Data fingerprinting reveals risk of giant earthquakes

A new model could help to identify locations around the world that may be at risk of particularly large earthquakes of magnitude 8.5 and above.

We hope that by using the geological data itself as the principal indicator of risk potential, we can tap into giant earthquake risk in areas with little historical data – and even for other hazards such as volcanoes and landslides. Dr Rebecca Bell Reader in Tectonics in the Department of Earth Science and Engineering

Drawing on data science, the method, developed by researchers in the Department of Earth Science and Engineering (ESE) at Imperial College London, reveals a ‘blueprint’ of the regions that can host these so-called giant earthquakes, regardless of the history of seismic activity – and potentially in places where none have occurred in the last 100 years.

These insights open the door to improving future preparedness and risk mitigation for not only earthquakes, but also other natural hazards and systems, based on their underlying signatures.

Prediction challenge

Giant earthquakes occur at subduction zones, where two tectonic plates are converging (moving towards each other), and one slips underneath the other. Energy builds up where the plates meet – called the fault line – and it is the sudden slip of the plates that causes an earthquake. Subduction zones are common along particular geological margins, such as the rim of the Pacific Ocean, although giant earthquakes themselves are rare.

It is this rarity – from both the lack of historical record and long intermittence times between each one – that has made it impossible to determine when and where they might occur along subduction margins.

Study first author Valerie Locher, Research Postgraduate in ESE, said: “Our observational record of them is limited, and future earthquakes may happen at locations where we have not observed them before.”

The team’s new model is now helping to address this shortfall.

Earthquake ‘fingerprints’

The study, published in Geology, used existing geological data taken from more than 1,500 segments of subduction margins, which measured factors like sediment thickness and roughness, and convergence rate (the speed at which the tectonic plates move towards each other).

Importantly, the researchers did not include any data on past earthquakes to prevent their model being influenced by historical records, and to keep the focus on the geology that underpins these hazards.

Using a technique called Principal Component Analysis (PCA) that identifies patterns in the data, the model was able to cluster the margins into different combinations, based on those that best explain the differences between margins.

As co-author Dr Parastoo Salah, Teaching Fellow Position in Geo-Energy in ESE, explains: “These patterns can be thought of as fingerprints that indicate where giant earthquakes are more likely to occur. Our model found that relationships between the studied properties are non-linear, highlighting that the size of earthquakes cannot simply be linked to just one geological property or even a simple linear combination of them.

“This complexity makes it more challenging to anticipate when and where large earthquakes might strike.”

Quiet and extreme

What the PCA analysis also identified were two critical categories of margins: one indicating ‘active and moderate’ for smaller, more frequent earthquakes, and one for ‘quiet and extreme’ – in which there may be very little seismic activity for centuries, but where the stress along the fault line may be building for a long time, eventually resulting in a release as a giant earthquake.

These ‘quiet and extreme’ areas include locations where giant earthquakes have been observed before – but, crucially, also ones that have been ‘quiet’ during the last century.

One such location is the North Island of New Zealand, which has not experienced an earthquake greater than magnitude 7.5 since the start of the 20^th century, yet, as revealed by the study, has a subduction zone akin to those that have slipped in magnitude 8.5 or above earthquakes.

Evidence from cores taken from lakes in the North Island lend further support to the possibility of giant earthquakes here, showing that the country was hit by a tsunami hundreds of years ago that would have required an earthquake of this scale.

The implication is that these regions may have slipped under the radar for their giant earthquake potential, yet still hold capacity for destruction.

Unknown earthquake zones

The next step – and where there is the most potential – is in applying this approach to regions where very little, if anything, is known about the incidence of giant earthquakes.

Study co-author Dr Rebecca Bell, Reader in Tectonics in ESE, said: “Our next piece of work is to take subduction zones where it is still a bit of a mystery as to whether very large earthquakes are possible – places like the Hellenic subduction zone in Greece and Lesser Antilles in the Caribbean – and use our model to investigate whether these areas look more like the 'quiet and extreme' zones or ‘active and moderate.’

“We hope that by using the geological data itself as the principal indicator of risk potential, we can tap into giant earthquake risk in areas with little historical data – and even for other hazards such as volcanoes and landslides.”