• The Listener
  • North & South
  • Noted
  • RNZ

The trouble with earthquake prediction

Although seismologists have made huge advances in their knowledge of how earthquakes happen, accurate forecasting still lies in the future.

 Above the Alpine Fault.




'Only fools, charlatans and liars predict earthquakes," said the 20th century American seismologist Charles Richter. Back in the 1990s, when I was studying earth sciences at Victoria University, I used this quote to open my honours dissertation, a seismic hazard and risk assessment for Featherston.

Today, I don't think I'd be so glib. Science has come a long way since Richter's day, and as well as the inevitable "fools, charlatans and liars", there are now mathematical modellers, statisticians, geoscientists and physicists all working on earthquake prediction.

Victoria University's John Townend, EQC fellow in seismic studies, says when it comes to earthquake prediction, "you have to do it in a reliable robust way; it's not good enough to say there's going to be an earthquake of some loosely specified size, in a broadly defined area, some time in the next week. And then after an earthquake occurs to say, 'That was the one I was talking about' - that's not a reliable basis for making emergency management, construction or planning decisions."

Although it remains controversial whether earthquakes are intrinsically predictable, the scientific community takes efforts to predict earthquakes seriously and the Collaboratory for the Study of Earthquake Predictability has so far tested more than 100 earthquake prediction models. But science already gives us some useful tools in the form of probability data and early warning systems.

For some time we've had big picture probabilities, like the GNS Science data that suggest that the probability of a rupture of the central section of the Alpine Fault, which typically produces a magnitude 8 earthquake, is about 17% in the next 20 years; or that in Wellington, the likelihood of a rupture of one of the region's major mapped faults, which typically produce earthquakes of magnitude 7 or greater, is about 3.5% in the next 20 years.

Now Matt Gerstenberger, a geological hazard modeller with GNS Science, is using mathematical modelling to provide this sort of probability data in a more immediate way. Geonet is now using his model, used in California since 2005, to publish maps showing the probability of strong shaking in Canterbury over the next 24 hours.

The maps are updated every hour, after taking into account the latest seismic data and, according to Gerstenberger, are performing as expected. The data is also presented in another way: for the four weeks from March 8 to April 4, 2011, for example, the model predicted four to 15 aftershocks of magnitude 4.0 to 4.9 and up to three aftershocks of magnitude 5 and above. Exactly where and how deep these earthquakes will occur is not specified, but they will all be "in the aftershock zone," says Gerstenberger.

So did the model predict the 6.3 quake of 2011, which was itself an aftershock of the 7.1 Darfield earthquake of September 4, 2010? "On February 1, we calculated a probability of around 25% of a magnitude 6 or greater in the next year for the entire aftershock region. I wouldn't say we predicted the earthquake but I would say we gave a probability for it to occur that was consistent with what happened," he says.

Good modelling, of the sort Gerstenberger is doing, relies on good data. One reason there are so many seismometers in Christchurch is that detecting seismic waves from movement on the Alpine Fault is the focus of a major investigation by an international team co-ordinated by Townend and GNS Science's Rupert Sutherland.

In 2011 the team drilled into the fault to obtain rock cores and install a permanent "fault observatory" to ­measure seismic activity, temperature and fluid pressure in the active fault zone. Scientists already know that an Alpine Fault earthquake will typically produce a magnitude 8 earthquake. "You can expect if you have a magnitude 8 earthquake that you can have at least one magnitude 7 aftershock and a whole load of magnitude 6 aftershocks spread right across a really large part of the South Island," says Sutherland. "So in terms of coming to grips with an Alpine Fault earthquake scenario, how we deal with the aftershocks might turn out to be more important than the actual Alpine Fault event itself."

But will we ever be able to warn people that the Alpine Fault is about to rupture? Seismic waves propagate out from the source of an earthquake at speeds of a few kilometres a second, depending on the rock or soil they've travelling through.

"If you define the time of an earthquake as being the time when the rupture starts, then we can't predict that," says Sutherland, "but, if you define an earthquake as being the time when the ground shakes where you are, then you can predict that." In Japan, early warning systems are set up to detect seismic shaking at one seismic monitoring station and alert organisations in nearby population centres in time to take precautions.

This was the principle used by Christchurch company Roam 3, which employed seismograms south and east of Christchurch to allow search and rescue personnel early detection of new aftershocks. The current aftershock sequence is not the ideal scenario for this sort of system, however. "If you are in Christchurch and the earthquake happens beneath Lyttelton, then the waves that shake your seismic sensors are affecting people and buildings surrounding Banks Peninsula at almost the same time," says Townend. But for a large Alpine Fault event, or a Wellington earthquake, Sutherland says, the system could work, with a possible "tens of seconds" of warning in some parts of the country for a Wellington event.

A quake on the Alpine Fault, says Sutherland, "might rupture 400km of fault. So the rupture may take 100 seconds or more to propagate through. You could potentially give more than a minute warning." Tens of seconds is enough time for sensitive infrastructure to shut down and for stop signs to be deployed on transit systems, and a minute might be enough time for people to get out of a potentially dangerous building or find a safe place to shelter. Although lots of seismometers are already in place, the network is sparse in many places, particularly in the South Island. For an early warning system to work, we need more seismometers and a way to get the information to the right people to take precautions."

As for the future, will we ever be able to accurately predict the time, position and magnitude of individual large earthquakes? "There have been so many big advances in science it's hard to know what might come but right now you'd have to say it seems unlikely," says Townend.

But even if we can't predict exactly when and where an earthquake will occur, we can take note of time-varying hazard assessments based on ongoing analysis of data recorded by Geonet. And early warning systems may eventually provide some warning of impending shaking.

But apart from that, "you have to develop your infrastructure ... so that when it happens you minimise the impact of it", says Sutherland. That will be a focus of efforts in New Zealand as we review our earthquake preparedness after the Christchurch earthquakes.