When Seamounts Subduct

See allHide authors and affiliations

Science  29 Aug 2008:
Vol. 321, Issue 5893, pp. 1165-1166
DOI: 10.1126/science.1162868

Volcanoes on the sea floor of ocean basins—called seamounts—migrate with the ocean plates as they subduct beneath continental plates. This process creates shear interfaces called subduction zones, where most of the world's earthquakes nucleate. It has been proposed that scraping a subducted seamount from the oceanic plate nucleates great subduction-zone earthquakes (magnitude 8 or above) (1). However, at crustal depths below 10 km, where great earthquakes nucleate, ship-based seismic techniques cannot image subducted seamounts. On page 1194 of this issue, Mochizuki et al. (2) use an array of seismometers on the sea floor to investigate these issues. They show that seamounts provide an opportunity to investigate causes for a transition from stable to the unstable slip that nucleates earthquakes and find a clear beginning limit of seismogenic behavior.

Numerous seamounts with heights of 2 to 3 km and basal widths of 20 to 50 km exist on oceanic plates that migrate toward continents. The converging plates meet at deep ocean trenches, where the ocean plate carrying the seamounts bends downward into trenches to subduct beneath the continental plate. When high seamounts collide with the wedge-shaped continental margin, they first plow open the thin apex of weak material, creating an embayment in the landward slope of the trench (see the figure). As the colliding seamount plows into an increasingly thick part of the continental wedge, the entire seamount tunnels beneath the continental framework. Insertion of the seamount produces a broad bulge in the overlying sea floor; collapse of the trailing flank layers sends debris slides toward the trench (see the figure). Removal of collapse debris produces a furrow in the sea floor for distances proportional to the seamount's height.

Seamounts off the central Costa Rica continental margin.

Seamounts in the Pacific Basin (five of which can be seen in the lower part of the image) typically have diameters of ∼20 km and heights of 2 to 2.5 km. At this location, the oceanic and continental plates converge at a rate of 90 km per million years. As the ocean crust flexes into the 4.5-km-deep Middle America Trench (middle), bend faults form the stepped topography of the trench axis. On the trench slope are two circular bulges above subducted seamounts. Across the seaward slopes of the bulges and down slope are furrows from slides as the sea floor steepens seaward.


Initially, the seamount collides with weak rock of the continental wedge apex, and plate interface friction is relatively low. Rock in the apex of the upper plate wedge contains ∼30% pore fluid pressured by the overburden weight that reduces subduction zone friction. Therefore, subduction produces few recordable earthquakes until fluid drains to 10 to 15% and the continental wedge is thick enough to accumulate the elastic strain released in earthquakes (3). Earthquakes of magnitude ∼3 or above can be recorded at stations on land. But with only distant land station records, the precise location of these offshore earthquakes cannot be determined.

Sea-floor seismic records indicate deep anomalous features along subduction zones that are associated with aftershock clusters beneath the shelf (46). However, it will be difficult to prove that seamounts nucleate these earthquakes without understanding the mechanism through which they do so. Mochizuki et al. now show that with two-dimensional data from an array of sea-floor seismometers, a subducted seamount at 10 km depth along the subduction zone can be outlined as a diffuse bump on the subducting plate. Leaving the array above the seamount for extended periods to record local earthquakes provides sufficient precision to resolve the relation between seismicity and the seamount. Surprisingly, seismicity around the studied seamount is concentrated in front of its leading flank, rather than over its crest. These data imply that friction over the seamount is less than in adjacent deeper areas. They also indicate a steady or stable sliding over the seamount, whereas the subduction zone in front of the seamount slides intermittently during earthquakes (referred to as unstable sliding).

Recent observations are consistent with the inferred low friction. In a study of a subducted Costa Rican seamount (see the figure), Sahling et al. found large volumes of fluid vent from sediment layers exposed by trailing flank collapse (7). The strata ramped upward over the subducting seamount will create a hydraulic gradient up its flanks, which will concentrate fluid above its crest and thus reduce friction. This can help explain the distribution of friction off Japan found by Mochizuki et al.

Whether scraping seamounts from the subducting plate produces great earthquakes is still speculative (1). Mochizuki et al. examined a seamount subducted to a depth where earthquakes first nucleate, so their experiment does not answer this question. Subducted seamounts at depths of 20 km are proposed to uplift the coast of Costa Rica (8), so they remain attached at least to these depths. Some detached fossil seamounts are exposed in outcrops on land, although a graveyard of many detached fossil seamounts is not commonly recognized in outcrops on land. The low friction indicated by Mochizuki et al. is consistent with seamounts remaining attached in shallow regions of the seismogenic zone. Perhaps lower friction at the beginning of seismogenesis increases deeper in the subduction zone to detach subducting relief. Although detachment must sometimes occur, its relation to great earthquakes remains unresolved.

Recording a grid of signals from a surface ship (commonly two or more intersecting lines of shots are recorded) could provide the required three-dimensional seismic coverage. Three-dimensional data can also be acquired from an array of seismometers in a drill hole, yielding vertical seismic profiles. From such data, physical properties in subduction zones can be derived (9). Such data will help to elucidate whether frictional behavior changes are a result of physical relief or changes in the physical properties of fault materials.


  1. 1.
  2. 2.
  3. 3.
  4. 4.
  5. 5.
  6. 6.
  7. 7.
  8. 8.
  9. 9.
View Abstract

Stay Connected to Science

Navigate This Article