A class project shines new light on an Iranian nuclear facility
A class project shines new light on an Iranian nuclear facility
Two women use machine learning and satellite images to analyze activity from afar
Arelena Shala and Katharine Leede chose a class project to help them learn how satellite imagery could be used to stop or slow the spread of nuclear weapons.
By the end of the quarter, the two women knew the technology so well that they could shed new light on a uranium enrichment complex in Iran and publish their findings in the July 2021 Jane’s Intelligence Review, a journal widely read by the intelligence community.
Shala and Leede met in the online classroom for “Verification for 21st Century Arms Control Treaties,” an international policy course taught by Rose Gottemoeller, who is the Payne Distinguished Lecturer at the Center for International Security and Cooperation and a former Undersecretary of State for Arms Control and International Security.
“I was so proud of Arelena and Katharine’s ability to take this new technology and run with it,” Gottemoeller said. “Their article was not only published in Jane’s, but also covered by the New York Times’ Visual Investigation Team and nominated as a project of distinction at the Geospatial World Conference 2021.”
Shala, a graduate student in the Master’s in International Policy program, had studied geospatial analysis as an undergraduate at the Military Academy at West Point. And Leede, a political science major entering her senior year, had studied open-source technology during her research with CISAC Senior Fellow Amy Zegart and her student fellowship at the Hoover Institution.
Gottemoeller asked the class to pair up and collaborate on original research related to nonproliferation. Shala and Leede decided to work together and reached out to Allison Puccioni, a CISAC affiliate who has provided imagery analysis for the military, tech and academic communities.
Puccioni met with the students, learned what they were hoping to do, and offered some ideas.
“She had been wanting to look at the site in Iran that had not been covered in a comprehensive way,” Shala said. “If we were interested, Allison said she could link us with the appropriate people at the satellite company, BlackSky.”
Shala and Leede were interested and met with experts at the Seattle-based geospatial intelligence and analytics company that uses a constellation of satellites to capture multiple images of earth each day. Ultimately BlackSky would provide both high-cadence imagery, or multiples images taken from multiple satellites over the course of a single day, and ‘burst collections,’ or images collected rapid-fire as a single satellite flies over the site.
“With the burst collections, we were able to stitch the images together into an animation,” Shala said, adding that the animation showed vehicles moving back and forth over a short distance. “We were able to make the claim that they were probably construction vehicles that were grading the road in front of the tunnel entrance to move something heavy into the tunnel.”
The high-cadence images of the site were used in different ways. Shala and Leede worked with experts from Orbital Insight to create machine learning algorithms to find the important information in the once-a-day images. This way, they could track features at the site from June 2020 until April of this year. They were able to see construction vehicles coming and going from the site, new roads being built and m(Text description of video's content is incorporated below.) ounds of dirt growing where the site was being excavated.
Video provides an animated sequence of BlackSKy burst imagery collected on 4 Match 2021 and shows a vehicle in motion at the eastern tunnel are of the Natanz South underground centrifuge assembly facility
The researchers determined the Iranians began construction on the new site last year between Aug. 30 and Sept. 14. They observed a nine-fold increase in vehicles at the site over the next three months, “indicating a significant growth in activity,” as they wrote in June in the journal Janes Intelligence Review.
The researchers determined the Iranians began construction on the new site last year between Aug. 30 and Sept. 14. They observed a nine-fold increase in vehicles at the site over the next three months, “indicating a significant growth in activity,” as they wrote in June in the journal Janes Intelligence Review.
In the article, Shala and Leede describe the essential role that artificial intelligence played in the analysis. “Orbital Insight’s object detection algorithm counted vehicles in 84 satellite images collected between May 2018 and May 2021, offering insight into activity at the existing and future facilities at Natanz,” they wrote. “The vehicle activity was tracked specifically in the parking lot outside the main Natanz site, as well as at the construction support facility at Natanz South, suggesting that the vehicles at each site were directly related to operations or construction activity.”
Ultimately, Shala and Leede could say with a high degree of confidence that they were seeing Iran construct an underground facility to assemble the advanced centrifuges used to enrich uranium for nuclear weapons.
The two women presented their findings in a virtual classroom and published their findings in Janes Intelligence Review. The article, which included dozens of images, attracted attention from industry experts who shared it online and tweeted notes of appreciation.
“It showed that the potential of the open-source community and government collaborating for a non-proliferation purpose,” said Leede, who will return to campus for her senior year as a CISAC Honors student with Gottemoeller as an advisor.
Shala is returning to campus to complete her master’s degree.
The two women are already brainstorming their next project, another collaboration at the intersection of imagery analysis and non-proliferation.
Gottemoeller will again be teaching “Verification for 21st Century Arms Control Treaties” in winter quarter and will be looking for new innovators.