• Open Access

Stress perturbations and seismic response associated with the 2011 M9.0 Tohoku-oki earthquake in and around the Tokai seismic gap, central Japan



[1] The M9.0 Tohoku-oki event is the largest earthquake in Japan's modern history. Social concerns and scientific interests require an urgent evaluation of this event's impact on other megathrust zones nearby Japan. Here we investigate the stress transfer from the Tohoku-oki event on the Tokai subduction zone, where an M8-class megathrust earthquake has been long-time anticipated. First we demonstrate that the clear increase of crustal seismicity around Izu Peninsula, near the Tokai gap, is the consequence of a Coulomb static stress increase due to the Tohoku-oki earthquake, calculated using a variable slip model for the mainshock and regional focal mechanisms of events occurred before and after the M9.0 earthquake. The largest stress increase is of about 1.0 bar, as estimated on earthquake nodal planes of maximum stress change. The time-decay characteristics of the activated seismicity favor the triggering by static stresses. Such validations of the Coulomb hypothesis support our stress perturbation assessment on the Tokai gap. To precisely calculate the stress changes on the Tokai source, we use the curved plate interface and plate-motion inferred rake directions. The computed stress changes are predominantly positive but have relatively small values, of less than 0.1 bar. The large afterslip (Mw8.6) and the aftershocks following the Tohoku-oki earthquake caused additional, but minor, stress increases. Our results imply that the stress-state on the Tokai plane did not change significantly after the Tohoku-oki event, however static stress driven seismic activation in neighboring areas can bring a secondary impact on the interplate seismicity.

1. Introduction

[2] The MJMA9.0 (Mw9.0) Tohoku-oki earthquake occurred on March 11, 2011, at the subduction of the Pacific plate under the Eurasian plate (Figure 1). The earthquake was followed by many large aftershocks, with magnitudes up to MJMA7.7 (the Japan Meteorological Agency (JMA) magnitude, MJMA, is denoted hereafter by M; M and the moment magnitude, Mw, are used throughout the paper). Besides the aftershock activity taking place on or close to the megathrust fault of the giant event, significant earthquake activity occurred at relatively large distances, in particular in inland areas [e.g., Toda et al., 2011].

Figure 1.

Map showing the fault area of the M9.0 Tohoku-oki earthquake with superposed slip distribution [Suzuki et al., 2011] and the seismogenic zone of interplate megathrust earthquakes along Nankai-Suruga trough. This seismogenic zone can be divided into three segments, Tokai, Tonankai and Nankai (green-colored areas). The violet rectangle indicates the study region. The three stars are the epicenters of the Tohoku-oki earthquake (M9.0, Mw9.0), its largest aftershock (M7.7, Mw7.8) and the Shizuoka earthquake (M6.4, Mw5.9). Red triangles denote active volcanoes. The map area is shown as a blue rectangle in the inset.

[3] Another source region that has the potential of a giant earthquake would be the Nankai-Suruga trough, in southwest Japan, where M8-class megathrust earthquakes occur at intervals of 100–200 years [Ando, 1975] at the subduction of the Philippine Sea plate under the Japanese Islands. The megathrust seismogenic zone is divided into three segments: Nankai, Tonankai and Tokai (Figure 1). The most recent megathrust earthquakes struck along the Tonankai and Nankai segments in 1944 and 1946, respectively. However, the Tokai segment did not rupture at that time and is still in a locked state, holding a considerable slip deficit [e.g., Ohta et al., 2004]. In contrast, all three segments ruptured in 1854 and 1707. Therefore, the occurrence of the anticipated Tokai event is of great concern for earthquake hazard in the region.

[4] Here we perform in-depth analyses of static Coulomb failure stress changes (ΔCFS) and seismicity in a relatively large region (Figure 1) around the Izu Peninsula, central Japan. We first discuss the increase of crustal seismicity in several areas located either on the Philippine Sea plate or Eurasian plate and relate this to a regional static stress increase for the predominantly strike-slip crustal earthquakes, as a result of the Tohoku-oki earthquake. Since the most hazard-prone earthquakes in the region are the megathrust events in the Tokai subduction zone, we next evaluate the impact (in terms of ΔCFS) of the Tohoku-oki earthquake on the Tokai source. In addition we estimate the static stress changes on the Tokai plane due to subsequent significant seismic activity, as well as afterslip following the M9.0 Tohoku-oki event.

2. Data and Methods

[5] We use the Tohoku-oki earthquake slip model ofSuzuki et al. [2011] (Figure 1), determined from the inversion of K-NET and KiK-net [Okada et al., 2004] strong-motion data, to resolve the coseismic ΔCFS either on nodal planes of regional earthquakes or on the Tokai megathrust plane. The post-seismic ΔCFS on the Tokai plane is evaluated using an afterslip model of the Tohoku-oki earthquake obtained from the inversion of geodetic data [Geospatial Information Authority of Japan (GSI), 2011; Ozawa et al., 2011].

[6] The stress change is computed [King et al., 1994] using the equation: ΔCFS = Δτ + μ′Δσ (1), where Δτ is the shear stress change on a given fault plane (positive in the direction of fault slip), Δσis the fault-normal stress change (positive when unclamped), andμ′ is the apparent coefficient of friction. We use μ′ = 0.4, which is an average value among those usually used in the literature [e.g., King et al., 1994; Toda et al., 2011, and references therein]. However, we have checked the robustness of our findings by using μ′ values from 0.0 to 0.8. In all cases the spatial distribution of ΔCFS values does not change significantly compared to the results obtained for μ′ = 0.4. Since some studies [e.g., Hasegawa et al., 2011] indicate that μ′ in subduction zones, on the plate interface, could be very small (i.e., close to 0), we have verified that using such small apparent friction coefficients for the Tokai plane does not change our conclusions.

[7] For the calculation of ΔCFS in areas of increased seismicity after the Tohoku-oki earthquake, we used P-wave first-motion polarity focal mechanism (FM) data (the focal mechanisms are shown and described in theauxiliary material, Section 1) from the Hi-net [Okada et al., 2004] earthquake database. We select FM data from January 2003 to December 2011, by requiring that each selected event has a magnitude M ≥ 2.0, at least 15 P-wave polarities, the number of inconsistent polarities is less than 10% and the standard deviations associated with the P- and T-axes orientations are less than 10°. All FMs are visually inspected and in some cases the routinely determined focal mechanisms were reanalyzed. The main assumption is that the FMs in a certain region reflect the tectonic background and do not change significantly as a function of time. Thus, by resolving the ΔCFS due to the mainshock on FM nodal planes in the areas of interest, we can assess whether the seismic activity is promoted as a result of stress increase, or inhibited due to stress decrease. Because of the nodal plane ambiguity, we consider as receiver-fault the nodal plane of maximum stress increase (or least stress decrease). In the case of the March 15, 2011, M6.4 (Mw5.9) Shizuoka earthquake (Figure 1), the ΔCFS is calculated on the nodal plane that is consistent with its aftershock distribution [Takeda, 2011]. The inclusion of post-Tohoku-oki FMs ensures a good coverage with FM data in the study area. We use polarity-based FMs, in contrast to the F-net moment-tensor solutions [Okada et al., 2004] employed in other Coulomb stress change studies related to Tohoku-oki earthquake [e.g.,Toda et al., 2011], to increase the number of available FMs.

[8] For the estimation of stress changes on the Tokai megathrust plane, we use a detailed plate configuration model. The “eggplant” shaped source region of the anticipated Tokai earthquake (Figure 1) and the depths contours of the upper surface of the subducting Philippine Sea plate are based on the study of plate configuration and historical earthquake records, and used in official seismic risk evaluation by the Japanese government (for a detailed discussion, see Aoi et al. [2010]). The strike, dip and rake at each location on the plate boundary were determined using the geometry and relative motion of the tectonic plates [see Aoi et al., 2010, and references therein, for explanations].

3. Results and Discussion

[9] We show in Figure 2athe distribution of seismicity (JMA earthquake catalog, M ≥ 0.5, depth ≤ 20 km) around Izu Peninsula that occurred from one month before to one month after the 2011 Tohoku-oki earthquake. At several locations, delimited in the figure by rectangles, the seismicity after the mainshock (blue circles) increases compared to the one before the mainshock (red circles) (see alsoauxiliary material). The cumulative number curves in Figure 2bshow increased seismicity in response to the Tohoku-oki event. The earthquake activity in the rectangles B, C and D increases within minutes after the mainshock, while in the areas A and E the increase appears delayed by a few days (see alsoauxiliary material, Section 2). Note that Toda et al. [2011] report the immediate activation of seismicity in a larger region that includes the areas B, C and D of this study. The activation in area A began with the 15th March M6.4 Shizuoka earthquake followed by its own aftershocks. Note that the seismicity in area A (Figure 2b) consists mainly in the aftershocks of the M6.4 event and thus we cannot directly relate such activation to the M9.0 Tohoku-oki earthquake. However, it is remarkable that the M6.4 event is one of the three large (M ≥ 6.0) inland earthquakes that occurred within four days from the Tohoku-oki earthquake [e.g.,Toda et al., 2011]. Since on average there is one earthquake of M ≥ 6.0 annually in Japan [Toda, 2008], the occurrence of three such events within a few days from the Tohoku-oki earthquake is unlikely to be due to chance. We therefore consider the M6.4 Shizuoka earthquake as a delayed aftershock of the Tohoku-oki megathrust event. A rather clear increase of seismicity can be noticed in area E, starting from 19th March, when clustered seismicity (with a maximum magnitude of 3.3) starts occurring. We present in theauxiliary material, further results that support the delayed activation of seismicity in area E due to the Tohoku-oki earthquake.

Figure 2.

(a) Seismicity (depth ≤ 20 km, M ≥ 0.5) in the study region, within a time window of 30 days before (red circles) and 30 days after (blue circles) the Tohoku-oki earthquake. The A, B, C, D and E rectangles correspond to the areas where seismicity activation is observed after the Tohoku-oki earthquake. (b) Cumulative number of events in the five rectangular areas A, B, C, D and E of Figure 2a.

[10] The computed ΔCFS resolved on the FM nodal planes of crustal events in the study region (Figure 3a) is predominantly positive (94%) within all five rectangles A to E, and reaches values larger than 0.2 bar. The percentage of “receiver-faults” (nodal planes) characterized by a positive stress change is 91% and 98% for the events occurred before and after the Tohoku-oki earthquake, respectively. Since the main faulting style in the study region is strike-slip (auxiliary material), with predominantly NS striking faults, we have also considered for each of the strike-slip FMs in our dataset the ΔCFS on the nodal plane that is closest to a NS striking direction, instead of the nodal plane receiving the maximum stress increase (or least stress decrease). In this case as well, the majority of ΔCFS (91%) are positive, reaching values of up to ∼0.9 bar. Our results indicate that the tectonic structures in the five areas ofFigure 3aare “favorable” to static stress triggering by the Tohoku-oki earthquake. As discussed in previous studies [Parsons et al., 2008, and references therein], stress increases larger than 0.1 bar commonly raise regional seismicity and thus potentially bring a major fault to failure with delays ranging from seconds to decades. The occurrence of large local events is responsible for the addition of secondary aftershocks nearby.

Figure 3.

(a) ΔCFS due to the Tohoku-oki earthquake, resolved on the nodal planes of earthquakes that occurred from January, 2003 to December, 2011, within the A to E rectangles (Figure 2a). The color of each event is coded with its corresponding ΔCFS. The inset map shows an enlarged distribution of the ΔCFS color-coded events occurred in the area A. The green-contour circles in the main figure and inset represent events occurred after the Tohoku-oki earthquake. (b) ΔCFS on the Tokai plane, due to the 2011 Tohoku-oki earthquake. The “eggplant” shaped source region is delimited by bold solid line. Contours indicate the depths of the subduction interface of the Tokai seismic gap. Thick arrow indicates average slip direction on the presumed Tokai fault plane and the six rectangles outlined by thick grey lines show presumed Tokai asperities. Red and blue colors indicate relative stress increase and decrease, respectively.

[11] The earthquake catalog completeness is an important factor in assessing whether the delayed activation observed in areas A and E is real. From the occurrence time of Tohoku-oki earthquake until 15th March 2011 Shizuoka earthquake, there is no seismic activity in area A. On the other hand, in the neighboring area B, there are many small events (M ∼ 0 and above) recorded in the same time period. The completeness magnitude after one day from the mainshock is around 1.0 for area B (auxiliary material). Under the reasonable assumption that a similar level of catalog completeness should characterize both areas A and B (which are geographically closely spaced), we conclude that at least above some small magnitude threshold (M ∼ 1.0) the observed delay of activation onset is real. Such activation delays support the static stress triggering mechanism rather than the dynamic one, due to the passage of seismic waves [Belardinelli et al., 2003]. Note that the dynamic triggering has been proposed to play a role in the initiationof triggering at Hakone volcano (located in the area B of our study) following the Tohoku-oki earthquake [Yukutake et al., 2011]. The rate-and-state dependent friction law [Dieterich, 1994] is widely used to model the time-dependent aftershock decay following a stress-step and can explain the earthquake triggering delay. Another physical process that has been proposed to explain the time delay of earthquake activation following a large mainshock is the time-dependent adjustment of local pore pressure changes on and around the main fault by the main rupture [e.g.,Nur and Booker, 1972]. This theory considers the fluid migration and diffusion as the driving mechanism of pore pressure changes.

[12] Another line of evidence that substantiates the triggering by static stress changes comes from the Omori-type decay of seismicity (auxiliary material). Toda et al. [2011]reported the activation of seismicity in various inland regions all-over Japan following the Tohoku-oki earthquake and interpreted this observation by static stress triggering, using essentially F-net moment-tensor solutions of pre-Tohoku-oki seismicity. The results presented here, based on a richer and extended regional dataset of crustal earthquakes, demonstrate that the static stress transfer plays the main role in the activation of seismicity around Izu Peninsula and thus can be applied to evaluate the Tokai gap.

[13] The ‘receiver fault’ resolved ΔCFS is now switched to the plate interface. The ΔCFS distribution on the Tokai plane (Figure 3b) clearly shows that most of the Tokai fault area is under a stress increase due to the 2011 Tohoku-oki event. However, the increase of stress is relatively small: the largest value is of 0.1 bar (10 kPa) and most of the Tokai plane experiences stress increases of less than 0.07 bar. We also show inFigure 3b the presumed asperities on the Tokai plane, inferred from the study of background seismic activities and focal mechanisms within the subducting Philippine Sea slab by Matsumura et al. [2008]. For a detailed discussion about these inferred asperities, with relevance to static stress changes and their related uncertainties, we refer to Aoi et al. [2010]. Most of the asperity areas experienced stress changes of less than 0.08 bar, which is far less than the ΔCFS of 0.5 bar on the fault plane of the M6.4 Shizuoka earthquake, located 20 km east of the Tokai source. The examination of the JMA earthquake catalog did not reveal any significant change of seismicity on or in the nearby of the Tokai megathrust fault plane, after the occurrence of the Tohoku-oki earthquake. We have also estimated the long-term impact of the stress transfer from the Tohoku-oki earthquake on the Tokai plane (auxiliary material). Our results show that the probability gain for the next Tokai megathrust event due to the Tohoku-oki earthquake is less than 0.1%. Thus, our findings indicate that the stress state on the Tokai plane did not change significantly due to the Tohoku-oki earthquake.

[14] In order to test the source fault model dependency, we also considered three other variable slip models (auxiliary material, besides that of Suzuki et al. [2011], and then confirmed that our results are robust and insensitive to the slip details of the M9.0 earthquake source. The auxiliary materialincludes also additional discussion related to the ΔCFS resolved on FMs of pre- and post-Tohoku-oki crustal events.

[15] Several large aftershocks occurred soon after the Tohoku-oki earthquake. The Mw7.8 (M7.7) aftershock (Figure 1) is the largest and closest to the Tokai region, thus could be a potential source of stress increase on the Tokai fault plane. The aftershock occurred on March 11, about 30 min. after the mainshock. We have used its moment-tensor solution reported by NIED, F-net, and select the nodal plane that has a dip consistent with the dip of the plate interface as the fault plane of the earthquake. The size of the fault plane was estimated using empirical relations [Wells and Coppersmith, 1994] and we use a uniform slip distribution, tapered at the fault edges. Another potential source of static stress changes on the Tokai plane is represented by the post-seismic deformation taking place on the Tohoku-oki megathrust fault. The afterslip moment release from March 11, 2011, 18:00 to October 31, 2011, 18:00 is equivalent with that of an Mw8.6 earthquake [GSI, 2011]. We have used an afterslip model inverted from geodetic data [Ozawa et al., 2011; GSI, 2011] to estimate the corresponding stress change on the Tokai plane. Finally, the Mw5.9 Shizuoka earthquake is another potential source of stress increase on the Tokai plane, due to its proximity to the megathrust fault. The fault plane of the Shizuoka earthquake was determined using the CMT solution and the precise aftershock distribution reported by Takeda [2011]. We present in Figure 4athe cumulative ΔCFS on the Tokai plane due to all these (three) sources and the Tohoku-oki co-seismic slip. We first point out that the stress change contribution of the Mw7.8 aftershock to the ΔCFS pattern shown in Figure 4ais negligible, being less than about 0.01 bar (i.e., about one tenth of the stress change due to the mainshock). As one can notice in the figure, two areas in the deeper portion of the Tokai plane (depth below ∼15 km) have cumulative stress changes of 0.1 to 0.15 bar. The exceedance of the 0.1 bar “threshold” is the effect of gradual, small incremental stress changes caused by the afterslip taking place on the Tohoku-oki megathrust plane in a period of ∼7 months. Note that the Mw5.9 Shizuoka earthquake is too far from these areas to have any significant effect. The lack of any (observable) earthquake activation on or close to the Tokai plane confirms that such stress changes were too small to trigger seismicity. Moreover, as we show in the auxiliary material, the probability gain for the next Tokai event associated with such a stress change is very small (less than 0.2%). By comparing Figures 3b and 4a, one can notice a relatively small area (indicated by a dotted ellipse in Figure 4a) in the shallow, north-eastern part of Tokai plane, where the ΔCFS “switches” from negative to positive. The area is close to the fault plane of the Mw5.9 Shizuoka earthquake (Figure 4a). The maximum stress change due to the Mw5.9 earthquake alone in that area is of about 0.18 bar, which explains these positive ΔCFS values. While Shizuoka earthquake is a moderate event and did not occur extremely close to the Tokai plane to cause large positive stress changes on the plate interface (or on the presumed Tokai asperities), it still warns about the possibility of subsequent significant seismicity in the region (including earthquakes of M ∼ 6–7), triggered by the Tohoku-oki earthquake, that could impact significantly the stress state on the Tokai subduction plane. Such activation could occur not only at crustal depths around Izu Peninsula, but also within the subducting Philippine Sea plate. Indeed, as we show in theauxiliary materialthe stresses within the Philippine Sea plate due to the Tohoku-oki earthquake, resolved on characteristic intraslab fault planes, show an increase of up to 0.8 bar.Aoi et al. [2010], for example, document a relatively large stress increase on one of the presumed Tokai asperities, as a result of the 2009 Mw6.4 (M6.5) Suruga-bay earthquake (Figure 4a) that occurred just below the Tokai source region.

Figure 4.

(a) Cumulative ΔCFS on the Tokai plane, due to the 2011 M9.0 Tohoku-oki earthquake (March 11, 14:46), the M7.7 aftershock (March 11, 15:15), the M6.4 Shizuoka earthquake (March 15, 22:31) and the afterslip occurred on the Tohoku-oki megathrust plane in the time period March 11, 2011, 18:00 to October 31, 2011, 18:00. The dashed green contours indicate areas on the Tokai plane where the cumulative ΔCFS has values between 0.1 to 0.15 bars. The dotted green ellipse indicates an area of stress increase, mainly due to the Shizuoka earthquake. The modeled fault plane of the Shizuoka earthquake is shown as a thick red line, with the corresponding focal mechanism to the right. The large star is the epicenter of the 2009 M6.5 (Mw6.4) Suruga-bay earthquake. (b) Histogram comparison of ΔCFS distributions on the Tokai plane: the green and purple bars correspond to the stress changes due to the mainshock alone (Figure 3b) and to the cumulative stress change (Figure 4a), respectively.

4. Conclusions

[16] We performed an in-depth investigation of static stress and seismicity changes in a relatively large region around the Izu Peninsula, central Japan. We have identified five distinct crustal areas, located either on the Philippine Sea plate or the Eurasian plate, where the earthquake activity increased either immediately or with some delay after the M9.0 Tohoku-oki event. The ΔCFS (of maximum 1.0 bar) resolved on focal mechanisms of regional earthquakes that occurred before and after the megathrust event support the Coulomb triggering hypothesis. Further support for the triggering by static stresses is provided by the delayed activation of earthquake activity in two of the five areas of the study region and the Omori-type decay with time of seismicity triggered directly by the Tohoku-oki earthquake. The static stress changes on the Tokai megathrust plane, resolved using the detailed plate configuration, while predominantly positive, are about 10 times smaller (maximum values of 0.1 bar). The afterslip on the Tokoku-oki fault and the large aftershocks following the M9.0 event did not change significantly the stress state on the Tokai plane either. However, the potential of relatively large events triggered by the static stress increase in the crustal regions around Izu Peninsula or inside the subducting Philippine Sea plate may pose an indirect triggering hazard for the plate boundary seismicity.


[17] We are grateful to all our colleagues for their hard work of restoring and maintaining the NIED facilities after the damaging earthquake. We thank Takuya Nishimura for providing the post-seismic slip model used in this study and Fred Pollitz, Gavin Hayes and Shenji Wei for providing their co-seismic slip models of Tohoku-oki earthquake. We acknowledge the use of the JMA earthquake catalog. We are grateful to Sandy Steacy and two anonymous reviewers for their thoughtful and constructive comments that helped improve substantially the present study.

[18] The Editor thanks Sandy Steacy and an anonymous reviewer for assisting in the evaluation of this paper.