The risky business of earthquakes

Earthquake damaged buildings

Earthquake damaged buildings in L'Aquila, Italy

23 December 2011 by Philip England

Most of us learned about plate tectonics in school, and how cities like San Francisco, built where two plates collide, are most vulnerable to earthquakes. But is it that straightforward? Philip England thinks not.

The melancholy trail of earthquakes running through the opening years of this century has raised public awareness of seismic hazard and made scientists reappraise our understanding of it.

The popular view is that the most damaging earthquakes occur on the boundaries between plates, that the largest earthquakes cause the greatest number of casualties, and that the prediction of earthquakes is currently impossible. The first two of those statements are incorrect. The third is correct, although as I write a friend of mine, and his colleagues, are being prosecuted in an Italian court, in essence for failing to predict the 2009 L'Aquila earthquake.

By definition, a plate boundary is a narrow (a few tens of kilometres) zone of faults - fractures through rock - that separates two rigid plates. The locations of those boundaries are precisely determined. But faulting doesn't just occur at plate boundaries. Continental interiors contain diffuse networks of faults that are hundreds or thousands of kilometres wide, and many of these are undetected.

Since the beginning of this century, four earthquakes of magnitude 8·5 or greater have occurred on plate boundaries. Two of them generated tsunamis that killed about 250,000 people, while the other two, between them, were responsible for about 2000 deaths. Over the same period, five much smaller earthquakes were responsible for the deaths of about 400,000 people. These did not occur on plate boundaries.

Map showing distribution of rates of strain in the Aegean region

The distribution of rates of strain in the Aegean region. The plate boundary is shown by the toothed lines and the dots show the epicentres of M > 5·8 earthquakes during the past 100 years.

Ever since man began living in cities, large numbers of people have died in small earthquakes in the continental interiors. But the problem is growing, as ever more people migrate into vulnerable megacities.

In contrast, the developed nations around plate boundaries of the Pacific rim - like the USA and Japan - have been increasing their resilience to larger earthquakes since the late 19th century, through systematic construction of earthquake-resistant buildings.

The magnitude 9 earthquake in Tohoku, Japan, in March 2011, for all its tragic death toll and dramatic aftermath, exemplifies that progress. The most remarkable events in this earthquake were those that did not happen: buildings, in general, did not fall down.

Over six million people were exposed to intense shaking; of these, only about 1,000 were killed by building collapse. Such resilience to shaking in major earthquakes is now commonplace in developed countries around the Pacific rim.

The Chilean earthquake in February 2010 went largely unremarked by the western media, even though it was the eighth largest ever recorded; perhaps because only 600 people died - out of four million affected by severe shaking.

When the long-anticipated 'Big One' hits California, only a couple of thousand deaths are expected, even though the most intense shaking may affect up to ten million people.

It may seem callous to talk of 'only' a few thousand deaths, but the proportion of deaths is a crucial consideration if we hope to reduce fatalities in future. The most significant difference between plate boundary earthquakes and those in continental interiors is the fraction of the population that died as a result of each.

In the 2003 earthquake in Bam, south-east Iran, more than 30 per cent of the people in the area subjected to severe shaking were killed, and death rates of 10 per cent or more are common in continental earthquakes.

The same GPS signals that are used in your sat nav can measure movements of the Earth's crust as slow as a millimetre per year.

The obvious message carried by these numbers is that continental earthquakes mostly affect the developing world, where building standards are lower. But, while this analysis is correct in broad terms, it does not help us find a solution.

The fundamental problem of seismic hazard in the continents is that devastating earthquakes usually happen on faults whose seismic hazard has been underestimated, or whose existence is undetected until it is revealed by earthquakes.

As with many aspects of the Earth system, the study of earthquakes is a forensic exercise. Nature leaves clues about past earthquakes littered over the surface of the continents, and we attempt to interpret these clues, using all the tools that we can bring to bear on each individual problem.

For example, we use high-resolution satellite data to measure the slow deformation of the crust between earthquakes. The same global positioning system (GPS) signals that are used in your sat nav can measure movements of the Earth's crust at speeds as slow as a millimetre per year. By mapping differences in movement between one part of the crust and another we can see where strain is accumulating that will lead to earthquakes.

We used 25 years of such measurements to create a strain map of Greece and western Turkey, one of the most significant areas of continental deformation. The map gives us some important clues about how the continents behave. It shows that the most rapidly deforming parts of the region are not near the plate boundary at all but lie several hundred kilometres to the north.

This pattern of strain is inconsistent with the rules that govern plate tectonics, but it can be explained if, rather than being a rigid plate, the region behaves like a very viscous fluid, with a thin skin of elastic crust on top of it, which breaks in earthquakes.

The fact that almost all the major earthquakes in the Aegean during the past hundred years took place in the regions where the rate of strain is highest strongly suggests that maps like these will play a central role in identifying 'hot spots' of seismic hazard in the continents.

Map showing global distribution and scale of earthquake-related deaths

Earthquake-related deaths for 1900-2011. The area of the circle is proportional to the number of deaths and the colour to the earthquake magnitude. Earthquakes represented by circles with black rims did not occur on plate boundaries.

But strain maps alone are not enough. Most deaths in continental earthquakes are caused by earthquakes of about magnitude 7, which are a thousand times weaker than the Tohoku earthquake.

It can be literally a matter of life and death whether your house is 1km, 10km or 30km from such an earthquake. Over the past couple of decades, geologists have become skilled in reading the signals that are left in the landscape by repeated earthquakes, and thereby identifying where the dangerous faults lie.

Sometimes these signals smack you in the face. For example, repeated movements of faults are responsible for the steep mountain fronts that cross the landscape of central Greece (most famously, the cliff against which the Battle of Thermopylae was fought).

In other places, the signal is far more subtle. For example, we've known for about 150 years that the remains of a 2000-year-old shoreline clings to the cliffs of western Crete as much as 10m above sea level, like a ring of grime on the edge of a bath tub.

A few years ago I was in Crete with University of Cambridge colleagues James Jackson and Beth Shaw, trying to understand this shoreline, when we suddenly realised that the shape of this shoreline was exactly what one would expect from uplift in a single earthquake, but the distance by which it is lifted up is unexpectedly large. Ten metres of uplift would require an earthquake of enormous magnitude; could such an event actually have happened?

Beth's sharp eyes discovered the remains of several tiny corals clinging to the cliffs, and when our colleagues in the Oxford radiocarbon lab dated them, it became clear they had all died within a few decades of AD 365.

The simplest explanation is that they were lifted out of the water in a single earthquake, which we calculated would have had to be around magnitude 8·3. According to historical sources, there was a large earthquake in July AD 365, accompanied by a tsunami that devastated the Egyptian port of Alexandria.

Matthew Piggott at Imperial College London calculated that the size of tsunami generated by an 8·3 magnitude earthquake would be entirely consistent with the historical reports. But how frequently do tsunamis like this happen? Faced with this question, we looked afresh at our GPS data.

Most of Greece is stretching, but a relatively small area around the plate boundary is contracting, in a fashion that suggested to us that earthquakes like the AD 365 event happen about once every thousand years in the eastern Mediterranean. A similar tsunami earthquake happened in AD 1303, and it would be unwise to ignore the possibility of other such events in the near future.

The past few years have been sobering for seismologists. The Japanese earthquake showed that we are still a long way from being able to map seismic hazard accurately and to forecast catastrophic events.

And earthquakes like those in Haiti, China, Iran and Christchurch emphasised that, often, we don't even know where the dangers lie. But we are getting closer, and the lessons we have learned will accelerate our progress towards mitigating the effects of such disasters in the future.


Professor Philip England is the professor of geology in the Department of Earth Sciences at the University of Oxford.

NERC supports the Dynamic Earth & Geohazards Group of the National Centre for Earth Observation, which is based in the universities of Oxford, Cambridge, Leeds, Reading and UCL. One goal of this group is to understand the physical processes that govern the deformation of the continents, and the earthquakes that result from it.