The legacy of Tohoku-oki
Ten years ago, the magnitude 9 Tohoku-oki earthquake rocked Japan and caused massive damage. The earthquake also generated a destructive tsunami, the impacts of which are still being managed today. Kodaira et al. review what was learned from the tremendous number of observations from the great earthquake that unexpectedly ruptured into a shallow part of the megathrust fault. Postseismic deformation is ongoing, as is the risk of another very large normal fault earthquake seaward of the Japan Trench.
Science, this issue p. eabe1169
Ten years have passed since the 2011 Tohoku-oki earthquake occurred in the Japan Trench, where the Pacific plate subducts beneath the continental plate. The earthquake and tsunami caused enormous damage along the coast of northeast Japan in the Tohoku region, and local communities are still recovering. Tsunami traces more than 10 m above sea level were observed along 530 km of coastline in central and northeast Japan, and runups higher than 20 m were observed over about 200 km of the Tohoku coast. The tsunami inundated an area of 561 km2, and its runup reached a maximum of 40 m in northern Tohoku. These statistics made it one of the largest tsunamis ever recorded in historical literature as well as in geological records.
The earthquake occurred in the vicinity of the world’s most densely instrumented seismic, geodetic, and tsunami observation networks, which clearly recorded the dramatic geodynamic effects of the earthquake. In addition, geophysical and geological data were acquired offshore, in the rupture zone, before and after the earthquake. These marine observations are decisive ground-truth data showing that coseismic slip exceeded 50 m in places and that the rupture reached the shallowest parts of the megathrust fault in the subduction zone.
Geodetic data from the seafloor before and after the earthquake, from the Global Navigation Satellite System–Acoustic combination technique, show that the seafloor near the epicenter underwent coseismic displacement of 31 m toward the southeast and had an uplift of 3 m. Differential bathymetric mapping, comparing the seafloor before and after the earthquake, inferred a coseismic seafloor displacement of more than 50 m at the trench axis. A rapid-response deep-sea drilling project successfully collected material from the earthquake’s rupture zone on the plate boundary fault near the trench and measured the thermal anomaly due to frictional heating of the fault during coseismic slip. This showed that the plate boundary fault is rich in weak layers of clay and suggested that thermal pressurization within the clay layers promoted the exceptionally large coseismic fault slip. These ground-truth data provided evidence to constrain a slip behavior in a shallow part of the subduction zone. Crucial progress in determining the coseismic slip character is represented by the evidence showing lateral variations of coseismic slip near the trench and possible structural factors to control the lateral variation. For example, no resolvable coseismic seafloor displacement from differential bathymetry was observed at the north and south of the main rupture zone, and seismic images showed lateral discontinuity of a pelagic clay layer due to subduction of petite-spots at the north of the main rupture zone.
Crustal deformation, including afterslip and viscoelastic relaxation of the mantle, continues in the Japan Trench 10 years after the earthquake. Viscoelastic relaxation is predominant in the central part of the trench, where the large coseismic slip extended to the trench. The relaxation displaces this seafloor westward, whereas afterslip displaces it eastward around the main rupture zone. These observations, along with aftershock activity from normal fault earthquakes in the incoming oceanic plate, indicate that trench-normal extension remains in the oceanic plate and seaward of the main rupture zone. The recurrance probability of a great earthquake (magnitude = ~9) in the Japan Trench in the near future is very low, but even 10 years after the Tohoku-oki earthquake, seismic activities in east Japan, including the trench outer-slope and surrounding areas of the main rupture zone, are still higher than those before the earthquake. Because past observations show that large normal fault earthquakes in the incoming oceanic plate can occur after great plate-boundary earthquakes, further investigations into the temporal changes in the stress state around the Japan Trench are necessary to evaluate the possibility of such a “follow-up” earthquake in the incoming Pacific plate.
Oblique view of northern Japan and the Japan Trench showing coseismic displacements (black and white arrows), tsunami heights (colored bars), and differential bathymetry profiles across the trench (multicolored lines). The scale of the onland displacements is five times as large as those of the seafloor displacements. Dashed areas indicate approximate areas where a set of three turbidite layers attributed to the Tohoku-oki earthquake and past large events are observed (ellipse) and where a deformed and altered pelagic clay layer by a petit-spot volcanism is inferred from seismic images (half-rectangle), respectively. A yellow circle indicates the location of the Japan Trench Fast Drilling Project (JFAST) site.
The 2011 Tohoku-oki earthquake occurred in the Japan Trench 10 years ago, where devastating earthquakes and tsunamis have repeatedly resulted from subduction of the Pacific plate. Densely instrumented seismic, geodetic, and tsunami observation networks precisely recorded the event, including seafloor observations. A large coseismic fault slip that unexpectedly extended to a shallow part of megathrust fault was documented. Strong lateral variations of the coseismic slip near the trench were recorded from marine geophysical studies, along with a possible cause of these variations. The seismic activities in east Japan are still higher than those before the earthquake, and crustal deformation is still occurring. Although the recurrence probability of a great earthquake (magnitude = ~9) in the Japan Trench in the near future is very low, a large normal fault earthquake seaward of the Japan Trench is a concerning possibility.
The coast of the Tohoku region of northeast Japan, where the 2011 Tohoku-oki earthquake occurred, has repeatedly suffered earthquakes and tsunamis due to subduction of the Pacific plate beneath the continental plate on which Japan sits. Historical records show that earthquake and tsunami hazards have varied greatly along the Tohoku coast, depending on the terrain features and seismic activity. The most recent great earthquakes and tsunamis before the 2011 Tohoku-oki earthquake were the 1896 Meiji Sanriku earthquake [moment magnitude (Mw) 8.0 to 8.2] and the 1933 Showa Sanriku earthquake (Mw 8.4), both of which caused catastrophic damage along the coast of the Sanriku region, in northern Tohoku (Table 1). In particular, the seismic intensity of the Meiji Sanriku earthquake on the Sanriku coast was 3 to 4 according to the Japan Meteorological Agency (JMA) seismic scale, indicating that the felt shaking was moderate. However, the tsunami from this earthquake was very large, with a maximum runup of 38.2 m above sea level (1), and 21,959 people were recorded as dead or missing after the tsunami (Fig. 1). Because the tsunami was larger than that expected from the seismic intensity, this earthquake is known as a “tsunami earthquake” (2). The 1933 Showa Sanriku earthquake is considered to have been an outer-rise earthquake, the rupture of a normal fault in the oceanic plate entering the Japan Trench. Ground shaking was felt with intensities of 4 to 5 in a wide area of eastern Japan, not only in Tohoku but also in the Kanto region to the south. This earthquake caused a tsunami with a runup as great as 29.2 m (3) that caused extensive damage along the Sanriku coast (Fig. 1).
(A) Distribution of seismic intensities (JMA scale) of the 2011 Tohoku-oki earthquake from QuiQuake (91). Black dots with numbers indicate approximate epicenters and dates of the earthquakes in Table 1. White dots indicate the locations of 2011 tsunami traces. (B) Tsunami trace distribution of the 1896, 1933, and 2011 events. The vertical axis corresponds to the coastline of (A). Tsunami heights were obtained from the Japan Tsunami Trace database of Tohoku University, and from the Nuclear Regulation Authority of Japan (92). The original tsunami data of the 1896, 1933, and 2011 earthquakes were reported by Iki (1), Matsuo (3), and the 2011 Tohoku Earthquake Tsunami Joint Survey Group (93), respectively.
Past earthquakes and tsunamis
Previous large earthquakes and tsunamis in the Tohoku region have been recognized from descriptions in historical literature and geological records of tsunami deposits (Table 1). One of the largest of these was the 869 Jogan earthquake, which caused severe damage in the Tohoku region. A historical account reports that the inundation from the ensuing tsunami reached the Tagajo Castle, on the Ishinomaki-Sendai plain on the central Tohoku coast. This description is consistent with the record from tsunami deposits (4–7). Since the 15th century, six tsunami disasters preceding the 1896 tsunami in the Tohoku region are attested by historical accounts and tsunami deposits. The 1454 Kyotoku earthquake (or possibly the 1611 Keicho earthquake) (8–10) caused a huge tsunami disaster on the Ishinomaki-Sendai plain. Other large earthquakes identified in the Tohoku region occurred in 1677 (two events), 1763, 1793, and 1856 (11). These five events caused tsunami height greater than 3 m along the Sanriku coast, but the tsunamis caused no major damage on the Ishinomaki-Sendai plain (Table 1).
The totality of the evidence shows that only two earthquakes in the historical record have caused widespread catastrophic tsunami disasters affecting the entire Tohoku region comparable to the 2011 Tohoku-oki earthquake: the 869 Jogan earthquake and the 1454 Kyotoku (or 1611 Keicho) earthquake (Table 1).
The shaking and the tsunami in the 2011 earthquake
On 11 March 2011, a magnitude 9 earthquake occurred off the Tohoku coast, and the resulting massive tsunami left 18,550 people dead or missing (as reported by the National Police Agency of Japan on 10 September 2020). The damage caused by the earthquake and tsunami to buildings, lifeline facilities, infrastructure, agriculture, forestry, fisheries, and other public facilities amounted to about 16.9 trillion yen, according to government estimates. It was the worst disaster in Japanese history, and it is still affecting the global environment, including through the crippling damage to the Fukushima Nuclear Power Plant. A large area of eastern Japan experienced seismic intensities of 6+ or 7, the highest levels on the JMA scale, and intensity 7 was measured at three locations in central east Japan (Fig. 1). Despite the strong shaking, building damage and collapses tended to be less than expected for an earthquake of magnitude 9 because the predominant seismic waves had periods shorter than those affecting low-rise buildings (12, 13), in addition to Japan’s earthquake risk mitigation effort to strengthen the built and infrastructure environment against seismic shaking.
Tsunami heights greater than 10 m were observed over a distance of about 530 km along the east Japan shoreline, and runups greater than 20 m were observed over about 200 km along the Tohoku coast (Fig. 1), inundating an area of 561 km2 (14). The maximum runup was 39.7 m, at Ryori on the Sanriku coast, exceeding the record there from the 1896 Meiji Sanriku event. On the Sendai plain, the tsunami reached inland more than 5 km from the coastline, causing extensive damage, including by tsunami-related fires. The extent and height of tsunami traces exceeded those of the 1896 Meiji Sanriku and 1933 Showa Sanriku tsunamis (Fig. 1). The tsunami trace height distribution tended to decrease monotonically away from the epicenter toward the north and south, but it reached a maximum along the Sanriku coast, north of the epicenter. Inundation extending far inland on the Ishinomaki-Sendai plain was also a major characteristic of the tsunami. These lateral variations of tsunami height and inundation are attributed to characteristics of tsunami wave behavior and topography of the cast along the Tohoku region. The tsunami along the Sanriku region was characterized by short periods and large amplitudes, whereas the tsunami along the Ishinomaki-Sendai Plain was characterized by long periods and large amplitudes (15). In addition, along the Sanriku region with characteristics of the ria coast, the tsunami energy was suggested to be expended in raising the tsunami height near the coastline, rather than inundating the area a long distance inland. By contrast, on the Ishinomaki-Sendai Plain, a tsunami wave more than 10 m high expended its energy in runup and inundation over a longer distance (Fig. 1).
Tsunami signals of the Tohoku-oki earthquake were recorded by offshore instruments in the Pacific Ocean as well as by coastal monitoring networks, and these records capture not only the propagation of the tsunami but also the progress of individual tsunami waves as they traveled in the Pacific Ocean. Records of sea surface fluctuations made in water depths greater than 100 m can be regarded as tsunami records directly from a tsunami source and are thus invaluable for estimating tsunami sources. For example, two cabled pressure gauges deployed at depths of 1000 m and 1600 m recorded multiple stages of tsunami waves as the water level gradually increased up to 2 m in the first 10 min, and then observations of impulsive tsunami waves 3 to 5 m high. Such signals obtained from offshore observations have provided critical information for deciphering the tsunami source characteristics. The enormous number of observation records has helped elucidate details of the earthquake’s generation mechanism [e.g., (15–18)].
Fault slip that reached the trench and its lateral variation
The Tohoku-oki earthquake was captured by the world’s most densely instrumented seismic, geodetic, and tsunami observation networks. The resulting body of evidence has supported many detailed studies, including review papers, reporting what happened at the causative fault zone [e.g., (19–27)]. We will not cover all these results because of our intended scope, but we focus on the trench-breaching nature of the fault rupture and its lateral variations, which are constrained by marine geophysical, geological, and geodetic data [see also a comprehensive review in (23)].
The distribution of aftershocks is commonly used to gain an overall picture of the size of a coseismic rupture zone; in this case, the area of aftershocks extends over a large area from 35°N to 40°N and from the coastline to the trench, and from 37°N to 40°N, it extends onto the Pacific plate on the seaward side of the trench. A notable observation is that most of these aftershocks were normal fault earthquakes occurring within both the overriding plate and the oceanic plate. This seismic activity contrasted with the steady activity before the earthquake, when reverse fault earthquakes on the plate boundary were dominant (28). Various models of the coseismic fault rupture process have been proposed and are still being debated; however, 43 published models compiled by Wang et al. (22) show that the coseismic slips are mostly far offshore, with peak values in the upper part of the megathrust fault (Fig. 2B). It was commonly thought that a shallow part of a megathrust would resist rupture and that coseismic slip would rapidly terminate propagation [e.g., (29)], although there were studies that showed stress is coseismically transferred to the shallow part of the megathrust to cause seismic slip [e.g., (30, 31)]. The seafloor observations and marine geological and geophysical surveys that have been conducted in rupture zones have provided convincing ground truth data to constraint a slip behavior in a shallow part of the megathrust (32–35) (Fig. 2).
(A) Oblique view of the Japan Trench and northern Japan showing coseismic displacements from the 2011 Tohoku-oki earthquake. White and black vectors indicate coseismic displacements obtained by GNSS (onshore) and GNSS-A (offshore), respectively (32, 34, 94). White and black circles indicate GNSS and GNSS-A stations, respectively. Note the different scales of these two sets of displacements. The scale of the onshore displacements is five times as large as those of the seafloor displacements. See a map view of the original GNSS and GNSS-A data in Iinuma et al. (48). Multicolored lines across the trench indicate differential bathymetry profiles (23, 33, 35, 39). Dashed areas delineated by a and b indicate approximate areas where a set of three turbidite layers attributed to the Tohoku-oki earthquake and the past two large events is observed (41), and where deformed or altered pelagic clay layer is inferred from seismic data (44), respectively. Yellow circle indicates the location of the JFAST site (36). (B) Oblique view of mean magnitudes of 43 published coseismic slip models (22). Dashed contours indicate depths of the plate interface below sea level. Vertical exaggeration of (A) is three times as large as that of (B). (C) Map view of the dashed quadrangular area in (A). White numbers indicate horizontal seafloor displacements inferred from the differential bathymetry data (23, 33, 35, 39). The topographic and bathymetric data used are SRTM+15 (95). Red relief imaging method (96) was used to draw the maps.
Magnitude 7 to 8 class earthquakes have repeatedly occurred in the Japan Trench. Before the Tohoku-oki earthquake, the Headquarters for Earthquake Research Promotion, Japan, had estimated the probability of a M 7.5 to 8 (but not M 9) earthquake in the middle of the Japan Trench within the next 30 years to be more than 90%. Therefore, even before 2011, the central part of the Japan Trench had been intensively studied through ocean-bottom earthquake observations, geodetic observations, bathymetric surveys, and plate boundary fault imaging. Seafloor geodetic data, which were obtained before and after the 2011 earthquake by using the Global Navigation Satellite System–Acoustic (GNSS-A) combination technique, showed coseismic seafloor deformation greater than 30 m (32, 34). For instance, Kido et al. (34) reported a coseismic seafloor displacement of 31 m horizontally toward the southeast and 3 m vertically at about 50 km landward from the trench (Fig. 2). A bathymetric profile extending from the epicenter across the trench to the incoming Pacific plate, surveyed before and after the earthquake, showed clearly that coseismic seafloor displacement had reached the trench axis (33). Fujiwara et al. (33) used differential bathymetric data to infer a coseismic horizontal displacement of 50 to 56 m and uplift of 7 m on the landward slope of the trench (Fig. 2). In addition, Kodaira et al. (35) produced distinctive images of disturbed trench sediment consistent with trench-breaching coseismic slip based on seismic surveys before and after the earthquake along the same profile as the differential bathymetry survey.
Although these studies showed convincingly that coseismic slip of more than 50 m reached the trench axis, they could not answer essential questions about the mechanism of this coseismic slip behavior. The Japan Trench Fast Drilling Project (JFAST) of the Integrated Ocean Drilling Program, which was conducted about a year after the earthquake, provided essential insights into this question (36). This project had two scientific objectives. The first was to collect and return a sample from the plate boundary fault to examine the fault’s internal structure and measure the physical properties of the fault rocks. The second was to measure the temperature anomaly caused by frictional heating on the plate boundary fault and determine the fault’s coefficient of friction during the 2011 earthquake. As part of JFAST, drilling vessel Chikyu drilled about 850 m into the seafloor at a water depth of 6900 m, penetrating the rupture zone at a location near the trench where coseismic slip had reached the trench. The cores retrieved from the plate boundary fault revealed that the fault was confined in a layer of pelagic clay less than 5 m thick (36). High-velocity friction experiments on the core material yielded evidence that the fault had undergone a small stress drop with very low peak stress and steady-state shear stress (37). Temperature measurements in the borehole for 9 months from July 2012 recorded a temperature anomaly of 0.31°C, from which an apparent friction coefficient of 0.08 was estimated (38). This value was close to the in situ apparent friction coefficient (0.03) determined by the friction experiments on the core material made under impermeable conditions (37). From these results, JFAST project scientists concluded that the plate boundary fault is rich in weak clay minerals and that thermal pressurization within the clay layers may have helped enable the large coseismic fault slip to reach the trench (37).
Studies about seafloor observations in the past few years provided crucial information about lateral variations of the coseismic slip and possible structural factors to control the lateral variations. A differential bathymetry profile 50 km north of the epicenter showed that coseismic seafloor displacement had reached the trench, just as a profile across the epicenter had shown (35) (Fig. 2). However, no resolvable displacement was apparent in bathymetric profiles around 37.5°N (19), 70 km south of the epicenter, or around 39.3°N, 110 to 120 km north of the epicenter (39) (Fig. 2). These observations delimit the area of exceptional slip, which corresponds well with the slip distribution estimated from seismic and geodetic data (Fig. 2) [e.g., (21, 22)]. It should be noted that the precision of the differential bathymetry is a few meters vertically and 20 m horizontally, so it does not capture smaller variations (33).
The sediment record from the seafloor offered more insight into the distribution of exceptional slip along the Japan Trench (40, 41). Shallow sediment samples were collected by piston coring between 37°N and 40°N along the Japan Trench after the Tohoku-oki earthquake. A fresh turbidite layer, attributed to the 2011 earthquake, was found on the seafloor, and turbidite layers from the 1454 Kyotoku and 869 Jogan earthquakes were found deeper in the cores. However, a set of these three turbidite layers were only observed in an area 120 km long in the central part of the Japan Trench, centered at 38°N.
Recent studies have suggested that the strong lateral variation of exceptional slip at the trench stems from discontinuities in the pelagic clay layers in the shallow part of the plate boundary fault that prevented the lateral propagation of the seismic rupture. For instance, seamounts covered with biogenic sediment are being subducted at around 37°N (42), and pelagic clay layers have been deformed and altered by young igneous activity, known as petit-spot volcanism (43), at around 39°N (44). These features have been proposed to disrupt the lateral continuity of the pelagic clay layer and thus effect the slip behavior of the shallow part of the plate boundary fault (42, 44). It should be noted that, recently, Nishikawa et al. (45) reported that low-frequency tremors and very low frequency earthquakes are concentrated in the southern (36°N to 37°N) and northern (39.5°N to 40.5°N) parts of the main rupture zone, where the subducted seamounts and the petite-spots are proposed to cause lateral discontinuity of the pelagic clay layer (42, 44). However, biomarker analyses of the JFAST core have detected multiple faults within many of the stratigraphic units that may be subject to seismic slip (46). Clearly, there is much still to be learned before we can conclude that the lateral discontinuity of pelagic sediments is a structural factor controlling slip to the trench.
The coseismic slip distributions estimated from seismic waveforms (47), GNSS and GNSS-A measurements (48), differential bathymetry (33, 39), time-lapse seismic imaging (35), and turbidite distribution (40, 41) consistently support a scenario in which coseismic slip greater than 50 m reached the trench in an area limited to between 38°N and 39.3°N, in the middle of the Japan Trench. However, to account for the observed tsunami heights, particularly in the Sanriku region, several tsunami modeling studies require another tsunami source near the trench axis at around 39° to 40°N, where seismic, geodetic, and bathymetry data do not indicate large slip on the plate boundary fault (15). A source at that location may have involved seafloor deformation other than the elastic response to the coseismic slip of the plate boundary fault. Two proposed explanations are uplift of the seafloor due to inelastic deformation of the seaward tip of the overriding plate (39) and the occurrence of a large submarine landslide (49). A recent high-resolution seismic study shows slump deposits, suggesting that slope failures have occurred in the past, in the area near the proposed additional tsunami source around 39° to 40°N (50). Nakamura et al. speculated that the slump might have contributed to tsunamigenesis; however, further geological sampling and quantitative estimation are necessary to examine whether the slump occurred during the Tohoku-oki earthquake and whether the volume of the slump can be equivalent with mass failure to generate the tsunami from the additional source, as predicted by the tsunami inversion. The existence and nature of such an additional tsunami source, including a reconciliation of the seismic and geodetic data with the tsunami data, are a major unresolved question related to the 2011 Tohoku-oki earthquake.
What is happening and what may happen next in the Japan Trench
Great earthquakes are typically followed by postseismic deformation due to afterslip and viscoelastic relaxation. Afterslip is slow, aseismic slip that occurs on the fault plane mostly around the area of coseismic slip to release the increased shear stress [e.g., (51)]. Viscoelastic stress relaxation owes mainly to the mechanical properties of the asthenosphere [e.g., (52)]. The 2011 Tohoku-oki earthquake caused large stress perturbations around the rupture area, triggering both afterslip and viscoelastic relaxation. Land-based geodetic observations have shown that large postseismic deformation occurred immediately after the mainshock (53, 54). Distinguishing afterslip from viscoelastic relaxation on the basis of land-based data alone is difficult, yet accurate knowledge of the afterslip distribution is necessary to evaluate the risks of forthcoming interplate earthquakes because afterslip releases strain energy accumulated on locked interplate boundaries. Fortunately, data from seafloor geodetic stations previously installed in the forearc region of the Japan Trench reveal distinctive features of the postseismic crustal deformation.
Notably, the GNSS-A observations detected landward movement of the seafloor immediately after the mainshock in the main rupture zone (55, 56) (Fig. 3). These landward movements, exceeding the plate convergence rate, can be accounted for by a combination of the recovery of interplate locking and viscoelastic relaxation. Several studies incorporated heterogeneous viscoelastic structure to explain the postseismic deformation observed onshore and offshore (56–58). They suggested that a weak layer of low viscosity exists beneath the subducting oceanic plate, although the viscosity of the oceanic asthenosphere was traditionally thought to be an order of magnitude higher than that of the continental mantle wedge (59). Owing to its nonlinear rheology, the low-viscosity layer may be a transient result of large stress perturbations from the huge coseismic slip of the Tohoku-oki earthquake near the trench (60).
(A) White and black vectors indicate displacement rates during the period from September 2012 to September 2016 (69). Dashed blue lines indicate surface traces of outer-rise faults (83). The yellow dashed line marks the location of the schematic profiles shown in (B) and (C). The topographic and bathymetric data used are SRTM+15 (95). Red relief imaging method (96) was used to draw the maps. (B and C) Schematic profiles of the continental and oceanic plate across the Japan Trench along yellow dashed lines a-b and c-d in (A). Gray arrows represent viscoelastic flow in the asthenosphere (60). Yellow squares and circles denote onshore and offshore geodetic observation sites, respectively, with their displacement rates indicated by black arrows. Focal-mechanism solutions are shown in (B) for the outer-rise earthquakes (75). The neutral zone between compression and extension stress state before and after the 2011 earthquake is outlined by violet dashed lines in (B). White harpoon arrows in (C) represent the afterslip on the plate interface (71).
These studies of the viscoelastic relaxation have led in turn to improved estimates of the afterslip distribution. Studies that estimated afterslip without considering viscoelastic relaxation have reported that large (up to 3 m) afterslip at the deep plate interface beneath the coastline of the Iwate and Miyagi prefectures in central and northern Tohoku had occurred by 7 months after the mainshock (54, 61–63). Other studies have estimated a similar afterslip distribution while taking viscoelastic relaxation into account; these used a simple layered structure to estimate viscoelastic relaxation, but their models did not fully explain the seafloor crustal deformation data (64–66). Two studies have successfully explained both onshore and offshore geodetic observations by taking heterogeneous viscoelastic structure into account (57, 67). They estimated small afterslip on the deep plate interface at areas that are separated from the area of coseismic large slip; however, they could not determine whether or not the afterslip occurred in the coseismic rupture area because seafloor geodetic observations were not dense enough to resolve the afterslip distribution in detail. Further discussions about the afterslip distribution on the deep plate interface were conducted by Wang et al. (22). Recently, two studies documented seafloor displacement fields in a wide area along the Japan Trench based on GNSS-A observations at 20 stations (68, 69) (Fig. 3). They proposed that viscoelastic relaxation is predominant in the large coseismic slip area where westward displacements are distinctive, whereas eastward displacements south of the main slip area indicate rapid afterslip. New seafloor geodetic data documenting the ongoing postseismic deformation will be useful not only for investigating the afterslip distribution but also for revising the coseismic slip distribution on the basis of better knowledge of the viscoelastic response [e.g., (66, 70, 71)]. Precise and realistic viscoelastic structure modeling is necessary to explain all coseismic and postseismic geodetic observations.
The change in seismic activity on the plate interface is another key observation to constrain the afterslip distribution. Three recent studies have estimated the afterslip distribution from the activity of small repeating earthquakes and the distribution of small thrust events (72–74). These results based on seismic data are independent of viscoelastic relaxation, but the method cannot be applied at the very shallow plate interface, where small earthquakes never occur.
The large coseismic slip during the Tohoku-oki earthquake also affected the stress state in the overriding continental and subducting oceanic plates. Immediate changes were reported in the focal mechanisms in and around the coseismic rupture after the mainshock (27, 28). Ocean-bottom seismograph observations near the trench axis have revealed that the earthquake changed the stress regime in the Pacific plate shallower than 30 to 40 km to trench-normal extension in both incoming and subducting regions (75, 76) (Fig. 3). Reflecting this change in the stress field, a doublet event consisting of a shallow normal-fault event (M 7.2) and a deep thrust event (M 7.1) occurred on 7 December 2012 within the Pacific plate near the Japan Trench (77). Kubota et al. (78) suggested that the downdip edge of the shallow normal-faulting seismic zone became substantially deeper (~30 to 35 km) than its depth in 2007 (~25 km) by analyzing tsunami waveform data from ocean-bottom pressure gauges. This growth in the area of tensional stress may play an important role in the occurrence of large normal-faulting earthquakes around the Japan Trench.
Although the recurrence probability of an occurrence of a great earthquake (M = ~9) in the Japan Trench in the near future is very low (79), seismic activities in east Japan, including the trench outer-slope region and surrounding areas of the main rupture zone, are still higher than those before the Tohoku-oki earthquake. The number of earthquakes greater than magnitude 4 in the outer trench-slope in 2019–2020 is about twice as high as an annual average of those between 2001 and 2010 (80). Moreover, a potential of tsunami hazard by a large earthquake in the outer trench-slope is concerning [e.g., (81)], because of the stress state after the earthquake and past observations showing that megathrust earthquakes in the shallow part of subduction zones are sometimes followed soon afterward by large normal-fault type earthquakes. For instance, on 13 January 2007, an Mw 8.1 extensional earthquake occurred within the Pacific plate beneath the seaward edge of the Kuril Trench 2 months after the interplate megathrust earthquake (Mw 8.3) of 15 November 2006 (82). In the northern Japan Trench, the 1933 Showa Sanriku outer-rise earthquake is considered to have a similar relationship with the 1896 Meiji Sanriku earthquake that ruptured the interplate fault near the trench. In the 10 years since the 2011 Tohoku-oki earthquake, a large normal-fault earthquake greater than magnitude 8 has not occurred in the outer trench. However, seafloor crustal deformation and numerical simulations suggest that viscoelastic relaxation is ongoing beneath the Pacific plate; therefore, extensional strain is still remaining within the plate. If such an earthquake occurs in this region, a large tsunami will surely strike the Tohoku coast again. A recent study based on the realistic fault geometry of the outer trench region (Fig. 3) has estimated that such an event could produce a maximum tsunami height of more than 20 m (83). Further investigations of the stress state around the Japan Trench are necessary to better evaluate the possibility of a potentially devastating “follow-up” earthquake.