Report

Stress orientations in subduction zones and the strength of subduction megathrust faults

See allHide authors and affiliations

Science  11 Sep 2015:
Vol. 349, Issue 6253, pp. 1213-1216
DOI: 10.1126/science.aac5625

Faults well oriented for failure

A deck of cards will remain motionless while pressed on from above, but easily separates when sheared. Similarly, the ease by which geological faults rupture depends on the geometry of the fault relative to the size and direction of stress. Hardebeck finds that faults are well oriented for failure in subduction zones worldwide, suggesting a low-stress environment (see the Perspective by Bürgmann). Subduction zone faults unleash powerful earthquakes. These estimates on the state of stress constrain potential generation mechanisms of destructive subduction zone earthquakes.

Science, this issue p. 1213; see also p. 1162

Abstract

Subduction zone megathrust faults produce most of the world’s largest earthquakes. Although the physical properties of these faults are difficult to observe directly, their frictional strength can be estimated indirectly by constraining the orientations of the stresses that act on them. A global investigation of stress orientations in subduction zones finds that the maximum compressive stress axis plunges systematically trenchward, consistently making an angle of 45° to 60° with respect to the subduction megathrust fault. These angles indicate that the megathrust fault is not substantially weaker than its surroundings. Together with several other lines of evidence, this implies that subduction zone megathrusts are weak faults in a low-stress environment. The deforming outer accretionary wedge may decouple the stress state along the megathrust from the constraints of the free surface.

Subduction zones pose a considerable hazard as the main source of great earthquakes (magnitude ≥8). Relative to other fault types, subduction zone megathrust faults may also have unique physical properties, due to the sediments and fluids that are entrained by the subducting oceanic plate (1). The presence of fluids, particularly at high fluid pressures, can weaken faults substantially (2). Direct evidence suggests that subduction zone megathrust faults slip with low frictional resistance near Earth’s surface (3). The strength of these faults at seismogenic depths is less clear, as they can only be studied indirectly.

The orientation of a fault with respect to the stress field can be an indicator of the fault’s strength. Faults optimally oriented for frictional failure make an angle of ~30° to the maximum compressive stress axis, σ1. Nonoptimally oriented faults can also be active, but faults with typical frictional strength should not slip at a high angle to σ1 (≥60°) (4). Faults operating at a high angle to σ1 must be very weak compared to the surrounding material to slip at relatively low applied shear stress. Similarly, faults oriented at a very low angle to σ1 (<10°) must also be weak.

The traction at Earth’s surface is zero, requiring an Andersonian stress state in which one principal stress axis is vertical (5). Most crustal faulting is consistent with an Andersonian stress state (6). The very shallow dip (~10°) of the upper portion of many subduction zone megathrusts (7) suggests that, if these faults exist within an Andersonian stress state, they operate at a high angle to one principal stress axis (and at a low angle to the other two) and are therefore frictionally weak. However, observations from some subduction zone locations indicate that the stress in these locations is not Andersonian (811).

I systematically investigate the stress orientations in subduction zones worldwide, to determine whether the stress state is generally non-Andersonian and whether megathrust faults are well-oriented for failure. I compile the moment tensors (12, 13) of earthquakes located within 20 km of the subduction zone interface (7) to represent the megathrust region, and more shallow events to represent the upper plate. I stack the events for each subduction zone to invert for stress orientation (14) as a one-dimensional function of subduction interface depth. For the several subduction zones near Japan, there are enough moment tensors to invert for the two-dimensional spatially varying stress field as well.

The stress state in the subduction zones near Japan is generally not Andersonian. The σ1 axis in most megathrust regions plunges systematically trenchward (opposite the direction of subduction zone dip), generally plunging between 10° and 50° (Fig. 1A). A similar plunge of the σ1 axis is observed within the upper plate above the Japan Trench, Kuril Trench, and Nankai Trough (Fig. 1G), whereas back-arc extension (near-vertical σ1) dominates above the Ryukyu and Izu-Bonin trenches.

Fig. 1 Orientation of the maximum compressive stress axis, σ1, for subduction zones near Japan.

(A) Spatial distribution of the plunge of σ1 near the megathrust, positive if σ1 plunges trenchward (i.e., opposite to the direction of subduction zone dip). (B) Spatial distribution of the angle of σ1 to the subduction zone interface, positive to drive reverse slip. (C) The plunge of σ1 near the megathrust as a function of subduction zone depth, shown separately for the subducting Pacific Plate (Kuril, Japan, and Izu-Bonin trenches) and the Ryukyu Trench. Dashed lines indicate 95% confidence range. (D) The plunge of σ1 near the megathrust for the Nankai Trough. (E) The angle of σ1 to the subduction zone interface, as a function of subduction zone depth, for the subducting Pacific Plate and the Ryukyu Trench. Dashed lines indicate 95% confidence range. Light shading indicates high or low angles implying a poorly oriented fault; dark shading indicates angles inconsistent with reverse slip. (F) The angle of σ1 to the subduction zone interface for the Nankai Trough. Shading as in (E). (G) Spatial distribution of the plunge of σ1 in the upper plate, positive if σ1 plunges trenchward.

Comparing the plunge of the σ1 axis in the megathrust region to the dip of the subduction interface (7) gives the angle of σ1 to the megathrust fault (Fig. 1B). For the Kuril, Japan, and Izu-Bonin trenches, the σ1 axis is typically at an angle of 20° to 50° to the fault. For the Ryukyu Trench, the σ1 axis makes somewhat larger angles of 30° to 80° to the fault. In the Nankai Trough, σ1 is oriented at 10° to 30° to the fault in the upper 20 km, while at deeper depths, the σ1 axis is near-vertical. The plunge of the σ1 axis is stable with depth (Fig. 1C), except in the Nankai Trough (Fig. 1D), and the subduction megathrust is generally well oriented for failure from the near-surface to at least 60 km depth (Fig. 1E). The stacked Kuril, Japan, and Izu-Bonin trenches (the subducting Pacific plate) show a megathrust fault that is optimally oriented for failure, at ~30° to σ1, over the full depth range. The megathrust at the Ryukyu Trench is generally well oriented for failure at <60° down to 30 km depth, and is marginally oriented at ~60° from 30 km to 60 km depth. The exception is the Nankai Trough (Fig. 1F), which is well oriented at ~20° to σ1 only in the upper 20 km. Below 30 km, the megathrust is not oriented for reverse faulting within the error bounds. This suggests that 30 km is the downward limit of coupling between the subducting and overriding plates.

For all global subduction zones studied, for almost all depths, the σ1 axis plunges systematically trenchward, with plunge generally between 10° and 50° (Fig. 2A). The plunge tends to decrease with depth, while the subduction zone dip increases (7), so that the angle of σ1 to the subduction interface is remarkably stable with depth (Fig. 2B). This angle is also remarkably similar across subduction zones, most falling within the range 45° to 60°.

Fig. 2 Orientation of the maximum compressive stress axis, σ1, as a function of depth for subduction zones worldwide.

(A) The plunge of σ1 near the megathrust, positive if σ1 plunges trenchward, as a function of subduction zone depth. Line colors and styles correspond to subduction zones as shown in the key; results from Japan are reproduced in black. The 95% confidence (not shown) is on average ±8°. (B) The angle of σ1 to the subduction zone interface as a function of subduction zone depth. Light shading indicates high or low angles implying a poorly oriented fault; dark shading indicates angles inconsistent with reverse slip. (C) Map with the subduction zones used indicated in red, excluding Japan. Numbers correspond to those in the key.

The Japan and South American subduction zones are anomalous in being more optimally oriented at 20° to 40°, as is the Nankai Trough in the upper 20 km. The Marianas Trench, similar to the Ryukyu Trench, is at a somewhat high angle at 20 to 40 km depth, although in both cases angles of <60° are within the error bounds. The Mexico Trench is the only case in which the megathrust is unambiguously at high angle, although only below 40 km, implying that this megathrust may be unusually weak at depth. These few anomalous stress orientations are not correlated with the age of the subducting plate (15) or the trench sediment thickness (16). The subduction zones at low angle (Japan, Nankai, South America) all exhibit strong seismic coupling (17). However, subduction zones at moderate to high angles exhibit a wide range of seismic coupling, so there is no clear correlation.

The most-vertical principal stress axis, σV, must be vertical at Earth’s surface. However, the systematic plunge of the σ1 axis implies that σV is generally not vertical. I invert for stress orientation the earthquakes near the main Japanese islands, separated into megathrust and upper plate regions and binned by hypocentral depth (14), to investigate whether σV becomes closer to vertical near the free surface. I include the aftershock sequence of the 2011 Tohoku earthquake, which contains many shallow events in the megathrust region. The aftershocks must be inverted separately because of the impact of the Tohoku earthquake on the local stress field (8, 9) and on the distribution of earthquakes in the upper plate (18). The megathrust zone shows no rotation of σV toward vertical at shallower depths near the trench, either before or after the Tohoku earthquake (Fig. 3, A and B). The upper plate, however, demonstrates a clear rotation of σV toward vertical with decreasing depth, over the whole depth range considered (Fig. 3, C and D). The upper plate appears to see the free surface at depth, but the megathrust zone does not. The difference may be due to the deforming outer accretionary wedge (19). The outer wedge permanently deforms in response to the applied stresses, including any difference in traction between the free surface and the base of the wedge, which may decouple the stress along the megathrust at greater depths from the constraints of the free surface.

Fig. 3 Orientation of the most-vertical principal stress axis, σV, for the main islands of Japan as a function of earthquake hypocentral depth.

Error bars indicate 95% confidence range. Color indicates which stress axis is most vertical. Depth is relative to sea level. (A) The plunge of σV in the Japan Trench megathrust region prior to the 2011 M9 Tohoku earthquake. (B) The plunge of σV in the upper plate of the main islands of Japan prior to the Tohoku earthquake. (C) The plunge of σV in the Japan Trench megathrust region for the 3 months following the Tohoku earthquake. (D) The plunge of σV in the upper plate of the main islands of Japan for the 3 months following the Tohoku earthquake.

The angle of σ1 to the subduction interface indicates the frictional strength of the megathrust fault. The observed optimal to moderate angles imply that the megathrust frictional strength is similar to that of its surroundings. These angles are inconsistent with models of a relatively weak megathrust in a strong crust (20). The observed stability of the angle of σ1 to the subduction interface suggests that the coefficient of friction of the megathrust is similar in most subduction zones worldwide and does not vary substantially from the near-surface to at least 60 km depth. The strength of a fault can be estimated from the angle of σ1 to the fault, using a Mohr circle construction (21), if assumptions are made about the strength of the surrounding material. If the region around the megathrust is assumed to be strong and critically stressed (all other faults have a typical friction coefficient of μ = 0.6, and pore pressure is hydrostatic), the megathrust coefficient of friction is implied to be μ = 0.35 to 0.6. If pore pressure is elevated only within the megathrust fault zone, the required megathrust pore pressure would be 0.3 to 0.6 of lithostatic pressure.

Several lines of evidence suggest that the megathrust is weaker than the above assumptions imply. A coefficient of friction of μ = 0.08 is estimated from the heat flow observed at shallow depths after the 2011 Tohoku earthquake (3), and the coefficient of friction inferred from heat flow for subduction zones worldwide is generally μ < 0.1 (22). Seismological imaging of fluids in subduction zones implies nearly lithostatic pore pressure at greater depths (23, 24). Rotations of the stress field due to large subduction zone earthquakes imply low deviatoric stress levels (8, 9). A more likely model, therefore, is that the entire region is at low deviatoric stress. In this model, all faults, including the megathrust, are weak and support only low shear stress. A similar model has been proposed for the San Andreas continental transform plate boundary (25), despite the very different tectonic setting.

If the entire subduction zone region is at low strength, this implies that high fluid pressures or low friction coefficients are pervasive. We rule out a model of a broad area of high fluid pressure as seismic observations limit them to a narrow plate interface (26). Therefore, other active faults in the subduction zone region must be weakened as a result of narrow zones of high fluid pressure or low friction coefficient. The stability of the observed stress orientations with depth implies that there are no large down-dip changes in the relative strengths of the megathrust and other faults, despite substantial changes in temperature and pressure. This suggests that all faults in the subduction zone region have similar physical mechanisms controlling their strength.

Supplementary Materials

www.sciencemag.org/content/349/6253/1213/suppl/DC1

Materials and Methods

Figs. S1 and S2

References (27, 28)

References and Notes

  1. See supplementary materials on Science Online.
  2. Acknowledgments: I thank P. McCrory, A. Michael, R. Bürgmann, and an anonymous reviewer for constructive comments on an earlier version of this manuscript. The Japan National Research Institute for Earth Science and Disaster Prevention (NIED) moment tensor catalog is available from the NIED at www.fnet.bosai.go.jp/fnet/event/search.php. The Global CMT (GCMT) moment tensor catalog is available from the GCMT project at www.globalcmt.org/CMTsearch.html. The SLAB 1.0 subduction zone model is available from the U.S. Geological Survey at http://earthquake.usgs.gov/data/slab. The SATSI code used for stress inversion is available from the U.S. Geological Survey at http://earthquake.usgs.gov/research/software/#SATSI.
View Abstract

Navigate This Article