Geophysical Research Letters

Seismic evidence for reactivation of a buried hydrated fault in the Pacific slab by the 2011 M9.0 Tohoku earthquake

Authors


Abstract

[1] We employ seismic tomography to estimate detailed 3D seismic velocity structures in the focal area of an intraslab earthquake (M7.1), which occurred on April 7, 1 month after the 2011 Tohoku earthquake (M9.0) near its source area. The results show a low-velocity zone around the focal area of the M7.1 event, and that the aftershock activity is limited to the upper 15 km of the oceanic mantle. The lateral extent of the low-velocity zone is comparable to the distribution of aftershocks, suggesting a concentration of fluids in the aftershock area. The angle between the aftershock alignment and the dip of the slab surface is estimated to be ∼60°, which is consistent with the dip of an oceanward-dipping normal fault observed at the outer-trench slope. These observations suggest that the M7.1 intraslab event occurred as a result of reactivation of a buried hydrated fault that formed prior to subduction. The upper ∼15 km of the oceanic mantle may be locally hydrated by bending-related tensional faulting at the outer-trench slope.

1. Introduction

[2] A M9.0 megathrust earthquake, the 2011 off the Pacific Coast of Tohoku Earthquake, occurred on 11 March 2011 on the plate boundary of northeastern (NE) Japan. The rupture associated with this event propagated along the upper surface of the subducting Pacific plate, resulting in a source area of 500 km by 200 km (Figure 1). The maximum slip was estimated to be >30 m around the hypocenter [e.g., Iinuma et al., 2011], and an extremely large tsunami (up to 30 m high) was generated, causing severe damage along coastal areas on the Pacific side and resulting in >20,000 fatalities and missing persons.

Figure 1.

(a) Hypocenter distribution (colored circles) after the 2011 M9.0 earthquake. Colors are proportional to the depth of hypocenters. Large circles denote six M > 7 earthquakes. The focal mechanisms of these earthquakes are determined by F-net of the National Research Institute for Earth Science and Disaster Prevention. Slip areas of the M9.0 event are shown by purple contours with an interval of 5 m [Iinuma et al., 2011]. Active volcanoes are shown by white triangles. (b) Across-arc vertical cross-section of the hypocenter distribution along line A-A′ in Figure 1a. Earthquakes shown are those that occurred within <20 km from the line. Gray dots represent earthquakes that occurred before the M9.0 event.

[3] The M9 earthquake has activated seismicity in the Pacific plate as well as along its upper surface. Two thrust-type interplate earthquakes (M7.4 and M7.7) and an outer-rise normal-fault-type earthquake (M7.5) followed the mainshock, and other numerous aftershocks occurred around the focal area of the M9 event. On 7 April, an earthquake (M7.1) occurred in the Pacific slab at a depth of 66 km, located near the down-dip limit of the large interplate slip of the M9 event (Figure 1). GPS data indicate slips of 2.5 m during the M7.1 event and the source area of 30 km by 30 km [Ohta et al., 2011]. The M7.1 event was probably triggered by stress change resulting from the large fault slip by the M9 event.

[4] However, earthquakes cannot be easily triggered without an effective mechanism to weaken the fault strength, because the static stress change due to fault slip of the M9 event is estimated to have been on the order of 1 MPa at most [e.g., Toda et al., 2011], which is too small for a brittle fracture under the high lithostatic pressure (∼2 GPa) at the focal depth of the M7.1 event. One mechanism of reducing the fault strength is the existence of overpressurized fluids, which results in enhanced pore-fluid pressure and reduced effective normal stress.

[5] In the present study, we perform travel-time tomography to characterize heterogeneous seismic velocity structures around the focal area of the 2011 M7.1 intraslab event. Although this study uses earthquakes and stations distributed throughout a wide region of NE Japan to increase the number of rays propagating in the Pacific slab and hence 3D velocity models are determined for the entire study area, we focus on detailed 3D seismic velocity structures in the Pacific slab and discuss the occurrence of the 2011 M7.1 event in terms of dehydration embrittlement hypothesis.

2. Data, Methods and Resolution Tests

[6] We applied the double-difference tomography method [Zhang and Thurber, 2003] to arrival-time data obtained at a nation-wide seismograph network in Japan. From the unified catalogue of the Japan Meteorological Agency for the period from January 2001 to February 2011, we selected 8,911 earthquakes (M > 1.0) (Figure 2) that satisfy both of the following criteria: (1) earthquakes with Hdep > Dmin, where Hdep is the depth of an earthquake and Dmin is the epicentral distance to the nearest station with P- and S-wave arrival pickings; and (2) earthquakes with 25 or more arrival-time data. These criteria keep earthquakes whose focal depth can be well constrained. Note that we only considered aftershocks with M > 4 for the 2003 Miyagi-oki intraslab earthquake (M7.1), because this event was followed by a large number of aftershocks [Okada and Hasegawa, 2003]. The arrival-time data from these earthquakes, recorded at 188 stations, comprised 247,504 P waves and 196,057 S waves. The distance between earthquake pairs was limited to 10 km, yielding 1,608,230 P-wave and 1,114,068 S-wave differential travel-time data.

Figure 2.

(a) Map and (b) cross-sectional views of the grid distribution (crosses) adopted in the inversion. Reverse green triangles and gray dots represent seismic stations and earthquakes used in this study, respectively. Red and blue stars denote hypocenters of the 2011 and 2003 intraslab earthquakes, respectively. An orange rectangle in Figure 2a indicates an area shown in Figure 3e. The grid distribution along line A-A′ is shown in Figure 2b.

[7] Grid intervals were set at 10–20 km in the along-arc direction, 5–10 km perpendicular to the arc, and 5–10 km in the vertical direction (Figure 2). Smaller grid intervals were adopted for the focal area of the 2011 M7.1 event. A 1-D velocity structure (JMA2001 [Ueno et al., 2002]) was adopted as the initial velocity model, for which P- and S-wave velocities within the Pacific slab were assigned to be 5% faster than those in the mantle based on the plate model proposed by Nakajima et al. [2009a]. The final results were obtained after eight iterations, which reduced the travel-time residual from 0.17 s to 0.11 s for P waves, and from 0.33 s to 0.19 s for S waves.

[8] We performed three resolution tests to assess the reliability of the results (see the auxiliary material for details). The results of the tests show that our data set can resolve heterogeneous structures in the subducting Pacific slab with a spatial resolution of ∼10 km. We also refer to values of the derivative weighted sum (DWS) [Thurber and Eberhart-Phillips, 1999], and areas with DWS > 500 are discussed herein, within which the checkerboard patterns around the Pacific slab are well recovered and hence the obtained velocity structures are reliable.

3. Tomographic Results

[9] This study focuses on S-wave velocity structures to characterize distinct features in the subducting slab, because S-wave velocities are more sensitive than P waves to the existence of fluids [e.g., Nakajima et al., 2009b] and are not affected by azimuthal anisotropy [Reynard et al., 2010]. P-wave velocity structures are shown in the auxiliary material. To enable a detailed discussion, we relocated the aftershocks of the 2011 M7.1 event (n = 121, up to 29 May 2011) using the 3D seismic velocity structures obtained in this study.

[10] Aftershocks of the 2011 M7.1 event are distributed in an area of 30 × 30 km (Figure 3e), and the hypocenter of the mainshock is located at ∼10 km below the oceanic Moho in the southernmost part of the aftershock area. Aftershocks in the southern part of the aftershock area are distributed on an east-dipping plane in a depth range of 55–68 km (line A-A′ in Figure 3). The angle between this plane and the dip of the slab surface is estimated to be ∼60°. In the northern part of the aftershock area, the distribution of aftershocks does not define a dipping plane; instead, they appear to be scattered below the oceanic Moho. The distribution of aftershocks suggests that the rupture associated with this event did not propagate into the oceanic crust beyond the Moho. These conclusions are also supported by the fault plane solution of the M7.1 event inferred from geodetic data [Ohta et al., 2011].

Figure 3.

(a) Across-arc vertical cross-sections of S-wave velocity structures along lines A-A′, B-B′, and C-C′ shown in Figure 3e. A white star represents the mainshock of the 2011 M7.1 event (lines A-A′) and of the 2003 M7.1 event (line C-C′). Repeating earthquakes (gray stars) are identified by Uchida et al. [2006]. Black solid and dashed curves denote the upper surface of the Pacific slab and the oceanic Moho, respectively. The thickness of the oceanic crust is assumed to be 7 km. A white rectangle in each panel in Figure 3a shows the area enlarged in Figures 3b and 3c. (b) S-wave velocity and (c) Vp/Vs structures within the area outlined by the white rectangle in Figure 3a. The straight black line in A-A′ and B-B′ shows the fault plane of the 2011 M7.1 event, as inferred from the aftershock distribution and focal mechanism solution. (d) Schematic showing fault slips of the M9.0 event (purple arrows) and the M7.1 intraslab event (white arrows) (modified from Ohta et al., 2011). (e) Map showing the aftershock distribution of the 2011 and 2003 intraslab events and the locations of three lines, A-A′, B-B′, and C-C′.

[11] The aftershock distribution of the M7.1 event appears to correlate with the lateral extent of a low-velocity zone in the oceanic mantle. The hypocenter of the mainshock is located within a zone of low-velocity oceanic mantle, and the distribution of aftershocks coincides with a low-velocity area (line A-A′ in Figure 3b). A high-Vp/Vs value of >1.9 is observed around the hypocenter of the mainshock, but values are moderate (1.7–1.8) throughout the distribution of aftershocks. The scattered aftershocks along line B-B′ occur in a relatively wide low-velocity zone observed below the oceanic Moho. The along-strike extent of aftershocks is ∼30 km, which is coincident with the lateral extent of a low-velocity zone in the upper 10 km of the oceanic mantle (Figure 4). It is noted that the rupture of the M7.1 event did not propagate further to the north (along-strike distance of >60 km in Figure 4), where seismic velocity is relatively low in the upper ∼20 km of the oceanic mantle. Because we did not use arrival-time data from the aftershocks of the 2011 M7.1 event in the tomographic inversions, the low-velocity anomaly in the oceanic mantle must have existed prior to the 2011 M7.1 event.

Figure 4.

(a) Along-fault cross-section of S-wave velocity structures along an eastward-dipping plane shown in Figure 4b. The dip and strike of the plane are the same as those of the fault plane of the 2011 M7.1 event. The relocated aftershocks of the 2011 event are projected onto the eastward-dipping plane (Figure 3). Other symbols are the same as in Figure 3. (b) Map and cross sectional views of the location of the eastward-dipping plane along which the seismic velocities are calculated.

4. Discussion and Conclusions

[12] Large normal-fault-type earthquakes in the shallow outer-trench slope region are relatively rare, but examples include the events at Sanriku (Mw8.4) in 1933 [e.g., Kanamori, 1971], Sumbawa (Mw8.3) in 1977 [e.g., Lynnes and Lay, 1988], Kuril (Mw8.1) in 2007 [e.g., Ammon et al., 2008], and Samoa-Tonga (Mw8.1) in 2009 [e.g., Lay et al., 2010]. The ruptures of these earthquakes extended throughout the oceanic crust and down to a depth of at least 30 km from the upper surface of the oceanic plate. Smaller outer-trench earthquakes (M∼7) ruptured the upper ∼20 km of the oceanic plate [e.g., Fromm et al., 2006; Hino et al., 2009]. In addition, bending-related faulting at the outer-trench slope tends to extend into the oceanic plate for at least 20 km [e.g., Ranero et al., 2003]. Such tensional faulting at the outer-trench slope may produce pervasive fractures and facilitate the hydration of the oceanic mantle before subduction [e.g., Faccenda et al., 2009], probably resulting in serpentinization along tensional faults. Bending-related faults dip at angles of 45–70° [e.g., Ranero et al., 2003].

[13] As an old plate subducts, hydrous minerals in the oceanic mantle are expected to break down. If dehydration occurs along a pre-existing hydrated fault and if the released fluids are trapped, pore-fluid pressure may become extremely high, approaching the lithostatic pressure. Such overpressurized fluids act to reduce the effective normal stress and weaken the strength of pre-existing faults, resulting in brittle failure. Consequently, fluids released by dehydration contribute to the development of a low-velocity anomaly around the faults. The aftershock alignment of the M7.1 event (at an angle of 60° with respect to the slab surface) can be explained by the reactivation of an oceanward-dipping fault that formed at the outer-trench slope. This discussion gives rise to the possibility that the 2011 M7.1 event occurred along a pre-existing weak fault, as a result of an increase in shear stress due to downward slip along the plate interface during the 2011 M9.0 event (Figure 3d).

[14] We consider overpressurized fluids in the oceanic mantle to be responsible for the observed low-velocity zone, but the velocity reduction may be related to serpentinization of the mantle peridotite. Indeed, Vp/Vs values in the focal area (1.7–1.9) are comparable to those of isotropic serpentine (1.70–1.86) [Bezacier et al., 2010], and pressure and temperature conditions in the focal area are expected to be ∼2 GPa and 400–500°C, respectively [e.g., Yamasaki and Seno, 2003]. Therefore, dehydration embrittlement of serpentine and brucite [e.g., Omori et al., 2004] may have directly facilitated the M7.l intraslab event. Although we cannot discriminate between the existence of overpressurized fluids and serpentine minerals based solely on the observed seismic velocities, the present results indicate that, in the region of seismicity, the oceanic mantle is locally hydrated to the upper ∼15 km. Large intraslab earthquakes may occur at intermediate depths as a consequence of dehydration of buried hydrated faults that formed in the outer-trench slope region.

[15] Mishra and Zhao [2004] argued that the mainshock and aftershocks of the 2003 M7.1 Miyagi-oki intraslab earthquake occurred in a P-wave low-velocity anomaly. The present study further reveals the existence of a distinct S-wave low-velocity zone at the hypocenter of the 2003 event (line C-C′ in Figure 3). For another large intraslab earthquake that occurred in the Pacific slab, the 1993 Kushiro-oki earthquake (M7.8), Nakajima et al. [2009c] reported a horizontal low-velocity zone in the oceanic mantle, spatially coincident with the aftershock distribution. For both the 1993 and 2003 earthquakes, the angle between the plane of aligned aftershocks and the dip of the slab surface is ∼60° [Ide and Takeo, 1996; Okada and Hasegawa, 2003]. The seismic velocity structures for the three recent intraslab earthquakes, in 1993 (M7.8), 2003 (M7.1), and 2011 (M7.1), indicate that the oceanic mantle is not uniformly hydrated, but that there are discrete low-velocity and high-Vp/Vs areas of pronounced hydration and dehydration where large intraslab earthquakes may occur.

Acknowledgments

[16] We used arrival-time data from the unified catalogue of the Japan Meteorological Agency, and focal mechanism data from the F-net seismic moment tensor catalogue of the National Research Institute for Earth Science and Disaster Prevention. Constructive and careful reviews by two anonymous reviewers improved the paper. This work was supported in part by the Ministry of Education, Culture, Sports, Science and Technology of Japan, under its Observation and Research Program for Prediction of Earthquakes and Volcanic Eruptions, and by the Global COE Program, Global Education and Research Center for Earth and Planetary Dynamics, Tohoku University.

[17] The Editor thanks the two anonymous reviewers for their assistance in evaluating this paper.

Ancillary