Magnetic fields seen as clue to quakes

Research focusing on energy in days before draws skeptics

A new study found an increase in magnetic field energy near some earthquakes in California one to three days before they hit, a finding that its authors hope is a step toward accurately predicting when quakes will strike — a goal both elusive and steeped in controversy.

“This study provides important evidence that a physical change can be observed in the days before an earthquake,” said Dan Schneider, director of research and development at QuakeFinder. The Palo Altobased company, which aims to find a way to forecast major earthquakes, collaborated on the study with the Applied Science team at Mountain View-based Google Research.

“While we still have much work to do to improve the isolation of the signals, this research supports our vision that earthquake forecasting using the magnetic field may one day be possible,” he said.

Magnetic fields have been part of seismological research for decades, but the notion that they might be reliable, measurable precursors to earthquakes is controversial.

The U.S. Geological Survey states flatly on its website that there is no way to predict a major earthquake. “We do not know how, and we do not expect to know how any time in the foreseeable future,” the agency says.

And Jeffrey Love, a USGS research geophysicist, expressed skepticism about the new study’s findings.

“As they say themselves, the results are not useful for earthquake prediction,” he said.

The study, published Sept. 1 in JGR Solid Earth, analyzed magnetic field data collected by

magnetometers at 125 sensor stations along major faults in California in advance of 19 earthquakes magnitude 4.5 or higher from 2005 to 2019.

After applying algorithms to 50% of the data set, the researchers found a signal indicating changes in the magnetic field 24 to 72 hours before the earthquakes struck.

The study notes that the “statistical signal is of modest size,” so it’s not yet possible to predict an earthquake with the findings.

And separating a meaningful signal prior to an earthquake from all of the rest of the “noise,” including solar weather and lightning, BART trains, passing cars, construction machinery and power lines, adds to the challenge. Schneider said the sensors pick up on such “wideranging changes.”

“In this study we find that there tends to be an increase in magnetic field energy (beyond the regular background levels) in the window 24 to 72 hours before quakes, but we don’t yet know what is causing the increase,” he said. “Several theories have been proposed, but evaluating them in light of this study’s result is a whole other effort.”

The study acknowledges the controversy around magnetic field changes, and says the process is further complicated by both the infrequency of large earthquakes and sparse magnetometer coverage along fault zones.

Love, the USGS geophysicist, said the results did not advance the case for magnetic fields as earthquake forecasters. “I can’t see that they are doing anything better than a random prediction,” he said. “The results do not appear to be doing better than if they were just flipping a coin.”

The real meaningful test, he said, would be to predict an earthquake before it happens, rather than analyze earthquakes that have already happened. The prediction would not only have to say when an earthquake will happen, but also where it’s going to occur and how big it is, he said — “otherwise, it’s not a meaningful prediction.”

Schneider said many previous studies were “anecdotal,” revolving around data from a single earthquake to “try to identify something that might count as a precursory signal.” That made it more likely that the signal was just a coincidence, he said.

The current research addressed that through “careful study design,” he said.

“Our study approached the problem statistically and looked at many quakes from different pairs of stations to determine if a signal was consistently detectable above the normal variations in the magnetic field at the stations,” he said. “Not only that, but the choices to combine signals from pairs of stations and only train the algorithms on half the data before evaluating on the other half (a train-test split) make this one of, if not the most, rigorous study in this area to date.”

Love said further study “might be worth” continuing, but he remains dubious.

“Maybe the methods can be refined in the future, but until this happens we don’t have anything significant,” he said. “Earth is complicated, and the magnetic field of Earth is complicated and changes all the time. To sort all that out and make it into a useful, practically applicable method for predicting earthquakes is really difficult.”

Schneider is hopeful that earthquake forecasting will one day become a reality, just as scientists can now predict weather events including tornadoes and hurricanes.

“To get there, first we need to be able to better isolate the signal from the noise,” he said.

It will also be important to investigate whether signals change according to magnitude, or vary depending upon fault type and local geography, he said.

Ultimately, he said, the scientific community will want a physical model that explains these signals and “thus improves our understanding of the geophysical effects of the earthquake process.”