The New Scientist is carrying a piece on the risk of earthquakes in seismically quiet areas entitled Roaming quakes: is New York or London due a big one? (registration required). Written by Ferris Jabr, the article focuses on intraplate earthquakes, ones that happen away from the more tectonically active plate margins where earthquakes are expected:
These "intraplate" earthquakes have long been a mystery. "They are the last frontier for plate tectonics," says Magnani. What we are finding out now, though, is giving us pause for thought. It might be that it's not just San Francisco and Los Angeles that are susceptible to significant earthquakes, but New York, Sydney and perhaps even London too. Should we be worried?
Indeed, should we be worried? It’s a question that the article fails to adequately answer.
The communication of risk is problematic. Us human beings aren’t very good at understanding risk, exaggerating it when the event in question is big and dramatic like an earthquake and downplaying it when it is mundane like a car crash, even though we’re more at risk of being injured in a car crash than an earthquake.
Earthquake map for a 2010 temblor in Illinois
Peter Sandman came up with an equation to describe how we react:
Risk = Hazard + Outrage
Where the perceive risk is a function of the actual risk (hazard) and the outrageousness of the event if it comes to pass.
I prefer to reword this equation to:
Perceived Risk = Actual Risk * Impact(Emotional + Actual)
Where the Impact is both the emotional impact of an event such as how dramatic or outrageous it is, and the actual impact in terms of number of lives affected, money lost etc.
When it comes to high-impact low-probability (HILP) events, impact is so big, in both emotional and actual terms, that it swamps our understanding of the actual risk, which is often very small or unknowable. But the more unknowable a risk is, the more it exaggerates the impact.
Risk is often communicated using percentages and probabilities, which most of us struggle to interpret. For example, Jabr says:
The United States Geological Survey (USGS) suggests that there is a 25 to 40 per cent chance of a magnitude 6 or larger quake hitting the New Madrid area in the next 50 years, with a 7 to 10 per cent chance of an event as big as the one two centuries ago [which reshaped the landscape].
What does that really mean? For all practical purposes, such words mean nothing. Should businesses take action now? In the future? Never? Without context, these numbers are meaningless.
Geological faults do not necessarily produce earthquakes in a nice, linear fashion over time. They wax and wane, and sometimes become completely inactive. The tectonic stresses shift and, as happened Christchurch, NZ, new or previously unknown faults produce temblors in unexpected places. Unless you know how the numbers were derived and have some understanding of their context, such probabilities help no one.
Another way to try and communicate risk is to use what’s called a return period. Jabr obligingly provides us with another example:
In 2008, Lynn Sykes of Columbia University in New York City catalogued all 383 quakes in a 39,000-square-kilometre area around New York City from 1677 to 2007 and estimated the future risk. He concluded that New York can expect a magnitude 5 quake once every century, a magnitude 6 quake every 670 years and a magnitude 7 quake every 3400 years (Bulletin of the Seismological Society of America, vol 98, p 1696).
According to this data, the return period for M5 quakes for New York is 100 years. This does not mean that M5 quakes happen 100 years apart, it means that the average time between M5 events in that area is about 100 years. We should not be looking at the centenary of the last M5 and assuming that the ground is going to shake.
Indeed, we could have two M5s within a few years of each other, then nothing for two centuries. Which is another reminder that geophysical events don’t happen in a nice, orderly fashion.
Ultimately, Jabr’s piece conflated two different issues: research into the mechanisms of intraplate earthquakes and the problems with predicting earthquakes. The inclusion of the latter seemed very much like a feeble attempt to justify the former and was, in my opinion, neither necessary nor illuminating.
Science journalists have a duty to explain the science clearly and that includes communicating risks in a way that readers can understand. HILP events are naturally difficult for us to wrap our brains around because they play against some of our in-built cognitive biases. That means that we need to think harder about when and how we communicate statistics like probabilities and return periods.
