Strength of stick-slip and creeping subduction megathrusts from heat flow observations

See allHide authors and affiliations

Science  29 Aug 2014:
Vol. 345, Issue 6200, pp. 1038-1041
DOI: 10.1126/science.1255487

Strong yet creeping megathrust faults

Powerful faults in subduction zones, called “megathrust faults,” produce the largest earthquakes on Earth. Gao and Wang use heat flow data to show that when the faults subduct jagged sea floor, they generate tamer earthquakes than do faults that subduct smooth sea floor. The rugged sea floor brings irregularities into the fault that cause it to deform slowly over time, which results in a comparatively higher fault strength and lower seismicity. The finding has a direct impact on assessing regional earthquake and tsunami hazards.

Science, this issue p. 1038


Subduction faults, called megathrusts, can generate large and hazardous earthquakes. The mode of slip and seismicity of a megathrust is controlled by the structural complexity of the fault zone. However, the relative strength of a megathrust based on the mode of slip is far from clear. The fault strength affects surface heat flow by frictional heating during slip. We model heat-flow data for a number of subduction zones to determine the fault strength. We find that smooth megathrusts that produce great earthquakes tend to be weaker and therefore dissipate less heat than geometrically rough megathrusts that slip mainly by creeping.

Subduction megathrusts that primarily exhibit stick-slip behavior can produce great earthquakes, but some megathrusts are observed to creep while producing small and moderate-size earthquakes. The relationship between seismogenesis and strength of subduction megathrust is far from clear. Faults that produce great earthquakes are commonly thought of as being stronger than those that creep (1). Megathrusts that are presently locked to build up stress for future great earthquakes are thus described as being “strongly coupled.” However, some studies have proposed strong creeping megathrusts because of the geometric irregularities of very rugged subducted sea floor (2, 3).

Contrary to a widely held belief, geodetic and seismic evidence shows that very rough subducting sea floor promotes megathrust creep (2). All the subduction zones that have produced moment magnitude (Mw) 9 or greater earthquakes feature rather smooth subducting sea floor, because the igneous crust is devoid of large seamounts or fracture zones and/or because of large amounts of sediments (4, 5). To know whether the smooth and highly seismogenic megathrusts are stronger or weaker than the rough and creeping megathrusts, we compared the amounts of frictional heat they dissipate. A stronger fault moves against greater resistance and, for the same slip distance, dissipates more heat per unit area.

We first compared two very different subduction zones in terms of seismicity and fault roughness. The Japan Trench (Fig. 1A) is a highly seismogenic subduction zone and has produced many large earthquakes (6), including the 2011 Mw 9 Tohoku-oki earthquake that generated a devastating tsunami. The megathrust moves primarily in a stick-slip manner, although various parts of it exhibit creep accompanied with small repeating earthquakes (7). The subducting sea floor is smooth, except for the northern and southern ends of the margin beyond the 2011 main rupture area (Fig. 1A). The northern Hikurangi subduction zone (Fig. 1B), however, is in sharp contrast with the Japan Trench. Among all the clearly documented megathrust earthquakes here, the largest is the Mw 5.6 Gisborn earthquake in 1966 (8). Most of the megathrust is creeping at the subduction rate (100% creeping ratio), as constrained by Global Positioning System observations from ~350 sites (9) (Fig. 1B). The observed creep is probably a persistent instead of transient behavior, given the lack of clear paleoseismic evidence for great megathrust earthquakes. The subducting sea floor is extremely rugged, featuring a number of subducting seamounts (10), but becomes smoother to the south owing to large amounts of sediments (Fig. 1B).

Fig. 1 Tectonic setting of the Japan Trench and northern Hikurangi subduction zones.

(A) Japan Trench. Coseismic slip distribution of the 2011 Mw 9.0 Tohoku-oki earthquake (table S2) is shown by color shading and epicenter by star. Heat-flow sites are shown by red symbols (circles, marine probe; squares, land borehole) (11). (B) Northern Hikurangi. Geodetically determined creeping ratio (ratio of creeping rate to subduction rate) is from (9). Star shows the location of the largest clearly documented megathrust earthquake (table S2). Heat-flow sites include those of subsea bottom-simulating reflectors (red dots) and bottom-hole temperature measurements (red squares) (11). In both (A) and (B), the thick blue line indicates thermal model profile. Heat-flow values in the rectangular area are projected on to the model profile and shown in Fig. 2.

To determine frictional heating to infer fault strength, we developed two-dimensional (2D) finite element models. At both the Japan Trench and northern Hikurangi, heat-flow data (Fig. 1) provide adequate model constraints (11). The long-term frictional strength of the shallow megathrust is represented by the apparent coefficient of friction μ′ = τF/σn, where τF is shear strength and σn is normal stress, approximately the weight of the overlying rock column. Parameter μ′ includes contributions from both the intrinsic friction and pore-fluid pressure (11). It is understood that μ′ represents the integrated strength of a shear band that accommodates fault motion. The band can be as thin as millimeters for seismic slip but wider by orders of magnitude for creeping motion (2, 12). The rate of frictional heating per unit area is thus q = τFV = μ′σnV, where V is the long-term rate of slip of the megathrust. The frictional heating contributes to surface heat flow, so that the value of μ′ can be estimated by comparing model predicted q with heat-flow measurements. Because heat is dissipated only when the fault is in motion, for stick-slip faults V is accomplished mainly by repeatedly having large earthquakes, and the μ′ value inferred from frictional heating is a representation of the average strength during coseismic slip. The actual stress and strength changes during individual earthquakes are much more complex (13), but the frictional heat seen in surface heat-flow measurements is their integrated effect over numerous earthquakes. Unlike with previous thermal models, we invoked a parameterization based on laboratory experiments (14) to model the gradual downdip transition along the subduction interface from shallow frictional slip to deeper viscous shearing with increasing temperature (11).

The model results show that the thermally defined frictional strength of the subduction faults is low, with μ′ ≈ 0.025 for the Japan Trench and 0.13 for northern Hikurangi (Fig. 2). In comparison, the effective friction of rocks based on laboratory-derived intrinsic friction (15) and at hydrostatic pore-fluid pressure is about 0.4. The low strength is consistent with the long-standing notion that subduction faults are weak, barely transferring enough stress to prevent forearc topography from collapsing under its own weight (16, 17). Of even greater importance to the present study is the difference between the apparent coefficients of friction for the two megathrusts. The difference cannot be reconciled by considering data uncertainties. For example, trading their preferred μ′ values fails to fit the heat-flow observations in both places (Fig. 2). Hydrothermal circulation within the subducted oceanic crust can lower the surface heat flow (11), but this effect is too small at the Japan Trench to explain the lower observed heat flows (18).

Fig. 2 Heat dissipation models for Japan Trench and northern Hikurangi.

(A and B) For each subduction zone, the top panel shows observed (symbols) and model-predicted (thin lines) surface heat flows and model interface shear stress (thick gray line). Circles and squares indicate marine and land heat-flow measurements, respectively (11). Bottom panel shows model subsurface temperatures (contours) and plate interface geometry (blue line) (11).

One reason for a lower μ′ for the Japan Trench is that it primarily represents coseismic fault strength (Fig. 3). In laboratory experiments, faults weaken by a factor of three to five when their slip accelerates to seismic rates of ~1 m/s, independent of their frictional behavior at lower rates (19). In real faults, such dynamic weakening may locally cause complete stress drop, but it cannot happen over very large fault areas, as evidenced by the very small average stress drops in great earthquakes (20). Even in the 2011 Mw 9.0 Tohoku-oki earthquake, which exhibited a greater stress drop than other great earthquakes, reported average stress drop of the rupture zone is still only about 4 to 7 MPa (21, 22), with peak value in a localized region as high as 40 MPa (23). At 20 km depth, roughly in the middle of the depth range of the Tokoku-oki rupture, σn is about 500 MPa, so that a 5-MPa static stress drop only requires a 0.01 decrease in μ′. In other words, a factor of 1/3 decrease in μ′—for example, from 0.035 to 0.025—is adequate to explain the Japan Trench results (24). The seemingly small decrease in fault stress is sufficient to reverse the stress state of the upper plate in the rupture area from margin-normal compression to tension (16), a change that indeed happened as a result of the Tohoku-oki earthquake (25).

Fig. 3 Schematic illustration of fault stress and frictional heating for stick-slip and creeping faults.

(A) Fault slip history. (B) Fault stress corresponding to slip in (A). (C) An enlarged portion of (B) showing partitioning of energy in seismic slip for stick-slip faults, in comparison with energy dissipated by strong and weak creeping faults. For stick-slip fault, τ0 is pre-earthquake stress, τp is peak stress, and τd is stress during seismic slip (13, 27). Embedded Imageand Embedded Image are yield stresses of smooth/weak and rough/strong creeping faults, respectively.

Therefore, despite the remarkable weakness of subduction faults, the average stress drop in great earthquakes is only a relatively small, although important, fraction of the average fault strength. Large dynamic weakening in parts of the fault must be accompanied by much less weakening or even strengthening in other parts, such that the average stress drop of the entire rupture zone is much less than localized peak values. For example, local strengthening at rupture boundaries is known to trigger well-documented afterslip following subduction earthquakes (26). An important implication is that the amount of frictional heat is several times the combination of the energy radiated as seismic waves and consumed in permanently deforming rocks around the rupture (13, 27) (Fig. 3C).

Because the difference in μ′ between the two subduction zones cannot be fully explained by dynamic weakening, it must reflect a difference in static fault strength. From observations (2) and theoretical reasoning (3), extreme ruggedness of the subducting sea floor such as at northern Hikurangi gives rise to heterogeneous stress and structural environments that promote creep and small earthquakes. But unlike creeping along a smooth fault facilitated by weak gouge [e.g., (28)] (shown in Fig. 3 as “weak creep”), rough faults creep by breaking and wearing geometrical irregularities in a broad zone of complex internal structure. The integrated resistance to creep is expected to be relatively high (shown in Fig. 3 as “strong creep”). Our results indicate that the creeping megathrust at northern Hikurangi is indeed stronger than the stick-slip, but much smoother, megathrust at the Japan Trench.

To investigate whether the difference in μ′ for the Japan Trench and northern Hikurangi reflects a systematic correlation between fault strength and predominant mode of fault slip, we analyzed another seven subduction zones where heat flow observations can reasonably constrain frictional heating and allow μ′ to be determined with our modeling (Fig. 4) (11). Most good-quality heat-flow data are from highly seismogenic subduction zones, which attract more research efforts. Therefore, northern Hikurangi, Manila Trench, and Kermadec are valuable examples for less seismogenic subduction zones. We also added Costa Rica as an example of intermediate degree of creeping, but its apparent friction is poorly constrained from previous studies (11).

Fig. 4 Apparent friction of megathrust versus maximum size of clearly documented interplate earthquake.

1, northern Hikurangi; 2, Manila Trench; 3, Costa Rica; 4, Kermadec; 5, Nankai; 6, Kamchatka; 7, northern Cascadia; 8, Japan Trench; 9, Sumatra; and 10, south-central Chile (table S2). Except for Costa Rica (11), the apparent coefficients of friction are obtained using thermal models developed in this study, with error bars based on numerical testing of model fit to heat-flow data (Fig. 2 and figs. S1 to S7) (11). Error bars for Mmax are based on publications on these earthquakes (table S2).

The estimated apparent coefficient of friction decreases with increasing Mmax, the maximum magnitude of clearly documented megathrust earthquake at each subduction zone (Fig. 4). The Mmax values provide a proxy of long-term seismic slip. In determining Mmax for northern Hikurangi, we do not consider two poorly recorded events in 1947 both with Mw ~ 7 (29), because they occurred at a very shallow depth of the plate interface and do not represent the general slip mode of the megathrust. For subduction zones with Mmax 8.3 to 9.5, the fault motion is primarily stick-slip. Coseismic slip in their largest earthquakes is comparable to slip deficits accumulated over typical interseismic intervals of several hundred years, and geodetic observations show a high degree of megathrust locking at present (table S2). For three of the other four subduction zones, the present locking/creeping state of the megathrust is constrained by modern geodetic measurements (table S2). They show large creep, with northern Hikurangi having the most active creep (Fig. 1B). At Kermadec, the present creeping/locking state of the megathrust cannot be adequately determined by geodetic measurements (30). All the highly seismogenic subduction zones in this suite feature smooth subducting sea floor, and the faults are weaker than the faults associated with the subduction of rugged sea floor. The general correlation between subducting sea floor ruggedness, creeping, and greater heat dissipation suggests that geomorphological and thermal observations may be useful in assessing earthquake and tsunami hazards for risk mitigation.

Supplementary Materials

Materials and Methods

Supplementary Text

Figs. S1 to S7

Tables S1 and S2

References (31109)

References and Notes

  1. Materials, methods, and other information are available as supplementary material on Science Online.
  2. In some earthquakes, the final fault stress at the end of the rupture may be somewhat smaller (overshoot) or larger (undershoot) than the fault strength during slip, but the deviation from the coseismic strength is expected to be a small fraction of the stress drop.
  3. Acknowledgments: We thank J. He for writing finite element code PGCTherm and implementing the line-element method for fault modeling, W. -C. Chi for making available digital heat-flow data for Manila Trench, and S. Wu and J. Zhang for discussions. X.G. was supported by Chinese Academy of Sciences’ Strategic Priority Research Program Grant XDA11030102 and Open Foundation of Key Laboratory of Marine Geology and Environment Grant MGE2012KG04, and K.W. was supported by Geological Survey of Canada core funding and a Natural Sciences and Engineering Research Council of Canada Discovery Grant through the University of Victoria. This is Geological Survey of Canada contribution 2014105. All heat-flow data used are from published sources as listed in the reference list. Modeling parameters and tabulated results are available in the supplementary materials.
View Abstract

Navigate This Article