Scientists Gain a New Tool to Listen for Nuclear Explosions – GWC Mag

Both natural and human-made events can shake the ground: Grinding tectonic plates send earthquakes through the crust, and weapons tests set off explosions. Telling the difference from afar is tricky but crucial, especially when it comes to enforcing international weapons agreements such as the Comprehensive Nuclear-Test-Ban Treaty.

In a new paper published in Geophysical Journal International, researchers present a new mathematical method that’s able to correctly classify 99% of explosions and 98% of earthquakes in a test data set of seismic events from the western United States.

It’s the latest tool in a growing arsenal. Scientists have developed several techniques to distinguish natural earthquakes from explosions, including studying different seismic waves and the directionality of the energy emanating from these seismic events.

“We need a realistic way to say, ‘This is an earthquake, and this is not an earthquake; it’s an explosion.’”

“But no single method has been completely successful,” said Mark Hoggard, an Earth scientist at the Australian National University and lead author of the study. Previous methods missed nuclear tests carried out by North Korea in 2017. “The goal is to have multiple options,” Hoggard said. The redundancy will increase trust that these methods are working, he explained.

The new technique was able to correctly classify explosions caused by North Korean nuclear tests over the past couple of decades that had been miscategorized by older classification methods.

That’s progress, according to Vipul Silwal, an Earth scientist at the Indian Institute of Technology Roorkee. “We need a realistic way to say, ‘This is an earthquake, and this is not an earthquake; it’s an explosion,’” said Silwal, who was not involved in the study.

A Sphere in Many Dimensions

At the center of the new method are seismic moment tensors, mathematical representations of how Earth is deformed at the source of seismic events. Using computational methods, scientists can transform the moment tensors into points projected on the surface of a multidimensional sphere or hypersphere—in this case a sphere in five dimensions. “It can be difficult for our brains to conceptualize more than three dimensions, but we can do it with mathematics,” Hoggard said.

Projecting seismic moment tensors onto a hypersphere allows scientists to analyze the distribution patterns of these points using statistics and classification algorithms and to differentiate between earthquakes and explosions with high accuracy.

“One way of thinking about these complex mathematical transformations is to imagine that we are adding additional knobs to play with,” Silwal said. “These knobs can help scientists really fine-tune their analyses of the moment tensors and understand what happened at the origin of the seismic events.”

Previously, it was assumed that the points representing moment tensors would cluster in a spherical shape on the surface of the hypersphere. But Hoggard and colleagues found that the points instead cluster in a ellipse. “This shows that seismic events are more complex than previously thought,” Hoggard said. “We think we have a more accurate representation of the seismic events in our test data set.” That, in turn, helped Hoggard and colleagues zero in on explosions versus natural earthquakes more precisely.

“This paper is a methodological improvement for differentiating between earthquakes and explosions,” Silwal said.

Work in Progress

“We want to minimize the chances of missing an actual explosion and thinking it’s a natural earthquake.”

The new study also adds a couple of other analytical wrinkles to increase the accuracy of its modeling. “We want to minimize the chances of missing an actual explosion and thinking it’s a natural earthquake,” Hoggard said.

The method performs well, correctly identifying nuclear tests carried out by North Korea in 2016–2017 that previous models had classified incorrectly. But according to Hoggard, there’s still work to be done; the new model misclassified a couple of explosions as natural earthquakes.

“One of the things we take for granted is the existence of seismic moment tensors,” Hoggard said. “But these are man-made mathematical functions and can have errors.” He is now trying to rebuild the seismic moment tensors with fewer errors and will then explore whether the rebuilt moment tensors increase the accuracy of the model.

In a similar vein, Silwal said he would like the new model to be tested against a more controlled data set. “In the current data set used to train the model, we kind of assume what’s being called an earthquake is actually a natural earthquake,” he said. “Using a synthetic data set where we specify natural earthquakes versus other events would be a next step for testing this model.”

—Adityarup Chakravorty (chakravo@gmail.com), Science Writer

Citation: Chakravorty, A. (2024), Scientists gain a new tool to listen for nuclear explosions, Eos, 105, https://doi.org/10.1029/2024EO240142. Published on 27 March 2024.

Text © 2024. The authors. CC BY-NC-ND 3.0
Except where otherwise noted, images are subject to copyright. Any reuse without express permission from the copyright owner is prohibited.

Related