European Geosciences Union

EGU logo

Can we predict earthquakes?

The Seismology and Natural Hazards Divisions of the European Geosciences Union (EGU) heard seismology experts to summarise the current status of earthquake prediction.

Earthquakes, such as the 2011 Tohoku , the 2011 Christchurch and the 2010 Maule earthquakes, or the 1999 Afghanistan and 1999 Turkey earthquakes, are often devastating, resulting in wide-spread infrastructural damage and high number of fatalities. But even moderate-size events, like the April 06, 2009, L’Aquila earthquake (Mw 6.3), may create local disasters and tragic losses of properties and human lives. The public and decision makers are, therefore, interested in information about the time and location of future events. However, despite considerable research effort, scientists are only capable of making estimates about where big quakes are most likely to happen in the long term (years to decades). There is yet no reliable way of predicting in the short term (days to weeks or months) when an earthquake of a given size will occur in a specific location.

More than 200,000 earthquakes are recorded globally each year. Global, regional, and local seismic monitoring provides information to understand earthquake occurrence and thus map regions of high and low seismic hazard. Dense seismic monitoring and advanced seismic analysis techniques have facilitated the discovery of so far unknown seismic phenomena (slow-slip events, multiple re-rupturing of faults, super-sonic rupture speed), which shed light onto the physics of earthquakes. Continuous seismic monitoring thus provides fundamental information for seismic hazard assessment and earthquake engineering.

Why are earthquakes hard to predict?

Earthquake generation is a very complex process deep in the Earth crust. The magnitude and timing of a large quake depends on various factors such as the size of a fault section and the amount of stress accumulated there. Measuring the stress is an engineering feat in itself, as it requires drilling into the ground for several kilometres. Furthermore, scientists do not know precisely how much stress it takes to break a fault. However, once enough stress has accumulated, even a low-magnitude earthquake may cascade into or trigger a large tremor.

A possible strategy to predict earthquakes would consist in finding a diagnostic precursor. This precursor should be an observable signal, such as a deformation of the ground, anomalous emission of radon gas from the Earth’s interior, or strange animal behaviour, detected before any tremors. This diagnostic signal then indicates – with high probability and strong statistical significance – where, when, and with what magnitude an earthquake will occur. However, this “silver-bullet” strategy of earthquake prediction has not yet provided a successful and statistically stable prediction scheme. The International Commission on Earthquake Forecasting for Civil Protection analysed a series of proposed precursors in 2011 and concluded that none of them offers a reliable diagnostic of an impending earthquake.

Operational earthquake forecasting

As an alternative to earthquake prediction, scientists focus on developing earthquake forecasts, that is, on quantifying the likelihood of an event occurring. Researchers determine these probabilities of occurrence to inform communities about seismic hazards who, in turn, can use the data to make decisions before an earthquake.

Earthquakes tend to occur closely spaced in time at a given location, as so called earthquake clusters or earthquake sequences (the aftershocks that follow the mainshock are an example). “Events that have happened in a region are analysed for clustering, and from this knowledge seismologists are able to construct a probability model for future events,” explained Charlotte Krawczyk, president of the EGU Seismology Division. These probability models then allow researchers to forecast the likelihood of future tremors occurring at a given location.

Using short-term models, scientists can determine if there is an added probability of an earthquake happening at a given location in a period of a few days to a week. Modelling can show that the likelihood of an event occurring in that period and location increased “from 1 in 10,000 or 1 in 100,000 to 1 in 100 or 1 in 1000,” according to Paolo Gasparini, professor of geophysics at the University of Naples. But the absolute probability of an earthquake happening remains low, “lower than 1%,” he adds.

Warner Marzocchi, chief scientist of Italy’s National Institute of Geophysics and Volcanology defended that it is up to decision makers to use these probabilities to take mitigation actions. He stressed, however, that “the mitigation actions should be light, in the sense that most of the times warnings turn out to be false alarms, and therefore decision makers have to be careful to minimise complains and lose credibility.” He added: “Each individual might be inclined to do what is in his or her best safety interest, being given an informative hazard advisory by authorities.”

Davide Miozzo expert in communication of risk prediction from the CIMA Research Foundation in Italy, warned about the cry-wolf syndrome: “If communication is excessively recurrent and redundant it will be disregarded. And once institutions lose the population’s trust, the population itself becomes more vulnerable and less resilient to natural adversities.”

Experts agree that the most effective way of reducing risk and preventing damage is to use operational earthquake forecasting to build long-term probability models for future events. These are then used to inform earthquake engineers on seismic hazard that can be used to devise proper seismic safety designs. “Long-term forecasting models are the base of hazard maps used for civil protection, establishing building codes and for retrofitting regulations,” said Prof Krawczyk.

“Most important,” Dr Marzocchi concluded, “every kind of mitigation actions in the short-term will not replace the best long-term action that is to build buildings that can resist earthquakes.”

Additional Information

The EGU Seismology Division acknowledges the help of the following scientists, who contributed to this document: Charlotte Krawczyk (Leibniz Institute for Applied Geophysics, Hannover, Germany; President of EGU Seismology Division), Stefano Tinti (Department of Physics, University of Bologna, Italy; President of EGU Natural Hazards Division), Paolo Gasparini (Department of Physical Sciences, University of Naples, Italy), Warner Marzocchi (National Institute of Geophysics and Vulcanology [INGV], Rome, Italy), Davide Miozzo (CIMA Research Foundation, Savona, Italy), Massimo Cocco (INGV; Vice-President of the EGU Seismology Division), Mourad Bezzeghoud (University of Evora, Portugal; Science Officer of the EGU Seismology Division), Martin Mai (Kaust; Science Officer of the EGU Seismology Division), and Raúl Madariaga (Geology Laboratory, Ecole Normale Supérieure, Paris, France; regular contributor at EGU General Assemblies).

This text is based in part on the report Operational Earthquake Forecasting: State of Knowledge and Guidelines for Utilization by the International Commission on Earthquake Forecasting for Civil Protection, published in 30 May 2011. The document has been endorsed by the International Association of Seismology and Physics of the Earth’s Interior, a world-leading seismology organisation.


Charlotte Krawczyk (President of EGU Division on Seismology)
Leibniz Institute for Applied Geophysics
Hannover, Germany
Tel: +49-511-6433518

Stefano Tinti (President of EGU Division on Natural Hazards)
Department of Physics, University of Bologna
Bologna, Italy
Tel: +39-051-2095025

Bárbara Ferreira (EGU Media and Communications Officer)
EGU Executive Office
Munich, Germany
Tel: +49-89-21806703

[ Back ]