Stanford University News Service



CONTACT: Stanford University News Service (650) 723-2558

Geophysicists find rumblings at start of earthquakes

STANFORD -- An earthquake has to start somewhere.

For years, geophysicists have wondered how an earthquake fault can sit still for decades, only to suddenly break into an earthquake, with rocks on each side of the fault slipping past each other at meters- per-second rates. Laboratory studies and theoretical models strongly suggest that there ought to be some sort of gradual beginning or "nucleation phase" before an earthquake, that a small rupture or a slow creep along the fault should precede that dramatic slip.

However, until now geophysicists have not found strong evidence for nucleation in nature. On their seismographs and other detection instruments, an earthquake looks like a sudden shake that seems to start without warning.

Now two seismologists at the U.S. Geological Survey (USGS) and Stanford University have taken a closer look at the seismograms recorded very close to 30 earthquakes around the world, ranging in magnitude from 2.6 to 8.1. They see a weak irregular rumbling, at the beginning of the measurable shaking for each quake, that may be the signature of the last seconds of the nucleation phase -- just as it grows powerful enough to trigger the full, fast-growing earthquake.

They also see differences in this seismic signature depending on the size of the eventual earthquake. As scientists study the evidence for earthquake nucleation more closely, those differences may shed light on two possible explanations of how earthquakes start. Depending on which theory turns out to be supported by the data, there may -- or may not -- be a physical basis to judge ahead of time which earth tremors are going to grow into big ones.

William Ellsworth, a geophysicist at the USGS in Menlo Park and a consulting professor at Stanford, and Gregory Beroza, associate professor of geophysics at Stanford, report their findings in the May 12 issue of the journal Science.

"The new data has spurred us and others to look at what starts a fault that has been stuck for a very long time, and moves it into high gear," said Ellsworth. We think this is a window into the physics of the whole process."

"In seismology, the data drives the theory," Beroza said. He expects research over the next several years to show which hypothesis about earthquake nucleation is closer to the truth.

Two concepts of nucleation

Because most earthquakes begin with a bang, it has been thought that nucleation might involve only a tiny area on the fault. A rupture as small as a dime, deep in the earth, could trigger a larger rupture, which triggers a larger rupture, until the earthquake is large enough to measure on seismographs -- and then large enough to shake everything on the surface.

In this hypothesis, the cascade model, there is no difference between the way a small earthquake and a large, destructive one start. At the beginning, in the nucleation phase, the eventual outcome cannot be foretold.

"As you can imagine, if this model turns out to be correct, that makes the prospect for short-term earthquake prediction pretty bleak," Beroza said.

So far, the data collected by Ellsworth and Beroza is consistent with the cascade hypothesis. But it also is consistent with another explanation of earthquake nucleation, the pre-slip model.

In the pre-slip model, the earthquake starts when the two surfaces straining against each other on the fault line begin to slip in a small region, perhaps a few yards or a few miles across. This quiet slipping may continue for weeks or months, at a level too small to detect with seismic waves, until the fault is weakened and starts its sudden, rapid break. In this model, the larger the area of pre-slip, the bigger the eventual earthquake will be.

Beroza and Ellsworth's data show that the weak seismic nucleation interval at the beginning of each earthquake lasts for a longer time when the earthquake is large. This scaling relationship is consistent with the pre-slip hypothesis. If that model is correct, and earthquakes are preceded by a nucleation phase that happens very slowly, then it may be possible to design instruments that could detect the precursory nucleation phase itself.

"If the line of investigations that we've begun pan out, we'll know whether earthquakes are predictable in principle," Beroza said. "Whether we'll ever be able to predict them in practice is a different, and much more difficult, question."

An anomaly at Loma Prieta

Ellsworth and Beroza's discovery is due partly to some improvements in the instruments that monitor earthquakes, and partly to their decision to look at a pattern in seismic data that seismologists had seen before but ignored.

Beroza said that he was studying the Loma Prieta earthquake, which struck the San Francisco Bay Area in 1989, when he noticed an anomaly. An earthquake is expected to grow steadily, from magnitude 3 to magnitude 4 to 5 and so on, until limiting factors slow it down. However, the Loma Prieta earthquake started with irregular shaking that did not increase much for the first one and a half seconds -- then it grew rapidly to magnitude 6.9.

Ellsworth had noticed similar anomalies, as long ago as the earthquakes near Hollister and Coalinga, Calif., in the 1970s and 1980s. "We realized that there were subtle things happening at the beginning of earthquakes, but the instrumentation at the time was not good enough to study them," he said. Even now, only a few seismic instrument stations have equipment to record both the magnitude of an earthquake and the motion of the quake from its measurable beginning.

Several such stations recorded the motions of the Loma Prieta quake. Because it was only one instance, Ellsworth and Beroza initially discounted the importance of the anomaly. Then they noticed it twice again -- when studying seismograms from the June 1992 Landers, Calif., earthquake and from the January 1994 earthquake in Northridge, Calif.

They began collecting data recorded near the epicenters of 30 earthquakes, from the magnitude 8.1 Michoacan earthquake, which damaged Mexico City in 1985, to a magnitude 2.6 quake recorded in a 2- kilometer-deep borehole drilled into the earth in Long Valley, Calif. They magnified the seismic signals to analyze the first few seconds of each earthquake and found the same irregular, limited shaking at the start, followed by rapid growth in power as the quake began its normal motion.

The researchers emphasize that much more data must be gathered to prove the seismic nucleation hypothesis and to show whether the cascade or the pre-slip model -- or some other model yet to be proposed -- is correct. So far, most of the data they have collected comes from a single recording station for each earthquake studied. To expand that information, their research groups are analyzing recordings from seismic arrays in Taiwan, Japan and California, and from earthquake foreshock sequences such as those that preceded the Landers and Joshua Tree earthquakes in Southern California.

In addition, Ellsworth is co-principal investigator with Stanford geophysicist Mark Zoback on a project to drill a deep borehole right into the San Andreas Fault. "There is a limitation to seismic studies done on the surface," Ellsworth said. "Even if we trap an earthquake with a dense network of instruments, we're only looking at the process once the fault begins to slip. To find out about nucleation processes, we must get our instruments down to the fault itself, where all the action is happening."



This is an archived release.

This release is not available in any other form. Images mentioned in this release are not available online.
Stanford News Service has an extensive library of images, some of which may be available to you online. Direct your request by EMail to

© Stanford University. All Rights Reserved. Stanford, CA 94305. (650) 723-2300.