University of Alaska Fairbanks research center monitors nuclear explosions worldwide

FAIRBANKS, Alaska (KTVF) - The Wilson Alaska Technical Center (WATC) at the University of Alaska Fairbanks (UAF) works to record, detect, and analyze nuclear explosions around the world.

According to Alex Witsil, Postdoctoral Researcher at the WATC, the facility works with other organizations to monitor global nuclear activity.

“One of our principal objectives is to increase our ability to record, and detect, and to otherwise monitor for nuclear detonations and nuclear explosions around the world,” Witsil explained. “We work closely with a lot of global agencies including the Comprehensive Test Ban Treaty Organization to deploy, maintain, and analyze data from a number of different sources - whether that be seismic sources, or acoustic sources - all with the effort of trying to, again, make these detections for non-proliferation.”

Alaska is especially poised to accomplish this task due to location and weather, according to Witsil. “Alaska is in a good location to do this because a lot of our sites are in remote and cold weather environments. Alaska offers a really good test bed to deploy these instruments and get a sense of how they’re going to work and the challenges that you face when you go to even more remote areas like Antarctica.”

Witsil continued, “Ultimately, we’re looking to not only detect large explosions, whether that be from a nuclear detonation or even large chemical explosions, but characterize that type of activity. We want to know where it came from, how big it was and how long, and if there were multiple explosions for instance.”

Various factors can make it difficult to analyze and compare data being recorded, but Witsil says he and his fellow researchers found a solution to this problem. “One of the primary focuses of our research was how to better compare those waveforms that are recorded at larger distances away, where that atmosphere might have changed that recording. We took a machine learning perspective to do that, basically training computers to recognize the subtleties. Even though those waveforms and those signals might look very different from one another, we can train these computers to find commonalities between them.”

Machine learning requires a large sample size to educate the computer, and researchers found a solution to this as well.

“One thing that machine learning algorithms need in order to be successful is a lot of examples of what, in this case large explosions from nuclear chemical detonations, look like,” Witsil said. “But thankfully there haven’t been that many nuclear detonations in the world. There’s been about maybe 500-600 I believe, and that’s not enough to really train a robust, machine learning algorithm. So what we decided to do at UAF was actually synthetically generate numerous thousands upon thousands of nuclear explosions within our computer. As geophysicists, we have the models and the skills to make relatively accurate and realistic explosions within this very safe world contained within our computer.”

Copyright 2022 KTVF. All rights reserved.