Great Earthquakes

Figure 1. Seismic coupling influences the dynamical, geological, and seismological setting of a subduction zone. As described first by Uyeda and Kanamori [1979], most of Earth's subduction zones lie within a range described by two extreme types of subduction. Seismically coupled subduction zones (top) typically exhibit great thrust earthquakes and back-arc compression. The seismic coupling is caused by compressive interaction between the subducting and overriding plates. Seismically decoupled subduction zones (bottom) do not produce great earthquakes and exhibit back arc extension in an overriding plate that moves away from the trench.

Figure 2. The ten subduction zones defined in this study. Arrow lengths represent the maximum pull force that the upper mantle slab can exert on the subducting plate (the excess weight of the upper mantle slab).

Figure 3. Comparison of (a) observed plate motions with (b) plate motions predicted by a model in which all upper mantle slabs pull on subducting plates with their maximum pull force and with (c) plate motions predicted by a model in which some slabs do not pull with their maximum pull force. In the case of (c), we use a genetic algorithm to invert for the best-fitting pull force at each subduction zone (Fig. 4), which improves the fit to plate motions significantly. Here colors and arrow lengths represent the the plate speed relative to the global average.

Figure 4. The fraction of the maximum pull force that operates at each subduction zone in order to generate a predicted plate velocity field (Fig. 3c) that best fits observed plate motions (Fig. 3a). We do not use the Philippine slab in this study because the Philippine plate motion is poorly constrained.

Figure 5. Correlation of the fraction of the maximum possible slab pull force with the moment of the largest earthquake that has occurred at each subduction zone (expressed as moment per unit length). Here we show that great earthquakes are more likely to occur at subduction zones in which the slab pull force is weak.

Figure 6. Correlation of the fraction of the maximum possible slab pull force with back-arc stress state of each subduction zone. Here we show that the slab pull force is stronger at subduction zones that exhibit back-arc extension. By contrast, back-arc compression is correlated to a weak slab pull force.

Figure 7. Correlation of the fraction of the maximum possible slab pull force with the trenchward velocity of the overriding plate. The slab pull force is stronger if the overriding plate is moving away from the subduction zone.

Figure 8. A new model for subduction zone dynamics showing the two end-member styles of subduction that explain the relationships outlined here (Figs. 5-7). If the motion of the overriding plate is toward the subduction zone, as in (a), the overriding plate is driven into the subducting slab. This compressional environment results in back-arc compression and strong seismic coupling that is expressed by the release of great earthquakes. We find that slab pull at these subduction zones is small, indicating that the slab is weakened by the compressional stresses that are exerted on it by the overriding plate. Alternatively, if the motion of the overriding plate is away from the subduction zone, as in (b), the seismic coupling between the slab and the overriding plate will be small, resulting in back-arc extension and only moderate-sized earthquakes. Because the slab moves smoothly into the mantle, slabs of this type are well attached to subducting plates, which allows effective transmission of the slab pull force.

Great Earthquakes, Slab Weakening, and Slab Pull

C.P. Conrad, S. Bilek, and C. Lithgow-Bertelloni, "Great earthquakes and slab pull: interaction between seismic coupling and plate-slab coupling," Earth and Planetary Science Letters, 218, 109-122, 2004. [abstract] [online version] [reprint]

Some subduction zones, such as those in Alaska, Chile, or Kamchatka, release great earthquakes (magnitude 9.0 and above) and exhibit compression in the back-arc (Fig. 1, top). Such subduction zones are considered "seismically coupled". On the other hand, "seismically uncoupled" subduction zones, such as Tonga or the Marianas, do not release earthquakes larger than about magnitude 7.5 earthquakes and exhibit back-arc extension (Fig. 1, bottom). Whether or not a subduction zone is coupled seismically has been shown to depend on the motion of the overriding plate (see Fig. 1). Certainly the motion overriding plates, and thus degree of seismic coupling, depends on the forces that driving plate motion. We looked for variations in the tectonic forces operating at different subduction zones in order to explain the observed variations in seismic coupling.

We examined ten different tectonically-defined subduction zones (Fig. 2). We have already shown (see plate-driving forces) that plate motions are best described by a model in which lower mantle slabs excite mantle flow that pushes on the base of plates ("slab suction") while upper mantle slabs pull directly on subducting plates ("slab pull"). This model produces plate motions (Fig 3b) in which subducting plates move about 4 times faster than overriding plates, as is observed (Fig. 3a). However, this model does not predict the directions of some plates accurately. For example, the Pacific plate move too rapidly northward, and the Nazca plate rotates into South America while South America itself remains too stationary. We found that we could obtain a better prediction for plate motions (Fig. 3c) if we allowed slabs (Fig. 4) at some subduction zones to only pull weakly on surface plates. The Chilean, Aleutian, and Northwest Pacific (Japan, Kurile, Kamchatka) slabs, in particular, must exert a weak pull force. Thus it seems that some slabs must be only weakly attached to their surface plates.

We can infer a physical mechanism for the apparent detachment of some slabs by examining correlations between the degree of plate-slab attachment (Fig. 4) and geological or geophysical observables at these subduction zones. For example, the incidence of great earthquakes (Fig. 5) seems to be inversely correlated with slab pull force. There is also a correlation with back-arc stresses, with decreased plate-slab attachment associated with back-arc compression (Fig. 6). Finally, decreased plate-slab attachment is also associated with faster trenchward motion of the overriding plate (Fig. 7). These patterns indicate that slabs are detached from subducting plates at seismically coupled subduction zones, while the slab pull force is strong for seismically decoupled subduction zones.

Detachment of slabs from from subducting plates might be expected if the frictional interaction associated with seismic coupling weakens the slab (Fig. 8a). In this case, the slab's weight is no longer supported from above by the subducting plate, and it falls under its own weight. This drives a circulation in the upper mantle that tends to push the overriding plate toward the subduction zones, thus increasing the compressional interaction at the subduction zone. This suggests a possible feedback mechanism (Fig. 8a, right side) that leads to increased seismic coupling and decreased slab pull at some subduction zones. Such subduction zones are characterized by back-arc compression and periodically release great earthquakes (Fig. 8a). By contrast, the slab pull force is strong at seismically uncoupled subduction zones (Fig. 8b).

We expect increased frictional interaction at a subduction zone to weaken the slab. Many rocks weaken when increased stresses are applied to them. If seismic coupling at a subduction zone increases, the feedback mechanism described in Fig. 8 could rapidly weaken the slab and diminish the slab pull force. This could dramatically change the character of the plate boundary (see release of methane hydrates) and could also rapidly change the motion of the subducting plate. If the Alaskan subduction zone became seismically coupled about 45 million years ago, the northward motion of the Pacific plate would have slowed. This mechanism could explain the apparent change in plate motion associated with the bend in the Hawaii-Emperor seamount chain.