The federal government plans to embark this year on a major research effort aimed at developing within the next decade, a reliable method for predicting all major earthquakes.

Two federal agencies, the National Science Foundation and the U.S. geological Survey, expect to spend more than $800 million over the next 10 years on earthquake studies, including research on geological processes and observations of "precursory phenomena," such as color changes in well water and unusual animal behavior, that may forecast quakes.

The first phase of the new initiative is reflected in the budget the Ford administation sent to Congress. It calls for $54 million in federal earthquake research spending in fiscal 1978, more than double current expenditures.

Budget officials say the new interest in earthquakes stems partly from experience in 1976, when quakes caused more than 700,000 deaths around the world - the highest total in more than four centuries.

But the major reason for the increased federal effort is a new confidence on the part of earthquake expers - geologists and seismologists - the accurate prediction methods are within fairly close reach.

An advisory panel concluded last fall in a report to former presidential science adviser H. Guyford Stever that "earthquake prediction . . . has emerged from the realm of soothsaying to become a serious scientific endeavor with a significant likelihood of a short-term technological breakthrough."

Stever said in a budget briefing this month that the breakthrough may well occur within the 10 years.

"The new federal funding is going to permit us to gather the data necessary for pretty reliable prediction," Stever said. "Within a decade, if the research goes well, we should be able to perdict the time, location, and magnitude of earthquakes in time to permit precautionary measures."

Geologists who share Stever's optimism point to seismological advances in China, where a corps of 5,000 to 10,000 government forecasters and millions of amateur "people's scientists" have accurately predicted two major earthquakes.

The most dramatic Chinese success came in the Heich'eng earthquake of 1975. The quake was forecast five hours before it occurred, providing time to evacuate nearly 1 million people.

The result, according to Dr. Frank Press, chairman of the Department of Earth and Planetary Sciences at the Massachusetts Institute of Technology, was that "they lost a couple of hundred lives in an event that could have been expected to take a couple of hundred thousand."

The Chinese experience bodes well for American efforts at prediction, Press says, because "they have a few gimmicks of their own, but basically they're working from the same hypotheses that we're using."

Current earthquake theory in the United States, China, and Japan, another center of seismological study, stems largely from the science of "plate tectonics," which holds that the earth's continents have all split off from a single land mass and are still drifting across the globe. Major faults in the earth, such as California's San Andreas fault, are considered to mark the borders between the drifting land masses, or "plates."

If plate movements are in fact a contributing cause of earthquakes, instruments that measure movements and pressure changes along fault lines could provide warnings when quakes are about to occur.

Oriental and occidental geologists have sesigned several new instruments in recent years that measure seismic waves within the earth and changes in magnetization, electrical resistance and other properties of the earth's crust.

In addition, researches in China and Western Europe have reported less exotic forecasting mechanisms.

The Chinese regularly measure the level and color of water in wells to gauge changes in subsurface pressure. They have also reported that animal behavior - dogs barking strangely, cats leading their kittens to open fields, snakes leaving their holes - are earthquake warning signs.

For all this progress, earthquake prediction is still a highly unreliable science. Although the Chinese are the world leaders at it, they failed to foresee the Tangshan earthquakes last July that devastated a major industrial region and killed an estimated 650,000 people.

"Most geologists believe that the knowledge we need for prediction exists already," says Robert hamilton, who heads earthquake research at the Geological Survey. "What we need is a body of empirical data to tell us which indicators are going to be reliable for forecasting time and magnitude.

"Basically, we need to follow 10 earthquakes of Richter magnitude five or greater to get a sufficient body of experience," he says. "In the United States, we're not going to get that many in a 10-year period. But if we get the funding to go elsewhere, we should meet the 10-year goal."

Besides research on methods of prediction, the new initiative on earthquakes will provide greatly increased federal funding for studies of earthquake proof building design.

In addition, the National Science Foundation plans to allot $5 million or more annually for legal and sociological studies of the impact of reliable earthquake prediction.

"We have to admit that an accurate forecasting capability is going to have a considerable effect on property values and insurance rates and the human psyche in areas where earthquakes are predicted," Hamilton said.

"If we're going to develop this new technology, we though we'd better take a look at some of the effects of our technology while we were at it." CAPTION:

Picture, An earial view of part of the San Andreas fault in California. U.S. Geological Survey; Map, Map shows sites of U.S. earthquakes and extent of the damage they caused.