Predicting earthquakes
Every time a major earthquake occurs, we wonder why scientists aren’t better at pinpointing where and when they will strike
Can you predict earthquakes?
No. As the US Geological Survey (USGS) puts it: “Neither the USGS nor any other scientists have ever predicted a major earthquake. We do not know how, and we do not expect to know how any time in the foreseeable future.” A proper prediction would have to define date and time, location, and magnitude. The location is the easiest part: most big earthquakes occur along the fault lines between the Earth’s 17 or so tectonic plates. More than 80% of large quakes occur around the “Ring of Fire” at the edges of the Pacific plate. Also very vulnerable are “triple junctions”, where three plates meet: southeastern Turkey sits on the junction of the African, Anatolian and Arabian plates. But even defining location is a vague science, since there are so many thousands of miles of fault lines across the world. And defining time and magnitude is even harder.
Why are these so hard to predict?
At their simplest, earthquakes happen because of a process known as “stick-slip”. Tectonic plates slowly slide past each other; some move away from, or bump into, each other. Friction caused by this movement (happening at perhaps the speed that fingernails grow), causes stress to build up. When the stress becomes too great, stick gives way to slip: energy is released in the form of waves that travel through the Earth’s crust and cause shaking. The problem is that there is little or no data for geologists to observe – until the slip, when it’s too late. Sometimes there are smaller “precursor” shakes first. In 1975, Chinese officials reacted to precursors by ordering the evacuation of the city of Haicheng, shortly before a quake, saving many lives. But minor tremors frequently occur without being followed by an earthquake; and big quakes often happen without any precursors at all (such as the 1976 Tangshan earthquake in China, which killed at least 240,000 people).
But don’t quakes follow some broader patterns?
Yes: scientists are able to map probabilities from the past frequency of earthquakes in given locations, and to make very broad forecasts. In the 1980s, researchers who had analysed reams of historical data concluded that a segment of the San Andreas Fault near Parkfield, California, was overdue a large event. They decided an earthquake would hit by 1993; but it didn’t happen until 2004. Scientists put the difficulty of making such predictions down to the sheer complexity of analysing the movements of the planet’s crust, and the fact that much activity occurs over very long periods, often at depths far beyond humans’ ability to observe them. Forecasting earthquakes would require detailed measurements deep underground over the course of decades, at the least, which would then have to be modelled in precise simulations. Even then, the best that could be hoped for is perhaps an hour’s worth of lead time, since there are so many variables at play, and few tools to analyse them meaningfully.
Will this situation ever change?
Scientists have examined a whole series of potential indicators of coming earthquakes, from electromagnetic anomalies in the atmosphere to increased radon levels. One group in Israel claims it can predict large quakes with more than 80% accuracy up to 48 hours before they strike, by examining changes in electron content in the Earth’s ionosphere. There are also reports of spikes in the concentrations of radon before major earthquakes; this has been attributed to pre-seismic stress, which can fracture rocks and release gas. The most promising area is artificial intelligence: AI could crunch seismic data far beyond what has previously been possible. But all such systems are for now highly theoretical, and many seismologists maintain that earthquake prediction is inherently impossible.
Can other steps be taken?
Yes: warnings can be issued after earthquakes have occurred. Seismic waves take time to travel from their epicentres; they typically move at speeds between 1 and 14km per second, which is slower than modern telecoms systems. So, for instance, during the 2011 Tohoku earthquake and tsunami in Japan, the country’s earthquake early-warning system, which uses data from more than 1,000 seismometers, sent out warnings of impending strong shaking to millions. These reached Tokyo, 232 miles from the epicentre, about a minute before the earthquake; this is said to have saved many lives. ShakeAlert, a system built by the USGS, can send a notification to a person’s phone, giving them about 20 seconds to a minute advanced notice before an earthquake. But providing a warning any further ahead is hard to do.
So what can be done?
The general advice is that, rather than trying to make short-term predictions, governments should concentrate on the long-term mitigation of the dangers. Over the past two decades, earthquakes have caused nearly 750,000 deaths globally, the World Health Organisation says – “more than half of all deaths related to natural disasters”. And a 1992 study found that 75% of earthquake deaths in the 20th century were caused by buildings collapsing. Engineering solutions already exist that can eliminate most of these deaths.
How effective are the solutions?
In Japan, which sits at the margins of the Eurasian, Philippine, Pacific and North American plates, some of the world’s strictest regulations stipulate that all structures must be at least to some extent resistant to earthquakes. High buildings are equipped with shock absorbers and motion dampers; older buildings have to be retrofitted. The Tohoku earthquake showed the value of such regulations. A magnitude 9.1 quake, the biggest in Japan’s modern history (compared with 7.8 in Turkey), it claimed 20,000 lives – but nearly all were killed by the tsunami. Fewer than 1,000 people were killed by falling buildings.