1

Accepted for publication in Geophys. Res. Lett., Apr. 21, 2010.

Copyright 2010 American Geophysical Union.

Further reproduction or electronic distribution is not permitted.

Deep roots of upper plate faults and earthquake generation illuminated by volcanism Ryosuke Ando1 Satoshi Okuyama2* 1Active Fault and Earthquake Research Center, Geological Survey of Japan, National Research Institute

of Advanced Industrial Science and Technology, Tsukuba, Japan 305-8567. email: ryo-ando@aist.go.jp

2Institute of Geoscience and Geoinformation, Geological Survey of Japan, National Research Institute of

Advanced Industrial Science and Technology, Tsukuba, Japan 305-8567.

*Now at Hokkaido University. Abstract For many years, the generation mechanism of disastrous intraplate earthquakes in the upper plate of

a subduction zone have remained unclear largely because the roots of upper plate intraplate faults

have particularly quiet inter-seismic nature and limited visibility. Here we propose that the 2008 Mw

6.9 Iwate-Miyagi Nairiku, Japan, Earthquake, occurred on a dipping fault in a volcanic region,

contains previously-unreported co-seismic evidence of a ductile shear zone (DSZ) present on the

down-dip fault extension beneath the seismogenic layer, and DSZ plays a dominant role in the

seismogenesis. We found the evidence in the spatial pattern of the co-seismic ground displacement,

which was well captured by synthetic aperture radar, reflecting geothermal anomalies there. A

dynamic forward model including the inter-seismic deformation of DSZ naturally reproduced this

displacement pattern without kinematic inversions and consistently explained other independent

observations. This finding should make breakthroughs in observational and theoretical studies of

earthquake generation.

Introduction In the last few decades, establishing a depth-dependent mechanical model of crustal faults has

been a central problem of earthquake studies [Sibson, 1977; Tse and Rice, 1986; Scholz, 1988;

Burgmann and Dresen, 2008]. Geologic field studies of exhumed faults have found bands of

mylonite, rock sheared in a ductile manner under high pressure and temperature [Sibson, 1977], and

laboratory experiments have explored the onset of rock plasticity at around 300–350 °C with

2

strain-softening behavior [Tullis et al., 1990] further intensified by hydrolytic effects [Kronenberg,

1994]. Those findings imply the existence of localized ductile shear zones (DSZs) on the deep

extension of active faults in the lower crust beneath the seismogenic layer (SGL) delimited by the

geothermal gradient [Sibson, 1986; Scholz, 1988]. Geodetic observations and kinematic analyses

targeting plate boundary transform faults [Bourne et al., 1998; Savage et al., 1999] have captured the

relevant inter-seismic ground deformation. However, it is intrinsically difficult to resolve the slip at

depth because of their vertical dip, hence the debates on this model is currently still ongoing [Fialko,

2009]. Furthermore, the validity of the DSZ model remains an open question for intraplate faults in

the upper plate of a subduction zone (also called inland faults). These upper plate faults are subjected

to a particular tectonic condition remote from the loading source, a plate interface, so that they are

immaturer and far less active though ubiquitous on shore, causing major seismic hazards directly to

nearby cities. Thus the model verification has unique importance for this particular case although it

should contain broader implication.

To this end, we targeted the 2008 Mw 6.9 Iwate-Miyagi Nairiku earthquake, eastern Japan, which

occurred in the immediate vicinity of the inland volcanic front associated with the subducting Pacific

plate. The earthquake occurred at shallow depth on a dipping fault, allowing us to infer precisely the

slip distribution on the fault by analysis of synthetic aperture radar (SAR) images, which can resolve

surface displacement over a large area with very high spatial resolution. Moreover, geothermal

including hydrologic anomalies existing in the focal area [Tanaka et al., 2004; Hasegawa et al.,

2005; Yoshida et al., 2005; Okada et al., 2010] make it possible to investigate deep below the surface,

because considerable variation in the base of the SGL is surprisingly visible in available co-seismic

observations.

Observational results Figure 1 shows the observed small seismic events, which clarify the relationship between the

extent of the 2008 main-shock focal region and the geothermal background. The aftershock

distribution (Figure 1a) suggests that the main-shock rupture expanded following a NNE-SSW trend,

but its expansion was obviously limited by areas of high geothermal anomaly, indicated by the

clusters of volcanic low-frequency earthquakes (LFEs), presumably surrounding magmatic bodies

[Hasegawa et al., 1991], at both the northern (near Mt. Yakeishi) and southern (near Onikobe

caldera) ends of the rupture area, as well as just inside its western margin (near Mt. Kurikoma). The

LFEs occurred underneath the SGL (Figure 1b), whose basal depth, inferred from the aftershocks

and regular shallow background seismicity, is notably shallow above the three LFE clusters, almost

certainly because of the higher thermal gradients there (Figure 1c). There is direct evidence of a

geothermal gradient as high as 200 °C/km near these volcanoes and as low as 30 °C/km away from

them [Tanaka et al., 2004]. Likewise, low seismic velocity anomalies [Okada et al., 2010] provides

3

indirect evidence for magmatic bodies. The moment tensor solution together with the identified

surface rupture locations [Toda et al., 2010] and seismic inversions [Suzuki et al., 2010] suggests that

the source was a predominantly westward dipping reverse fault. Supposing a primarily west-dipping

fault, then the geometry of the western verge of the aftershock area (Figure 1a) corresponds well to

the shape of the bottom of the SGL (Figure 1b).

To obtain the detailed 3-dimensional ground-surface displacement pattern associated with the

main-shock, we performed a SAR pixel offset analysis [Tobita et al., 2001] using images taken by

PALSAR (Japanese spaceborne SAR). Figure 2 shows the resulting displacement map. In particular,

we found this pattern to have three characteristics (illustrated in the figure inset with the

corresponding numbers): (1) a crescent-shaped uplift area that overlaps the aftershock area (Figure

1a), and a continuous zone of peak uplift that approximately follows its western margin over tens of

kilometers, suggesting larger slip on the deeper portion of the west-dipping fault; (2) eastward

motion around Mt. Kurikoma with relatively large subsidence, changing to uplift towards the east,

but with no sudden changes in the horizontal motion; and (3) obvious contraction in the horizontal

components along the eastern verge of the uplift area, but not along the western verge, supporting

the fault model of a predominantly west-dipping reverse fault setting. These detailed patterns

captured by SAR are significant signals that allow us to verify the physics-based forward model

described next.

The model To obtain a physics-based explanation for the various observations described above, we carried

out forward modeling of the whole earthquake generation process, from slow tectonic loading to

dynamic earthquake rupturing on the fault, which will be an unprecedented application. This

approach provides much richer insights into the physical background than kinematic inversions do.

For simplicity, to identify the essential features, we assumed a simple dipping planar fault with

depth-dependent fault properties embedded in a 25-km-thick lithospheric elastic plate subjected to

remote stressing at a constant, slow rate. The thickness of the DSZ is neglected, as is usual in such

models [Tse and Rice, 1986]. We employed the efficient boundary integral equation method, which

enables full dynamic numerical modeling [Tada, 2006; Ando et al., 2007; Ando and Yamashita,

2007] with a numerically included ground free surface. In our model, the fault obeys a simplified but

essential constitutive law that depends on the slip and slip velocity and takes into account common

rock properties such as apparent slip-weakening friction at high velocities [Wibberley et al., 2008],

effective healing at low velocities [Dieterich, 1979] and plastic flow at high temperatures [Tullis et

al., 1990]. Figure 3a shows the assumed parameter distribution on the fault, differentiating the brittle

and ductile regions (the strength parameters are the same within each layer). The upper layer is

stronger but exhibits slip- and velocity-weakening behavior, which induce instability, whereas the

4

lower layer is weaker [Burgmann and Dresen, 2008] but held at a constant state, leading to stability.

It is important to note that the geometry and depth of the brittle–ductile transition are approximately

constrained only by the micro-seismicity distribution (shown in Figure 1b), not by the observed

ground-surface displacement (Figure 2).

Simulation results Figures 3b–h illustrate how the fault motion evolves in the simulation, from a long, quiet

inter-seismic period to a dramatic co-seismic period. During the inter-seismic period (Figure 3b and

c), shear stress continuously accumulates in the locked upper layer, especially along its base (see

right panels), whereas the lower layer stably slides, releasing the remote load (see left panels). In

general, the stress in the locked upper layer σupper is as follows,

ξξξσσ dSGlower

remoteupper )();()()( xxx ∫Γ+=, (1)

where the first and second terms on the right-hand side denote the stress at location x, respectively

representing the contributions of the remote load and of the slip S, represented by spatial convolution

through ξ with Green’s function G over the fault plane in the lower layer Γlower. Although the first

term is homogeneous everywhere on the fault, the second term introduces heterogeneity, showing a

concentration of stress along the base of the upper layer. This stress heterogeneity determines the

subsequent co-seismic slip pattern.

When enough stress has accumulated (Figure 3c), slip accelerates slightly at the deepest part of

the upper layer, where the stress is at maximum owing to the convex geometry of the layer. Note that

the condition for this rupture nucleation is met by the increment of the second term of Equation (1)

while the value of the first term remains lower than the residual strength of the upper layer.

In the co-seismic period (Figures 3d–h), the rupture starts to expand dynamically with seismic

wave radiation. Initially, the rupture propagates bilaterally, but soon it becomes unilateral, towards

the right (south). The dominant unilateral propagation is caused by the asymmetrical geometry of the

SGL base, which is longer to the right. Finally, the rupture is arrested and the co-seismic slip

becomes V-shaped (Figure 3g) reflecting the initial high stress area (Figure 3c) and the area of

negative stress drop around the ruptured area (Figure 3h). The simulation explains the observed

main-shock hypocenter location near where the base of the SGL is deepest (Figure 1b). Unilateral

southward propagation was also observed by seismic inversions [Suzuki et al., 2010]. Moreover, the

observed surface offsets were larger near the northern and southern ends of the fault than in its

middle section (Figures 1 and 2), consistent with the V-shaped slip.

Figure 4 shows the simulated co-seismic ground-surface displacement, which reproduces well the

observed pattern overall (Figure 2). This result is remarkable although the simulated rupture

overshoots somewhat, considering that we performed forward modeling without inversions to

5

constrain the source fault slip. Three aforementioned characteristics are clearly seen in the result.

The first two are especially important: (1) the crescent shape is indeed a manifestation of the

V-shaped fault slip distribution, and (2) the relatively larger motion around Mt. Kurikoma is seen to

result from the particularly thin SGL where the fault is cut by the volcano. Given these similarities, it

is clear that the shape of the western margin of the crescent is defined by the basal geometry of SGL.

They are evidence of the DSZ being manifested by the thinning of the SGL near volcanoes.

One might wonder whether the crescent shape is really evidence for the presence of a DSZ. If one

supposes, however, that the thinning of the SGL occurred without a DSZ (SGL occupies lithosphere),

then the stress would be concentrated only near the volcanoes, which would make it impossible to be

an equally large slip between the volcanoes, along the deep portion of the fault. Therefore, it would

be difficult to account for the crescent shape of the surface deformation without a DSZ. Although not

all detailed features including rather abrupt vertical displacement change near Mt. Kurikoma were

recovered by our simple model, its overall ability to explain the observations indicates the presence

of a DSZ. Consideration of secondary effects such as subsidiary faulting and any heterogeneity will

necessarily improve the capability of the model, with additional data required to evaluate these

aspects.

Conclusions Several geometric characteristics of the detailed ground displacement pattern constitute promising

co-seismic evidence for the presence of a DSZ. This evidence was verified by physics-based forward

modeling constrained by seismicity distributions reflecting existing geothermal anomalies.

Observations of the main-shock hypocenter, rupture directivity, lateral extent of the rupture area and

surface rupture profiles are all reasonably explained by our simulation. This simulation revealed the

role of DSZs through the inter-seismic fault loading process. The model verification in our study was

made possible by SAR data and could not be achieved with the sparser GPS or seismic network data

available at present. We anticipate that subsequent studies will identify similar clue in various active

faults by using volcanism owing to the clearer visibility due to existing anomalies; e.g., the

Atostugawa fault, central Japan, is well known as delimited by two volcanoes with the correlated

SGL depth variation [Ito et al., 2007]. Our methods and finding will lead to deeper understanding of

the structure and mechanics of crustal faults beyond upper plate faults.

References Ando, R., N. Kame, and T. Yamashita (2007), An efficient boundary integral equation method applicable

to the analysis of non-planar fault dynamics, Earth Planets Space, 59, 363-373.

Ando, R., and T. Yamashita (2007), Effects of mesoscopic-scale fault structure on dynamic earthquake

ruptures: Dynamic formation of geometrical complexity of earthquake faults, J. Geophys. Res.,

6

112, doi:10.1029/2006JB004612.

Bourne, S. J., T. Arnadottir, J. Beavan, D. J. Darby, P. C. England, B. Parsons, R. I. Walcott, and P. R.

Wood (1998), Crustal deformation of the Marlborough fault zone in the South Island of New

Zealand: Geodetic constraints over the interval 1982-1994, J. Geophys. Res., 103, 30147-30165.

Burgmann, R., and G. Dresen (2008), Rheology of the lower crust and upper mantle: Evidence from rock

mechanics, geodesy, and field observations, Annu. Rev. Earth Planet. Sci., 36, 531-567.

Dieterich, J. H. (1979), Modeling of Rock Friction .1. Experimental Results and Constitutive Equations, J.

of Geophys. Res., 84, 2161-2168.

Fialko, Y. (2009), Self-consistent Models of Postseismic and Interseismic Deformation due to Mature

Strike-slip Faults, Eos Trans. AGU, Fall Meet. Suppl., Abstract T12C-01.

Hasegawa, A., J. Nakajima, N. Umino, and S. Miura (2005), Deep structure of the northeastern Japan arc

and its implications for crustal deformation and shallow seismic activity, Tectonophys., 403,

59-75.

Hasegawa, A., D. P. Zhao, S. Hori, A. Yamamoto, and S. Horiuchi (1991), Deep-Structure of the

Northeastern Japan Arc and Its Relationship to Seismic and Volcanic Activity, Nature, 352,

683-689.

Ito, K., T. Ueno, H. Wada, and K. Matsumura (2007), Crustal structure from seismic surveys and

seismicity in the northern Chubu district with special reference to the Atotsugawa fault area, in

Geodynamics of Atotsugawa Fault system, edited by M. Ando, pp. 65-78, Terrapub, Tokyo.

Kronenberg, A. K. (1994), Hydrogen speciation and chemical weakening of quartz in Silica; physical

behavior, geochemistry and materials applications, edited by P. J. Heaney, C. T. Prewitt and G. V.

Gibbd, pp. 123-176 Mineral. Soc. of Amer., Washington, DC.

Okada, T., N. Umino, and A. Hasegawa (2010), Deep structure of the Ou mountain range strain

concentration zone and the focal area of the 2008 Iwate-Miyagi Nairiku earthquake, NE

Japan—seismogenesis related with magma and crustal fluid, Earth, Planet and Space, 62,

347-352.

Savage, J. C., J. L. Svarc, and W. H. Prescott (1999), Geodetic estimates of fault slip rates in the San

Francisco Bay area, J. Geophys. Res., 104, 4995-5002.

Scholz, C. H. (1988), The Brittle-Plastic Transition and the Depth of Seismic Faulting, Geol. Rundsch., 77,

319-328.

Sibson, R. H. (1977), Fault rocks and fault mechanisms, J. Geol. Soc. Lond., 133, 191-213.

Sibson, R. H. (1986), Earthquakes and Rock Deformation in Crustal Fault Zones, Annu. Rev. Earth Planet.

Sci., 14, 149-175.

Suzuki, W., S. Aoi, and H. Sekiguchi (2010), Rupture Process of the 2008 Iwate-Miyagi Nairiku, Japan,

Earthquake Derived from Near-Source Strong-Motion Records, Bull. Seismol. Soc. Amer., 100,

256-266.

7

Tada, T. (2006), Stress Green's functions for a constant slip rate on a triangular fault, Geophys. J. Int., 164,

653-669.

Tanaka, A., Y. Yano, M. Sasada, and M. Yamano (2004), Geothermal Gradient and Heat Flow Data in

and around Japan, P-5 ed., Geological Survey of Japan, AIST, Tsukuba.

Tobita, M., M. Murakami, H. Nakagawa, H. Yarai, S. Fujiwara, and P. A. Rosen (2001), 3-D surface

deformation of the 2000 Usu eruption measured by matching of SAR images, Geophys. Res.

Lett., 28, 4291-4294.

Toda, S., T. Maruyama, Y. Masayuki, Y. Awata, T. Yoshioka, R. Ando, and H. Kaneda (2010), Surface

Rupture Associated with the 2008 Iwate-Miyagi Nairiku, Japan, Earthquake and its Implications

to Evaluation of Active Faults, Zisin, 62, 153-178 (Japanese text with English abstract and

captions).

Tse, S. T., and J. R. Rice (1986), Crustal Earthquake Instability in Relation to the Depth Variation of

Frictional Slip Properties, J. Geophys. Res., 91, 9452-9472.

Tullis, J., L. Dell'Angelo, and R. A. Yund (1990), Ductile shear zones from brittle precursors in

feldspathic rocks; the possible role of dynamic recrystallization, in The Brittle–Ductile

Transition in Rocks, edited by A. Duba, W. Durham, J. Handin and H. Wang, pp. 67-82, Amer.

Geophys. Union, Washington, DC.

Wibberley, C. A. J., G. Yielding, and G. Di Toro (2008), Recent advances in the understanding of fault

zone internal structure: a review, in The Internal Structure of Fault Zones: Implications for

Mechanical and Fluid-Flow Properties, edited by Wibberley, W. Kurz, J. Imber, R. E.

Holdsworth and C. Collettini, pp. 5-33, The Geol. Soc. London.

Yoshida, T., J. Nakajima, A. Hasegawa, H. Sato, Y. Nagahashi, J. Kimura, A. Tanaka, O. D. A. Prima, and

K. Ohguchi (2005), Evolution of Late Cenozoic Magmatism in the NE Honshu Arc and Its

Relation to the CrustoMantle Structures, Quatern. Res., 44, 195-216.

Acknowledgements

We thank Teruo Yamashita and Christopher Scholz for helpful comments. This work was supported

by a Grant-in-Aid for Scientific Research, MEXT, Japan. PALSAR level 1.0 data were provided by

ERSDAC, Japan. PALSAR data are owned by METI and JAXA. The unified hypocenter catalogue

and focal mechanisms are maintained by JMA in cooperation with MEXT.

8

Figure 1. Seismicity pattern and geothermal conditions. a, Map view of seismic event along with

the locations of active Quaternary volcanoes and the identified discontinuous surface ruptures. The

light blue rectangle approximately shows the fault plane of the main-shock; its top side is indicated

by the solid line. The beach ball shows the focal mechanism. b, Along-strike section of seismic

events within the gray rectangular region in (a). c, Schematic illustration of the geothermal

conditions inferred from (b); LFEs are indicated by dark red circles.

9

Figure 2. Detailed pattern of co-seismic ground-surface displacement and geothermal

conditions. The color contours and vectors denote the spatial distribution of surface displacement

obtained by SAR pixel offset analysis. The other symbols are as in Figure 1. The inset shows the

locations of the three characteristic features discussed in the text.

10

Figure 3. Physics-based simulation result for the inter-seismic and co-seismic sequence. a,

Perspective view showing spatial distribution of the assumed frictional properties of the fault plane,

comprising a brittle upper layer (blue) and a ductile lower layer (orange). b, c, (Left) slip velocity V

normalized to the steady slip velocity in the lower layer Vo and (right) the shear stress on the fault. b,

The inter-seismic stable phase showing stable sliding in the lower layer and the associated

accumulation of shear stress along the base of SGL. c, The quasi-static nucleation phase showing the

gradual acceleration of slip from the deepest point of the SGL. d–f, (Left) slip velocity and (right)

vertical displacement of the ground surface above the fault during the co-seismic period. d, Dynamic

rupture acceleration phase, during which seismic waves do not reach the ground surface. e, Bilateral

propagation phase and the associated ground motion. f, Unilateral propagation phase. g, Final

co-seismic slip showing the characteristic V-shaped distribution. h, Co-seismic stress drop. Note that

an area of negative stress drop (gray) is seen along the margin of the ruptured area (surrounding

areas not ruptured are shown in black).

11

Figure 4. Simulation result for co-seismic ground surface displacements. The horizontal and

vertical axes denote the along-strike distance and the distance perpendicular to the strike,

respectively, as in Figure 3d. The color contours and vectors denote the spatial distribution of

co-seismic surface displacement. The north arrow corresponds to the north direction in Figure 2.

12

Introduction

This auxiliary material includes the illustration of the model configuration called Figure A1 and

information for the PALSAR data set used for SAR pixel offset analysis called Table A1. The

caption to Figure A1 would read: Imposition of a free surface by placement of a numerically

included virtual crack over the source fault.

Figure A1. Imposition of a free surface by placement of a numerically included virtual crack over

the source fault.

Table A1. PALSAR data used for the pixel offset analysis Obs.date of image pairs

Path / Orbit Off-Nadir angle (º)

19/06/2006 24/06/2008 053 (Descending) 41.5

06/07/2006 11/07/2008 054 (Descending) 41.5

29/08/2007 16/07/2008 057 (Descending) 34.3

02/08/2006 22/06/2008 061 (Descending) 21.5

21/06/2007 23/06/2008 402 (Ascending) 34.3

03/02/2007 08/02/2009 402 (Ascending) 34.3