Wed | Jul 24, 2024

Can we get better at predicting earthquakes?

Published:Sunday | September 24, 2023 | 12:05 AMMichel Campillo and Rob van der Hilst - Contributors
Men ride a motorcycle past destroyed buildings in Samandag, southern Turkey, on February 22.
Men ride a motorcycle past destroyed buildings in Samandag, southern Turkey, on February 22.

A man is observed sitting among the wreckage that was caused by earthquake in the town of Imi N’tala, outside Marrakech, Morocco, on September 12.
A man is observed sitting among the wreckage that was caused by earthquake in the town of Imi N’tala, outside Marrakech, Morocco, on September 12.

An earthquake measuring 6.2 rocked central Italy in the early hours of August 24, 2016, leaving more than 200 dead and hundreds missing in the rubble of the disasters.

Given the devastation earthquakes cause, seismologists and public officials have long wanted to know when earthquakes will happen, and after the powerful 1964 Alaska earthquake, US scientists proposed a worldwide research programme on earthquake prediction.

In covering the initiative, Science magazine emphasised that “carrying out the proposal (i) would offer a fair chance to develop a method of giving warnings ‘hours to days’ in advance of major earthquakes and (ii) would, through engineering research, provide means of minimizing loss of life and property damage, even if a warning system were not achieved”.

The public has been interested more in prediction than in mitigation, however. And in spite of major progress among scientists in understanding the earthquake process and what causes disastrous shaking, it seems there is substantial disappointment in the apparent failure of ‘earthquake prediction’.

To some extent, this is an issue of semantics and objectives. Is the goal to predict an earthquake occurrence, predict ground motion due to an earthquake, or predict a disaster? Considering all of these, what is it that seismologists can and cannot do when it comes to predicting earthquakes?


Fortunately, all earthquakes do not lead to disasters and, therefore, understanding where and why disasters are produced is the first goal of earthquake seismology.

But in our efforts to better predict earthquakes, we have to be precise about the timescale: is it a prediction that an earthquake is imminent – that is, within seconds, hours, or even days before the shaking? Or that it is likely to happen within years or tens of years?

Each of these predictions could be useful, and the type of action this information would lead to depends on the location of the earthquake and the sociogeographic and economic circumstances.

For example, even if one cannot predict earthquakes themselves, the ability to predict ground motion shortly after the onset of an earthquake could allow one to send warnings seconds or minutes before the main shaking is expected to occur. This means that some critical infrastructure could be automatically switched off safely, such as trains and nuclear power plants, and the public could be alerted.

This form of early warning, produced after the earthquake (or at least after the beginning of the earthquake), is already in practice in several of the most populated and earthquake-prone places in the world.


The vibrations we feel during earthquakes, and which could destroy buildings and other infrastructure, are the effect of the elastic waves that propagate from the earthquake source through the rocks in the Earth’s interior and along its surface. Earthquakes produce different types of waves, some more destructive than others.

These waves propagate with velocities of the order of kilometres per second, but the most damaging waves often arrive after the first waves. This means there is a time lag between the electronic signals sent by our instruments and the most damaging waves, creating opportunities for action. This field is called real-time seismology, and operational systems have been set in different regions of the world, including Japan, Italy, Mexico, and California. These systems do not predict the time of occurrence, but act as a warning of activity.

For long-term intervals (years to decades), geophysics provides the basis for rational management of affected resources and seismic safety. The quantitative analysis of actual earthquakes allows one to decipher the local conditions prone to disastrous shaking for specific buildings. Therefore, it provides the basis for including land management in seismic risk-mitigation plans. In that sense, seismologists already provide the information required to adopt development policies oriented towards public safety.

Large improvements have been and will continue to be made, but it is difficult for the general public to appreciate the effectiveness of this approach (known as probabilistic seismic hazard analysis) because this type of ‘prediction’ is expressed in terms of long-term probabilities, whereas the visible parts of earthquakes are the dramatic images of disasters in the news.

The public view of earthquake prediction typically concerns a deterministic (yes/no) assessment of whether or not an earthquake will happen on the intermediate timescale (hours to days). And it is precisely this field for which seismologists have not yet made definitive operational progress. The reasons for this are related to the complexity of the physical processes that cause earthquakes.


Earthquakes result from instabilities at locations where the resistance to slipping (the friction) of a fault is close to the forces that come from the slow plate movements of the Earth.

Although not felt in our everyday experience, the solid Earth is an evolving system. Its inner parts are subject to thermochemical convection. That is, cold and heavy materials at the surface tend to move downwards and, conversely, hot material from the deep interiors moves upwards, and these motions result in plate tectonics.

Typical velocity of plate motion is of the order of centimetres per year. During an earthquake, the two sides of a fault slide at velocity of the order of a meter per second – about a billion times (!) faster than the steady-state background motion.

In other words, damaging earthquakes occur in seconds but have typically been in the making for tens, hundreds or sometimes even thousands of years. The time for us to observe the Earth is so short that we have no hope of assessing with fine precision when a critical state will be reached.

Nevertheless, it has been recognised for a long time in experimental and theoretical studies that instabilities could be preceded by a change in a short time period relative to the long tectonic/geologic preparation. Such a preparatory phase, often referred to as initiation, is not systematically observed with present-day geophysical methods, and one would be forgiven to think that the case is hopeless, given the uncertainties about duration and amplitude of the initiation process. However, discoveries in recent years are giving a more encouraging prospective.


Geodesy and seismology have seen substantial progress in detecting subtle changes in rocks below the surface. Continuous GPS recorders and advanced processing techniques allow for the detection of smaller and smaller motions. Both GPS and seismometers are increasingly deployed in large dense arrays, producing antennae with unprecedented detection capabilities.

In the past few years, these efforts have led to the discovery of new types of deformation processes. The most spectacular of these concerns the widespread observation of so-called slow earthquakes. Slow earthquakes are slipping episodes at depth with deformation velocities that are in-between plate tectonic motion and regular earthquake slip.

The largest known slow earthquakes have magnitude of more than 7.5 when measured as regular earthquakes, and there is increasing evidence that, just as for regular earthquakes, slow earthquakes occur on a wide range of magnitudes. This suggests that deformation of the Earth occurs on a broad spectrum of timescales. That is, times in between the extremes of slow motions from plate tectonics and mantle convection and the ultrafast and disastrous seismic ruptures.

The newly discovered transient deformations can be studied with geodetic measurements and by a refined analysis of seismic records. These have indicated that slow deformation is accompanied by a characteristic weak grind.

There is, therefore, hope that one day we could detect and monitor extremely slight changes in the rocks that would precede earthquakes. This is, indeed, a long way from ‘prediction’ of precisely when and where a disaster will occur, but geophysicists will persevere and continue to make new discoveries about the changing Earth.

For now, knowing earthquakes is one way to live with them, to be prepared, to know the vulnerability of our communities, and to adopt sound policies for earthquake-safe environments.

This article was originally published in French. Michel Campillo is sismologue, professeur à l’Université Grenoble-Alpes, Institut universitaire de France, Université Grenoble Alpes (UGA) and Rob van der Hilst, sismologue, professeur au Massachusetts Institute of Technology, Massachusetts Institute of Technology (MIT). This article was first published in The Conversation, and is republished under Creative Commons Licence. Read original article here: