Susan E. Hough
US Geological Survey, Pasadena, California
A seismologist considers a new method of earthquake prediction.
I am acutely aware that numerous methods of earthquake prediction at one time held great promise, but fell apart under proper scrutiny. In recent years, I have heard about many studies purporting to uncover evidence of electromagnetic precursors, almost all of which involved weak or non-existent statistical analysis.
But occasionally I come across research that is not so easy to dismiss. For example, data from the French micro-satellite DEMETER, which was launched in 2003 to investigate electromagnetic perturbations in the ionosphere, have been analysed by a team of French and Czech researchers (F. Nmec et al. Geophys. Res. Lett. doi:10.1029/2007GRL032517; 2008). These authors find that there are very-low-frequency electromagnetic fluctuations in the ionosphere above the epicentres of moderate and large earthquakes that occur a day or two before the ground starts to shake.
Nmec and colleagues' results could be fatally flawed. If electromagnetic disturbances are generated when earthquakes occur, what are apparently true signals of one earthquake could actually be signals related to a preceding shock. Or the analysis might go awry because of subtle data-selection biases. But if there are fatal flaws, they are not obvious.
In any case, as the authors themselves emphasize, the significance of the DEMETER results can be demonstrated only when data from many earthquakes are averaged. This highlights a key point: it is entirely possible for precursors to be real but of no use for prediction. If earthquake scientists can separate consideration of earthquake precursors from the highly charged debates about earthquake prediction, the research community might just learn something about earthquake processes.
Comments
While the DEMETER observations look promising, it should be pointed out that it is not just the statistical testing that can be questioned in electromagnetic earthquake prediction methods, present or past. In the early 1970's, when shallow earth current anomalies were put forward as precursors to incipient earthquakes, I installed similar ground probes and measuring instrumentation on the Eastern Caribbean island of Trinidad - in an area of moderate seismicity. We recorded lots of lovely electric potential transients of various shapes and durations, every day, until one night the mains electricity generation failed and there was an island-wide blackout lasting many hours. In that time, the peak electric "signal" variations dropped from 100's milliV/m to well below the microV/m level, and variations were no longer detectable. We concluded that for the technique to have any chance of measuring "natural" signals, if such exist, one would have to be a long way away from any form of main electricity with ground return loops - perhaps 100's of km would be necessary. The experiment was rapidly abandoned.
Posted by: Willy Aspinall | February 22, 2009 09:25 AM
To be a useful predictor these precursors must occur relatively more frequently before earthquakes than before non-earthquakes. If all 9000 earthquake events had this precursor, but 27000 such phenomena occurred without being followed by earthquakes, the reliability of prediction would be only 25 percent.
Posted by: David Perkins | February 24, 2009 08:16 PM