Earthquake prediction has a notoriously bad track record. The underlying geophysical mechanisms are so complex (in both the "difficult" sense and the "emergent" sense) that successful backcasting -- using the models to "predict" past quakes -- is no guarantee that forecasting will be any better than random guessing. Last year, UCLA professor Vladimir Keilis-Borok got a bit of press for a method that successfully forecast two quakes... but failed from then on. Now researchers in Sweden are giving it a shot.
The Swedish Defense Research Agency has come up with a method that was able to successfully backcast the Sumatran-Andaman earthquake last December (which triggered the deadly tsunami), as well as a more recent earthquake in the same area. Given the history of quake prediction, the Swedish scientists are rushing to say that this method is nowhere near ready to be used to look ahead. Still, coupled with increasingly useful seismic and geological data coming from new monitor technologies, it's inevitable that the Swedish method will be employed for forecasts. As before, if it succeeds, we'll be in a new world of disaster management... and if it fails, it just goes on the heap of previous attempts.