Correction: An earlier version of this story said the Japanese government spends more than $100 billion annually on earthquake prediction. In fact, the amount is far lower. The Japanese government spends about 10 billion yen ($123 million) annually on quake research, according to the Nikkei news agency. The story has been updated.

When two University of Tokyo seismologists released a study last month forecasting that a major earthquake would hit this city of 13 million sometime in the next four years, their prediction grabbed front-page headlines and led television news programs.

But the forecast brought a backlash, as other earthquake researchers — including some at the same university — said that such predictions shouldn’t be attempted in the first place, because, at least based on Japan’s history, they’re more likely to be wrong than right.

Within days of the news reports, the university’s Earthquake Research Institute posted a lengthy disclaimer on its Web site, saying the forecast was just the opinion of two of its researchers and noting that the study had a “large margin of error.” Colleagues said the report jumped to conclusions that other earthquake experts didn’t support.

“Many seismologists think this kind of study is too simple,” said Naoyuki Kato, a seismologist at the institute.

The growing skepticism here about earthquake forecasts is the byproduct of Japan’s long, expensive and fruitless effort to figure out where and when the big ones will strike. Last year’s March 11 mega-earthquake, unforeseen by experts, only reinforced those scientific shortcomings: All of Japan is vulnerable to huge quakes, some experts say, but that’s about all we know for sure.

The skepticism is shared by seismologists worldwide, but Japan is the world’s earthquake capital, and no country spends more on forecasting temblors.

A search for clues

For decades, researchers in this island nation have studied historical data, dug through fault lines, attached instruments to the sea floor and pulverized rocks in laboratories — all in a search for clues about how the earth behaves before it slips, shifts and shakes.

Unlock that mystery, scientists say, and they could make predictions (“Tokyo will have a 7.0 quake tomorrow”) with enough warning time to allow a city to be evacuated. They could also make broad and accurate forecasts (“There’s a 70 percent chance Tokyo will have a 7.0 quake within the next four years”) that could influence everything from building codes to government planning.

Experts have spent decades looking for earthquake precursors by examining changes in the levels of water and radon gas, electrical current and animal behavior. They have frequently had their hopes raised. In the laboratory, for instance, rocks under pressure show small signs of rupture before a dramatic slip, suggesting the possibility that “pre-slip” deformations could signal a developing earthquake. But most scientists say that pre-slip movement, if it exists in real life, is too inconspicuous to serve as a clue.

There’s also the common but fraught practice of forecasting earthquakes by using apparent historical patterns. In some regions of the country, major quakes usually happen every 100 or 150 years. But even those areas can remain calm for three centuries.

In Tokyo, which had its last major quake in 1923, centuries of records produce no clear pattern. Sometimes, most notably with last year’s 9.0 magnitude temblor, the problem is history itself: Scientists either lack enough data to see the pattern, or they don’t see one until it’s too late.

According to the government-issued earthquake hazard map that Japan updates annually as part of its public safety efforts, the northeastern coast of Japan was thought to be among the least likely places in the country to experience a major jolt. Then, March 11, miles off the northeastern shoreline, the earth slipped 100 to 160 feet.

In the weeks after, Japanese seismologists looked for missed signals. The northeastern coast, it turned out, was indeed susceptible to major quakes. But the region hadn’t seen one since the year 869. Clues of that quake were buried deep in the ground, in the form of sediment left behind by the resulting tsunami. Some experts in Japan knew of the 869 quake well before last year, but only after March 11 was there enough data to even guess at a cycle.

“And that’s where we got surprised with Tohoku,” said Jim Mori, an earthquake researcher at Kyoto University, referring to the region affected by last year’s quake. “Over 400 years up there, there were maybe a dozen magnitude-7s. People thought they understood that. But it turned out you didn’t see the pattern over 400 or 500 years. You had to go back 1,000 years to see a pattern.”

None of the major earthquakes to hit Japan since 1979 have occurred in areas that government seismologists describe as top danger zones.

Some experts think the Japanese government — which devotes about $10 billion yen ($123 million) annually to earthquake research, according to the Nikkei news agency — should deemphasize the forecasts, because they pressure researchers to draw conclusions from the data they gather.

“The problem is, people are getting lots of money for predictions,” said Robert Geller, a University of Tokyo professor of geophysics. “So it’s like an arms race to exaggerate the practical benefits of one’s own research.”

‘It could be random’

The flap over the University of Tokyo study, however, underscored a common problem for quake scientists: If you want sober research to grab headlines, couple it with a forecast.

As study co-author Shin-ichi Sakai recounts, he and his colleague, Naoshi Hirata, set out with modest goals. They simply wanted to crunch some numbers and chart seismic activity.

In the months following the March 11 earthquake, the pair noticed that there had been a spike in the number of minor earthquakes centered near or around Tokyo. Perhaps, Sakai reasoned, shifting caused by the March 11 quake had also changed pressure on plates around Tokyo. Maybe these small quakes hinted at something bigger.

That’s how he and Hirata arrived at their forecast. They used something called the Gutenberg-Richter law, which describes a correlation between small quakes and major ones. In Japan, for instance, magnitude-3 earthquakes occur almost every hour. Magnitude-5s occur every three days; magnitude-6s, about once a month.

Based on that correlation, the rise in small quakes upped the chances of a large one, Sakai and Hirata concluded. They plotted their data and came up with their prediction.

“The purpose of our study was not really to make a forecast,” Sakai said. “It was to show that seismic activity has intensified. So why do we have all these new earthquakes in an area so far away from the [March 11] epicenter? It could be random. There is no way to resolve whether our forecast is accurate.”

Special correspondent Ayako Mie contributed to this report.