GPS could predict earthquakes two hours ahead, but there's a catch

Earthquakes could theoretically be predicted 2 hours before they occur, saving countless lives – but we must first develop GPS sensors that are 100 times more precise than those in use today.

Over the past few decades, expert opinion has shifted on whether any telltale seismic activity exists prior to earthquakes, or if they are inherently chaotic and unpredictable events. Now, Quentin Bletery and Jean-Mathieu Nocquet at Côte d’Azur University in Nice, France, may have settled the debate.

The pair have used GPS data to identify a gradual, accelerating slip between tectonic plates in the lead-up to an earthquake. These slips are too small to appear on seismographs, but could – if detected – indicate when earthquakes are about to begin. Such an approach has been tried before, but Bletery says previous research has only looked at a handful of earthquakes and produced warning signs that are also seen when no earthquake follows, or that are observed an uncertain amount of time before the quake.

The researchers used GPS measurements gathered over a period of 5 minutes, making them accurate to within 1 centimetre, taken during the 48 hours prior to 90 separate earthquakes. With a combined data set of over 3000 measurements, they compared recorded ground movements with the expected direction of movement that each site would see during an earthquake.

In each case, they found that the largest movement in the expected direction occurred just prior to the earthquake. They also found that the last 23 data points showed a gradually increasing movement in the expected direction, and the final seven were higher than any others during the entire 48-hour period.

Bletery says that this is indicative of a gradual, slow and otherwise undetectable slip between tectonic plates starting around 2 hours before earthquakes – something that could lead to a reliable earthquake detector.

But there is a problem. Bletery says that the noise levels of current GPS sensors means that detection is only possible on the big data set, and not from any one site. That would require GPS sensors able to detect movements of just 0.1 millimetres, he says.

“We can’t detect at the scale of one earthquake, so we cannot make predictions,” says Bletery. “But it tells us there’s something going on, and if we make significant progress in measurement – either the sensor itself, improving its sensitivity, or by just having more of them – we could be able to perceive things and make predictions.”

Roland Bürgmann at the University of California, Berkeley, says the work looks promising, but the proposed signals will need to be confirmed by further research. “There have been quite a few retrospective observations of various types of earthquake precursors in the past – foreshocks, deformation, etc – however, they are not unique in character from similar things happening at other times,” he says. “As Bletery and Nocquet see this two-hour-long precursor candidate looking at dozens of earthquakes, this looks somewhat promising.”