The Seismogenic Zone of Subduction Thrust Faults

The Seismogenic Zone of Subduction Thrust Faults originated from a MARGINS Theoretical Institute on The Seismogenic Zone Revisited, held at Snowbird, Utah, March 16 to 21, 2003. More than 100 scientists from North, South, and Central America, Asia, and Europe attended this meeting. Theoretical Institutes strive to assemble active workers and young scientists, including graduate students, to review the subject matter, and foster interaction between the active fields of a discipline. A particular goal is to communicate the state of the art to a broader community, especially developing young professionals. The meeting itself involved spirited interaction and communication among participants. This volume attempts to capture some of that excitement and to also provide a broader perspective on the Seismogenic Zone of Subduction Thrusts to a larger audience. The volume evolved from manuscripts submitted by interested scientists after the meeting. Although not an exact image of the meeting, the papers provide representative snapshots and an excellent overview of the subject. We appreciate the efforts of the authors and reviewers.

Subduction zones, with their devastating earthquakes and tsunamis, are a principal type of plate boundary and commonly occur along continental margins. Continental margins are the focus of human activity, replete with natural resources and geologic hazards. The MARGINS Program, sponsored by the U.S. National Science Foundation (NSF), was developed to coordinate research activity along continental margins. The Seismogenic Zone Experiment or SEIZE is one initiative of the MARGINS Program and is responsible for organizing the Theoretical Institute that led to this volume. The MARGINS program sponsored the participation of U.S. scientists in the Theoretical Institute, while similar programs in other nations provided travel funds for non-U.S. scientists. We particularly thank Bilal Haq, the NSF MARGINS Program Director, for his support. The MARGINS Office (Julie Morris, Meredith Berwick, and Paul Wyler) tirelessly facilitated the assembly of this volume. We thank the MARGINS Publications Series (Neal Driscoll, Gary Karner, Julie Morris, and Eli Silver) for their support. We are also indebted to the editorial staff at Columbia University Press for their work in final production of The Seismogenic Zone of Subduction Thrust Faults.

This volume is being published on the cusp of Integrated Ocean Drilling Program penetrations into active subduction zones in Japan and Central America. The drilling activity involves diverse related scientific investigations and will culminate in holes into the seismogenic zone of subduction thrusts, the first direct observation of these important features. We hope this volume will provide a partial foundation for this work, and look forward to significant advances in the study of subduction zones during the next decade.

Subduction zones such as the circum-Pacific ring of fire are characterized by a dipping, relatively narrow (~100–200 km width) and shallow (<50 km depth) surface capable of generating large (M > 7) and great (M > 8) earthquakes. This surface represents the interface between subducting and overriding plates and is often termed the seismogenic zone (fig. 1.1). For reasons that are fundamental to the subduction process, including depression of isotherms and consequent broadening of that part of the plate interface where seismic processes can occur, these zones generate Earth’s largest earthquakes and most destructive tsunamis. The earthquakes and subsequent tsunamis often impact densely populated coastal areas, causing large numbers of fatalities, as tragically demonstrated by the 26 December 2004 Sumatra earthquake and tsunami, believed to have killed more than 200,000 people.

Although we have a general understanding of the processes that occur in the seismogenic zone, many critical details remain obscure. For example, we know that earthquakes reflect the rapid release of strain associated with prior locking of the shallow plate interface and strain accumulation during inter-seismic periods lasting tens to hundreds of years. However, we do not understand why some interplate subduction earthquakes are relatively modest in size (M 7.5 or less), rupturing relatively small areas with limited along strike (trench-parallel) rupture length (<100 km), while others, such as the great (M > 9) 1960 Chile [Plafker, 1972], 1964 Alaska, and 2004 Sumatra events [Lay et al., 2005; Ammon et al., 2005; Stein and Okal, 2005], can rupture ~1200 km or more along strike. The effective rupture area for such events is huge, ~10⁵ km². For comparison, the longest known rupture on the San Andreas fault, the 1906 San Francisco earthquake (400 km), had a rupture area of ~6 × 10³ km².

Critical details of the strain accumulation process are also obscure. Why are some subduction zones fully locked, accumulating elastic strain at essentially the full plate convergence rate, while others appear to be only partially coupled or even freely slipping, accumulating little or no elastic strain? Are they locked all the way to the surface, or mainly locked at depth, perhaps in patches [Wang and Dixon, 2004; Lay and Schwartz, 2004]? Reasoning by analogy with the San Andreas fault in California, complete locking from seismogenic depths for this fault (~15 ± 5 km) all the way to the surface is common, but surface creep is also observed in places such as the Parkfield segment, implying that more complex patterns of locking and strain accumulation can occasionally occur. However, as pointed out by Underwood (this volume), sediments deposited on the downgoing plate of a subduction zone evolve as they are exposed to increasing pressure and temperature during the subduction process, leading to profound downdip changes in porosity, permeability, and pore-fluid pressure on and near the seismogenic plate interface. This means that the physical environment of subduction thrust faults is fundamentally different from transform and normal faults, perhaps limiting the utility of the San Andreas comparison and suggesting the possibility of complex spatial locking patterns (fig. 1.1).

Figure 1.1     Perspective view of the seismogenic zone. The seismogenic zone represents the failure surface of large and great earthquakes and may be represented by aftershock distribution. The seismogenic zone is characterized mechanically by large areas or asperities that fail in a frictionally unstable or velocity-weakening manner. Failure by accelerating slip is the essence of an earthquake. Conversely, aseismic areas both updip and downdip of the seismogenic zone or within it may fail in a frictionally stable or velocity-strengthening manner. The absence of runaway slip in these areas precludes earthquakes or limits their size. Modified from Bilek and Lay [2002], Lay and Schwartz [2004], and Lay and Bilek (this volume).

Temporal variation is another important and poorly understood factor. Are patterns of strain accumulation on the seismogenic plate interface constant during a given earthquake cycle? If so, do these patterns translate into patterns of strain release during the next earthquake? During subsequent earthquakes? If patterns change, what are the controlling factors? Are variations in the frictional properties of subducted sediments critical or are variations in pore-fluid pressure and corresponding changes in effective normal stress more important? What role do diagenetic and metamorphic reactions play in generating and modulating fluid flow and permeability along subduction thrust faults?

The MARGINS program of the U.S. National Science Foundation began in 1998 and includes the Seismogenic Zone Experiment (SEIZE) to study such questions. In March 2003, MARGINS hosted a 5-day workshop in Snowbird, Utah, to evaluate progress to date. The papers in this volume are based on talks presented at the workshop and subsequent discussion; they provide a comprehensive review of key seismogenic zone processes.

The seismogenic zone is that portion of the plate interface that produces earthquakes via stick-slip sliding [Brace and Byerlee, 1966]. For simplicity, in this volume we focus on interplate subduction fault zones capable of producing large (M > 7) and great (M > 8) earthquakes. Other regions, including intraplate fault zones such as the outer rise of subduction zones or upper crustal earthquakes within the overriding plate, are also interesting but generally are associated with smaller events.

In a simple stick-slip frictional mechanics model, stress and strain accumulate during the interseismic period (stick), subsequently failing in a catastrophic manner (slip). More sophisticated dynamical friction models are probably more relevant: that part of a fault zone undergoing stick-slip behavior is thought to be dominated by unstable or conditionally stable frictional slip [Scholz, 1998] (fig. 1.1). Frictionally unstable regions fail with velocity-weakening behavior. Once a small rupture initiates, slip acceleration occurs because frictional strength decreases as slip velocity increases (e.g., Marone and Saffer, this volume). In contrast, aseismic portions of the fault, usually considered to be shallower or deeper than the seismic portion, are dominated by stable frictional failure (aseismic creep). These regions are considered to be velocity strengthening: frictional strength increases as slip velocity increases. They are therefore incapable of runaway slip, and earthquakes cannot nucleate. New results show that aseismic portions of the fault can also occur embedded within portions of the fault generally considered seismic [e.g., Norabuena et al., 2004].

The seismogenic zone as defined above corresponds roughly to a temperature range of ~100°–150°C at the updip end to ~350°C–450°C at the down-dip end. The temperature limits of the seismogenic zone are established by comparing thermal models to the depth extent of earthquake ruptures and their aftershocks [Currie et al., 2002; Hyndman and Wang, 1993; Hyndman et al., 1995, 1997; Oleskevich et al., 1999; Wang et al., 1995] and hence have uncertainties associated with model limitations. This temperature range corresponds roughly to a depth range of 5–10 km (updip) to 40–50 km (downdip), but it is important to realize that significant inhomogeneities occur within this region in terms of frictional and seismic behavior.

The zone of seismic slip for a given major earthquake is often assumed to be approximately equivalent to the location of aftershocks (e.g., Schwartz and DeShon, this volume). The extent of earthquake dislocation on the fault surface may be estimated from seismic or tsunami waveform data and geodetic data measuring coseismic displacement [e.g., Sataki, 1993]. The location of the seismogenic zone may also be inferred from geodetic measurements during the interseismic period because the locked plate interface causes measurable strain accumulation at the surface [e.g., Savage, 1983; Thatcher and Rundle, 1984; Dixon, 1993; Lundgren et al., 1999; Norabuena et al., 1998, 2004]. It is usually assumed that the fault area that accumulates strain during the interseismic period is the same area that will subsequently rupture in the next large or great earthquake, although this remains to be fully tested. Microseismicity during the interseismic interval may only partially represent the extent of subsequent rupture, in part because this type of seismicity may actually represent slippage and partial or full strain release [Norabuena et al., 2004; Schwartz and DeShon, this volume].

While the seismogenic zone is often referred to as being locked through most of the seismic cycle, locking may actually be both spatially and temporally heterogeneous (fig. 1.1). In terms of spatial complexity, seismic data have long suggested, and geodetic data have recently confirmed, that seismogenic zones may not be fully locked and slip at some fraction of the full plate rate during interseismic intervals [e.g., McNally and Minster, 1981; Norabuena et al., 1998]. The mechanical basis for this behavior is not clear, but one model involves patches of frictionally unstable fault surface surrounded by conditionally stable and stable surfaces [Bilek and Lay, 2002; Lay and Schwartz, 2004]. The frictionally stable portions of the fault are subject to creep, while the frictionally unstable portions remain locked through most of the seismic cycle. Surface geodetic monitoring with low spatial density might see this fault surface as being partially locked. Recent geodetic observations in Costa Rica, where the Nicoya peninsula enables surface observations quite close to the trench, have confirmed that strain accumulation in the seismogenic zone during the interseismic period can indeed be patchy [Norabuena et al., 2004]. The locked, frictionally unstable portions of the fault surface presumably fail during the seismic intervals, and their displacement catches up with the surrounding creeping, frictionally stable portions. The frictionally unstable locked patches would thus constitute asperities.

However, even this spatially complex picture of the seismogenic zone may be oversimplified. Observations of transient slip events on or near the seismogenic zone imply that locking and frictional properties can also vary in time throughout the seismic cycle. Such events have long been suspected on the basis of seismic and classical (terrestrial) geodetic data [e.g., Sacks et al., 1978; Linde and Silver, 1989], but the development of space geodetic techniques such as GPS has provided a much better observational framework. Such events are now well known in Japan [e.g., Kawasaki et al., 1995; Heki et al., 1997; Ozawa et al., 2002], Cascadia [Dragert et al., 2001; Miller et al., 2002], and Mexico [Iglesias et al., 2004; Larson et al., 2004] and are likely common in most subduction zones. As they become better studied, they will likely have important implications for our evolving picture of frictional conditions on the seismogenic plate interface.

The papers in this volume document significant progress in understanding the issues discussed above since the initiation of the SEIZE program, but many important problems remain. Hyndman (this volume), who pioneered many of the topics discussed in the volume, sets the stage for our review by summarizing current knowledge and outstanding issues. His comprehensive review includes references to some of the key early work on seismogenic zone processes, providing historical context. The remainder of the volume is organized in five sections:

•   The Incoming Plate

•   Convergent Margin Structure, Fluids, and Subduction Thrust Evolution

•   Laboratory Studies

•   Seismic and Geodetic Studies

•   Regional-Scale Deformation

The Incoming Plate

Knowledge of the nature of materials entering the subduction thrust fault is one of the advantages of studying seismogenesis at subduction zones. By looking at features of the incoming plate just outboard of the trench and assuming recently subducted lithosphere has the same or similar characteristics, it may be possible to draw conclusions about the seismogenic process and material properties in the fault zone through comparison of adjacent segments of subduction zones with contrasting characteristics. Three papers in this section discuss the nature of incoming sediments, the thermal state of the incoming lithosphere, and the effects of subducting topography on seismogenesis.

Most subducting seafloor is covered by a layer of sediment. Underwood (this volume) emphasizes that this sediment is essential in defining major hydrogeologic and geotechnical-mechanical aspects of the subduction system. Therefore it is essential to understand its lithologic architecture in order to predict its influence in the seismogenic zone. Underwood further points out that the simple model of sediments that coarsen and thicken upward as the plate approaches the trench is violated at most well-studied subduction zones. The specific documentation of the nature of incoming sediment will be essential at any locality for which deep drilling is planned to evaluate the seismogenic zone.

Hutnak et al. (this volume) discuss existing and new heat flow data and corresponding thermal models for the incoming Cocos plate offshore northern Costa Rica. They focus on the well-known but poorly understood contrast between relatively high heat flow characteristic of lithosphere created at the Cocos-Nazca spreading center (CNS) and anomalously low heat flow in the immediately adjacent East Pacific Rise (EPR) lithosphere. The strong heat flow contrast in oceanic crust off northern Costa Rica and the possible role of hydrothermal cooling were first identified by Langseth and Silver [1996]. New bathymetric, magnetic, seismic reflection, and heat flow data allow considerable refinement of this picture. These data illustrate the role of seamounts and other topographic features in guiding large fluxes of hydrothermal fluids. The new data support a model whereby EPR lithosphere is regionally cool because of efficient hydrothermal circulation in the uppermost basement.

The contrast in thermal state of the incoming lithosphere offshore Costa Rica provides a rare opportunity for a variety of comparative studies, because the northern half of the peninsula subducts cool EPR-generated lithosphere, while the southern half subducts warm CNS lithosphere. For example, New-man et al. [2002] note a change in the depth extent of interplate microseismicity, while Norabuena et al. [2004] and Schwartz and DeShon (this volume) point out correlated changes in the updip extent of locking on the plate interface as inferred from inversion of GPS geodetic data.

It has been known for some time that large subduction zone earthquakes exhibit considerable heterogeneity in slip and moment release across the fault plane, which may relate to strength contrasts (asperities) on the fault plane. Bilek (this volume) reviews the history of the asperity model and examines the role of bathymetric features such as seamounts in controlling regions of high versus low moment release in earthquakes.

Convergent Margin Structure, Fluids, and Subduction Thrust Evolution

Marine geophysical studies, drilling and associated laboratory results, and investigations of ancient accretionary prisms all provide a view of the nature of the subduction thrust, the overlying and underlying materials, and inferences on what controls seismogenesis. The papers noted below address these issues from the frontal part of the margin to greater depth as imaged by seismic reflection or revealed by exhumed rocks.

Bekins and Screaton (this volume) review the state of hydrogeologic knowledge for the Barbados accretionary complex, an intensively studied subduction-related sediment accretion zone. Data acquired from multiple drill holes and two long-term borehole observatories (CORKs) allow observations of fluid flow and transient events. These data show that pore pressures in the upper few kilometers of the accretionary complex tend to be high, especially below major faults, and can exceed lithostatic values. Also, fluid flow tends to be channelized. Fluid budgets and models suggest that transient high-flow events are relatively rare and are spatially restricted to major faults and other channels.

Fluid pressure is a significant factor controlling fault strength and sliding behavior and hence is critical to understanding seismogenic zone behavior. Saffer (this volume) reviews fluid-pressure distribution and magnitude within underthrust sediments at the well-studied Nankai and Costa Rican margins. Techniques used to ascertain fluid pressure within underthrust sediments at modern convergent margins are discussed. A compilation of existing data is also presented. Comparison of the fluid-pressure observations to end member models affirms a system in which the décollement acts as a major fluid drain. Saffer shows that spatial patterns of underconsolidation suggest upward drainage to the décollement and outward toward the trench. Documentation of elevated fluid pressures within underthrust sediments indicates that the rate of geological forcing by burial exceeds the ability of sediments to drain, limited by low permeability and drainage path lengths of several hundred meters. The magnitude of elevated pore pressures is consistent with results from hydrologic modeling studies and suggests that sediment permeability is the primary control on fluid pressure. This implies that permeability ultimately exerts a strong control on the strength of subduction systems by modulating effective stress.

Morgan et al. (this volume) provide an extensive range of evidence showing that sediments incoming to the Nankai subduction zone along the Muroto transect undergo diagenetic changes that increase their strength. These authors argue that some of the porosity anomalies that might be interpreted as a result of overpressuring (Saffer, this volume) could also be explained by diagenetic cementation. They further suggest that the cementation and strengthening of the underthrust sediments protect it from deformation as it continues to be underthrust beneath the accretionary prism.

McIntosh et al. (this volume) present new seismic reflection data for the Nicaragua margin. They image the plate interface down to a depth of 40 km and are able to observe a number of subducted seamounts. They suggest that such seamounts played a critical role in promoting shallow rupture during the 1992 Nicaragua earthquake. This event was a tsunami earthquake, defined as an earthquake that generates a tsunami larger than expected given its seismic moment [Kanamori, 1972].

Moore et al. (this volume) summarize features of exhumed accretionary prisms that may explain their seismogenic behavior. These authors note that, in the presumed temperature ranges of seismic activity (Hyndman, this volume), quartz and carbonate veining localized along faults tend to introduce materials to fault surfaces that can increase their cohesion or alter their frictional properties. These changes may induce stick-slip behavior and fundamentally control the position of the seismogenic zone.

Laboratory Studies

As the lithologic composition and pressure-temperature conditions of subduction thrusts become better known, experimental studies of the frictional properties of key materials become important. In the past, much experimental work has focused on materials and conditions typifying continental plate boundaries, such as the San Andreas fault. Papers in this volume provide an overview of recent experimental efforts to understand the behavior of materials at a range of P-T-H2O conditions relevant to subduction thrusts.

Moore and Lockner (this volume) review laboratory data on the smectite clay montmorillonite, a key mineral in many subduction thrust faults. This clay is stable to temperatures of ~150°C, whereupon transition to illite occurs. The authors point out that published literature on the frictional strength of montmorillonite gives coefficients of friction from 0.06 to 0.78, i.e., more than an order of magnitude variation, and that both velocity-weakening as well as velocity-strengthening behavior have been reported. They emphasize that some early experiments may not have controlled the water-saturation conditions very accurately and that variable water saturation probably accounts for most of the variation in frictional behavior.

Marone and Saffer (this volume) review some recent experiments on illite, smectite, and quartz. They demonstrate that the smectite-illite phase transition is unlikely to control the upper aseismic to seismic transition in subduction zones because illite is velocity strengthening and will not foster stick-slip behavior (Saffer and Marone, 2003). They note the tendency of unconsolidated materials including gouge to be velocity strengthening whereas lithified, especially quartz-rich rocks, tend to be velocity weakening, promoting stick-slip behavior. Overall they favor increasing consolidation/lithification, associated increase in effective stress, and quartz cementation as key processes controlling the upper aseismic-seismic transition.

Beeler (this volume) reviews laboratory results on faulting in weak materials, focusing on frictional strength, seismic coupling, dilatancy, and pore-fluid pressure. Over the last 4 decades, tectonophysicists have shown that, in general, it is not the strength of a fault that is the key to seismic slip but rather its stability or rate of change of strength with slip rate [e.g., Scholz, 1998]. Beeler carefully examines the basis of rate-dependent frictional parameters. He concludes that, in the absence of chemical weakening, the nominally stronger and more dilatant materials have a greater tendency toward velocity weakening and therefore unstable seismic slip compared to weaker materials. The reduction of contact area across asperities during dilation is the key to reducing friction. Accordingly, the instability that fosters accelerating slip is rooted in fault strength. Subduction thrusts initially develop in intrinsically weak, overpressured sediments (Saffer, this volume). Something must change at depth to spawn large and great earthquakes along these subduction faults.

Seismic and Geodetic Studies

Hasegawa et al. (this volume) examine rupture patterns for interplate earthquakes in northeast Japan, focusing on regions where successive earthquakes have occurred in the same location. They compare areas of maximum moment release (maximum slip) to the locked and slipping zones of the plate boundary during the corresponding interseismic period, as measured by inversion of geodetic data. They show that regions of maximum seismic slip in a given event tend to repeat in the next earthquake and also correspond to regions of maximum locking during the interseismic period. Thus interseismic geodetic data have significant value in hazard assessment, in the sense that regions of maximum locking will likely correspond to regions of subsequent high moment release.

Lay and Bilek (this volume) review seismic and tsunami data on moderate and large subduction zone earthquakes in the upper portion of the seismogenic zone. Numerous events with source depths <10 km tend to have anomalous faulting properties, typically involving either low rupture velocities or low stress drops. Tsunami earthquakes involve very large events that rupture the shallowest part of the seismogenic zone, with displacements extending all the way to the trench. Direct measurements for the 1992 Nicaragua tsunami event demonstrate that the rupture velocity was low, possibly because of low effective rigidity around the shallow fault zone. The variability of ruptures at shallow depths leads Lay and Bilek to infer that multiscale frictional heterogeneity, with patches of stable, conditionally stable and unstable friction, is a general attribute of shallow subduction zones.

Heki (this volume) summarizes key crustal deformation results from the dense array of continuous GPS stations in Japan, currently the best-instrumented subduction zone in the world. These results include an improved kinematic description of the major rigid blocks and bounding plates, secular subsidence in part of the fore-arc region, which may reflect subduction erosion of part of the upper plate, detailed coseismic rupture patterns on some major faults, and a variety of transient events. These include slow postseismic fault slip, which may be aseismic, slow-slip events independent of major earthquakes that may be aseismic or accompanied by microseismicity, and seasonal movements related to surface loading from groundwater and snow. Separating true transient deep crustal motion from station noise (e.g., from mismodeled orbits) and surface loading effects is important not only from a signal-versus-noise perspective but also because, in some cases, the transient surface loading or unloading may actually stimulate deeper transient processes.

Wang (this volume) reviews rheological models that may be combined with geodetic data to study subduction earthquakes and the earthquake cycle. He describes common elastic half-space and viscoelastic coupling models, the latter with an elastic layer overlying and coupled to one or more underlying viscoelastic layers, often implemented via finite element techniques. Wang points out that the way such models are implemented, including their kinematic boundary conditions, may influence results; hence caution is warranted. A case study of the Cascadia subduction zone is presented, illustrating some common trade-offs among model parameters.

Schwartz and Deshon (this volume) review seismic and geodetic data for the northern Costa Rica margin. The combination of unique geographic setting (the Nicoya peninsula is located quite close to the trench, essentially perched over the seismogenic zone) and a wealth of data from MARGINS-funded studies is leading to significant advances in our understanding of the behavior of seismogenic zones that subduct young oceanic lithosphere. The authors demonstrate that ongoing interseismic microseismicity is not spatially synonymous with the area of the seismogenic zone suggested by geodetic studies or failure in large earthquakes. They compare the thermal state of the incoming plate to the mechanical transitions inferred from seismic and geodetic data to investigate the role of sediment diagenesis and low-grade metamorphic reactions in controlling seismogenic behavior.

Regional-Scale Deformation

Most of the papers in this volume address the set of conditions and processes directly related to the earthquake cycle, a relatively short-term (<10³ year) process. However, if the plate interface is highly coupled (locked) during the interseismic part of the cycle, relatively high stresses may develop and propagate through the upper plate, which can act as a stress guide. Assuming this condition repeats through many earthquake cycles, geologically significant permanent strain and block rotation and translation may result. For example, oblique subduction may drive trench-parallel motion of fore-arc sliver blocks [Fitch, 1972; Jarrard, 1986]. In parts of Central America, coastal terrains migrate northwest at rates between 5 and 14 mm/yr [DeMets, 2001; McCaffrey, 2002; LaFemina et al., 2002; Norabuena et al., 2004]. Trench-normal compressive stresses related to a locked plate interface in southern Costa Rica may be responsible for out-of-sequence (nonplate boundary) thrust faulting and folding in both the fore-arc and back-arc regions of southern Costa Rica [Norabuena et al., 2004]. If continued over millions of years, this can result in significant crustal shortening and thickening, i.e., Andean-style mountain building.

Seno (this volume) contrasts the subduction and collision process, and the connections between processes on the main subduction thrust fault and deformation in the back arc and hinterland. He emphasizes the importance of slab dehydration in controlling rheology and hence the subsequent response to compressive stresses, in particular in determining whether delamination of the lower crust or lithospheric mantle may occur.

Kley and Vietor (this volume) investigate the Cenozoic growth of the Andes. They focus on the interplay between factors controlling the direction and magnitude of horizontal compressive stress on the plate interface (e.g., global plate kinematics, slab dip, degree of mechanical coupling) and factors controlling the rheology of the upper plate.

Acknowledgments

We thank the authors of this volume for contributing comprehensive and insightful papers on the seismogenic zone and the reviewers for thoughtful and timely comments. We also thank NSF’s MARGINS program, in particular Bilal Haq, for supporting the workshop that led to this volume. THD and JCM thank Thorne Lay and Susan Schwartz for comments on this paper. Finally, we thank the MARGINS Office (Julie Morris, Meredith Berwick, and Paul Wyer) and series editors (Neal Driscoll, Gary Karner, Julie Morris, and Eli Silver) as well as the staff of Columbia University Press for assistance with the publication of this volume.

References

Ammon, C. J., et al. (2005), Rupture process of the 2004 Sumatra-Andaman earthquake, Science, 308(5725), 1133–1139.

Beeler, N. M. (2007), Laboratory-observed faulting in intrinsically and apparently weak materials: Strength, seismic coupling, dilatancy and pore fluid pressure, this volume.

Bekins, B. A., and E. J. Screaton (2007), Pore pressure and fluid flow in the northern Barbados accretionary complex: A synthesis, this volume.

Bilek, S. L. (2007), Influence of subducting topography on earthquake rupture, this volume.

Bilek, S. L., and T. Lay (2002), Tsunami earthquakes possibly widespread manifestations of frictional conditional stability, Geophys. Res. Lett., 29(14), 1673, doi:10.1029/2002GL015215.

Brace, W., and J. D. Byerlee (1966), Stick-slip as a mechanism for earthquakes, Science, 153(3739), 990–992.

Currie, C. A., R. D. Hyndman, K. Wang, and V. Kostoglodov (2002), Thermal models of the Mexico subduction zone: Implications for the megathrust seismogenic zone, J. Geophys. Res., 107(B12), 2370, doi:10.1029/2001JB000886.

DeMets, C. (2001), A new estimate for present-day Cocos-Caribbean plate motion: Implications for slip along the Central American volcanic arc, Geophys. Res. Lett., 28, 4043–4046.

Dixon, T. (1993), GPS measurements of relative motion of the Cocos and Caribbean plates and strain accumulation across the Middle America Trench, Geophys. Res. Lett., 20, 2167–2170.

Dragert, H., K. Wang, and T. S. James (2001), A silent slip event on the deeper Cascadia subduction interface, Science, 292(5521), 1525–1528.

Fitch, T. J. (1972), Plate convergence, transcurrent faults, and internal deformation adjacent to Southeast Asia and western Pacific, J. Geophys. Res., 23, 4432–4460.

Hasegawa, A., N. Uchida., T. Igarashi., T. Matsuzawa, T. Okada, S. Miura, and Y. Suwa (2007), Asperities and quasi-static slip on the subducting plate boundary east off Tohoko, northeast Japan, this volume.

Heki, K. (2007), Secular, transient and seasonal crustal movements in Japan from a dense GPS array: Implication for plate dynamics in convergent boundaries, this volume.

Heki, K., S. Miyazaki, and H. Tsuji (1997), Silent fault slip following an interplate thrust earthquake at the Japan Trench, Nature, 386, 595–598.

Hutnak, M., A. T. Fisher, C. A. Stein, R. Harris, K. Wang, E. Silver, G. Spinelli, M. Pfender, H. Villinger, R. MacKnight, P. Costa Pisani, H. DeShon, and C. Diamente (2007), The thermal state of 18-24 Ma upper lithosphere subducting below the Nicoya Peninsula, northern Costa Rica margin, this volume.

Hyndman, R. D. (2007), The seismogenic zone of subduction thrust faults: What we know and what we don’t know, this volume

Hyndman, R. D., and K. Wang (1993), Thermal constraints on the zone of major thrust earthquake failure: The Cascadia subduction zone, J. Geophys. Res., 98, 2039–2060.

Hyndman, R. D., K. Wang, and M. Yamano (1995), Thermal constraints on the seismogenic portion of the southwestern Japan subduction thrust, J. Geophys. Res., 100, 15,373–15,392.

Hyndman, R. D., M. Yamano, and D.A. Oleskevich (1997), Seismogenic zone of subduction thrust faults, Island Arc, 6, 244–260.

Iglesias, A., S. K. Singh, A. R. Lowry, M. Santoyo, V. Kostoglodov, K. M. Larson, and S. I. Franco-Sanchez (2004), The silent earthquake of 2002 in the Guerrero seismic gap, Mexico (Mw=7.6): Inversion of slip on the plate interface and some implications, Geofis. Int. 43(3), 1–9.

Jarrard, R. D. (1986), Terrane motion by strike-slip faulting of fore-arc slivers, Geology, 14(9), 780–783.

Kanamori, H. (1972), Tectonic implications of the 1944 Tonankai and 1946 Nankaido earthquakes, Phys. Earth Planet. Inter., 5, 129–139.

Kawasaki, I., Y. Asai, T. Tamura, N. Sagiya, Y. Mikami, M. Okada, M. Sakata, and M. Kasahara (1995), The 1992 Sanriku-Oki, Japan, ultra-slow earthquake, J. Phys. Earth, 43, 105–116.

Kley, J. and T. Vietor (2007), Subduction and mountain building in the Central Andes, this volume.

LaFemina, P., T. H. Dixon, and W. Strauch (2002), Bookshelf faulting in Nicaragua, Geology, 30(8), 751–754.

Langseth, M. G., and E. A. Silver (1996), The Nicoya convergent margin—A region of exceptionally low heat flow, Geophys. Res. Lett., 23, 891–894.

Larson, K. M., A. Lowry, V. Kostolodov, W. Hutton, O. Sanchez, K. Hudnut, and G. Suarez (2004), Crustal deformation measurements in Guerrero, Mexico, J. Geophys. Res., 109, B04409, doi:10.1029/2003JB002843.

Lay, T., and S. Bilek (2007), Anomalous earthquake ruptures at shallow depths on subduction zone megathrusts, this volume.

Lay, T., and S. Y. Schwartz (2004), Comment on Coupling semantics and science in earthquake research, Eos Trans. AGU, 85(36), 339–340.

Lay, T., et al. (2005), The great Sumatra-Andaman earthquake of 26 December 2004, Science, 308(5725), 1127–1133.

Linde, A., and P. Silver (1989), Elevation changes and the great 1960 Chilean earthquake: Support for aseismic slip, Geophys. Res. Lett., 16, 1305–1308.

Lundgren, P., M. Protti, A. Donnellan, M. Heflin, E. Hernandez, and D. Jefferson (1999), Seismic cycle and plate margin deformation in Costa Rica: GPS observations from 1994 to 1997, J. Geophys. Res., 104, 28,915–28,926.

Marone, C., and D. Saffer (2007), Fault friction and the upper transition from seismic to aseismic faulting, this volume.

McCaffrey, R. (2002), Crustal block rotations and plate coupling, in Plate Boundary Zones, Geodyn. Ser., vol. 20, edited by S. Stein and J. Freymueller, pp. 101–122, AGU, Washington, D. C.

McIntosh, K. D., E. A. Silver, I. Ahmed, A. Berhorst, C. R. Ranero, R. K. Kelly, and E. R. Flueh (2007), The Nicaragua convergent margin: Seismic reflection imaging of the source of a tsunami earthquake, this volume.

McNally, K. C., and J. B. Minster (1981), Nonuniform seismic slip rates along the Middle America trench, J. Geophys. Res., 86, 4949–4959.

Miller, M., T. Melbourne, D. Johnson, and W. Sumner (2002), Periodic slow earthquakes from the Cascadia subduction zone, Science, 295(5564), 2423.

Moore, D. E., and D. A. Lockner (2007), Friction of the smectite clay montmorillonite: A review and interpretation of data, this volume.

Moore, J. C., C. Rowe, and F. Meneghini (2007), How accretionary prisms elucidate seismogenesis in subduction zones, this volume.

Morgan, J. K., E. B. Sunderland, and M. V. S. Ask (2007), Deformation and diagenesis at the Nankai subduction zone: Implications for sediment mechanics, décollement initiation and propagation, this volume.

Newman, A. V., S. Y. Schwartz, V. Gonzales, H. R. DeShon, J. M. Protti, and L. Dorman (2002), Along-strike variability in the seismogenic zone below Nicoya Peninsula, Costa Rica, Geophys. Res. Lett., 29(20), 1977, doi:10.1029/2002GL015409.

Norabuena, E., L. Leffler-Griffin, A. Mao, T. H. Dixon, S. Stein, I. S. Sacks, L. Ocola, and M. Ellis (1998), Space geodetic observations of Nazca-South America convergence across the central Andes, Science, 279(5349), 358–362.

Norabuena, E., et al. (2004), Geodetic and seismic constraints on some seismogenic zone processes in Costa Rica, J. Geophys. Res., 109, B11403, doi:10.1029/2003JB002931.

Oleskevich, D. A., R. D. Hyndman, and K. Wang (1999), The updip and downdip limits to great subduction earthquakes; thermal and structural models of Cascadia, south Alaska, SW Japan, and Chile, J. Geophys. Res., 104, 14,965–14,991.

Ozawa, S., M. Murakami, M. Kaidzu, T. Tada, T. Sagiya, Y. Hatanaka, H. Yarai, and T. Nishimura (2002), Detection and monitoring of ongoing aseismic slip in the Tokai region, central Japan, Science, 298(5595), 1009–1011.

Plafker, G. (1972), Alaskan earthquake of 1964 and Chilean earthquake of 1960: Implications for arc tectonics, J. Geophys. Res., 77, 901–925.

Sacks, I. S., S. Suyehiro, A. T. Linde, and J. A. Snoke (1978), Slow earthquakes and stress redistribution, Nature, 275(5681), 599–602.

Saffer, D. (2007), Pore pressure within underthrust sediments in subduction zones, this volume.

Saffer, D., and C. Marone (2003), Comparisons of smectite and illite-rich gouge, frictional properties: Applications to the updip limit of the seismogenic zone of subduction thrusts, Earth Planet. Sci. Lett., 215, 219–235.

Satake, K. (1993), Depth distribution of cosesmic slip along the Nankai trough, Japan, from joint inversion of geodetic and tsunami data, J. Geophys. Res., 98, 4553–4563.

Savage, J. C. (1983), A dislocation model of strain accumulation and release at a subduction zone, J. Geophys. Res., 88, 4984–4996.

Scholz, C. H. (1998), Earthquakes and friction laws, Nature, 391, 37–42.

Schwartz, S. Y., and H. R. DeShon (2007), Distinct up-dip limits to geodetic locking and micro-seismicity at the Northern Costa Rica seismogenic zone: Evidence for two mechanical transitions, this volume.

Seno, T. (2007), Collision versus subduction: The importance of slab dehydration, this volume.

Stein, S., and E. Okal (2005), Speed and size of the Sumatra earthquake, Nature, 434, 581–582.

Thatcher, W., and J. B. Rundle (1984), A viscoelastic coupling model for the cyclic deformation due to periodically repeated earthquakes at subduction zones, J. Geophys. Res., 89, 7631–7640.

Underwood, M. B. (2007), Sediment inputs to subduction zones: Why lithostratigraphy and clay mineralogy matter, this volume.

Wang, K. (2007), Elastic and viscoelastic models of crustal deformation in subduction earthquake cycles, this volume.

Wang, K., and T. H. Dixon (2004), Coupling semantics and science in earthquake research, Eos Trans. AGU, 85(18), 180.

Wang, K. H., R. D. Hyndman, and M. Yamano (1995), Thermal regime of the southwest Japan subduction zone, Tectonophysics 248, 53–69.

There have been great advances recently in characterizing and understanding earthquakes on subduction thrust faults; this paper discusses some of the many questions that remain. Important seismic characteristics of subduction thrust faults and their physical associations include the following: (1) The maximum thrust earthquake magnitude, Mx, is highly variable among subduction zones; Mx may be related to the downdip seismogenic width, i.e., up-dip and downdip rupture limits, or to the physical characteristics and stress on the fault. (2) The term seismic coupling, i.e., fraction of relative motion that is accommodated seismically, needs careful definition. Meaningful use of the term requires specification of the downdip seismogenic width. Some subduction zones appear to be completely locked, with no aseismic slip between megathrust events; others have mostly aseismic slip. (3) The term seismic asperity also needs careful definition; it commonly describes fault regions that had especially large slip in a great earthquake. However, inferences that such areas always have larger earthquake displacement and that they are associated with fault physical characteristics are not yet firmly established. (4) Subduction thrust faults are concluded to be weak. The commonly favored explanation is regionally elevated fluid pressures, but weak fault zone materials and dynamic rupture processes also have been proposed. (5) Most subduction thrusts have consistent updip and downdip seismogenic limits, i.e., an updip aseismic zone tens of kilometers wide commonly limited by a temperature of 100°–150°C. There is not yet agreement on the mechanism responsible. The downdip limit is frequently the intersection of the thrust with the fore-arc Moho, i.e., ~40 km for continent subduction, less for island arcs. However, deeper thrust events have been observed in some regions. For very hot subduction zones a critical seismogenic temperature limit of ~350°C is reached at a shallower depth. (6) The reflection character of subduction thrust faults appears to change from a usually strong negative reflection in the updip aseismic zone, to a thin, sharp but weaker interface in the seismic portion, to a broad shear zone for the deeper aseismic zone. (7) Displacements on subduction thrust faults occur over a range of speeds, from earthquake rupture (seconds), to rates that generate tsunamis (minutes), to slower slip seen only in geodetic data. The speed controls are still unclear. (8) Immediately downdip of the seismogenic zone, slip on the aseismic zone in some areas occurs in slow slip events lasting a few weeks to months with intervals of a year to a few years. There are associated seismic tremors with no clear onset.

Most of the world’s great earthquakes (M ≥ 8), many intermediate magnitude events, and most large tsunamis are generated by rupture on the seismogenic zone of subduction thrust faults (fig. 2.1). In this discussion, I outline some of important seismic characteristics of subduction thrust faults and their physical associations; what I think we know and what we don’t know. Most of what we know has come from remote observation. There are no boreholes as yet into the seismogenic portion of subduction thrusts, although there has been drilling through the updip aseismic portion by the Ocean Drilling Program [e.g., Moore et al., 2001, and reference therein]. There also has been limited drilling through active land faults, and there is important information from exhumed ancient subduction thrust faults [e.g., Heermance et al., 2003; Hashimoto et al., 2002, and references therein]. In the next few years we hope to have much new data from deep drilling by the Japanese research ship CHIKYU. Seafloor precision geodetic systems now being developed also should make a very important contribution. What we know about the behavior of subduction thrust seismogenic zones, and associations of that behavior with the physical composition and state of the thrust, comes mainly from (1) great and smaller thrust earthquakes, (2) land geodetic data, (3) ocean drilling and exhumed subduction thrusts, (4) seismic reflection and wide angle data, (5) thermal data and models, (6) electrical sounding and other geophysical data, and (7) stress indicators near the plate boundary.

Figure 2.1     The seismogenic zone of subduction thrust faults. These generate most M > 8 earthquakes and large tsunamis globally.

The seismic behavior of subduction thrust faults is highly variable, both regionally and locally. Some subduction zones generate events with a maximum magnitude of M ~ 7; others have great earthquakes of magnitude over M 9. Some subduction thrusts are largely aseismic between infrequent great earthquakes and are inferred to be fully locked. Some have very frequent but only small to intermediate magnitude earthquakes. Some subduction zones have very large seismic moment release; others have very little. Below, I discuss some of the subduction thrust observations and their physical explanations, along with some of the more important questions about the seismic behavior of subduction faults. I do not discuss the seismological characteristics of great earthquakes in any detail; there are a number of excellent reviews of that subject [cf. Kanamori, 1986, 1983; Nishenko, 1991; Pacheco and Sykes, 1992; Ruff, 1996] and the nature of the great earthquake cycle and associated elastic and viscoelastic deformation [Wang, this volume]. I focus on the physical associations on the thrust with the characteristics and variations of great and smaller thrust earthquakes.

Some subduction zones generate thrust earthquakes of magnitude greater than M 9, whereas others have maximum earthquake magnitudes of M ~ 7 (fig. 2.2). The largest historical earthquakes, M ~ 9, have occurred in the subduction zones of southern Chile, Cascadia, southern Alaska, the Kuril trench, and north Sumatra [e.g., Plafker, 1969, 1972; Abe and Kanamori, 1980]. In contrast, only earthquakes of magnitude less than 7 or 7.5 have been recorded for most of the southwest Pacific island arcs; the smallest maximum, Mx, appears to be for most of the Mariana-Izu-Bonin subduction zones [e.g., Pacheco and Sykes, 1992; Pacheco et al., 1993].

Since the magnitude of earthquakes increases systematically with the fault rupture area [e.g., Wells and Coppersmith, 1994], this difference means that subduction thrusts producing Mx < 7.5 earthquakes probably have seismic behavior only in small patches, at most a few tens of kilometers across, whereas M 9 earthquakes have seismic rupture that may be over areas of ~100 km downdip and ~1000 km along strike, such as Alaska 1964 [e.g., Plafker, 1969, 1971], Chile 1960 [e.g., Plafker and Savage, 1970], and northern Sumatra 2004 events. The conclusion that subduction thrusts with only Mx ~ 7 earthquakes have only small patches that are seismic is supported by their generally small seismic moment release rate; their seismic efficiency (seismic coupling, see below) is very small [e.g., Pacheco et al., 1993]. Most of the plate convergence is inferred to be accommodated aseismically. Although they may be very infrequent, the large events represent a much greater seismic moment release rate than the many M < 7 earthquakes. The subduction zones that have exhibited great M ~ 9 earthquakes are found to have seismic efficiency close to 1 [e.g., Pacheco et al., 1993]. Plate convergence is accommodated mainly seismically over a defined downdip width. For most of these great earthquake regions, land geodetic data also require almost complete thrust locking between earthquake events; Cascadia, southwest Japan, and Chile are examples [e.g., Hyndman and Wang, 1995; Hyndman et al., 1995; Brooks et al., 2003].

Figure 2.2    Some subduction thrusts produce M 9 earthquakes, others only less than M 7; what are the controls: maximum fault area; seismic coupling; seismic versus aseismic slip?

Seismic coupling is an expression that needs careful definition [Wang and Dixon, 2004]. It usually refers to the fraction of the plate convergence rate at a subduction zone that is accommodated in thrust earthquakes [e.g., Ruff and Kanamori, 1983; Peterson and Seno, 1984; Jarrard, 1986; Pacheco et al., 1993; McCaffrey, 1997]. The remainder is inferred to be accommodated by some form of aseismic slip. In a kinematic description, a locked or fully coupled fault has no or low slip between great earthquakes. It is important to recognize that this definition does not involve any inferences of stress condition or fault properties [Wang and Dixon, 2004]. Partial coupling usually refers to the fraction of plate convergence that is accommodated seismically. The aseismic motion may occur in post seismic transient slip, in steady creep motion, or in slow slip events (discussed below). In estimating the seismic component of plate convergence, it is important to recognize that some component of plate convergence may be accommodated by crustal shortening in the back arc and possibly the fore arc [e.g., Hindle et al., 2002; Mazzotti and Hyndman, 2002; Norabuena et al., 2004]. Also, transient slip must be considered, including tsunami earthquakes and post seismic slip of great events that are not included in the seismic moment [Wang, this volume].

A critical parameter in the calculation of seismic coupling that often is not emphasized is the downdip seismogenic width. The calculated coupling is inversely proportional to the assumed seismogenic width. A less ambiguous expression may be seismic efficiency, the fraction of relative convergence motion that is accommodated seismically over a defined downdip fault width. However, in the section below I will follow the common use of seismic coupling. It probably is preferable to use the less ambiguous quantity, the average seismic moment release rate per unit length of subduction zone, because the downdip seismogenic width is often poorly known. It is also important to recognize that subduction zones with only infrequent great earthquakes have poor statistical sampling. There may have been no such events in the historical record.

To calculate the seismic coupling, the seismic slip from the average seismic moment release rate is compared to the convergence rate to give a seismic coupling efficiency α [e.g., Pacheco et al., 1993, and references therein]. The coupling α = 1 if all convergence is accommodated in earthquakes over the defined downdip seismogenic width. However, it is important to recognize that the calculation of α involves a number of poorly know parameters, especially the critical variable, the downdip width of the seismogenic zone "W." The computed seismic coupling is inversely proportional to the choice of W. The width W has often been fixed, taking the updip limit near the trench and the downdip limit at depth 40–50 km (i.e., width of ~100 km) based on a common maximum rupture depth of great earthquakes [e.g., Pacheco et al., 1992; Tichelaar and Ruff, 1993]. In the discussion below I will use the expression apparent seismic coupling for use where the downdip seismogenic width is unconstrained and is assumed.

Three explanations have been proposed for the apparent variation in seismic coupling, one related to seismogenic fault area, one to stress state, and one to frictional conditions. The first is that the downdip width of the seismogenic or coupled zone is wide where there are great earthquakes and very narrow where there are only smaller thrust earthquakes [e.g., Hyndman et al., 1997]. Within the defined downdip width W, the thrust then is fully or almost fully seismically coupled in both cases (seismic efficiency of 1). The rupture length of great earthquakes along the margin is commonly 2 to 3 times the down-dip width [e.g., Tichelaar and Ruff, 1993] (there are important exceptions with very long ruptures along strike, such as Cascadia, southern Chile, and Sumatra) so the maximum magnitude decreases strongly with decreasing downdip width. The maximum downdip depth may be the most important variable (see below). This explanation is supported by a maximum earthquake depth of ~10 km for the subduction zones such as Mariana with Mx ~ 7, compared to 40–50 km for the subduction zones with great earthquakes such as southern Chile and Alaska, Mx ~ 9 [e.g., Pacheco et al., 1993; Zhang and Schwartz, 1992; Tichelaar and Ruff, 1993]. If, for subduction zones like Mariana, the updip and downdip seismogenic limits approximately coincide, only a few patches will be seismic. This explains, at least in part, the very small seismic moment release rate and is one limiting case explanation. The other limiting case is that thrust earthquakes are distributed over a broad width to a depth limit of ~40 km but that only a few patches are seismic, producing only intermediate maximum magnitude earthquakes (see discussion of Wang, this volume). The up-dip and downdip seismogenic limits generally are nearly constant along strike for each subduction zone segment where there are no significant changes in plate age, plate convergence rate, and plate dip. The limited local variation in these limits suggests that they are related to thrust physical conditions or state common to significant subduction zone lengths along strike. These limits are discussed in the next section.

The second explanation for variations in apparent seismic coupling and for subduction zone maximum earthquake magnitude focuses on differences in the stress regimes, for example between southwest Pacific arcs and east Pacific continental subduction zones [e.g., Scholz and Campos, 1995]. The former are inferred to be in extensional regimes with the arcs receding from the trenches and subducting slabs, and the latter in compression with the continents overriding the subducting slab [Hyndman, 1972; Uyeda and Kanamori, 1979]. Scholz and Campos [1995] argue that extensional regimes have only small earthquakes; in contrast, compressional regimes have large megathrust earthquakes. However, the west Pacific subduction zones of Kuril, northeast Japan, and southwest Japan all have had great earthquakes. Also, the apparent seismic coupling factor is high in the southwest Japan subduction zone and lower in the northeastern Japan subduction zone [Astiz et al., 1988], whereas the force interaction as reflected by upper plate stress is opposite [Wang and Suyehiro, 1999; Wang and Dixon, 2004]. For the areas with large earthquakes, subduction occurs beneath thick continental crust. A more consistent association seems to be that small maximum magnitudes are associated mainly with island arcs whereas great earthquakes are associated with subduction beneath continental crust (i.e., Andean type subduction). The latter association is at least partly explained above; the fore-arc crust in island arcs is thin so the fore-arc Moho and aseismic fore-arc mantle behavior are reached at a shallow depth.

The third explanation for variations in apparent seismic coupling is variations in fault properties, i.e., frictional state, or force required for rupture. An association that has not been well quantified is that very smooth incoming oceanic plates have infrequent but very large events (up to M 9), whereas rough incoming plates with seamounts, fracture zones, etc. that cause stress concentrations usually have smaller more frequent thrust earthquakes (M < 8) [e.g., Bilek et al., 2003]. Thrust irregularities such as seamount chains and aseismic ridges may especially limit the along-strike lengths of subduction thrust earthquakes and their character [e.g., Kodaira et al., 2002]. A related association is that great earthquakes commonly occur where there are large accretionary sedimentary prisms [e.g., Ruff, 1989]; the latter may smooth the subduction thrust such that there are few stress concentrations and strain can build up to very large ruptures. An example is the larger apparent coupling for the southwest Japan, Nankai trough, compared to the northeast Japan trench [e.g., Astiz et al., 1988; Pacheco et al., 1993]. Other examples of subduction zones with great thrust earthquakes and large accretionary prisms are southeastern Alaska, Cascadia, and southern Chile [e.g., Oleskevich et al., 1999]. However, large accretionary prisms only occur for continental subduction zones where there is a large source of sediment, so the main association of large earthquakes may be with subduction beneath continents rather than with accretionary prisms.

Seismic asperity is another expression that is used in a number of different ways and needs careful definition (see also Wang [this volume]). The simple and unambiguous use is for patches or regions within the overall area of rupture in a specific great earthquake that has especially large displacement [e.g., Lay and Kanamori, 1981; Lay et al., 1982]. The earthquake slip distribution may be mapped through seismic waveform modeling, and the updip region in a few cases by tsunami modeling [e.g., Johnson, 1999; Toshitaka et al., 2002]. Confusion comes with the extension of this definition to associations with spatial variations in the physical properties of the thrust interface (fig. 2.3). Asperities are often inferred to be stronger than the surrounding region of the thrust, and therefore they accommodate most of the plate convergence seismically, whereas adjacent areas have more aseismic slip (cf. reviews by Scholz [1990] and Ruff [1992]). Although a number of clear associations have been suggested between the patterns of subduction thrust seismic behavior and physical features on the incoming plate such as seamounts, aseismic ridges, and fracture zones [e.g., Bilek et al., 2003; Kodaira et al., 2002], only few associations of local variations in rupture displacement in great earthquakes with localized physical features on thrusts have been conclusively established [e.g., Igarashi et al., 2003; Zweck et al., 2002].

Figure 2.3    What is the nature of asperities that are inferred to be stronger patches on the thrust?

If there is a physical connection with areas of large rupture, we expect that there will be more slip in these asperity regions for all great earthquake ruptures that include them and that adjacent areas will always have more aseismic slip. This behavior is opposite to the concept of seismic gaps in which areas of little or no recent rupture displacement are expected to have greater rupture in future events [McCann et al., 1979; Wyss and Wiemer, 1999, and references therein]. In the latter case, the long-term average earthquake slip may be nearly constant along a subduction zone. In one model, there may be fully locked patches surrounded by regions that are freely slipping. A number of repetitions of great earthquakes in the same region are required to establish which view is more correct. There have been only a few reported repeat ruptures of the same area by historical great earthquakes [e.g., Wyss and Wiemer, 1999; Schwartz, 1999], and the first of the two repeated events is usually too old for the high-quality seismic data required to model the rupture distribution. New geodetic data may allow connection between areas on the thrust that are locked and large earthquake rupture asperities.

It has been recognized for many years that shallow angle thrust faults must be weak to accommodate large subhorizontal displacements [Hubbert and Rubey, 1959; Raleigh and Evernden, 1981; Davis et al., 1983]. Other arguments based on regional stress and force balance [e.g., Wang and He, 1999], and on the lack of thermal anomalies due to frictional heating, have been developed subsequently [e.g., Wang et al., 1995, and references therein] (fig. 2.4). On some margins the maximum principal stress is margin parallel rather than in the direction of convergence [e.g., Wang et al., 1995]. Margin-normal stress on several margins is determined to be less than or equal to the vertical stress [e.g., Wang and Suyehiro, 1999; Wang et al., 1995]. The conclusion of weak large faults also applies to continental strike-slip faults [e.g., Zoback et al., 1987; Hickman, 1991; Lachenbruch and McGarr, 1990]. The arguments for weak strike-slip faults comes first from the maximum horizontal principal stress approaching orthogonal (65°– 85°) to the fault as the fault is approached, suggesting that the fault is moving at very low levels of shear stress. Second, weak strike-slip faults are concluded from the lack of the thermal anomaly expected for motion on strong faults. The inferred strength of large faults is <20 MPa, similar to the stress relieved in large earthquakes (stress drop). In contrast, the overall strength of the brittle upper lithosphere is estimated to be much greater, >50–100 MPa, as predicted by friction laws [e.g., Hickman, 1991; Zoback and Townend, 2002].

The reasons for the weakness of large faults have been much discussed. The three main possibilities are (1) elevated fluid pressures [e.g., Magee and Zoback, 1993]; (2) low coefficients of friction due to the fault zone material, i.e., clay, serpentinite, silica gel [e.g., Vrolijk, 1990; Toro et al., 2004]; and (3) dynamic weakening, i.e., slip shear heating, propagation of dilation waves along the fault, and fluidization of fault-zone material [e.g., Brune, 1993; Shi et al., 1998; Brodsky and Kanamori, 2001; Ma et al., 2003; Melosh, 1996; Lachenbruch, 1980]. The elevated fluid-pressure explanation appears favored by most authors [see Davis et al., 1983], especially for subduction thrust faults, following the well-established acceptance of this explanation for the low angle of thrust faults in sedimentary fold and thrust belts [Hubbert and Rubey, 1959]. For subduction thrusts, there is an ample supply of fluid from the consolidation of underthrust material and from dehydration reactions in the downgoing slab. The fluid supply for continental transcurrent faults is less obvious, but some suggestions have been made [e.g., Kirby et al., 2002]. Secure confirmation awaits drilling of the seismogenic portion of subduction thrust faults by the International Ocean Drilling Program (IODP) drill ship CHIKYU and of active continental faults such as the San Andreas fault by the San Andreas Fault Observatory at Depth (SAFOD) project.

Figure 2.4    Subduction thrust faults are very weak based on regional stress, earthquake, and thermal data.

Both great subduction earthquakes and smaller thrust events usually do not extend to the trench (fig. 2.5); there is an updip aseismic zone commonly tens of kilometers wide (fig. 2.4) [e.g., Byrne et al., 1988; Byrne and Fisher, 1990]. This updip limit, or upper stability transition depth, is defined by (see summary by Oleskevich et al. [1999]) (1) the updip rupture limit in great earthquakes, as determined by waveform modeling; (2) the updip limit from modeling of the tsunamis generated by great earthquakes; (3) the updip limit of great earthquake aftershocks; and (4) the updip limit of small thrust earthquakes on the subduction thrust between great events. In several subduction zones the coast is close enough for land geodetic data to provide some constraint to the updip limit of the locked zones, although the resolution is low [e.g., Lundgren et al., 1999; Norabuena et al., 2004]. Seafloor geodetic measurements that are in progress by United States and Japanese groups should soon give additional information [e.g., Spiess et al., 1998]. All of these definitions are based on different measures and sometimes give different locations [e.g., Norabuena et al., 2004], but usually