Many of the world's largest earthquakes take place at subduction zones where tectonic plates collide, forcing one under the other. When pressure builds and the plates slip past one another, an earthquake results. The severity of that earthquake depends on how far and how widely each plate slips. In most cases, the plates slide past one another just a few inches every year. But sometimes they slide tens of feet, along hundreds of miles. New research points to sediment deposits as the difference between the two types of earthquakes.

Christopher Fuller and his colleagues at the University of Washington and at Yale University designed a mathematical model of a subduction zone that moved at a rate of 31 miles every million years. They then ran two simulations: one in which sediment did not accumulate on top of the subduction zone and one in which it did. In the model lacking sediment, a dip or basin formed where one plate slipped under the other and a bump rose at the front of the top plate. In addition, stress built up in the rock like snow in front of a plow. But in the other model, sediment evened out the depression and prevented the stress from forming a wavelike pattern in the plate.

"The sediment will stop the deformation of the upper plate," Fuller explains. "The simplest way to think of it is that the increased weight of the sediment stops the deformation from occurring."

The deformation, however, serves a purpose, allowing the plates to slip past one another in small steps. With the weight of the sediment smoothing the subduction zone, the plates slip less often. But when they do, it is more violent, according to Fuller. In effect, the sediment reinforces the position of the upper plate in the short term but leads to bigger earthquakes when the plate does move in the long term.

If correct, the theory could help predict where in the 32,000 miles of subduction zones riddling the earth major earthquakes are most likely to occur. For example, the Cascadia subduction zone off the northwestern coast of the U.S. has sediments as deep as 1.5 miles in several such basins. But the researchers caution that such sediment-filled basins vary from place to place and that the model may not be widely applicable. "You have to understand the nature of basins and how they work in each area before you can use them as an interpretive tool," Fuller notes. "You can't just apply these correlations everywhere." The research appears in the February issue of Geology.