A scientist who predicts a magnitude-6.4 or larger quake will strike the Southern California desert by Sept. 5 said his group has made other similar forecasts but that he would not disclose them publicly.

UCLA's Vladimir Keilis-Borok said Thursday doing so could trigger "disruptive behavior" while he and other scientists attempt to assess the validity of their largely untested prediction method.

"I would rather not talk about it," Keilis-Borok told a news conference at the annual meeting of the Seismological Society of America.

Other earthquake experts said the sort of predictions made by the group are of great scientific interest, but remain unproven.

"They have to understand the hypothetical nature of this research," Tom Jordan, director of the Southern California Earthquake Center, said of the public. "These types of predictions are made with tremendous amounts of uncertainty."

They also are of limited utility, whether disclosed or not, since they cover broad areas already known to be seismically active and large spans of time, experts added.

"You can argue people should do earthquake preparedness anyway and this is another reason to do it," said Egill Hauksson, a California Institute of Technology geophysicist.

Keilis-Borok and his colleagues predict there is a 50-50 chance that an earthquake will occur within a 12,000-square-mile area east of Los Angeles by Sept. 5.

The zone includes a large swath of the Mojave Desert, the Coachella Valley, the Imperial Valley and eastern San Diego County. It has experienced nine magnitude-6.4 or larger quakes over the last 70 years.

Some of those quakes killed people and toppled buildings, while others passed almost unnoticed, said William Ellsworth, chief scientist of the U.S. Geological Survey's earthquake hazards team.

The new prediction gives quake scientists a bully pulpit to remind people of the very real seismic risk in the region, Ellsworth added.

"We know there will be earthquakes," he said. "The advice is out there and not always is it followed or taken seriously."

Keilis-Borok's team successfully forecast two earthquakes last year, the magnitude-6.5 San Simeon quake in Central California and the magnitude-8.1 quake off Japan's Hokkaido island. As with the current Southern California prediction, the group had set wide limits in place and time for both quakes.

It's "only two, which is emphatically not enough to justify the methodology," Keilis-Borok said.

Beyond California and Japan, Keilis-Borok said his group also focuses on Italy and the Middle East. He would not say what regions his other predictions touched upon.

However, Ellsworth, of the USGS, said he knew of no other predictions for California quakes made by the group.

Keilis-Borok has said previously that he believes the combination of pattern recognition, geodynamics, seismology, chaos theory and statistical physics allows earthquakes to be predicted as hitting within a nine-month window.

He said he would disclose his other predictions at the close of the window or after the predicted quake occurred.