The controversial idea that earthquakes can be predicted by monitoring tiny fluctuations in Earth's magnetic field is to be tested by two new satellites.

Although many seismologists see little merit in the idea, NASA and the US Air Force are together contributing about $1 million to provide data analysis and ground instrumentation to support experiments with the first satellite, the privately funded QuakeSat. Built by QuakeFinder of Palo Alto, California, the craft is now returning data from orbit after its 30 June launch. A second more expensive and ambitious satellite, funded by the CNES, France's national space agency, will follow next April.

Both craft are designed to search for small variations in Earth's magnetic field, which some scientists believe precede earthquakes. According to their theories, extremely low-frequency magnetic fluctuations are produced by compression of crystalline rocks or the movement of water in fault zones before a quake. Such fluctuations could potentially be detected from an orbiting satellite.

The idea of using magnetic activity to predict earthquakes is not entirely without precedent. Antony Fraser-Smith, a geophysicist at Stanford University in California who is among those who will receive NASA funds to work on QuakeSat data, reported elevated ground-based magnetic signals several days before the 1989 Loma Prieta earthquake, which damaged San Francisco and other cities in California (A. C. Fraser-Smith et al. Geophys. Res. Lett. 17, 1465–1468; 1990). Most Earth scientists consider these results to be interesting but inconclusive.

There is also tentative evidence that such fluctuations could be detected from space. Russia's Cosmos-1809, for example, managed to record signals from a 1989 earthquake in Armenia. Although some other satellites have turned up no such evidence, the subject intrigues geophysicists.

QuakeSat, built in collaboration with Stanford's Space Systems Development Laboratory, will now test these ideas. The satellite, which will collect data on magnetic-field strength from orbit for about six months, will have greater coverage than ground-based magnetometers — which are QuakeFinder's main product.

The company's president, Marijean Seelbach, admits that “the jury's still out” on whether the device can predict earthquakes. Stephen Park, a geophysicist at the University of California, Riverside, is typical of those who doubt that it can. Magnetic fields during the main shock of an earthquake generally measure billionths of a tesla at the fault line, he points out. Precursor fluctuations are likely to be lower, and would lose energy as they travel to the satellite. Seelbach agrees that measuring such tiny signals is a daunting task, but says that QuakeSat will be sensitive to trillionths of a tesla.

NASA is also “agnostic” about whether the satellites will prove useful, says John LaBrecque, manager of the agency's Solid Earth and Natural Hazards research programme in Washington. He admits that NASA would have had difficulty being the sole funder for such a controversial project, but says he feels the agency is justified in making a contribution to a private scheme.