Some of the largest and deadliest earthquakes in recent years hit where earthquake hazard estimates didn't predict massive quakes.
A detailed computer model of large earthquakes in Japan and Taiwan helps explain why.
Contrary to decades of geologic wisdom, creeping fault segments, thought to slide smoothly past one another, can suddenly switch to a different mode during an earthquake, the model finds. Instead of acting as a barrier to rupture, the creeping segment loses its resistance, most likely through frictional heating.
"These segments are where earthquakes would tend to die," said Nadia Lapusta, a professor of geophysics and mechanical engineering at Caltech. "But during an earthquake, you generate frictional heat, just like when you're rubbing your hands, and the properties of the fault can change," she told OurAmazingPlanet.
The study showing the potential scenario in which creeping fault segments rupture was published online Wednesday by the journal Nature.
In geologic parlance, creeping faults are rate-strengthening. The arrival of earthquake slip (or the offset along a fault) increases friction between the fault's two sides, causing them to stick together and stopping a speeding quake from passing through. [The 10 Biggest Earthquakes in History]
But with sudden heating, such as the friction from an earthquake, fluids in the pulverized minerals lining the fault zone can switch a creeping fault to rate-weakening, Lapusta said. This means the earthquake slip dynamically weakens the fault as it moves along its fracture. And when faults act in rate-weakening mode, they generate earthquakes.
The model helps explain puzzling observations from the magnitude-9.0 Tohoku earthquake that struck Japan in 2011, as well as the 1999 Chi Chi quake in Taiwan, Lapusta said.
Japan's devastating tsunami was triggered by the fault's extremely large offset — about 165 feet (50 meters) at shallow depths on the subduction zone where the earthquake struck on March 11, 2011. A subduction zone is where two of Earth's tectonic plates meet and one slides beneath the other.
But there was less fault offset deeper in the subduction zone. This deep segment also produced higher-frequency shaking, and reached its maximum energy release faster than the shallow segment.
The time delay between the deep and shallow fault segments was due to the earthquake slip forcing its way into a creeping fault, the model suggests. The earthquake had to bash its way through via frictional heating.
"When the earthquake penetrated this area in our model, it started dying, but then it survived because of dynamic weakening. It's just like we observed in Tohoku," Lapusta said. "It's not proof, but it's an indirect confirmation that this model is what may have happened."
The new model suggests that creeping fault segments have lulled some scientists into a false sense of security.
"We have found a plausible physical explanation as to how these stable segments can support large seismic events, so the seismic hazards may be larger in some areas than anticipated," said Lapusta, who created the model with colleague Hiroyuki Noda of the Japan Agency for Marine-Earth Science and Technology in Yokohama.
The brief earthquake record is partly to blame for the oversight: seismic monitoring goes back only a century. Seeking older evidence requires digging trenches in the ground or drilling boreholes in the ocean floor, where layers of sediment preserve hints of past temblors.
But bad assumptions may also be at fault. In previous decades, researchers generally assumed faults had characteristic, repeatable earthquakes whose size was determined by the velocity plate of Earth's tectonic plates as they smash into one another. In Parkfield, Calif., where two of Earth's tectonic plates slide past one another along the San Andreas Fault, scientists determined a segment of the San Andreas there experienced earthquakes at an average of every 22 years. So they wired up the region with monitoring equipment and waited five, 10, then 20 years before Parkfield finally had its quake in 2004.
The idea of consistent behavior for faults is appealing, because it makes forecasting earthquakes much easier. But it can lead to underestimates of earthquake hazard, geologist Chris Goldfinger of Oregon State University writes in the Jan. 7 issue of Earth magazine.
Experts had predicted the largest likely quake for the Tohoku region was a magnitude 8.4. While still a massive earthquake, that is eight times weaker than the quake that did strike. The estimate was based on the past 100 years of quakes in the area and studies of current seismic strain, which come from GPS measurements. Few researchers accounted for prehistoric tsunami records, such as the 869 Jogan superquake, which produced a tsunami as large as Tohoku's. [In Pictures: Japan Earthquake & Tsunami]
Superquakes like those in Japan (and the 2004 Sumatra quake), and recent studies of sediment records of past earthquakes in Washington and California, reveal that earthquake patterns vary on long-term cycles.
Some earthquakes may be smaller, and strike more often than expected. And massive earthquakes, bigger than evidence predicts for a particular fault, are also possible.
Lapusta and Noda's model shows one way these massive, unexpected earthquakes can occur.
"It does stop and make you think," said Paul Segall, a Stanford professor of geophysics who was not involved in the study. "This idea has been on my mind for a while, that you could have dynamic weakening into these areas that are nominally stable, and they've done fabulous work," he told OurAmazingPlanet, referring to Lapusta and her team. "They're the first people to do careful, detailed calculations to show that this could happen."
The big question is whether what happened at Tohoku can repeat at other creeping faults, such as the Cascadia subduction zone off the coast of Washington and Oregon, and the middle section of the San Andreas Fault in California, Segall said. An earthquake that included the creeping sections of either fault would be a nightmare scenario for the West Coast.
"I think the only way we’re going to answer this is to look at the geologic evidence of past slip," Segall said.
Look to the past
Historical accounts collected by geologist Kerry Sieh of the Earth Observatory Institute of Singapore suggest the 1857 Fort Tejon earthquake, the last great San Andreas Fault earthquake in southern California, ruptured through the creeping San Andreas segment.
The San Andreas Fault varied its slip from earthquake to earthquake in the Carrizo Plain, near the creeping segment, geologists Nathan Toke of Utah Valley University and Ramon Arrowsmith of Arizona State University recently discovered. This hints that the fault may follow a variable cycle. They are now looking for evidence of past earthquakes on the creeping segment.
On the Cascadia subduction zone, sediments on the ocean floor show an earthquake in 1700 was the most recent in the area, but not the largest. The largest event hit about 5,800 years ago, and may have had three times the energy of the 1700 shaker, found Goldfinger and colleagues from Oregon State University.
As researchers collect more data about past fault behavior and the properties of faults, modelers will get better at predicting fault behavior, Lapusta said.
"As we keep exploring, we can put these measurements into models like ours and keep exploring what happens," Lapusta said.
Reach Becky Oskin at . Follow her on Twitter . Follow OurAmazingPlanet on Twitter . We're also on and.