These phenomena are generally non-seismic, so seismologists use geodetic observations such as measurements of tilt and land deformation to predict when an earthquake might be about to take place. Satellite data from the Global Positioning System (GPS) will improve the accuracy of geodetic measurements, but reliable short-term earthquake prediction is currently very difficult.

An alternative approach is to use electromagnetic rather than seismic information. Since the 1980s Panayiotis Varotsos and co-workers at the University of Athens have been predicting earthquakes in Greece based on measurements of electric currents in the earth.Their method – which is called VAN after the initials of the Athens team – is based on the potential difference between electrodes that are buried at several sites. By continuously monitoring these voltages, the researchers have detected anomalous transient signals just before earthquakes that they call seismic electric signals (SES).

However, the seismic electric signals exhibit some puzzling behaviour. First, they travel much further – between 10 km and 100 km – than expected. Second, they are only detected at a limited number of observation sites, which suggests that they only travel through narrow, highly conducting paths.Now the Athens team has found that the long transmission distances can be explained if the source of the seismic electric signals and the observation site are both close enough to these conducting paths (P A Varotsos 2003 Phys. Rev. Lett. 91 148501). Since earthquake faults are known to be highly conductive, this seems a reasonable assumption.

In the February issue of Physics World Seiya Uyeda and Haruo Tanaka in the Earthquake Prediction Research Center at Tokai University in Japan describe this work in more detail.