Offshore double-planed shallow seismic zone in the NE Japan forearc region revealed by sp depth phases recorded by regional networks

Transcription

1 Geophys. J. Int. (2009) 178, doi: /j.16-26X x Offshore double-planed shallow seismic zone in the E Japan forearc region revealed by s depth phases recorded by regional networks Shantha S.. Gamage, 1 orihito Umino, 1 Akira Hasegawa 1 and Stephen H. Kirby 2 1 Research Center for rediction of Earthquakes and Volcanic Eruptions, Graduate School of Science, ohoku University, Sendai , Japan umino@aob.geophys.tohoku.ac.jp 2 US Geological Survey, Menlo ark, CA 902, USA Accepted 2008 ovember 10. Received 2008 ovember 6; in original form 2007 ovember 22 SUMMARY We detected the s depth phase at small epicentral distances of about 10 km or more in the seismograms of shallow earthquakes in the E Japan forearc region. he focal depths of 1078 M > earthquakes that occurred from 2000 to 2006 were precisely determined using the time delay of the s phase from the initial -wave arrival. he distribution of relocated hypocentres clearly shows the configuration of a double-planed shallow seismic zone beneath the acific Ocean. he upper plane has a low dip angle near the Japan rench, increasing gradually to 0 at approximately 100 km landward of the Japan rench. he lower plane is approximately parallel to the upper plane, and appears to be the near-trench counterpart of the lower plane of the double-planed deep seismic zone beneath the land area. he distance between the upper and lower planes is 28 2 km, which is approximately the same as or slightly smaller than that of the double-planed deep seismic zone beneath the land area. Focal mechanism solutions of the relocated earthquakes are determined from -wave initial motion data. Although -wave initial motion data for these offshore events are not ideally distributed on the focal sphere, we found that the upper-plane events that occur near the Japan rench are characterized by normal faulting, whereas lower-plane events are characterized by thrust faulting. his focal mechanism distribution is the opposite to that of the doubleplaned deep seismic zone beneath the land area. he characteristics of these focal mechanisms for the shallow and deep doubled-planed seismic zones can be explained by a bending unbending model of the subducting acific plate. Some of relocated earthquakes took place in the source area of the 19 Mw8. Sanriku earthquake at depths of 10 2 km. he available focal mechanisms for these events are characterized by normal faulting. Given that the 19 event was a large normal-fault event that occurred along a fault plane dipping landward, the earthquakes that currently occur just beneath or oceanwards of the Japan rench are probably its aftershocks, suggesting that aftershock activity continues to the present day, 70 years after the main shock. GJI Seismology Key words: Seismicity and tectonics; Body waves; Wave propagation; Subduction zone processes; Intra-plate processes; Asia. 1 IRODUCIO Knowledge of the accurate locations of earthquakes and their focal mechanisms in subduction zones constitutes important information in terms of gaining a better understanding of the subduction process of oceanic plates beneath continental plates. Some of the difficulties encountered in determining focal depths for small offshore shallow earthquakes using regional network arrival times recorded at onland stations arise from a lack of seismic network detectability and the inadequate distribution of seismic stations. he hypocentres of suboceanic earthquakes are typically mislocated when using conventional location methods based on a simple one-dimensional seismic velocity structure model (e.g. Engdahl et al. 1982). he degree of mislocation depends on the velocity heterogeneity of the lithosphere and mantle in the subduction system, the location of the hypocentres in relation to the subducting plate and the configuration of the seismic network (Engdahl et al. 1982; McLaren & Frohlich 198). Both relative location methods and methods that utilize seismic ray tracing through a laterally inhomogeneous Earth have been developed to improve the accuracy of solutions. Focal depths can be determined accurately even for earthquakes that occur outside an observation network, provided that the depth phases from the events can be detected. eleseismic depth phases C 2009 he Authors 19

2 196 Shantha S.. Gamage et al. such as p, pw, s and sw have been used in many studies for relocating interplate and intraplate earthquakes. Short-period teleseismic depth phases (e.g. Yoshii 1979; Engdahl & Billington 1986) and long-period teleseismic depth phases (e.g. Herrmann 1976; Forsyth 1982) have been used for relocating shallow earthquakes in subduction zones; however, these teleseismic depth phases can normally be used only for relatively large earthquakes because they require sufficient radiated energy to generate detectable depth phases recorded at teleseismic distances. Umino et al. (199) found that a distinct s depth phase can be observed in seismic recordings from regional seismic networks beneath ohoku, E Japan, even for small earthquakes at epicentral distances of about 10 km or more. he authors used this depth phase for redetermination of depths of shallow offshore events off the Sanriku region of E Japan, thereby estimating the configuration of the subducting acific plate. Umino & Hasegawa (199) detected this s depth phase in seismograms of aftershocks of the 199 Hokkaido-ansei-oki earthquake, and obtained a detailed hypocentre distribution of aftershocks. Mishra et al. (200), Wang & Zhao (200, 2006) and Zhao et al. (2007) used the s depth phase for seismic tomography imaging of the E Japan forearc region, and discussed the relation of the obtained tomographic images with interplate coupling. Studies of the seismic structures of several subduction zones have confirmed that earthquakes that occur near trenches generally have shallow focal depths. Frohlich et al. (1982) and Engdahl and Billington (1986) showed that earthquakes that occur near the central Aleutian rench have focal depths shallower than about km. Yoshii (1979) and ishizawa et al. (1992) showed that earthquakes located near the Japan rench in the E Japan subduction zone, especially those located off the Sanriku region, occur at depths shallower than 20 km. Seno & Gonzalez (1987) found an earth- quake that occurred in the deeper portion (1 km) of the acific plate beneath the Japan rench by long-period seismic wave form analysis. he focal mechanism of this event in the trench-outer rise region of the E Japan subduction zone is a thrust fault type with an approximately horizontal axis. Seno & Yamanaka (1996) pointed out that subduction zones with double-planed deep seismic zones, such as the E Japan, are generally characterized by the occurrence of compressional deep events at the trench-outer rise region, as with the event described by Seno & Gonzalez (1987). In the present study, we relocated the focal depths of many offshore earthquakes in the E Japan forearc region using the s depth phase, and determined their focal mechanisms. his analysis was performed with the aim of understanding the detailed structure of seismic activity in the region. 2 DEECIO OF HE s HASE AD DEERMIAIO OF FOCAL DEH he E Japan subduction zone marks the zone along which the acific plate subducts beneath the orth America/Okhotsk plate, generating not only interplate earthquakes but also intraplate and intraslab events. he detailed seismic structure of the E Japan arc has been revealed by many studies based on seismic observations (e.g. Matsuzawa et al. 1986, 1990; Hasegawa et al. 1991, 199; Zhao et al. 1992, 199; Zhao & Hasegawa 199). Fig. 1 shows the locations of seismic stations in E Japan established and maintained by ohoku University, Hirosaki University, Hokkaido University, University of okyo, he Japan Meteorological Agency (JMA) and Hi-net. We selected shallow offshore M >.0 earthquakes that occurred in E Japan during the period from 2000 January to 2006 December 2006 and investigated their seismograms recorded by 10E 12E 1E 16E 10E 1E 10E 1E 2 A plate/okhotsk plate A B C D E F G I H J acific plate -000 K L M DK Figure 1. Map of the study area and locations of cross-sections A to shown in Fig. 8. he width of each cross-section is 0 km. Solid squares represent seismic stations operated by ohoku University, Hokkaido University, Hirosaki University, University of okyo, he Japan Meteorological Agency (JMA) and the ational Research Institute for Earth Science and Disaster revention (IED). Contour lines indicate the depth of the ocean floor. Large arrows show the convergence direction of the acific plate. DK is the Dai Ichi Kashima seamount. C 2009 he Authors, GJI, 178, 19 21

3 Offshore double-planed shallow seismic zone 197 (a) :9 7.00s E 12.km M Hz YKB EW 21km 297 YKB S 21km 297 YKB UD 21km 297 s S ime (sec) (b) : s E 7. M Hz JH EW 2km 22 JH S 2km 22 (c) JH UD 2km 22 s S ime (sec) Station s Sediments rench Ocean Crust [s] Mantle late Earthquake Figure 2. Examples of three-component seismograms for earthquakes with magnitudes of.6 (a) and. (b); the epicentral distances are 21 and 2 km, respectively. Hypocentre parameters are shown along the top of each seismogram. Focal depths were determined from the delay of the s depth phase (shown by solid circles) relative to the wave first arrivals (shown by dashed lines). (c) Schematic ray paths of s and [s] phases. Solid and dashed lines denote ray paths of and S wave, respectively. three-component 1-Hz seismographs at the seismic stations shown in Fig. 1. A remarkable phase is often observed between - and S-wave arrivals at epicentral distances of about 10 km or more for many shallow earthquakes that occur in the offshore region. Fig. 2 provides examples of three-component seismograms that show this distinct phase, while Fig. provides examples of vertical-component seismograms arranged in order of epicentral distance. Solid circles marked on the vertical-component seismograms represent this distinct phase. his phase appears between the arrivals of the direct and S waves; its characteristics are the same as those of the s depth phase detected by Umino et al. (199) as follows: (1) the amplitudes are predominant on vertical-component records, (2) the directions of wave approach are almost the same as those of the direct wave, () the apparent velocities are slightly lower than those of the direct waves and higher than those of direct S waves, () the predominant periods tend to be slightly longer than those of direct waves and approximately the same as those of direct S waves and () the maximum amplitudes range from about 0 to 200% of those of direct waves. Following Umino et al. (199), we interpret this phase as an s phase: an upgoing S wave from the focus that is then reflected and converted to a wave at the Earth s surface/the bottom of ocean before diving into the Earth again and finally arriving at seismic stations. Ray path of the s phase is schematically shown in Fig. 2(c). he difference in arrival time between the direct wave and the s phase is sensitive to the focal depth of the event and the seismic-velocity structure, especially the structure of V p /V s just C 2009 he Authors, GJI, 178, 19 21

4 198 Shantha S.. Gamage et al. (a) :9 7.00s E 12.km M.6 GB UD 218km 29 (b) : s E 10.6km M.9 H.OWHUD 18km 281 YKB UD 21km 297 H.KAKHUD SDF UD 21km 296 H.RZHUD 17km 291 UGI UD 207km 26 U.KSUD 17km 289 IW UD 207km 29 H.SZGHUD 172km 277 HMK UD 197km 298 [s] H.KKWHUD 162km 288 MYR UD 186km 26 (c) s S ime (sec) : s E 7. M. U.EUD 18km 267 (d) [s] s S ime (sec) : s E.km M. MO UD 18km 2 KMF UD 2km 28 AS UD 18km 2 JMK UD 18km 281 [s] HMK UD 299km 2 ESA UD JKZ UD 287km 26 KGJ UD 172km 299 JA UD 26km 26 KGL UD 172km 299 JH UD 2km 22 [s] MM UD 172km 296 FD2 UD 2km 2 KS UD 18km 28 s S s S ime (sec) ime (sec) Figure. Examples of vertical-component seismograms aligned with -wave first arrivals and displayed in order of epicentral distance. Distinct s phases (solid circles) are observed between direct and S waves. ossible [s] phases (open triangles) are also observed between direct waves and s phases. hese four earthquakes occurred near the Japan rench. he focal depths of the events shown in (a) (d), as determined from the s phase, are 16.8, 1.1, 1. and 9. km, respectively. above the earthquake focus. In this study, we assumed the same velocity structure model as that of Umino et al. (199), which is modified from the standard seismic velocity model of the ohoku University network (Hasegawa et al. 1978a) under the land area and a velocity model obtained by airgun-obs seismic profiling across the Japan rench (Suyehiro et al. 1990). Based on this seismic velocity model, theoretical s times can be calculated using a twodimensional (2-D) seismic ray tracing method (Cerveny & sencik 198; Iwasaki 1988). here are other potential candidates for this distinct later phase, such as p or [s] phases. he [s] phase is an upgoing S wave that is subsequently reflected and converted to a wave at the sediment basement boundary on the inner trench slope and then recorded at land stations (Engdahl & Billington 1986). Ray path of the [s] phase is schematically shown in Fig. 2(c). Examples of possible [s] phase are shown in Fig. by open triangles. Umino et al. (199) investigated the theoretical travel times of both p and [s] phases based on a plausible seismic velocity model and found that neither the p nor [s] phases are consistent with this observed phase in the E Japan arc. Observed time differences between precursory [s] and s phases are rather consistent with theoretical ones estimated by Umino et al. (199) and are large enough, so that we can avoid C 2009 he Authors, GJI, 178, 19 21

5 Offshore double-planed shallow seismic zone E 11E 12E 1E 1E 1E 0 (a) Japan rench - Onland seismic stations - emporary OBS stations - Epicenters located by OBS stations - Epicenters located by onland stations s- time (sec) (b) Depth by s- time (km) (c) Depth by OBS/ Onland (km) Depth by OBS/ Onland (km) Figure. Distribution of s times and comparison between focal depths estimated from the s phase and those determined by OBS/onland networks. (a) Map showing the locations of seismic stations and earthquakes: open and solid squares denote the locations of OBS stations and onland stations, respectively. he epicentres of suboceanic earthquakes determined by OBS and those of coastal earthquakes determined by onland stations are shown by circles and inverted triangles, respectively. (b) Distribution of s times: s times from suboceanic earthquakes and those from coastal earthquakes are shown by circles and inverted triangles, respectively. (c) Comparison of focal depths: open circles and inverted triangles denote the focal depths of oceanic and coastal earthquakes estimated from the s phase, respectively. he horizontal and vertical axes show the focal depths determined by OBS/onland networks and those estimated from the s phase, respectively. misidentifying s phase. According to Umino et al. (199), this phase can be identified uniquely as the s depth phase. o confirm this phase identification, we undertook comparisons with other observations. A temporary observation network with ocean bottom seismographs (OBSs) was installed in the Miyagi-oki region (e.g. Hino et al. 2007; Suzuki et al. 2007). he locations of OBS stations in this network are shown by open squares in Fig. (a). he hypocentres of suboceanic earthquakes are well determined by this OBS network, and are shown by circles in Fig. (a). Distinct s phases were detected in the vertical-component seismograms of these events at some of the onland stations (solid squares in Fig. a). Focal depths determined by the OBS network are plotted against s times observed at onland stations (circles in Fig. b). Focal depths estimated from the s phase and those determined by the OBS network are shown by circles in Fig. c. hese figures support the present interpretation that the phase of interest is the s depth phase. he hypocentres of earthquakes that occur close to the coastline of E Japan are well determined by the onland seismic network (e.g. Hasegawa et al. 199); therefore, we also used these earthquakes for further confirmation of the phase identification. he epicentres of coastal earthquakes with distinct s phases are shown by inverted triangles in Fig. (a). he distributions of s times and focal depths determined by the onland network are also shown by inverted triangles in Fig. (b). Focal depths estimated from the s phase and those determined by the onland network are also shown by inverted triangles in Fig. (c). It is clear that the focal depths estimated from the s times are again very consistent with those determined by the onland network. Since 1997 October, JMA has been determining earthquake hypocentres using all of the seismic wave data (including both the and S phases) observed at the ational University networks, Hi-net and the JMA network, as well as publishing a unified hypocentre catalogue. Assuming that the epicentres listed in the unified hypocentre catalogue of the JMA are accurately determined, we attempted to determine their focal depths using the s depth phase obtained in the present study. he wave speed structure and seismicity in E Japan can be spatially regarded as being well characterized by a 2-D structure; consequently, we calculated the theoretical travel times of direct and s phases using 2-D seismic ray tracing methods C 2009 he Authors, GJI, 178, 19 21

6 200 Shantha S.. Gamage et al. 80 (a) 10E 11E 12E 1E 1E 1E 16E 2 Individual depth (km) E Japan Average depth (km) 8 Figure. Relation between the averaged focal depth and focal depths estimated at individual stations for each analysed earthquake. (Cerveny & sencik 198; Iwasaki 1988). Given that the azimuth of acific-eurasian convergence in the study area is estimated to be 297 based on the UVEL 1A model (DeMets et al. 199), we used only those seismograms recorded at stations having azimuths from the earthquake hypocentre ranging from 20 to 00 and epicentral distances less than 00 km. his range of azimuths also brackets azimuths trench normal of the subduction system. hree-component seismograms of M >.0 earthquakes within the E Japan forearc region (Fig. 1) were examined by applying a band-passedfilterbetween1andhzandwerecheckedtodetectthe s depth phase by taking into consideration the features described above. Given that theoretical s times are calculated using a 2-D seismic ray tracing method, station selection is critical in obtaining accurate results. Only earthquakes with a distinct s phase recorded at three or more stations were selected, and focal depth was estimated from the s delay recorded at each station. By calculating the average of the focal depths estimated from s times at individual stations, we determined the focal depth of each earthquake. Fig. shows how the individual depths from s delays vary in relation to the average depth for a given event. o assess the scatter of the estimated focal depths, we plotted the averaged focal depths against those estimated at individual stations for each earthquake (Fig. ). he scatter in the focal depths estimated from s phases at individual stations is about km. he standard deviations of estimated focal depths are indicated by thin lines in the vertical cross-sections of depth-relocated earthquakes shown in Fig. 6. Assuming focal depth estimated from s times, we determined the origin time and the location of epicentre using the standard seismic velocity model of the ohoku University network (Hasegawa et al. 1978a). Fig. 7 shows the differences between original epicentres listed in the JMA catalogue and the ones relocated in this study. Since the area covered by the onland seismic network is broader in the northsouth direction than the eastwest direction (see Fig. 1), the distribution of epicentre shifts is narrower in latitudes (Fig. 7a) than in longitudes (Fig. 7b). Consequently, Fig. 7(c) shows that 90% of the events have small epicentre shifts which are less than 1 km. he relocated epicentres and focal depths estimated by s phase are shown in the epicentre maps and vertical cross sections in the following sections. (b) 10E 11E 12E 1E 1E 1E 16E 2 Region W E Japan Region E Japan rench upper plane lower plane 7 lower routine [Mag] [Mag] Figure 6. Epicentre distribution of relocated and routinely determined offshore earthquakes during the period from 2000 January 1 to 2006 December 1. (a) Upper (orange) and lower (blue) plane earthquakes relocated using the s phase: green circles represent lower-plane earthquakes in region W determined by the ohoku University routine location procedure. he dashed line denotes the boundary between regions W and E. Red triangles and the open star represent the locations of active volcanoes and the epicentre of the 19 Mw8. Sanriku earthquake, respectively. (b) Epicentre distribution of all M >.0 earthquakes determined by onland seismic networks: other symbols are the same as those in (a). Open star and dashed rectangle denote the locations of the epicentre and source region of the 19 Mw8. Sanriku earthquake, respectively. C 2009 he Authors, GJI, 178, 19 21

7 Offshore double-planed shallow seismic zone (a) 100 (b) umber of events umber of events Shift in latitude (Degree) Shift in longitude (Degree) 10 (c) umber of events Distance (km) Figure 7. Difference between epicentres listed in the JMA catalogue and the ones relocated in this study. (a) Epicentre shifts in latitudes. ositive shift denotes that a relocated epicentre migrated northward. (b) Epicentre shifts in longitudes. ositive shift denotes that a relocated epicentre migrated eastward. (c) Distance between JMA epicentres and relocated ones. HYOCERE DISRIBUIO We determined the focal depths of 1078 offshore events, representing about 0% of all the M >.0 earthquakes located by the JMA in the study area; the locations of the relocated events are shown in Fig. 6(a). he depth-relocated events are distributed in the region from the acific coastline to the Japan rench. he study area is divided into two (regions W and E) by the dashed line in Fig. 6. he locations of cross-sections A to are shown in Fig. 1. he cross-sections are oriented approximately perpendicular to the Japan rench and approximately parallel to the subduction direction of the acific plate in this region. In region W of each cross-section, most of the relocated hypocentres are aligned upon curved surfaces that are inclined westward at depths ranging from 20 to 60 km. he hypocentres overlap and coincide with the upper plane of the double-planed deep seismic zone (Umino & Hasegawa 197; Hasegawa et al. 1978a,b, 1991, 199), which is clearly evident in the hypocentre distribution determined from the onland seismic network (grey dots). Given that the hypocentres of earthquakes in the double-planed deep seismic zone are well determined by the ohoku University network, it is possible that the locations of the depth-relocated events that extend into the upper plane of the double-planed deep seismic zone provide an approximate indication of the upper surface of the subducting acific plate. he depth-relocated earthquakes in region E tend to form two subhorizontal planes, as is clearly seen in cross-sections D H in Fig. 9. he two planes are approximately parallel to each other, being separated by a distance of 28 2 km. he upper plane of this shallow offshore double seismic zone appears to represent the shallow-level extension of the upper plane of the double-planed deep seismic zone located beneath the land area. he locations of relatively shallow earthquakes suggest that they occur along the plate boundary or in its vicinity; consequently, the deeper events are located within the acific plate. hese findings demonstrate a clear double-planed shallow seismic zone in the offshore forearc region of E Japan. he double-planed shallow seismic zone is clearly evident even in the vertical cross-section along line A (Fig. 8a) near the junction between the Kuril arc and the E Japan arc. We should mention that the limitations encountered in identifying very short s times for very shallow events prevented us from locating events whose focal depths were shallower than about km below the seafloor, as shown in each cross-section in Fig. 8. DISRIBUIO OF FOCAL MECHAISMS Focal mechanism solutions for the presently relocated events were determined from the initial motions of waves. he best-fit solution for each focal mechanism was obtained using the grid search method developed by Horiuchi et al. (1972). A minimum number of initial motion data was selected as 20 to ensure better station C 2009 he Authors, GJI, 178, 19 21

8 202 Shantha S.. Gamage et al. A 0 km 1 B 0 km 1 C 0 km 1 D 0 km 1 E 0 km 1 Figure 8. Across-arc vertical cross-sections of earthquakes relocated in the present study (open circles) and those determined routinely by the ohoku University network (grey dots) for the period from 2000 January to 2006 December. he locations of the cross-sections are shown in Fig. 1. he inverted triangle and upright triangle at the top of each figure indicate the locations of the Japan rench and the easternmost seismic station within each cross-section, respectively. he solid square at the top of each cross-section shows the location of the boundary between regions W and E (see Fig. 6). hick, horizontal lines along the tops of cross-sections D, E, F and G denote the location of the aftershock area of the 19 Mw8. Sanriku earthquake. Standard errors for focal depths estimated in this study and those for epicentres determined by the routine procedure of the ohoku University network are shown by thin vertical and horizontal lines, respectively, within the open circles. he ocean floor and outer rise region are shown by a solid line and thick grey line in each cross-section. C 2009 he Authors, GJI, 178, C 2009 RAS Journal compilation

9 Offshore double-planed shallow seismic zone 20 F 0 km 1 G 0 km 1 H 0 km 1 I 0 km 1 J 0 km 1 Figure 8. (Continued.) coverage on the focal sphere. hese focal mechanism solutions are shown in map view in Fig. 9. Given the unavailability of the s depth phase for many of the lower-plane events in region W, earthquakes whose focal depths were determined by the ohoku University rou C 2009 he Authors, GJI, 178, C 2009 RAS Journal compilation tine locations were also selected for the determination of focal mechanisms (Fig. 9d). he focal mechanisms of upper-plane earthquakes in region W (Fig. 9c) tend to be low-angle thrust fault events, whereas those

10 20 Shantha S.. Gamage et al. K 0 km 1 L 0 km 1 M 0 km 1 0 km 1 Figure 8. (Continued.) located adjacent to the Japan rench in the upper plane of the double-planed shallow seismic zone at depths of less than 2 km (Fig. 9a) are normal fault events. In and around the aftershock area of the 19 Mw8. Sanriku earthquake (Kanamori 1971), as shown by the dashed rectangle in Fig. 9(a) and (b), all of the earthquakes located near the Japan rench are normal-fault events. Fig. 10 shows a cross-sectional view of relocated earthquakes and their focal mechanisms along lines D, F and K shown in Fig. 1. he upper-plane events located near the Japan rench are generally normal-fault events, whereas the lower-plane events are reverse fault events. he events in region W, located near the coast, have the opposite focal mechanisms to those of events in the double-planed shallow seismic zone. Offshore OBS stations located close to crosssections F and K (Fig. 1) were also used in determining the focal mechanisms. Four focal mechanisms that are typical of the observed upperand lower-plane events are shown in Figs 11 and 12, respectively. We assessed the effect of focal depth error on the determination of the focal mechanisms (Fig. 1). Although the epicentres of the two earthquakes in the figure are relatively close, the focal mechanisms are entirely different: a normal fault event for the M.8 upper-plane event and a thrust fault event for the M.9 lower-plane event. he focal depths of the upper- and lower-plane events estimated using the s depth phase are 18.1 and 6. km, respectively. o investigate possible biases in the focal mechanisms arising from uncertainties in focal depth, the focal depths of these two earthquakes are varied from 1 to km at 10 km intervals. he relevant focal mechanism for each focal depth is projected onto the lower focal hemisphere in Fig. 1. Even though we varied the assumed focal depths from 1 to km, the focal mechanisms of C 2009 he Authors, GJI, 178, C 2009 RAS Journal compilation

11 Offshore double-planed shallow seismic zone 20 (a) E 11E 12E 1E 1E 1E 16E (b) 10E 11E 12E 1E 1E 1E 16E Region E Upper Eq Region E Lower Eq (c) E 11E 12E 1E 1E 1E 16E (d) E 11E 12E 1E 1E 1E 16E Region W Upper Eq Region W Lower Eq Figure 9. Distribution of the focal mechanism solutions of relocated earthquakes in regions E and W. he focal mechanisms of earthquakes in region E (a) have a focal depth equal to or shallower than 2 km; those in (b) have focal depths greater than 2 km. he focal mechanisms of upper-plane earthquakes in region W are shown in (c); those in the lower plane are shown in (d). ote that the focal depths of lower-plane events in region W were determined by the ohoku University routine location procedure (grey-and-white focal spheres). Focal mechanisms are projected on the lower focal hemisphere using an equal area projection. he open star shows the epicentre of the 19 Mw8. Sanriku earthquake. he number above each focal sphere is the earthquake number assigned to each event. C 2009 he Authors, GJI, 178, 19 21

12 206 Shantha S.. Gamage et al (a) D (b) F (c) K Figure 10. Across-arc vertical cross-sections of relocated earthquakes and their focal mechanisms along lines D, F and K shown in Fig. 1. Focal mechanisms are shown as a wall-side equal area projection. Focal depths determined by the s phase (black-and-white focal spheres) were used to determine the focal mechanism solutions. Focal mechanisms shown by grey-and-white focal spheres were determined using the focal depths of ohoku University routine locations. he location of outer rise region is shown by thick grey line. the two events are clearly maintained as a normal fault type and reverse fault type, respectively. he orientations of the, and ull axes for all of the focal mechanisms determined in this study are plotted on the focal spheres shown in Fig. 1. A low-angle thrust fault event, which is the typical focal mechanism for interplate earthquakes, can be seen in the upper-plane of region W (Fig. 1a). In contrast, down-dip extensional type events, which are typical of lowerplane events of the double-planed deep seismic zone (Hasegawa et al. 1978a), can be seen in the lower plane of region W (Fig. 1c). As shown in Fig. 1(b) and (d), the upper- and lower-plane events in region E have approximately horizontal and axes, respectively. In the aftershock area of the 19 Mw8. Sanriku earthquake, both the upper- and lower-plane events have approximately horizontal axes and are characterized by a normal fault mechanism. o assess depth variations in stress within the acific plate, we plotted the and axes of earthquakes in region E and the aftershock area of the 19 Sanriku earthquake against focal depth (Fig. 1). he axes of the upper-plane events are mostly horizontal, not only in region E but also in the aftershock area of the 19 Sanriku earthquake; the axes of lower-plane events are approximately horizontal. hese depth distributions of and axes raise the possibility that a neutral bending surface is at 20 2 km depth within the acific plate near the Japan rench (region E). C 2009 he Authors, GJI, 178, C 2009 RAS Journal compilation

13 Offshore double-planed shallow seismic zone 207 o. 22 o : s E 1.km M=.0 =1 o :9 1.6s E 11.km M=.7 =86 o : 9.8s E 1.7km M=.2 = : 9.66s E 1.6km M=.8 =129 Figure 11. Examples of focal mechanism solutions for upper-plane earthquakes within the offshore doubled-planed shallow seismic zone located near the Japan rench. Focal mechanisms are projected on the lower focal hemisphere using an equal area projection. he number on the top of each focal sphere is the earthquake number assigned to each event, as in Fig. 9. Hypocentre parameters and the number of polarity data are listed at the bottom of each focal sphere. he focal depths listed beneath each focal sphere were determined based on readings of the s phase. DISCUSSIO Fig. 8 shows arc-normal vertical cross-sections of relocated earthquakes. he configuration of upper-plane events at shallow depths beneath the acific Ocean can be traced continuously up to the Japan rench. he overall depth distribution of the upper-plane events reveals that the upper plane has an extremely low dip angle near the Japan rench, increasing gradually at about 100 km from the trench axis. Examples of vertical-component seismograms for upper- and lower-plane events were shown in Fig. (a) (d). he s times for the upper and lower earthquakes are different, even though all of these events occurred near the Japan rench. he differences in s delays for the seismograms of the four offshore events reflect the depth differences between the upper and lower planes of the double-planed shallow seismic zone. Moreover, the seismograms of the lower-plane events (Fig. c and d) appear to have more weakly attenuated waveforms than those of the upper-plane events (Fig. a and b). he seismic attenuation factor Q 1 for ray paths within the acific plate is lower than that through the overlying plate (e.g. Hasegawa et al. 199; sumura et al. 1996, 2000). Rays for lower-plane events received by land stations travel much longer distances through the subducting plate than those for upper-plane events. his finding is in agreement with the observation that the seismograms of the lower-plane events show more weakly attenuated high-frequency waveforms (Fig. a d). he lower plane of the double-planed shallow seismic zone in region E appears to continue smoothly to the lower plane of the double-planed deep seismic zone in region W at depths of 60 km. he double-planed seismic zone is therefore recognized not only within the deeper portion of the subducting acific plate under ohoku, but also within the shallower portion in the forearc region. Moreover, it should be noted that many of the cross-sections show seismic gaps in the lower plane and some show reduced seismicity in the upper plane near the boundary between the regions E and W. hese seismic gaps are also apparent from the distribution of C 2009 he Authors, GJI, 178, 19 21

14 208 Shantha S.. Gamage et al. o. 6 o :19.77s E 2.6km M=. = o : s E 7. M=.7 =7 o :2 8.10s E 1.8km M=.9 = : s E 0.8km M=.1 =26 Figure 12. Examples of focal mechanism solutions for lower-plane earthquakes within the offshore double-planed shallow seismic zone located near the Japan rench. See Fig. 11 for an explanation of symbols. earthquake hypocentres routinely determined by the seismic network of ohoku University, as shown by grey dots in Fig. 8(a) (n) and grey circles in Fig. (b). he typical focal mechanisms of upper- and lower-plane earthquakes of the double-planed shallow seismic zone are normal fault types and reverse fault types, respectively (Figs 9 and 1). his characteristic of the focal mechanism distribution is the opposite to that of the double-planed deep seismic zone beneath E Japan (Umino & Hasegawa 197; Hasegawa et al. 1978a,b, 1991, 199) that records down-dip compression and down-dip extension focal mechanisms in the upper and lower planes, respectively. Seno and Gonzalez (1987) and Seno and Yamanaka (1996) also reported compressional events in the deeper portion in the trenchouter rise region of offshore ohoku in addition to shallow normal fault events; they explained this pattern in terms of bending of the subducting plate. Our observations confirm their hypotheses of the occurrence of compressional deep events. hus, not only are the typical shallow normal-faulting earthquakes and the deeper compressional events in the outer-rise/outer-trench-slope region reported by Seno et al. confirmed by the present study, but we also show that these same classes of shallow and deeper events occur as much as 120 km west of the Japan rench under the forearc where the plate boundary curvature is known to sharply increase (Fujie et al. 2006). Outer-rise/outer-trench-slope events are consequences of flexural deformation associated with the bending moments that accompany the negative buoyancy of sinking slabs as well as slabparallel loads due to frictional and viscous resistance to slab motion (see review and discussion in Watts 2001, pp ). Based on the westward continuation of this type of double-zone seismicity from the outer-trench slope to under the inner-trench slope, this fundamental explanation evidently applies to the offshore intraslab seismicity to the west of the Japan rench as well. Similarly, a simple model of plate bending at the trench helps to account for the distribution of focal mechanisms obtained in the present study. A schematic figure of the bending-unbending plate model is shown in Fig. 16. he cold acific plate, which has travelled a large distance from the East acific rise, starts to subduct at the Japan rench by downward flexure at this point. As illustrated in the accompanying focal mechanism diagrams, this bending leads to the earthquake-generating stresses beneath the Japan rench and the source region of the 19 Sanriku earthquake. Based on wide-angle seismic experimental data, sharp bending points of the subducted C 2009 he Authors, GJI, 178, 19 21

15 Offshore double-planed shallow seismic zone 209 Upper plane event - o. 16 ( : 9.66s E 1.6km M.8 =129) h=1km h=2km h=km h=km 10E 11E 12E 1E 1E 1E 16E 1 Lower plane event 0 Upper plane event 9 Lower plane event - o. 21 ( :2 8.10s E 1.8km M.9 =79) h=1km h=2km h=km h=km Figure 1. Details of a test of the sensitivity of focal mechanism solutions to the assumed focal depth of events. Variations in focal mechanism solutions with assumed focal depth are shown for an upper-plane earthquake (M.8) and a lower-plane earthquake (M.9). he focal depths of these events determined using the s depth phase are 18.1 and 6. km, respectively. Solid and open circles denote compression and dilatation of initial -wave motions, respectively. he focal depths of these events are assumed to vary from 1 to km at a 10-km interval. he inset map shows the locations of the upper- and lower-plane events. acific plate were found both off Aomori (0 ) (Ito et al. 200) and off Miyagi (8 ) (Ito et al. 200). Fujie et al. (2006) found a plate bending point off Iwate (9 ) using wide-angle seismic data, and delineated the plate bending line in the E Japan forearc region by combining three bending points off Aomori, Iwate and Miyagi regions. his plate bending line is located about 70 km landward of the Japan rench and is shown by an open reversetriangle on the top of Fig. 16. his sharp bending can produce the earthquake-generating stresses within the acific plate shown in Fig. 16. Consequently, the bending stress state dominates continuously from the outer rise region to the sharp bending line (Fujie et al. 2006). With continued subduction, the acific plate enters the low-v and low-q mantle where it begins recovering its original shape, resulting in unbending of that part of the plate that was previously bent downward, beginning at km from the trench. his unbending provides a plausible explanation of the earthquakegenerating stress of the double-planed deep seismic zones present in subduction zones (Engdahl & Scholz 1977; Isacks & Barazangi 1977). he proposed bending and unbending may also provide insight into the lower seismicity in the upper plane and the seismicity gap in the lower plane near the boundary between regions E and W. rogressive unbending of the plate would reverse the sign of focal mechanisms and generate a transient region of low bending stress. As explained above, the bending and unbending of the subducting acific plate is a plausible model that explains the observed features of the distribution of focal mechanisms throughout the double-planed seismic zone, yet it remains difficult to explain why earthquakes occur in the lower plane of the shallow double-planed seismic zone where conventional models of plate bending predict that bending deformation should occur by plastic yielding rather than brittle deformation and faulting (Watts & Burov 200). he nature of fluid pressure in the source region of the lower plane may provide an answer to this problem. Recent studies suggest that intermediate-depth and deep earthquakes are caused by dehydration and/or the release of CO 2 from plume sources that accompanies the devolatilization of hydrated and carbonate minerals (Kirby 199; Kirby et al. 1996; Seno & Yamanaka 1996; eacock 2001; Yamasaki & Seno 200). It would be reasonable to combine our bending unbending model with other features related to these material reactions. We identified a seismic gap in the lower plane of the double seismic zone immediately beneath the location of the bending of the acific plate near the boundary between regions E and W (see cross-sections A, B, C, F and G in Fig. 8). If the bending unbending model is correct, there should be a zone of no stress or minimal stress within the acific plate. C 2009 he Authors, GJI, 178, 19 21

16 210 Shantha S.. Gamage et al. (a) (c) Region W : : : axis axis axis (b) (d) (e) Region E Upper Lower 19 Sanriku aftershock area Figure 1. Distribution of the, and ull axes of focal mechanisms for events relocated in the present study. (a) Upper-plane earthquakes in region W; (b) upper-plane earthquakes in region E; (c) lower-plane earthquakes in region W; (d) lower-plane earthquakes in region E; and (e) earthquakes in the aftershock area of the 19 Sanriku earthquake for all depth ranges. he location of each axis is projected on a lower focal hemisphere using an equal area projection. Open circles, grey circles and plus symbols denote, and ull axes, respectively. he dashed great circle shows the approximate orientation of the upper surface of the subducting acific plate in this region. As mentioned above, earthquakes in the source region of the 19 Mw8. Sanriku earthquake have normal fault mechanisms (e.g. earthquake #7 in Fig. 10f). hese shallow events extend to depths of at least 20 2 km. his region between the upper and lower planes is more complex than other sections that show clear separation between the upper and lower zones east of the Japan rench, as shown in vertical cross-sections E, F and G (Fig. 8e g). resent-day seismicity in the source region of the 19 Sanriku earthquake (Mw8.) is relatively active, with repeated earthquake swarms. Fig. 17(a) shows the magnitude frequency distribution of earthquakes that occurred in the source region of the 19 event (dashed rectangle in Fig. 6b) during the period from 2000 to 2006 listed from JMA catalogue. he Gutenberg-Richter s b-value is 1.08, which is very similar to the b-value (b = 1.1) of aftershocks within 6 days after the main shock (Mw8.) (Utsu 1970). Assuming that cut-off magnitude and b-value are 2.7 and 1.08, the standard rate of occurrence of aftershocks per day n(t) can be estimated using the following equation (Utsu 1970): C 2009 he Authors, GJI, 178, 19 21

17 Offshore double-planed shallow seismic zone 211 Depth (km) Depth (km) 0 (a) Region E (c) Region E (b) 19 aftershock area (d) 19 aftershock area axis axis Figure 1. Depth distributions of the inclinations of and axes: (a) axes for upper-plane events and axes for lower-plane events in region E; (b) axes for upper-plane events in the aftershock area of the 19 Sanriku earthquake; (c) axes for upper-plane events and axes for lower-plane events in region E; and (d) axes for upper-plane events in the aftershock area of the 19 Sanriku earthquake. n(t) = 10b(M 0 M s) d (t + c) p, where b, p, M 0 and M s are Gutenberg-Richter s b-value, Oomori s p-value, magnitude of the main shock and cut-off magnitude, respectively. he constant d and c are assumed to be 1.8 and 0., respectively (Utsu 1970). he earthquake occurrence rate n(t) are calculated for case of p = 1.0, 1.1 and 1.2, which are shown by dotted, solid and dashed lines, respectively, in Fig. 17(b). he observed daily occurrence rate of earthquakes during the period from 2000 to 2006 is shown by solid circle in Fig. 17(b). As shown in this figure, the most suitable p-value can be estimated as to be 1.1 which is very similar to the median of p-value (p = 1.1) of the earthquakes in and around Japan (Matsu ura 199). he focal depths of these earthquakes, as determined from the s phase, range from 8 to 0 km beneath the Japan rench. he focal mechanisms are normal fault type for events down to 21 km. Given that the 19 earthquake was a normal fault type event (Kanamori 1971), at least the shallowest earthquakes in the 19 source area are likely to represent aftershocks of the 19 event, thereby indicating that aftershocks continue today, more than 70 years since the main shock. A scattered focal depth distribution is evident above the upper plane of the double-planed seismic zone in cross-sections K, L, M and (Fig. 8k n). One possible explanation for the scattered seismicity recorded off the Fukushima region, east of the southern part of the E Japan arc, is the influence of subducting seamounts in this region. Many seamounts are located on the acific plate far offshore from Fukushima prefecture, to the east of the Japan rench (Fig. 1). Dai Ichi Kashima seamount is one of a chain of seamounts located far offshore from Fukushima prefecture that is beginning to be subducted at the Japan rench (suru et al. 2002). he subduction of these seamounts may explain the complex and scattered distribution of focal depths above the slab in this region. E Japan acific ocean Sharp bending point Japan rench orth American plate/okhotsk plate Low Q, Low V High Q, High V 19 Mw8. acific plate Asthenosphere 0 km 1 Figure 16. Schematic illustration of the bending unbending model, showing earthquake-generating stresses within the acific plate. ypical focal mechanisms are projected on the wall-side focal hemisphere using an equal area projection. Small open circles and thin arrows show hypocentre and earthquake-generating stress, respectively. hick arrows show the subduction motion of the acific plate. he location of outer rise region is shown by thick grey line. Open and solid reverse-triangles show the location of a sharp bending line of the acific plate (Fujie et al. 2006) and the Japan rench, respectively. Solid curve shows the location of the boundary between regions E and W (see Fig. 6). C 2009 he Authors, GJI, 178, 19 21

18 212 Shantha S.. Gamage et al. (a) (b) umber of earthquakes Daily frequency of earthquakes b= cumulative number number of EQs 10 Day after main shock earthquakes ( ) p=1.0 p=1.1 p=1.2 Figure 17. (a) frequency distribution of earthquakes occurring in the source region of the 19 Mw8. Sanriku earthquake during the period from 2000 to Crosses and circles denote the frequency per 0.1 magnitude unit and cumulative frequency, respectively. (b) Daily frequency of earthquakes in the source region of the 19 event. Dotted, solid and dashed lines represent the daily frequency of earthquakes estimated by the equation for the standard rate of occurrence of aftershocks (Utsu 1970) with Oomori s coefficient of 1.0, 1.1 and 1.2, respectively. 6 COCLUSIOS Using the s depth phase often observed in seismograms of shallow earthquakes in the E Japan forearc region, we determined the focal depths of those M >.0 earthquakes that occurred over the period from 2000 January to 2006 December. We used the method developed by Umino et al. (199) to determine the focal depths. he focal mechanisms of the relocated shallow offshore earthquakes were determined from wave initial motion data. he results of this study are summarized as follows. he focal depth distribution of depth-relocated earthquakes clearly shows the configuration of a double-planed shallow seismic zone throughout the entire region of the offshore E Japan subduction zone. he upper seismic plane roughly delineates the upper surface of the acific plate, thereby revealing that the plate subducts initially at an extremely low angle before increasing gradually to 0 dip at about 100 km from the Japan rench. Lower-plane earthquakes that occur within the subducting acific plate show the same spatial pattern as that of upper-plane earthquakes, although separated by about 0 km. ypical focal mechanisms for upper- and lower-plane events are normal fault type and reverse fault type, respectively. Farther from the trench, focal mechanisms show the opposite pattern that is consistent with that found under the land area of the E Japan arc. he bending and subsequent unbending of the subducting plate provides a plausible explanation for the observed distribution of focal mechanisms within the acific plate. he lower-plane events require special conditions for seismogenic faulting. his patterns of hypocentre distribution and focal mechanisms found in this study are essentially the same as that found under the outer-rise/outer-trench slopes (OR/OS) of subduction zones by previous investigators (Stauder 1968a,b; Chapple & Forsyth 1979; Christensen & Ruff 1988; Seno & Yamanaka 1996). his is not surprising since the curvature of the seafloor oceanwards of trenches and of the plate boundary landwards of trenches are both ultimately caused by slab buoyancy forces. It is this common concave downward curvature that leads to the bending stresses and deformation, which are common to the acific late both seawards and immediately landwards of the Japan rench. In this sense, these two seismotectonic regions are identical except one has water above it and the other has the tip of the forearc of the upper plate above it. Based on not only the characteristics of focal depths and focal mechanisms but also relatively high seismic activity rate of earthquakes in the aftershock area of the 19 Sanriku earthquake, those shallow earthquakes can be regarded as aftershocks of the 19 event, suggesting that the aftershock activity continues to the present day, 70 years after the main shock. he present study is the first to describe a double-planed shallow seismic zone in the forearc region of E Japan; however, much work remains to be done to understand the subduction process in the forearc region. We only investigated those earthquakes in the offshore region of E Japan using the s delays. Given that the double-planed deep seismic zone has been detected over a much wider area, from Hokkaido to Kanto, future studies are necessary to investigate the distributions of the focal depths and focal mechanisms of shallow earthquakes in the offshore region of this wider area. In the present study, we assumed a 2-D seismic wave velocity model for the relocation of earthquake depths. he use of a -D seismic wave velocity structure model will provide more precise results than those obtained in the present study. Although we determined the focal mechanisms of many earthquakes from -wave initial motion data, more data are required to fully understand the distribution of earthquake generating stress within the acific plate. his requirement justifies further investigations of such earthquakes over a longer time period. ACKOWLEDGMES We thank members of the Research Center for rediction of Earthquakes and Volcanic Eruptions, Graduate School of Science, ohoku University for their discussions and assistance. We are also grateful to the staff of the ational Research Institute for Earth Science and Disaster revention (IED), Hokkaido University, C 2009 he Authors, GJI, 178, 19 21