Earthquakes, Networks, and the Tricky Topic of Quake Prediction
One of the goals of earthquake research is to provide warnings that can mitigate the effects of a disaster. At present, these attempts are limited to long range warnings which estimate the risk of significant damage over a period of years or decades, and to very short range warnings, on the order of a few seconds.
But high quality warnings that a quake is imminent in the next few days or weeks–a period that might allow large-scale evacuation–still elude earthquake scientists.
It may be that these kinds of warnings are not possible in principle. But that hasn’t stopped scientists looking. The study of earthquakes reveals all kinds of hidden patterns in the way they occur. Much of this work has compared the properties of specific earthquakes themselves, things like their magnitude and the time between successive quakes.
This has been rewarding, revealing all kinds of power laws governing things like the number of events of a specific magnitude and the difference between the main shock and its biggest aftershock.
But none of these patterns has yet turned out to be particularly useful for predictions on the scale of days or weeks. Perhaps, say the optimists, all that’s needed is a new way of thinking about earthquakes.
Today, Gene Stanley and pals at Boston University present just such a new approach. Instead of studying the properties of individual earthquakes, these guys have compared the patterns of quakes at different locations in Japan. They then create a network in which they link locations with similar patterns (see picture above).
That could turn out to be a powerful approach. One reason why earthquake science is so complex is that future quakes depend crucially on the history of quakes in that location.
To understand why, a good analogy is with forest fires, which also follow a power law in their size distribution. It’s obvious that the size of a forest fire does not depend on the size of the match that starts it. Instead, the way the fire spreads is determined largely by the network of connections between the trees. If there is no connection, the fire cannot spread.
So the size of a forest fire depends crucially on the history of tree growth (something that could be measured in principle but not in practice).
Many seismologists believe a similar process explains the size distribution of earthquakes. An earthquake becomes large if, at the moment it begins, the network of faults allows it to spread. So the size of an earthquake depends on the history of the fault network.
But while this network approach has revolutionised ideas about how earthquakes occur, it has done little for earthquake prediction on the scale of days.
Of course, seismologists have long studied whether regions with similar pasts will have similar futures. In the language of physics, these guys want to know whether the time series of events in the past is a predictor of the times series in the future.
The answer is a qualified yes. If you live in a region that has experienced big earthquakes in the past then it’s good bet you’ll get them in the future. However, the data does not allow predictions on the scale we’re interested in here.
What Stanley and co have done is to apply a network approach to the study of these time series. So they’ve identified regions in Japan with similar earthquake histories and then mapped out how these areas are linked to each other geographically.
The result is a network that reflects the geographical structure of the fault zone it describes. That’s never been done before using network science.
The question it raises, of course, is whether a network approach to earthquake histories will be any more predictive than the traditional analysis of time series.
Stanley and co raise the idea of improving earthquake prediction early in their paper but they studiously avoid discussing the impact their approach may have on earthquake forecasts.
It’s an omission that speaks volumes. But this approach may still help clarify and reveal other secrets of earthquake science.
Ref: arxiv.org/abs/1105.3415: Earthquake Networks Based On Similar Activity Patterns
The inside story of how ChatGPT was built from the people who made it
Exclusive conversations that take us behind the scenes of a cultural phenomenon.
How Rust went from a side project to the world’s most-loved programming language
For decades, coders wrote critical systems in C and C++. Now they turn to Rust.
Design thinking was supposed to fix the world. Where did it go wrong?
An approach that promised to democratize design may have done the opposite.
Sam Altman invested $180 million into a company trying to delay death
Can anti-aging breakthroughs add 10 healthy years to the human life span? The CEO of OpenAI is paying to find out.
Get the latest updates from
MIT Technology Review
Discover special offers, top stories, upcoming events, and more.