters of the 1949 magnitude 7.1 south Puget Sound, Washington, earthquake as determined from long-period body waves and strong ground motions, Bull. Seism. Soc. Am., 77, 1530-1557, 1987. Barazangi, M. and B. L. Isacks, Spatial distribution of earthquakes and subduction of the Nazca plate beneath South America, Geology, 4, 686-692, 1976. Burbach, G. V., C. Frohlich, W. D. Pennington, and T. Matumoto, Seismicity and tectonics of the subducted Cocos plate, J. Geophys. Res., 89, 7719-7735, 1984. 18 Cardwell, R. K., Isacks, B. L., and Karig, D. E., The spatial distribution of earthquakes, focal mechanism solutions and subducted lithosphere in the Philippine and northeastern Indonesian islands, TheTectonic and Geologic Evolution of Southeast Asian Seas and Islands, Part 1, ed. by D. E. Hayes, Geophys. Monogr., AGU, Wasington D. C., 23, 1-35, 1980. Charvis, P., and B. Pelletier, The northern New Hebrides back-arc troughs: history and relation with the North Fiji basin, Tectonophysics, 170, 259-277, 1989. Coblentz, D. D., and R. M. Richardson, Analysis of the South American intraplate stress field, J. Geophys. Res., 101, 8643-8657, 1996. Comte, D., and G. Suarez, An inverted double seismic zone in Chile: Evidence of phase transformation in the subducted slab, Science, 263, 212-215, 1994. Comte, D., and G. Suarez, Stress distribution and geometry of the subducting Nazca plate in northern Chile using teleseismically recorded earthquakes, Geophys. J. Inter., 122, 419-440, 1995. Comte, D., L. Dorbath, M. Pardo, T. Monfret, H. Haessler, L. Rivera, M. Frogneux, B. Glass, and C. Meneses, A double-layered seismic zone in Arica, Northern Chile, Geophys. Res. Lett., 26, 1956-1968, 1999. Coudert E., Isacks B. L.,Barazangi M., Louat R., Cardwell R., and Chen A.,, Spatial distribution and mechanisms of earthquakes in the southern New Hebrides arc from a temporary land and ocean bottom seismic network and form wo, J. Geophys. Res., 86, 5905-5925, l98l. Creager, K C, L. A. Preston, Q. Xu, and R. S. Crosson, The 2001 Mw 6.8 Nisqually intraslab earthquake: An along-strike elongated fault rupturing entirely within the subducted cceanic crust?, Eos, Trans. Am. Geophys. Union, S42D-10, 2001. Crosson, R. S., T. J. Owens, Slab geometry of the Cascadia subduction zone beneath Washington from earthquake hypocenters and teleseismic converted waves, Geophys. Res. Lett., 14, 824-827, 1987. Dalmayrac, B., and P. Molnar, Parallel thrust and normal faulting in Peru and constraints on the state of stress, Earth Planet. Sci. Lett., 55, 473-481, 1981. Davies, G.F., On the emergence of plate tectonics, Geology, 20, 963-966, 1992. DeMets, C. R., R. G. Gordon, D. Argus, and S. Stein, Current plate motions, Geophys. J. Inter., 101, 425-478, 1990. Eguchi, T., Seismotectonics around the Mariana trough, Tectonophysics, 102, 33-52, 1984. Elsasser, W. M., Sea-floor spreading as thermal convection, J. Geophys. Res., 76, 1101-1112, 1971. Engdahl, E. R., and C. H. Scholz, A double Benioff zone beneath the central aleutians: An unbending of the lithosphere, Geophys. Res. Lett., 4, 473-476, 1977. 19 Fleitout, L., and C. Froidevaux, Tectonics and topography for a lithosphere containing density heterogeneities, Tectonics, 1, 21-56, 1982. Froidevaux, C., S. Uyeda, and M. Uyeshima, Island arc tectonics, Tectonophysics, 148, 1-9, 1988. Froidevaux, C., and B. L. Isacks, The mechanical state of the lithosphere in the Altiplano-Puna segment of the Andes, Earth Planet. Sci. Lett., 71, 305-314, 1984. Gough, D. I., Mantle upflow under north America and plate dynamics, Nature, 311, 428-432, 1984. Hamburger, M. W., and B. L., Isacks, Diffuse back-arc deformation in the southwestern Pacific, Nature, 332, 599-604, 1988. Hansen, R. A,. and N. A. Ratchkovsky, The Kodiak Island, Alaska Mw 7 earthquake of 6 December 1999, Seism. Res. Lett., 72, 22-32, 2001. Hasegawa, A., Umino, N., and A. Takagi, Double-planed structure of the deep seismic zone in the northeastern Japan arc, Tectonophysics, 47, 43-58, 1978. Hasegawa, A., and I. S. Sacks, Subduction of the Nazca plate beneath Peru as determined from seismic observations, J. Geophys. Res., 86, 4971-4980, 1981. Hasenaka, T., Size, distributionÅCand magma output rate for shield volcanoes of the Michoacan-Guanajuato volcanic field, central Mexico, J. Volcanol. Geotherm. Res., 63, 13-31, 1994. Herron, E. M., Sea-floor spreading and the cenozoic history of the east-central Pacific, Geolog. Soc. Am. Bull., 83, 1671-1692, 1972. Hsui, A. T., and M. N. Toksoz, Back-arc spreading: Trench migration, continental pull or induced convention?, Tectonophysics, 74, 89-98, 1981. Ichikawa, M., Reanalyses of mechanism of earthquakes which occurred in and nea,r Japan and statistical studies on the nodal plane solutions obtained 1926-1968, Geophys. Mag., 35, 207-273, 1971. Isacks, B. L., and M. Barazangi, Geometry of Benioff zones: Lateral segmentation and downwards bending of the subducted lithosphere, in Island Arcs, Deep Sea Trenches and Back Arc Basins, Maurice Ewing Series, edited by M. Talwan and W. C. Pitman III, AGU, 1, 99-114, 1977. Jones, C. H., J. R. Unruh, and L. J. Sonder, The role of gravitational potential energy in active deformation in the southwestern United States, Nature, 381, 37-41, 1996. Katsumata, K., M. Ichiyanagi, M. Miwa, and M. Kasahara, Aftershock distribution of the October 4, 1994 Mw 8.3 Kurile islands earthquake determined by a local seismic network in Hokkaido, Japan, Geophys. Res. Lett., 22, 1321-1324, 1995. Kawakatsu, H., Double seismic zones: kinematics, J. Geophys. Res., 91, 4811-4825, 1986. Kikuchi, M., and H. Kanamori, The Shikotan earthquake of October 4, 1994: Lithospheric earthquake, 20 Geophys. Res. Lett., 22, 1025-1028, 1995. Kimura, G., Tectonic evolution and stress field in the southwestern margin of the Kurile Arc, J. Geol. Soc. Jpn, 87, 757-768, 1981. Kirby, S. , Intraslab earthquakes and phase changes in subducting lithosphere, Rev. Geophys. , Suppl., 287-297, 1995. Kirby, S., E. R. Engdhal, and R. Denlinger, Intraslab earthquakes and arc volcanism: Dual phyisical expressions of crustal and uppermost mantle metamorphism in subducting slabs, SUBCON volume, edited by G. Bebout, D. Scholl, and S. Kirby, Geophys. Monogr., 96, 195-214, 1996. Klitgord, K., and J. Mammerickx, Northern East Pacific Rise: Magnetic anomaly and bathymetirc framework, J. Geophys. Res., 87, 6725-6750, 1982. Kostoglodov, V., and W. Bandy, Seismotectonic constraints on the convergence rate between the Rivera and North American plates, J. Geophys. Res., 100, 17977-17989, 1995. Kosuga, M., Stress distribution in northern Japan derived from crustal earthquakes, The Earth Monthly, Extra Volumes, 27, 10 7-112, 1999. Kosuga, M., T. Sato, A. Hasegawa, T. Matsuzawa, S. Suzuki, and Y. Motoya, Spatial distribution of intermediate-depth earthquakes withi horizontal or vertical nodal planes beneath northeastern Japan, Phys. Earth Planet. Inter., 93, 63-89, 1996. Kotake, Y., Study on the tectonics of western Pacific region derived from GPS data analysis, Bull. Earthq. Res. Inst., Univ. of Tokyo, 75, 229-334, 2000. Kubo, A., and E. Fukuyama, Extensional field along Ryukyu arc revealed by earthquake focal mechanisms, Abstr. Seism. Soc. Jpn, P114, 2001. Liu, C., Curray, J. R., and J. M. Mcdonald, New constraints on the tectonic evolution of the eastern Indian ocean, Earth Planet. Sci. Letters, 65, 331-342, 1983. Liu, M., and Y. Young, Extensional collapse of mountain belts: Insights from geodynamic modeling, Recent Res. Devel. Geophys., 3, 107-124, 2000. Louat, R., and B. Pelletier, Seismotectonics and prsent-day relative plate motions in the New-Hebrides North Fuji Basin region, Tectonophysics, 167, 41-55, 1989. Lowman, J. P., and G.T. Jarvis, Mantle convection flow reversals due to continental collisions, Geophys. Res. Lett., 20, 2087-2090, 1993. Ma, L., R. Crosson, and R. Ludwin, Western Washington earthquake focal mechanisms and their relationship to regional tectonic stress, USGS Professional Paper "Asssessing Earthquake Hazards and Reducing Risk in the Pacific Northwest", Vol. 1, 1560, 257-284, 1996. 21 Malahoff, A., Feden, R. H., Fleming, H. S., Magnetic anomalies and tectonic fabric of marginal basins north of New Zealand, J. Geophys. Res., 87, 4109-4125, 1982. Malgrange, M.,and R. Madariaga, Complex distribution of large thrust and normal fault earthquakes in the Chilean subduction zone, Geophys. J. R. astr. Soc., 73, 489-505, 1983. Martinez, F., P. Fryer, and N. Becker, Geophysical characteristics of the southren Mariana trough, 11ÅK 50'N-13ÅK40'N, J. Geophys. Res., 105, 16591-16607, 2000. Matsumura, S. , Focal zone of a future Tokai earthquake inferred from the seismicity pattern around the plate interface, Tectonophysics, 273, 271-291, 1997. Meijer, P. Th., R. Govers, and M. J. R. Wortel, Forces controlling the present-day state of stress in the Andes, Earth Planet. Sci. Lett., 148, 157-170, 1997. Mikumo, T., Normal fault type earthquakes within the subducted slab beneath Mexico and the damages associated with them, Earthquake Eng. News, 171, 15-25, 2000 (in Japanese). Moriya, T., Tectonics of Hokkaido as viewed from shallow seismic activities and focal mechanisms, Special Rep. Chidanken, 31, 475-485, 1986 (in Japanese). Nakakuki, T., and S. Honda, 2-D dynamical models for shallow- and deep-angle subduction, Proc. Nat. Acad. Jpn, submitted, 2001. Nakamura, K., G. Plafker, K. H. Jacob, and J. N.Davies, A tectonic stress trajectory map of Alaska using information from volcanoes and faults, Bull. Earthq. Res. Inst., 55, 89-100, 1980. Nakamura, K., and S. Uyeda, Stress gradient in arc-back regions and plate subduction, J. Geophys. Res., 85, 6419-6428, 1980. Nakamura, M., Historical earthquakes near Kii Peninsula and off Nankaido as viewed from the microearthquake observations: Comparison with recent earthquakes accompaning disasters, Misc. Articles Wakayama Seism. Obs., ERI, Univ. of Tokyo, 1, 18-39, 1997. Nakamura, M., H, Watanabe, T. Konomi, S. Kimura, K. and K. Miura, Characteristics activities of subcrustal earthquakes along the outer zone of southwestern Japan, Annuals of the Disaster Prevension Research Instutute, 40, 1-20, 1997. Nakanishi, M., K. Tamaki, and K. Kobayashi, A new Mesozoic isochron chart of the northwestern Pacific Ocean: paleomagnetic and tectonic implications, Geophys. Res. Lett., 19, 693-696, 1992. Newcomb, K. R. and W. R. McCann, Seismic history and seismotectonics of the sunda arc, J. Geophys. Res., 92, 421-439, 1987. Ohkura, T., Structure of the upper part of the Philippine Sea plate estimated by later phases of upper mantle earthquakes in and around Shikoku, Japan, Tectonophysics, 321, 17-36, 2000. 22 Okino, K., Y. Shimakawa, and S. Nagano, Evolution of the Shikoku Basin, J. Geomag. Geoelectr., 46, 463-479, 1994. Ozawa, T. Tabei, and S. Miyazaki, Interplate coupling along the Nankai Trough off southwest Japan derived from GPS measurements, Geophys. Res. Lett., 26, 927-930, 1999. Pardo, M., and G. Suarez, Shape of the subducted Rivera and Cocos plate in southern Mexico: Seismic and tectonic implications, J. Geophys. Res., 100, 12357-12373, 1995. Parsons, B., and F. M. Richter, A relation between the driving force and geoid anomaly associated with mid-oceanic ridges, Earth Planet. Sci. Lett., 51, 445-450, 1980. Peacock, S. M., and K. Wang, Seismic consequences of warm versus cool subduction metamorphism: Examples from southwest and northeast Japan, Science, 286, 937-939, 1999. Peacock, S. M., Are the lower planes of double seismic zones caused by serpentine dehydration in subducting oceanic mantle?, Geology, 29, 299-302, 2001. Pitman III, W. C., R. L. Larson, and E. M. Herron, The age of the ocean basins, Geolog. Soc. Am. Map and Chart MC-6, 1974. Raleigh, C. B., and M. S. Paterson, Experimental deformation of serpentinite and its tectonic implications, J. Geophys. Res., 70, 3965-3985, 1965. Riddihough, R. P., Recent movements of the Juan de Fuca plate system, J. Geophys. Res., 89, 6980-6994, 1984. Sandiford, M., and D. Coblentz, Plate-scale potential-energy distributions and the fragmentation of ageing plates, Earth Planet. Sci. Lett., 126, 143-159, 1994. Sebrier, M., J. L. Mercier, F. Megard, G. Laubacher, and E. Carey-Gailhardis, Quaternary normal and reverse faulting and the state of stress in the Central Andes of South Peru, Tectonics, 4, 739-780, 1985. Seno, T., Northern Honshu microplate hypothesis and tectonics in the surrounding region: When did the plate boundary jump from central Hokkaido to the eastern margin of the Japan Sea, J. Geodet. Soc. Jpn., 31, 106-123, 1985. Seno, T., Tectonic evolution of the West Philippine Basin, Modern Geology, 12, 481-495, 1988. Seno, T., Syntheses of the regional stress fields of the Japanese islands, The Island Arc, 8, 66-79, 1999. Seno, T., Tectonic significance of the 2001 Geiyo earthquake, Abstr. Joint Meeting Earth Planet. Sci. Jpn, 2001. Seno, T., and K. Kurita, Focal mechanisms and tectonics in the Taiwan-Philippine region, J. Phys. Earth, 26, s249-s263, 1978. Seno, T., and Y. Yamanaka, Double seismic zones, deep compressional trench - outer rise events and 23 superplumes, in Subduction Top to Bottom, edited by G. E. Bebout, D. W. Scholl, S. H. Kirby, and J. P. Platt, Geophys. Monogr, AGU, Washington D. C., 96, 347-355, 1996. Seno, T. and Y. Yamanaka, Arc stresses determined by slabs: Implications for mechanisms of back-arc spreading, Geophys. Res. Lett., 25, 3227-3230, 1998. Seno, T., and S. K. Singh, Siminarities between southwest Japan and Mexico, Abstr. Joint Meeting Earth Planet. Sci. Jpn, 2001. Seno. T., S. Stein, and A. E. Gripp, A model for the motion of the Philippine Sea plate consistent with NUVEL-1 and geological data, J. Geophys. Res., 98, 17941-17948, 1993. Seno, T., T. Sakurai, and S. Stein, Can the Okhotsk plate be discriminated from the North American plate?, J. Geophys. Res., 101, 11305-11315, 1996. Seno, T., D. Zhao, Y. Kobayashi, and M. Nakamura, Dehydration in serpentinized slab mantle: Seismic evidence from southwest Japan, Earth Space Planet, 53, 861-871, 2001. Shimizu, H., K. Uyehira, and K. Goto, The detailed structure of the deep seismic zone and focal mechanism solutions from the western end of Chugoku-Shikoku district to the Kyushu district, Abstr. Seism. Soc. Jpn, B39, 2000 (in Japanese). Singh, S. K., G. Suarez, and T. Dominguez, The Oaxaca, Mexico, earthquake of 1931: Lithospheric normal faulting in the subducted Cocos plate, Nature, 317, 56-58, 1985. Singh, S. K., and M. Pardo, Geometry of the Benioff zone and state of stress in the overriding plate in central Mexico, Geophys. Res. Lett., 20, 1483-1486, 1993. Singh, S. K., M. Ordaz, and L. E. Perez-Rocha, The great Mexican earthquake of 19 June 1858: Expected ground motions and damage in Mexico City from a similar future event, Bull. Seism. Soc. Am., 86, 1655-1666, 1996. Slancova, A., A. Spicak, V. Hanus, and J. Vanek, How the state of stress varies in the Wadati-Benioff zone: Indications from focal mechanisms in the Wadati-Benioff zone beneath Sumatra and Java, Geophys. J. Inter., 143, 909-930, 2000 . Stauder, W., Mechanism and spatial distribution of Chilean earthquakes with relation to subduction of the oceanic plate, J. Geophys. Res., 78, 5033-5061, 1973. Stauder, W, Subduction of the Nazca plate under Peru as evidenced by focal mechanisms and by seismicity, J. Geophys. Res., 80, 1053-1064, 1975. Suarez, G., Molnar, P., and B. C. Burchfiel, Seismicity, fault plane solutions, depth of faulting, and active tectonics of the Andes of Peru, Ecuador, and southern Columbia, J. Geophys. Res., 88, 10403-10428, 1983. 24 Suarez, G., T. Monfret, G. Wittlinger, C. David, Geometry of subduction and depth of the seismogenic zone in the Guerrero gap, Mexico, Nature, 345, 336-338, 1990. Suter, M., State of stress and active deformation in Mexico and western Central America, in Slemmons, D. B., E. R. Engdahl, M. D. Zoback, and D. D. Blackwell, eds., Neotectonics of North America: Boulder, Cololado, Geological Society of America, Dacade Map Volume 1., 401-421, 1991. Suter, M., O. Quintero and C. A. Johnson, Active faults in Trans-Mexican Volcanic Velt, Mexico 1. The Venta de Bravo fault, J. Geophys. Res., 97, 11983-11993, 1992. Suzuki, S., T. Sasatani, and Y. Motoya, Double seismic zone beneath the middle of Hokkaido, Japan, in the southwestern side of the Kurile arc, Tectonophysics, 96, 59-76, 1983. Tanahashi, M., Tectonics of the spreading center in the North Fiji Basin, Bull. Geol. Surv. Japan, 45, 173-234, 1994. Tanioka, Y., L. Ruff, and K. Satake, The great Kurile earthquake of October 4, 1994 tore the slab, Geophys. Res. Lett., 22, 1661-1664, 1995. Tanioka, Y., K. Satake, and L. Ruff, Analysis of seismological and tsunami data from the 1993 Guam earthquake, Pure Appl. Geophys., 144, 823-837, 1995. Taylor, B., D. E. and Hayes, The tectonic evolution of the south China Basin, Geophys. Monog., AGU, 23, 89-104, 1980. Tsumura, N., H. Ikawa, T. Ikawa, M. Shinohara, T. Ito, K. Arita, T. Moriya, G. Kimura, T. Ikawa, Delamination-wedge structure beneath the Hidaka collision zone, central Hokkaido, Japan inferred from seismic reflection profiling, Geophys. Res. Lett., 26, 1057-1060, 1999. Ukawa, M., Lateral stretching of the Philippine sea plate subducting along the Nankai-Suruga Trough, Tectonics, 1, 543-571, 1982. Umino, N., A. Hasegawa, A. Takagi, S. Suzuki, Y. Motoya, S. Kametani, K. Tanaka, and Y. Sawada, Focal mechanisms of intermediate-depth earthquakes beneath Hokkaido and nothern Honshu, Jisin, 37, 521-538, 1984 (in Japanese). Utsu, T., Tables of earthquakes with magnitude 6.0 or larger and earthquakes associated with damage near Japan: 1885-1980, Bull. Earthq. Res. Inst., 57, 401-463, 1982 (in Japanese). Wang, K., T. Mulder, G. C. Rogers, and R. D. Hyndman, Case for very low coupling stress on the Cascadia subduction fault, J. Geophys. Res., 100, 12,907-12,918, 1995. Wang, K., Stress gradients in the forearc-backarc system revisited, J. Geophys. Res., submitted, 2001. Weissel, J. K., Watts, A. B., Lapouille, A., Evidence for late Paleocene to late Eocene seafloor in the southern New Hebrides Basin, Tectonophysics, 87, 243-251, 1982. 25 Wells, R. E., C. S. Weaver, and R. J. Blakely, Fore-arc migration in Cascadia and its neotectonic significance, Geology, 26, 759-762, 1998. Wilson, D. S., Confidence intervals for motion and deformation of the Juan de Fuca plate, J. Geophys. Res., 98, 16-53-16071, 1993. Yamamoto, J., Z. Jimez, and R. Mota, El temblor de Huajuapan de Leon, Oaxaca, Mexico, del 24 de Octubre de 1980, Geofis. Int., 23, 83-110, 1984. Yoshii, T., Proposal of the "aseismic front", Jisin, 28, 365-367, 1975 (in Japanese). Yu, S.-B., L.-C. Kuo, R. S. Punnngbayan and E. G. Ramos, GPS observation of crustal deformation in the Taiwan-Luzon region, Geophys. Res. Lett., 27, 923-926, 1999. Zoback, M. L., First- and second-order patterns of stress in the lithosphere: the world stress map project, J. Geophys. Res., 97, 11703-11728, 1992. 26 Figure Captions Figure 1 Focal mechanisms of large shallow slab earthquakes listed in Table 1 are plotted with event numbers (lower hemispheres in an equal area projection). For Events 2, 4 and 5, only epicenters are shown because no reliable mechanism solutions are available for these events. The dotted lines labeled with alphabets show sections along which P- and T-axes of smaller events from the Harvard centroid moment tensor catalogue are plotted in Figure A1 (Appendix). (a) Western and South Pacific regions (b) North and South American regions. Figure 2 σHmax trajectories in Hokkaido. The gray and broken lines indicate σHmax directions for reverse and strike-slip fault type stress regimes, respectively. The short bars are P-axes of the focal mechanisms of shallow earthquakes occurring within the crust of the upper plate (Moriya, 1986; Kosuga, 1999), which are used for drawing the stress trajectories. The thick and thin bars represent reverse and strike-slip fault type mechanisms, respectively. The solid circle shows the epicenter of the 1994 slab event. The solid triangles show active volcanoes. Convergence velocities are indicated by the arrows with numerals representing rates in mm/yr. Figure 3 σHmax trajectories in Kyushu-SW. Japan (Seno, 1999). The gray, broken and dotted lines indicate σHmax directions for reverse, strike-slip, and normal fault type stress regimes, respectively. The solid circles show the epicenters of the 1899, 1905, 2001, and 1931 slab events. The solid triangles show active volcanoes. Figure 4 σHmax trajectories in S. Mariana drawn using geological normal fault strikes in Martinez et al. (2000). The dotted lines indicate σHmax directions for a normal fault type stress regime. The solid circle shows the epicenter of the 1993 slab event. Figure 5 σHmax trajectories in Vanuatu - North Fiji Basin. Symbols for trajectory lines are the same as Figure 3. The short bars are P-axes of the focal mechanisms of shallow earthquakes occurring within the crust of the upper plate (Charvis and Pelletier, 1989), which are used for drawing the stress trajectories. The thick, thin, and thinnest bars represent reverse, strike-slip and normal fault type mechanisms, respectively. The solid circles show the epicenters of the 1981 and 1994 slab events. 27 Figure 6 σHmax trajectories in S. America. Symbols for trajectory lines are the same as Figure 3. The short bars are σHmax directions of the stress data within the crust of the upper plate (Assumpcao, 1992; Zoback, 1992), which are used for drawing the stress trajectories. The thick, thin, and thinnest bars represent reverse, strike-slip and normal fault type stress regimes, respectively. The solid triangles show active volcanoes. The solid circles show the epicenters of the 1970 and 1965 slab events. Figure 7 σHmax trajectories in SW. Mexico (Seno and Singh, 2001). Symbols for trajectory lines are the same as Figure 3. The small dots indicate monogenic volcanic centers (Hasenaka, 1994) and the thin lines with ticks indicate active normal faults (Suter, 1992), which are used for drawing the stress trajectories. The solid triangles and stars indicate active composite or silisic volcanoes and clusters of small monogenic volcanoes, respectively. J. M. G and O in the inset right bottom denote the Jalisco, Michoacan, Guerrero, and Oaxaca blocks. The solid circle shows the epicenter of the 1997 slab event. Figure 8 σHmax trajectories in Alaska (Nakamura et al., 1980). Symbols for trajectory lines are the same as Figure 3. The solid triangles show active volcanoes. The solid circle shows the epicenter of the 1999 slab event. Figure 9 Two types of stress gradients in the arc. The dotted lines in (b) and (d) indicate σHmax directions for a normal fault type stress regime. (a) and (c) represent the change of the three principal stresses across the arc. The differential stresses σxx - σzz and σyy - σzz at the aseismic front is determined by the coupling at the thrust zone and the gravitational collapse of the topography (Seno and Yamanaka, 1998; Seno, 1999; Wang, 2001). In both cases of (b) and (d), a weak coupling is assumed and σxx - σzz and σyy - σzz are assumed to be negative at the aseismic front. In (a) and (b), there is no mantle drag at the upper plate base. Since the plate thickness does not change much, neither of σxx or σyy changes much, and only σzz changes significantly, producing more tension in the thicker crust area. In (c) and (d), there is mantle drag at the upper plate base producing a change in the magnitude of σxx. This results in the change of the σHmax direction between x and y as shown in (c) and (d). Figure 10 Tectonic situations where large shallow slab earthquakes tend to occur. The stress gradient observed in the upper plate implies mantle drag forces beneath it. The slab shows a down-dip tensional stress, but its age is mostly young. (a) The arc is driven by a small scale convection current toward the trench. (b) The continental plate is driven by a large scale convection current toward the trench. 28 Figure 11 Illustrations of the kinematics and dynamics occurring in (a) Kyushu - SW. Japan and (b) Cascadia as examples of the tectonic situation shown in Figure 10a. The double arrows, converging and diverging, show compressional and tensional stresses, respectively. The single arrows indicate motions of the arc, upwelling plume or oceanic plate. (a) The mantle upwelling has been suspected in the South China Sea west of Kyushu (Seno, 1999). Kyushu and SW. Japan migrate to the E and collide with C. Honshu, probably dragged by the lateral flow from the upwelling. (b) The mantle upwelling has been suspected in the Basin and Range (Gough, 1984). The W. Oregon fore-arc block migrates to the NW and collides with the Canadian Coastal Mountains (Wells et al., 1998). Figure 12 Numerical simulations of mantle convection with drifting continents. (a) The supercontinent breakups (top) and each broken piece migrates toward the trench (bottom) (Lowman and Jarvis, 1993). T and φ indicate the temperature contours and stream lines, respectively. As the continent is driven oceanward and overrides the oceanic plate, streamline intervals in the shallow portion of the subducting slab become narrower. We call this phenomena "sucking" by a continental plate, contrary to the sucking by an oceanic plate proposed by Elsasser (1971). (b) The shallow part of the slab becomes tensional as the continent migrates (Nakakuki and Honda, 2001). Figure A1 P- and T-axes of HCMT solutions (Mw>=5.8) are plotted in a cross-section along a great circle passing through a large shallow event in the across-arc direction; sections lines are shown in Figure 1. 29 Table 1. List of large shallow slab earthquakes Region Event Date Epicenter °N °E Mw Depth (km) Strike/Dip/Rake (°) Age (Ma) E. Hokkaido 1 Hokkaido-toho-ok i Kyushu-SW. Japan 1994 10 04 43.42 146.81 8.3 33 158 41 24 123 2 Kii-Yamato* 1899 03 07 34.1 136.1 7.0 45 3 Geiyo 2001 03 24 34.13 132.71 6.8 47 181 57 -67 15-30 4 Geiyo* 1905 06 02 34.1 132.5 7.2 50 5 Hyuganada* S. Mariana 1931 11 02 32.2 132.1 7.1 40 6 Guam Manila 1993 08 08 12.98 144.80 7.7 45 238 24 82 164 7 Manila Sumatra 1999 12 11 15.87 119.64 7.2 35 112 13 -169 22 8 Sumatra Vanuatu 2000 06 04 -4.73 101.94 7.8 44 92 55 152 66 9 Vanuatu 1994 07 13 -16.50 167.35 7.1 25 272 42 02 52 10 Vanuatu N. Chile 1981 07 06 -22.31 170.90 7.5 58 345 30 -179 35 11 Taltal C. Peru 1965 02 23 -25.67 -70.79 7.0 60 16 86 -78 48 12 Peru El Salvador 1970 05 31 -9.18 -78.82 7.9 43 160 37 -90 44 13 El Salvador 1982 06 19 12.65 -88.97 7.3 52 102 25 -106 >37 14 El Salvador Mexico 2001 01 13 12.97 -89.13 7.7 56 121 35 -95 15 Oaxaca 1999 09 30 15.70 -96.96 7.4 47 102 42 -103 16 Oaxaca 1931 01 15 16.4 -96.3 7.7 40 90 34 -90 17 Michoacan N. Cascadia 1997 01 11 18.34 -102.58 7.1 40 175 18 -28 18 Nisqually 2001 02 28 47.14 -122.53 6.8 47 176 17 -96 19 Olympia** Alaska 1949 04 13 47.17 -122.62 7.1 54 14 82 -135 1999 12 06 57.35 -154.35 7.0 36 357 63 -180 20 Kodiak Island 10-17 10 55 *Hypocenters and magnitudes are from Utsu (1982), except for the depths of the 1899 and 1905 events which are estimated in this study. Magnitudes are referring to the Japan Meterological Agency (JMA) magnitude. **Hypocenters and magnitudes are from Baker and Langston(1987). Hypocenters for other events are from the Harvard University centroid moment tensor catalogue (HCMT), except for the depths by individual studies mentioned in the text. Strike, dip and rake are from HCMT except for the 1993 Guam event: Tanioka et al. (1995), 1965 Taltal event: Malgrange and Madariaga (1983), 1970 Peru event: Abe (1972), 1931 Oaxaca event: Singh et al. (1985), and 1949 Olympia event: Baker and Langston (1987). (a) 90˚E 50˚N 100˚E 110˚E 120˚E 130˚E 140˚E 150˚E 160˚E 170˚E 180˚ 1 Asia a 40˚N 3 b 30˚N 5 20˚N 4 7 2 Pacific 6 d c 10˚N 0˚ 8 10˚S e 9 f 20˚S km 0 1000 2000 Australia g 10 30˚S Fig. 1a (b) 160˚W 150˚W 140˚W 130˚W 120˚W 110˚W 100˚W 90˚W 80˚W 70˚W 60˚W 60˚N 20 n 50˚N 19 18 40˚N m N. America 30˚N 17 Pacific 20˚N 16 l k 10˚N 15 14 0˚ j S. America 13 12 10˚S i 20˚S km 0 30˚S 1000 2000 h 11 Fig. 1b s V< s Eurasian plate s Okhotsk plate H < s < Vs Hmin Hmax 45°N s Hokkaido s s s s s l s 1994 Japan Sea s s s Hidaka s ch 41° en s r il T ur K Oshima Peninsula 79 79 Pacific plate N. Honshu 140° 78 145°E Fig. 2 s V< s s Vit 10°S iaz H < s <Vs s H< s V Tre Hmin Hmax nch North Fiji Basin D'Entrecasteaux Ridge New 100 l 1994 ench s Tr ride Spreading Center Heb 20° 1981 l 170°E . rF Z. nte Hu 180° Fig. 5 85°W 10 ° N 35°W 10 ° N s s s s s s s s Cordillera Blanca 1970 l Altiplano ss s 77 South America s s s s -20° -20° 1965 l 84 s s s s s s s s Atlantic s Pacific s s s s V< s s s -50° H < s <Vs s Hmin H <s Hmax V Fig. 6 s c s s s s <s ★ 小単成火山の集合 s H< s 成層火山 Mexico s s V 単成火山の火口 Tra ns s H N V s Mexican Volcanic Velt c c Jal isc oB c loc k c c 20° cs c s s Mic hoac M .A 19 m er ica Sierr a ch lock del S ex en 1997 Madr l e M Tr an B o ic ur Rivera Plate TMVB 47 Cocos Plate 104° J M 100 km 102° 18° G O 100° W Fig. 7 s V< s s 140°W H 70°N < s < Vs Hmin s H< s Hmax Alaska V s s s 172°W 62°N s s Bering Sea s s s s s s s s s s 54°N s s s s s s s s s s s s s s s s s s s s s 1999 l 58°N 58 Aleutian Trench 500 km 65 Fig. 8 (c) s X Back-arc s zz s xx yy Stress magnitude Stress magnitude (a) Aseismic Front s X Back-arc s zz xx s yy Aseismic Front (d) (b) s s Hmax s s Asthenosphere Z Y on t t arc on Fr re- ic is Fo m Ax eis As ter Tr en ch Fo r ne In Crust Ou X s arc s re- Fr Thrust upling Plate Oceanic Co Weak s lca is ch en Tr Continental Plate s Ax ter eis Fo m re- ic Vo Fr arc on t nic reFo r ne Crust Ou X As s In s Y arc s s Vo lca nic Fr o nt Hmax Continental Plate Mantle Drag Asthenosphere Weak st ng Thru Coupli late P ic n Ocea Z Fig. 9 (a) Compression Tension Oceanic plate Arc Tension (b) Compression Tension Oceanic plate Continental plate Tension Fig. 10 (a) pan Ja W. Korea - S shu C. Honshu Kyu E. China Sea Philippine Sea plate (b) Washington Basin & Range Juan de Fuca Plate Fig. 11 E. Hokkaido (a,1) depth [km] 0 NW SE 0 100 100 200 200 NW P-axis 300 T-axis 300 0 100 200 300 SE 0 100 200 300 Kyushu - sw. Japan (b,3) depth [km] 0 W 0 E 100 W E 100 P-axis 200 T-axis 200 0 100 200 300 0 100 200 300 Mariana (c,6) depth [km] 0 NW SE 0 100 100 200 200 NW P-axis 300 T-axis 300 0 100 200 300 SE 0 100 200 300 Manila (d,7) depth [km] 0 W 0 E 100 W E 100 P-axis 200 T-axis 200 0 100 200 0 100 200 Sumatra (e,8) depth [km] 0 SW NE 100 0 SW NE 100 P-axis 200 T-axis 200 0 100 200 300 0 100 200 300 Vanuatu (f,9) depth [km] 0 SW NE 100 0 SW NE 100 200 200 P-axis 300 T-axis 300 0 100 200 300 horizontal distance [km] 0 100 200 300 horizontal distance [km] Fig. A1 Vanuatu (g,10) depth [km] 0 SW 0 NE 100 100 200 200 SW P-axis 300 T-axis 300 0 100 200 300 NE 0 100 200 300 N. Chile (h,11) depth [km] 0 W 0 E 100 W E 100 P-axis 200 T-axis 200 0 100 200 0 100 200 Peru (i,12) depth [km] 0 SW 0 NE 100 SW NE 100 P-axis 200 T-axis 200 0 100 200 300 0 100 200 300 El Salvador (j,13,14) depth [km] 0 SW 0 NE 100 SW NE 100 P-axis 200 T-axis 200 0 100 200 0 100 200 Mexico (k,15,16) depth [km] 0 SW NE 100 0 SW NE 100 P-axis 200 T-axis 200 0 100 200 300 400 0 100 200 300 400 Mexico (l,17) depth [km] 0 SW NE 100 0 SW NE 100 P-axis 200 T-axis 200 0 100 200 300 horizontal distance [km] 0 100 200 300 horizontal distance [km] Fig. A1 (continued-1) depth [km] 0 0 SW 100 SW 100 P-axis 200 T-axis 200 0 100 200 300 0 100 200 300 El Salvador (j) depth [km] 0 0 SW 100 100 200 200 SW P-axis 300 T-axis 300 0 100 200 0 100 200 Cascadia (m,18,19) depth [km] 0 W 0 E 100 W E 100 P-axis 200 T-axis 200 0 100 200 300 0 100 200 300 Alaska (n,20) depth [km] 0 NW SE 100 0 NW SE 100 P-axis 200 T-axis 200 0 100 200 300 horizontal distance [km] 0 100 200 300 horizontal distance [km] Fig. A1 (continued-2)