Six large tsunamis struck the eastern Aleutians over the past 1,700 years, usually at intervals of about 300 years, an analysis of the geological record shows.

The findings are changing the way seismologists understand how earthquakes happen along some faults -- and how to assess the threat those faults pose.

A study led by the U.S. Geological Survey, published in the journal Geophysical Research Letters, adds evidence to support the idea that "creeping" subduction zones -- where geologic plates move gradually without striking each other -- are not as safe as had been assumed.

The study site, Sedanka Island, is near Unalaska and between the epicenters of major tsunami-generating earthquakes in 1946 and in 1957. The 1946 tsunami was particularly destructive, causing 165 deaths, 159 of them in Hawaii.

Those tsunamis left their mark on Sedanka Island -- as did the big tsunamis that preceded them.

The most telling evidence came from pits dug into the earth as far inland as half a mile, in an area facing the Pacific Ocean. Those pits revealed layers of ocean sand stacked between island soil deposits. Researchers determined the sand was deposited at intervals of about three centuries on average.

The sand was identified as clearly oceanic, holding marine remains of diatoms and settling in the typical pattern created by tsunami-driven seawater surges, with finer grains resting above coarser grains, said lead author Rob Witter, a USGS research geologist. Researchers established a timeline by carbon-dating material from the layers of earth between sand deposits, and -- for the soil layer between the 1946 and 1957 events -- from measuring Cesium-137 levels. The presence of Cesium-137 in soil deposits is linked to atomic-weapons testing conducted around the planet in the mid-20th century.

While it was what they found in the pits that created the picture of past events, it was clues found above ground -- deposits of driftwood carried far inland and high on ridges -- that first prompted researchers to excavate on Sedanka Island.

For seismologists, the findings make the already uncertain business of predicting earthquake risk still more unpredictable.

In the past, scientists believed that only locked plates posed earthquake risks because some kind of clash between plates was needed to create the quakes. Conventional wisdom held that creeping faults, where plates move along each other without that friction, were safer, Witter said.

But that model may be wrong.

It is possible that such earthquakes -- and the destructive tsunamis that follow them -- could happen along creeping subduction zones such as the eastern Aleutians, Witter said. One theory is that subduction zones can switch between gradual creep and pressure-building lock, he said.

"Just because a fault is creeping today doesn't mean it wasn't locked in the past," Witter said. "It seems to be that if you're in the right place at the right time, this locking can turn on and off."

He cited research conducted by Jeff Freymueller at the University of Alaska Fairbanks, who works to track the movements of tectonic plates.

In a new paper, published online this month in the Journal of Geophysical Research, Freymueller and his colleagues analyze GPS data to show how subduction plate slips in the Cook Inlet region started and stopped abruptly.

Another possibility, Witter said, is that the creep of subduction zones releases only some of the geologic pressure along a fault line, meaning that accumulated pressure that remains could force a tsunami-triggering earthquake.

Subduction zones are areas where tectonic plates slide beneath one another. The Aleutian Islands -- and much of coastal Alaska -- lie over a major zone, the site where the Pacific plate moves beneath the North American plate. When those movements are sudden and forceful, the result can be major earthquakes, such as the 1964 Alaska earthquake, one of the most powerful ever recorded.

In subduction zone areas where the plates creep, however, the movements are continual and not sudden, and ground movements can be measured over time. Sedanka Island, where gradual ground movement has been measured, is believed to lie in such an area, in a gap between areas where the plates rub against each other.

Creeping subduction zones or faults are not unique to Alaska. One of the most famous is the Hayward Fault in California, which moves from year to year and has displaced the stadium at the University of California at Berkeley, Witter said.

Correction: This story originally misidentified Jeff Freymueller as John Freymueller. It also originally implied the Hayward Fault was a subduction zone; clarifying language has been added to this story.