Twenty-three hundred years ago, hordes of mice, snakes, and insects fled the Greek city of Helike on the Gulf of Corinth (map). "After these creatures departed, an earthquake occurred in the night," wrote the ancient Roman writer Claudius Aelianus. "The city subsided; an immense wave flooded and Helike disappeared."

Since then, generations of scientists and folklorists have used a dizzying array of methods to attempt to predict earthquakes. Animal behavior, changes in the weather, and seismograms have all fallen short. (Watch: home video footage and the science of earthquakes.)

The dream is to be able to forecast earthquakes like we now predict the weather. Even a few minutes' warning would be enough for people to move away from walls or ceilings that might collapse or for nuclear plants and other critical facilities to be shut down safely in advance of the temblor. And if accurate predictions could be made a few days in advance, any necessary evacuations could be planned, much as is done today for hurricanes.

Scientists first turned to seismology as a predictive tool, hoping to find patterns of foreshocks that might indicate that a fault is about to slip. But nobody has been able to reliably distinguish between the waves of energy that herald a great earthquake and harmless rumblings.

Seismologists just can't give a simple yes or no answer to the question of whether we're about to have a large earthquake, said Thomas Jordan, director of the University of Southern California's Southern California Earthquake Center at a meeting of the American Geophysical Union (AGU) in San Francisco in December.

So some scientists have turned their attention to other signals, including electricity, that might be related to activity occurring below ground as a fault prepares to slip. (How to build earthquake-ready homes—cheaply.)

Like Underground Lightning

One theory is that when an earthquake looms, the rock "goes through a strange change," producing intense electrical currents, says Tom Bleier, a satellite engineer with QuakeFinder, a project funded by his parent company, Stellar Solutions, of Cambridge, Massachusetts.

"These currents are huge," Bleier said at the AGU meeting. "They're on the order of 100,000 amperes for a magnitude 6 earthquake and a million amperes for a magnitude 7. It's almost like lightning, underground."

To measure those currents, Bleier's team has spent millions of dollars putting out magnetometers along fault lines in California, Peru, Taiwan, and Greece. The instruments are sensitive enough to detect magnetic pulses from electrical discharges up to 10 miles (16 kilometers) away.

"In a typical day along the San Andreas fault [in California], you might see ten pulses per day," he told National Geographic News. "The fault is always moving, grinding, snapping, and crackling."

Before a large earthquake, that background level of static-electricity discharges should rise sharply, Bleier said.

And that is indeed what he claims he's seen prior to the half dozen magnitude 5 and 6 earthquakes whose precursors he's been able to monitor.

"It goes up to maybe 150 or 200 pulses a day," he said.

The number of pulses, he added, seems to surge about two weeks before the earthquake then drop back to background level until shortly before the fault slips. "That's the pattern we're looking for," he said.

False Alarms

But magnetic pulses could be caused by a lot of other things, ranging from random events within the Earth to lightning, solar flares, and electrical interference from highway equipment, lawn mowers, or even a nearby farmers' tractor engine. And that's not the only thing that can interfere with sensitive electrical equipment. "Spiders got inside of our instruments once, so we had to put screens in front of it," Bleier remembers.

Bleier also saw that charged particles called ions produced from currents deep within the Earth eventually migrate to the surface. "So we added a negative ion sensor and a positive ion sensor," he said.

And because rainy weather can also produce spikes in ion concentrations, his team also added humidity sensors to help rule out this possible cause of false alarms.

Finally, he noticed that when the ions reach the air, the positive and negative charges neutralize. This produces a burst of infrared radiation that can fool weather satellites into thinking the ground near the fault is warming up, even when ground-based weather stations say it isn't. This is easily visible by GOES weather satellites, he says.

"If all of these things happen, then we think there's going to be an earthquake greater than magnitude 5 about two days after," he said.

His team hasn't yet monitored enough large earthquakes for him to be sure that what he's found is valid for all quakes. "But the patterns look really interesting," he said.

But he does feel they have enough good clues to move ahead. Starting in January, his team will try to start making forecasts. "Instead of looking backwards in time, we're going to start looking forwards," he said.

Other scientists are contributing laboratory analysis to support the magnetic field theory. Robert Dahlgren, an electrical engineer at the SETI Institute, has spent 16 months working with other scientists to squeeze rocks under high pressures to see if they produce electrical currents.

He confirmed that dry rocks indeed produce pressure-dependent current and voltage signals. But he found no discharges from rocks soaked in the type of brine found at earthquake epicenter depths, presumably because the salty brine short circuits the current.

What does this say for earthquake prediction? He has no idea. "I'm the instrument guy," he says. But he notes that the signals he measured in the laboratory could indeed generate magnetic fields under the right conditions.

It's a very painstaking type of research. "It takes a year to prepare the brine-saturated rock samples," he said. "It's like breeding elephants. It takes a long time to get results."

Nor is earthquake prediction a field rife with successes.

Sorting the Wheat From the Chaff

A few years ago, some scientists thought earthquakes might be predictable by changes in the Earth's ionosphere, a layer of the upper atmosphere a couple hundred miles (300 kilometers) above the Earth's surface. The theory was that ions produced by the soon-to-slip fault could disturb the ionosphere.

But an analysis of five ionosphere disturbances recorded before earthquakes found that each could have been caused by something else—usually the sun. That's "a space-physics signal, not an earthquake signal," says Jeremy Thomas, a space plasma physicist at Northwest Research Associates and the Digipen Institute of Technology in Redmond, Washington. Thomas also presented his findings at the AGU meeting.

It's particularly telling, Thomas says, that the same disturbances in the ionosphere could be found far away from the earthquake epicenter. "If it's related to the earthquake, you wouldn't have the signal thousands of kilometers away," he said.

This lack of success doesn't mean that earthquake prediction is quackery.

"It is a field that is very much alive," says Michael Blanpied, executive director of the National Earthquake Prediction Evaluation Council, whose scientists assess the believability of proposed earthquake prediction methods and report findings to the U.S. Geological Survey.

"There are a lot of people attacking the problem from a lot of angles and trying to sort out the wheat from the chaff—to find out if there is some wheat in there, which is still not clear," he said.

"The heart of the issue is that the field contains people who are working at a very highly professional level, people who come from other professional fields, and people who don't have a scientific background but think they have something to contribute."

And sorting that out, he added, is what earthquake prediction science is all about.