This study determined > 20,000 events in a subduction zone around the subducting seamount with a high-density (~ 6-km spacing) OBS network. Most of the events were determined within ±2 km error bar. To the best of our knowledge, this is an unprecedentedly large number of events by using temporal OBS experiments. This is owing to the dense OBS array, the occurrence of quite a few aftershocks, and the development of an effective event location workflow. This high-density event distribution allows us to identify the local spatial variation of the seismicity with a 1-km grid interval (Fig. 8). After the event detection capability analysis, low-seismicity zones are securely identified in the range from approximately M1 to M4 in the study area. Further, the resultant event distribution should be barely biased in space due to the optimization of \({K}_{1}\); therefore, the overall event depth distribution tends to be correct. In addition, the geometry of the interface was reasonably figured out comprising of two distinct seismic interfaces.
These event data allow us to discuss the local seismicity pattern in time and space. The unbiased event distribution enables us to compare with other geophysical survey results. Figure 10 shows the event distribution with the featured tectonics at the off-Ibaraki region: the subducting seamount, relocated hypocenter of the Mw7.9 event by the present study, shallow tectonic tremor from 2016 to 2018, and acoustic GPS (A-GPS) from 2012 to 2016.
The along-dip depth profiles of the event distribution along the seismic survey lines from past research are shown in Figs. 11 and 12. Figure 11 presents the seismic profile of Line EW reported by Mochizuki et al. (2008). Figure 12 shows the seismic survey of Line 13 reported by Tsuru et al. (2002), located ~ 30 km south of Line EW. Both seismic profiles clarify the depth of plate boundary at the updip limit of the seismogenic zone. Each figure exhibits that the shallow tectonic tremors and local events are spatially separated at the updip limit of the seismogenic zone. Most remarkably, it is shown that the majority of the events are distributed several kilometers deeper than the plate interface. The error ellipsoids of each seismic line presented in Figs. 11 and 12 are shown in Figures S8 and S9, respectively. Around the updip limit of the seismogenic zone, the maximum error bar as the 68% confidence interval is ~ 0.4 km, which is sufficiently smaller than the event depth offset from the plate interface. Note that all of the tectonic features shown in Fig. 10 come from the ocean bottom or marine seismic surveys. No results from a sole land seismic network are used in Fig. 10 to avoid the misinterpretation of the spatial interrelationships. Further details of the tectonics at the off-Ibaraki region are discussed in the subsequent subsections.
5.1 Seismicity overview concerning surrounding tectonics
The event distribution of this study showed that the high seismicity zone is concentrated at the front-end side of the seamount (Figs. 10–12). By contrast, the seismicity is quite low around the top or back-end side of the seamount. In this low-seismicity zone, the shallow tectonic tremors are distributed with little spatial gap with small earthquakes. We focus on the spatial relationship between the seismogenic zone and the subducting seamount.
In the plan view of the seismicity shown in Fig. 10, this high-seismicity zone showed a horizontal variation along the rim of the front-end side of the subducting seamount. The seismicity on the northern side is higher than that on the eastern or southern side. Nakatani et al. (2015) suggested that this zone is a part of the seamount. The spatial seismicity pattern in this study is consistent with this previous study. Nakatani et al. (2015) discussed that this horizontal and vertical seismicity variation along the rim of the seamount may be a consequence of a stress field change by the Mw7.9 event. On top of this consistency, this study can further discuss the event depth variation. The event depth variation clarifies that, temporally bounded by the occurrence of the Mw7.9 event, the depth variation of the seismicity changed considerably from a monotonic planar distribution (Fig. 9a) to a depth-variant heterogeneous distribution (Fig. 13). This seismicity suggests that subsurface structures are illuminated by the small earthquakes. Particularly, the presence of depth-variant subfaults and/or microfractures is shown around the seamount, including inside the oceanic crust.
The updip limit of the seismogenic zone corresponds to ~ 10 km plate interface depth. This updip limit is located around the top of the subducting seamount. Using numerical modeling, Sun et al. (2020) showed that at around the top of the seamount, the effective normal stress along the plate interface is considerably smaller than the one along the front-end side. This updip limit located around the top of the seamount may be explained, at least partly, by this normal stress reduction that is incapable of generating stick-slip events along with the plate interface of the seamount (Sun et al., 2020). However, the interpretation of the resultant seismicity in this study is complicated due to the presence of two seismically active interfaces (Fig. 13). To clarify the plate interface, in the next subsection, 5.2, we discuss the details of the gently sloped planar downdip interface—the most prominent interface identified in this study.
5.2 Planar downdip interface
The depth of the planar downdip interface is ~ 18 km at the updip limit of the seismogenic zone (Fig. 13). This planar interface is also discernable by the hypocenter reported by Shinohara et al. (2011, 2012), which is around the same depth as that in the present study (Figure S7). This planar downdip interface had been active before the Mw7.9 event throughout the entire OBS experiment period. The latest large earthquake in this off-Ibaraki region before the 2011 Mw7.9 event was the M7.0 event on 8 May 2008 (Takiguchi et al., 2011). This M7.0 event in 2008 also occurred at the deeper portion, and its rupture propagated from relatively deeper to a shallower direction (Takiguchi et al., 2011). The aftershock distribution of this M7.0 event in 2008 was examined by Yamada et al. (2011). The comparison of the event distribution is shown between the aftershocks of this M7.0 event in 2008 and the event distribution before the Mw7.9 event in 2011 determined in this study (Fig. 14). In the plain map view of Figs. 14a and 14b, the aftershocks of the M7.0 event in 2008 exhibit a similar spatial seismicity pattern as seismicity analyzed in the present study before the occurrence of the Mw7.9 event. The cross sections of both seismicities show a clear planar downdip trend. The aftershocks of Mw7.9 in 2011 reported by Shinohara et al. (2011, 2012) are consistent with these results. That is, the hypocenters reported by Yamada et al. (2011), Shinohara et al. (2011, 2012) and this study showed a consistent geometry for this planar downdip interface in spite of the different dataset and the different event location methods. Therefore, we conclude that these planar downdip interfaces between 2008 and this study from 2010 to 2011 are the same seismic interfaces. If this is true, then it is natural to interpret that as an overall tendency, this planar downdip interface has been stably sliding for years from 2008 to 2011. The depth of this planar downdip interface is ~ 18 km at around the updip limit of the seismogenic zone.
Meanwhile, Yamada et al. (2011) also reported that there is a low-seismicity zone in the study area of the present study, which overlaps with the low-seismicity zone of the 2011 Mw7.9 event in the present study (Fig. 14). This low-seismicity zone appears to be seismically inactive. Hence, this zone might be an exception to a stable sliding, which we will further discuss in section 5.5.
From the viewpoint of the geometry of this planar downdip interface and its temporal stability of the seismicity, this gently sloped planar downdip interface appears to be a plate interface of a subducting slab. However, it is questionable to conclude that this planar downdip interface is the plate interface at the top of the oceanic crust. As shown in Figs. 10–12, the active source seismic surveys revealed that the depth of the plate interface at the top of the oceanic crust is ~ 10 km (Mochizuki et al., 2008; Tsuru et al., 2002) and not 18 km. Nishizawa et al. (2009) also performed a seismic survey close to the Line EW of Mochizuki et al. (2008) and showed that the plate interface depth at around the top of the seamount was ~ 13 km, a few kilometers deeper than that of Mochizuki et al. (2008). On one hand, Mochizuki et al. (2008) applied an active source seismic tomography to determine the velocity structure. Moreover, an arrival time migration method was applied to identify and determine the depth of the plate interface validated by a synthetic waveform. Nishizawa et al. (2009) applied a wide-angle refraction survey method to obtain the depth of the plate interface. Accordingly, it is difficult to directly examine the depth difference of the plate interface between these studies. However, the depths of the plate interface from these seismic surveys are considerably shallower than the event distribution of the planar downdip interface in this study.
The depth offset between the plate interface from the seismic profile and the planar downdip interface from the small earthquakes is more evident at Line 13 (Fig. 12) than one at Line EW (Fig. 11). At Line 13, the depth offset is ~ 8 km at around the top of the subducting seamount. This appears to be a discrepancy between the depth of the plate interface inferred from the event distribution and those obtained from the active source seismic surveys.
Because the Vp/Vs ratio was optimized for the event location in the present study, we believe that the event depth is hardly biased. To further examine the effect of velocity model error against the depth of the event location, we performed a set of event location tests using different velocity models. The test conditions and results are presented in Table 1. The test result shows that the average event depth shift is at most 1.3 km. Even the nonoptimal Vp/Vs ratio of 1.78 does not explain the departure of event depth from the plate interface. In addition, the event location of the P-only dataset without using the S-wave hardly changed the average event depth below the OBS network. This supports that the S-wave velocity structure is accurate enough to constrain the event depths to our final results. Consequently, we conclude that the error of the velocity structure model is not the cause of the discrepancy between the depth of event location and the depth of plate interface using the active source seismic survey.
Table 1
Test Conditions of the Velocity Model Error Effect and Results.
| Use of seismic phases | Vp/Vs ratio below the basement | Mean event depth [km] | Mean event depth shift [km] | Mean coherence |
Reference velocity model | P&S | 1.74 (optimal) | 19.36 | - | 0.766 (best) |
Reference velocity model | P-only | - | 18.09 | −1.27 km | 0.833 (note: P-only) |
Vp/Vs change | P&S | 1.78 | 18.26 | −1.10 km | 0.764 |
The remaining possibility that can cause the event depth error is the presence of an extremely low-velocity anomaly in a real velocity structure that was not incorporated in the velocity model, especially for the S-wave around the plate interface. However, we believe that such an anomaly is unrealistic. First, the P-only event location did not have such a shift. Second, to result in the 8 km of the depth shift, ~ 1.0 s of the S − P time error must be accounted for throughout the study area (roughly assuming Vp = 6 km/s and Vs = 3.4 km/s). However, past studies using active source seismic surveys did not identify such a low-velocity layer (Mochizuki et al., 2008; Nakahigashi et al., 2012, Nishizawa et al., 2009; Tsuru et al., 2002). In this way, the velocity model error effect is difficult to explain this departure between the depth of the plate interface and the depth of events. Consequently, errors in the velocity structure make it hard to explain this prominent depth offset between the top of the oceanic crust and the planar downdip interface.
5.3 High-angle dipping plane above the downdip planar interface
As presented in Fig. 13, a high-angle dipping plane is identified above the downdip planar interface. The depth of this plane around the updip limit of the seismogenic zone is ~ 10 km. This depth appears to agree with the depth of the plate interface from seismic profiles. Therefore, this high-angle dipping plane could be a part of a plate interface. However, to avoid any mis-conclusion, we discuss the following two cases: 1) the planar downdip interface is the plate interface as the top of the oceanic crust (Fig. 15a) and 2) the high-angle dipping plane is a part of the plate interface (Fig. 15b).
5.3.1 Case 1: Planar downdip interface is the plate interface
In this case, the high-angle downdip interface is the subsurface structure above the plate interface. Wang and Bilek (2011, 2014) suggested that the subduction of the seamount causes microfractures in the overriding plate. As an alternative scenario, a cutting-off of the seamount from its base may be the other candidate for the consequence of the seamount subfuction (Cloos, 1992; Cloos & Shreve, 1996). Further, if this high-angle downdip interface is shallower than the top of the seamount, an out-of-sequence fault or accretionary wedge is perhaps the other candidate for a subsurface structure (e.g., Park et al., 2000). However, these structures are not identified in the off-Ibaraki region (Tsuru et al., 2002). Accordingly, in the particular case shown in Fig. 15a, either microfractures or the cutting-off of the seamount may be the potential causes of the fracturing of the overriding plate which we further discuss below.
In microfractures and cutting-off scenarios, the planar downdip interface is supposed to be a plate interface. As discussed in the previous subsection, we suppose that the case presented in Fig. 15a is less likely to occur. We raise a few additional factors that need further explanations for each microfracture and cutting-off scenario. First, the microfracture scenario does not explain why the shallow high-angle interface was activated only after the Mw7.9 event rather than a continual stable seismic activity. According to Wang and Bilek (2011), the microfracture is the consequence of a compressional stress against the overriding plate by a seamount subduction. The microseismicity associated with the microfractures is expected to occur continuously other than the aftershock. However, the observation in this study showed no such significant events above the plate interface. Second, if the planar downdip interface is the plate interface at the top of the oceanic crust, an explanation is needed as to why this interface exhibits no topological signature of the subducting seamount. Particularly in the case of the cutting-off scenario, an extra discussion is required if the base of the cutting-off interface is topologically smooth enough to cause stable seismic activity even before the Mw7.9 event.
5.3.2 Case 2: High-angle downdip plane is the plate interface
The second case is that the high-angle dipping plane is a part of the plate interface (Fig. 15b). The depth of this plane around the updip limit of the seismogenic zone is ~ 10 km. This is in reasonable agreement with the depth of the plate interface from the seismic survey (Figs. <link rid="fig11">11</link> and 11). This case suggests that the events are dominant along or below the plate interface and not above. Conversely, previous studies on seamount subduction anticipated that the seismicity on the overriding plate would be enhanced by developing microfractures (e.g., Sun et al., 2020; Wang and Bilek, 2011, 2014). No reasonable models appear to exist to explain the occurrence of the events below the plate interface in previous studies. Previous seamount subduction studies implicitly suggested that the oceanic plate is not fractured (see review by Wang & Bilek, 2014). Perhaps, the subducting oceanic plate is already fractured as reported in recent studies (e.g. Hino et al., 2009, Obana et al., 2021)
Most importantly in this case, one open question arises, i.e., how is the stable high seismicity of this planar dipping interface accounted for if it is deeper than the plate interface? As stated, this planar downdip interface seems stably sliding for years and it is a challenge to explain how such stable sliding of this interface persists for years below the oceanic crust. This topic is beyond the scope of this study and we cannot provide an answer for this question here. Further study is required, such as investigating a double-difference relocation and a seismic tomography for determining both P- and S-wave velocities to reveal the precise geometry of these interfaces and corresponding velocity structures.
5.4 Spatial boundary between earthquakes and shallow tectonic tremors
Shallower than the updip limit of the seismogenic zone, the tectonic tremors were identified using S-net (Nishikawa et al., 2019). These tremors were found after the deployment of S-net from 2016 to 2018 after the OBS experiment of this study. Meanwhile, shallow tectonic tremors were not identified during the OBS observation period. This is partly because of the difficulty in discriminating the signals of the tectonic tremors and those of the regional aftershocks of the Tohoku-oki earthquake. Here, we assume that the tremor distribution is temporally steady enough not to invade the seismogenic zone.
The noticeable feature of the shallow tectonic tremor distribution is that this tremor is spatially complementary with the normal earthquakes bounded by the updip limit of the seismogenic zone (Fig. 10–12). Kubo and Nishikawa (2020) showed that the rupture area of the Mw7.9 event and the subducting seamount are spatially complementary. Sun et al. (2020) showed that small-to-moderate earthquakes occur between the rupture area and the tectonic tremor. The present study agrees with Sun et al. (2020) in a much more precise manner that the spatial gap between the rupture area of the Mw7.9 event and the tremor is filled by small earthquakes located at the front-end of the seamount. The rupture area of the Mw7.9 event is discussed in the next subsection 5.5.
This spatial continuity between the small earthquakes and shallow tectonic tremors naturally suggests that the locations of these activities would be smoothly connected with the same or nearby interfaces. If this is true, it would be interesting and important to discuss what controls the boundary between this tectonic tremor and small earthquakes. The answer may not be as simple because as discussed in the previous subsection, the depth profile of the seismicity exhibited a variation along with the depth below the oceanic crust.
Regarding the shallow tectonic tremor, understanding the tremor generation mechanism is still in progress (e.g., Ide, 2021); however, extensive research is ongoing in subduction zones worldwide. Previous studies revealed that tectonic tremor comprises swarms of low-frequency earthquakes (LFEs) (Beroza & Ide, 2011; Nishikawa et al., 2019; Shelly et al., 2007). The duration of the tectonic tremor is approximately tens of seconds or longer (e.g., Nakano et al., 2019). The characteristic frequency content of LFEs is 1–8 Hz (Ide et al., 2007), which overlaps with those of the small earthquakes located in the study area (> 4 Hz). This indicates that the tremor region is also seismogenic in the sense of radiating elastic energy at these high-frequency bands in the order of Hertz. From tectonic implications, the shallow tectonic tremors were shown to occur at the forefront of an accretionary wedge (Obana & Kodaira, 2009) or a shear zone around the décollement on the top of the oceanic crust (Hendriyana & Tsuji, 2021, Sugioka et al., 2012). This study follows these previous studies proposing that shallow tectonic tremors occur along or in the vicinity of the plate interface.
However, these tectonic structures and tremor locations do not fit with the off-Ibaraki region because this region is characterized by the lack of décollement or an accretionary prism (Tsuru et al., 2002). According to the multichannel seismic survey, this earthquake-tremor boundary corresponds to the top of the seamount. No such subsurface structures are identified herein (Figs. 11 and 12). Perhaps, other tectonic structures or mechanisms may be required to account for the tremor generation at the off-Ibaraki region.
Instead of the accretionary wedge or a shear zone around décollement, this study considers a case where the morphology of the subducting seamount surface gives control to define this seismogenic-tremor boundary. This discussion below is based on the numerical modeling study reported by Sun et al. (2020), showing that the effective normal stress around the top of the seamount is considerably smaller than that at the front-end side of the seamount.
As aforementioned, the updip limit of the seismogenic zone is located along ~ 10 km of isocontour of the plate interface depth. This 10-km contour is close to the top of the subducting seamount. According to Sun et al. (2020), a stress shadow may be generated along the plate interface at the shallow ward in the seamount’s wake owing to the variation of the slope of the seamount morphology. Because this stress shadow is the region where the effective normal stress is considerably reduced, a shear slip around the top of the seamount’s back-end side is easier to initiate than that at the front-end side (Sun et al., 2020). In contrast with the back-end side of the seamount which can be a stress shadow region, at the front-end side of the seamount, the effective normal stress will be considerably larger compared with that in the shallow tremor region; hence, it is harder for the tectonic tremor to be initiated. Accordingly, it is implicated that the 10 km of the isodepth plate depth contour corresponding to a spatial boundary between the earthquakes and tectonic tremors is a boundary between the non-stress shadow and the stress shadow. This model presented by Sun et al. (2020) explains the boundary of the seismogenic and tremor region observed in this study even without the presence of an accretionary wedge or a décollement along the plate interface.
5.5 Semicircular low-seismicity zone and the largest Mw7.9 aftershock event
As shown in Fig. 10, a large semicircular low-seismicity zone was identified. The size of this low-seismicity zone is ~ 30 km × 25 km along the strike and dip direction, respectively. The seismicity of this zone has been continuously inactive since the aftershock of the 2008 M7.0 event as per the OBS observation (Fig. 14a). The event detection capability is quite good in this low-seismicity zone; the event-detection lower limit at ~ 20-km depth is M0.5 (Fig. 6). In this low seismicity zone, a tectonic tremor was not identified in this zone, especially before the Mw7.9 event. Because of these reasons, an extremely weak coupling condition along the fault plane is very unlikely in this low-seismicity zone.
Meanwhile, it is well known that the aftershocks occur around the rim of the main coseismic rupture area (e.g. Mendoza & Hartzell, 1988, Yagi et al., 1999). In the present study, the hypocenter of the Mw7.9 event was relocated around the western rim of the semicircular low-seismicity region underneath the OBS network (Fig. 8). Nakatani et al. (2015) reported a consistent hypocenter of the Mw7.9 event. The rupture direction of the Mw7.9 event is known to have propagated toward updip from the hypocenter (Kubo et al., 2013; Suzuki et al., 2020). These results suggest that the semicircular low-seismicity zone corresponds to a part of the coseismic rupture area of the Mw7.9 event, possibly the main rupture area. The A-GPS survey (Honsho et al., 2019, Tomita et al., 2017) showed that in contrast to the Tohoku-oki region, the southern Japan trench region including the off-Ibaraki region is characterized by an afterslip region after the 2011 Tohoku-oki earthquake. The A-GPS data in this off-Ibaraki region from 2012 to 2016 are shown in Fig. 10. This A-GPS result suggests that the afterslip of the Mw7.9 event may have continued for years.
One may argue that the fault plane depth of the Mw7.9 thrust event is still controversial because there are two dipping planes of a planar downdip interface and a high-angle downdip interface (Fig. 15), hence the coseismic fault plane cannot be unambiguously specified. Actually, this study cannot provide a constraint regarding the depth of the fault planes. Further characterization such as a delta CFF analysis will provide a better insight into the fault plane of this Mw7.9 event.