With surprisingly little fanfare, the Japan Meteorological Agency, which keeps tabs on tens of thousands of earthquakes a year, has been setting up a network of ultra-sensitive electronic motion detectors that will pick up on the kind of minute seismic quivering that heralds a major quake.

If everything goes according to plan, the kinkyu jishin sokuho (emergency earthquake alert) system will broadcast warnings to wired-up homes and offices before heavy rumbling sets in.

It will be Japan's first public early warning system for earthquakes and the envy of tremor-prone countries across the globe. But it will not be able to do what every Japanese who has ever felt the ground shake, rattle and roll really wishes it could -- predict that next monster earthquake before the tremors start. And preferably days before, thank you.

But there are plenty of people who claim to be able to forecast temblors, and growing numbers of Japanese, still spooked by the Dec. 26 tsunami, quakes in Niigata Prefecture in October and a recent government assessment that the death toll from a big tremor directly below Tokyo could exceed 10,000, are turning to them. Web sites posting quake warnings days before they are expected to arrive -- often with surprising accuracy -- are proving extremely popular.

The prediction methods in use at these jishin yochi saito (earthquake prediction Web sites) are as exotic as they are diverse, and they are universally irksome to officials who maintain that reliable nationwide quake prediction is impossible.

There are groups who monitor erratic catfish behavior, long held by Japanese folk wisdom to be an earthquake omen, and others who scrutinize the inner workings of the nemunoki ("sleeping tree," or Albizzia julibrissin) -- a tall, deciduous tree found in warmer parts of Japan and believed by jishin yochi enthusiasts to be sensitive to seismic activity.

Some of the quake-watchers boast technical degrees and sophisticated equipment, but most are just dedicated amateurs armed with compasses, binoculars or cameras. Whatever their forecasting method, they almost all share the view that quakes can be predicted by monitoring electromagnetic disturbances around the Earth, or secondary effects like the shift in "bioelectric potential" in the sleeping tree.

The researchers at the Yatsugatake South Base Observatory Earthquake-precursor Ionosphere Observation Research Center in Yamanashi Prefecture, west of Tokyo, belong to the equipment and credential-bearing camp.

They claim that a decade of tracking disturbances in the Very High Frequency, or VHF, electromagnetic waves that precede temblors has convinced them that "with a certain margin of error, it is possible to estimate the location, magnitude and time period of earthquakes."

Another popular Web site called Kumo ni Kiko! Chikyu no Koe (Let's listen to clouds! The voice of the Earth) purports that areas where major earthquakes are about to erupt emit huge magnetic fields along which airborne water droplets and other particles gather to form telltale "earthquake clouds," examples of which are posted online.

"Anybody who looks up and sees those strange-looking clouds right before a quake will figure it out," said the site's founder, 63-year-old writer Yoji Sasaki, after describing his first experience 27 years ago spotting "a mushroom-shaped cloud and seven or eight others shaped like jet exhaust over Miyagi Prefecture" shortly before a strong temblor struck there.

Mother Nature's temperament being what it is, Shinobu Uehara, a computer professional who runs a quake site called S.O.S. in her free time, diligently cautions readers against placing too much confidence in the forecasts by members of the public that she posts online. The recent quakes in the Indian Ocean and Niigata Prefecture, she pointed out, both went sadly unforeseen by S.O.S.

But there have also been many bull's-eyes.

The site, for instance, warned of a quake in or around Hokkaido of Magnitude 6 between Jan. 16-20, and another for an M5-grade quake around Ibaraki Prefecture between Feb. 15-18.

Both apparently proved accurate. At 11:09 p.m. on Jan. 18, an M6.4 quake struck Kushiro Oki, just off the coast of eastern Hokkaido and at 4:46 a.m. on Feb. 16, an M5.4 quake rattled southern Ibaraki, with the shockwaves from it being felt as far away as Tokyo, some 50 km distant.

For the happily uninitiated, an M5.4 quake can topple vending machines, while an M6.4 will shatter windows, crush non-reinforced structures and make it impossible for people to stand -- or, at worst, stay alive.

Uehara told The Japan Times that "more than half" of the forecasts posted on S.O.S. "hit."

Record of successes

The editor of the Kumo ni Kiko cloud site said give or take three days and 50 km for small quakes and more for bigger ones, it has a hit rate of 70-80 percent.

The Yatsugatake site posts a record of past successes that lists 17 M5-plus quakes that occurred within two days and 160 km of their forecasts.

The seeming accuracy of these predictions poses problems for the authorities. If it were conclusively shown that quakes are indeed predictable, the public would hold officials in Tokyo far more responsible for lost lives.

But before that happens, officials at the Meteorological Agency say the jishin yochi crowd must come up with much more proof.

"It's like a weekly magazine horoscope saying you'll have good luck this week because you're the Chinese zodiac sign of the sheep," said Mitsuyuki Hoshiba, deputy director of the agency's Earthquake Prediction Information Division. "Without an explanation of cause and effect, it doesn't amount to science."

Scientific principles

Hoshiba appeared exasperated when recalling the telephone queries he's fielded from ardent jishin yochi callers.

"I try my best to explain the scientific principles actually at work, but some go on believing what they want to believe and hang up," he said, echoing complaints even within the jishin yochi set that passion for earthquake prediction sometimes flares into zealotry. Despite their differences, there are signs of growing interaction between the unorthodox and establishment schools.

Cloud-watcher Sasaki said that last spring, a government official working in earthquake crisis control acknowledged using Sasaki's information to keep an eye out for quakes.

Uehara, too, says that although officials haven't contacted her in person, she has heard indirectly that at least one respected expert thought S.O.S. uncannily prescient.

Giant pool of data

Also, she said professional "earthquake researchers" demanding anonymity alert her ahead of big quakes -- another sign that quake prediction holds credence with more officials than the authorities want to let on.

Such cooperation could lead to fruitful results. Who knows? Cell-phone wielding jishin yochi types, say, could transmit a giant pool of empirical data from which the government creates an "earthquake threat index" in the fashion of the color-coded terror alerts used in the United States.

"It is known that if 10,000 people around the country chipped in with whatever observation they could, quake forecasts would improve in accuracy," Uehara wrote in an e-mail. "Maybe then we could avoid disaster when the Big One comes."