Inside the little-known group that knows where toxic clouds will blow

WHEN A NUCLEAR-POWERED satellite crashes to Earth, whom do the authorities call? What about when a derailed train spills toxic chemicals? Or when a wildfire burns within the fenceline of a nuclear-weapons laboratory? When an earthquake damages a nuclear power plant, or when it melts down? 

Though its name isn’t catchy, the National Atmospheric Release Advisory Center (NARAC) is on speed dial for these situations. If hazardous material—whether of the nuclear, radiological, biological, chemical, or natural variety—gets spewed into the atmosphere, NARAC’s job is to trace its potentially deadly dispersion. The center’s scientists use modeling, simulation, and real-world data to pinpoint where those hazards are in space and time, where the harmful elements will soon travel, and what can be done.

The landscape of emergency response

NARAC is part of Lawrence Livermore National Laboratory in California, which is run by the National Nuclear Security Administration, which itself is part of the Department of Energy—the organization in charge of, among other things, developing and maintaining nuclear weapons. 

Plus, NARAC is part of a group called NEST, or the Nuclear Emergency Support Team. That team’s goal is to both prevent and respond to nuclear and radiological emergencies—whether they occur by accident or on purpose. Should a dirty bomb be ticking in Tempe, they’re the ones who would search for it. Should they not find it in time, they would also help deal with the fallout. In addition, NEST takes preventative measures, like flying radiation-detecting helicopters over the Super Bowl to make sure no one has poisonous plans. “That’s a very compelling national mission,” says Lee Glascoe, the program leader for LLNL’s contribution to NEST, which includes NARAC. “And NARAC is a part of that.”

And if a suspicious substance does get released into the atmosphere, NARAC’s job is to provide information that NEST personnel can use in the field and authorities can use to manage catastrophe. Within 15 minutes of a notification about toxic materials in the air, NARAC can produce a 3D simulation of the general situation: what particles are expected where, where the airflow will waft them, and what the human and environmental consequences could be. 

In 30 to 60 minutes, they can push ground-level data gathered by NEST personnel (who are out in the field while the NARAC scientists are running simulations) into their supercomputers and integrate it into their models. That will give more precise and accurate information about where plumes of material are in the air, where the ground will be contaminated, where affected populations are, how many people might die or be hurt, where evacuation should occur, and how far blast damage extends. 

Modeling the atmosphere

These capabilities drifted into Lawrence Livermore decades ago. “Livermore has a long history of atmospheric modeling, from the development of the first climate model,” says John Nasstrom, NARAC’s chief scientist.

That model was built by physicist Cecil “Chuck” Leith. Leith, back in the early Cold War, got permission from lab director Edward Teller (who co-founded the lab and was a proponent of the hydrogen bomb) to use early supercomputers to develop and run the first global atmospheric circulation model. Glascoe calls this effort “the predecessor for weather modeling and climate modeling.” The continuation of Leith’s work split into two groups at Livermore: one focused on climate and one focused on public health—the common denominator between the two being how the atmosphere works. 

In the 1970s, the Department of Energy came to the group focused on public health and asked, says Nasstrom, whether the models could show in near real time where hazardous material would travel once released. Livermore researchers took that project on in 1973, working on a prototype that during a real event could tell emergency managers at DOE sites (home to radioactive material) and nuclear power plants who would get how much of a dose and where.

The group was plugging along on that project when the real world whirled against its door. In 1979, a reactor at the Three Mile Island nuclear plant in Pennsylvania partially melted down. “They jumped into it,” Nasstrom says of his predecessors. The prototype system wasn’t yet fully set up, but the team immediately started to build in 3D information about the terrain around Three Mile Island to get specific predictions about the radionuclides’ whereabouts and effects.

After that near catastrophe, the group began preemptively building that terrain data in for other DOE and nuclear sites before moving on to the whole rest of the US and incorporating real-time meteorological data. “Millions of weather observations today are streaming into our center right now,” says Nasstrom, “as well as global and regional forecast model output from NOAA [the National Oceanic and Atmospheric Administration], the National Weather Service, and other agencies.” 

NARAC also evolved with the 1986 Chernobyl accident. “People anticipated that safety systems would be in place and catastrophic releases wouldn’t necessarily happen,” says Nasstrom. “Then Chernobyl went wrong, and we quickly developed a much larger-scale modeling system that could transport material around the globe.” Previously, they had focused on the consequences at a more regional level, but Chernobyl lofted its toxins around the globe, necessitating an understanding of that planetary profusion.

“It’s been in a continuous state of evolution,” says Nasstrom, of NARAC’s modeling and simulation capabilities. 

‘All the world’s terrain mapped out’

Today, NARAC uses high-resolution weather models from NOAA as well as forecast models it helped develop. Every day, the center brings in more than a terabyte of weather forecast model data. And those 3D topography maps they previously had to scramble to make are all taken care of. “We already have all the world’s terrain mapped out,” says Glascoe. 

NARAC also keeps up-to-date population information, including how the distribution of people in a city differs between day and night, and data on the buildings in cities, whose architecture changes airflow. That’s on top of land-use information, since whether an area is made up of plains or forest changes the analysis. All of that together helps scientists figure out what a given hazardous release will mean to actual people in actual locations around actual buildings.

Helping bring all those inputs together, NARAC scientists have also created ready-to-go models specific to different kinds of emergencies, such as nuclear power plant failures, dirty bomb detonations, plumes of biological badness, and actual nuclear weapons explosions. “So that as soon as something happens, we can say, ‘Oh, it’s something like this,’ that we got something to start with.” 

Katie Lundquist, a scientist specializing in scientific computing and computational fluid dynamics, is NARAC’s modeling team lead. Her team helps develop the models that underlie NARAC’s analysis, and right now it is working to improve understanding of how debris would be distributed in the mushroom cloud after a nuclear detonation and how radioactive material would mix with the debris. She’s also working on general weather modeling and making sure the software is all up to snuff for next-generation exascale supercomputers. 

“The atmosphere is really complex,” Lundquist says. “It covers a lot of scales, from a global scale down to just tiny little eddies that might be between buildings in an area. And so it takes a lot of computing power.”

NARAC has also striven to improve its communications game. “The authorities make the decision, but in a crisis, you can’t just give them all the information you’ve generated technically,” Glascoe says. “You can’t give them all sorts of pretty images of a plume.” They want one or two pages telling them only what the potential impact is. “And what sort of guidelines might help their decision making of whether people should shelter, evacuate, that sort of thing,” says Glascoe. 

To that end, NARAC has made publicly available examples of its briefing products, outlining what an emergency manager could expect to see in its one to two pages about dirty bombs, nuclear detonations, nuclear power plant accidents, hazardous chemicals, and biological agents.

The sim of all fears

Recently, the team has been assisting with radioactive worries in Ukraine, where Russia has interfered with the running of nuclear power plants. It also previously kept an analytical eye on the 2020 fires in Chernobyl’s exclusion zone and the same year’s launch of the Mars Perseverance rover. The rover had a plutonium power source, and NARAC was on hand to simulate what would happen in the event of an explosive accident. Going farther back, the team mobilized for weeks on end during the partial meltdown of the Fukushima reactors in Japan in 2011. 

But one of the events Glascoe is most proud of happened in late 2017, when sensors in Europe started picking up rogue radioactive activity. Across the continent, instruments designed to detect elemental decay saw spikes indicating ruthenium-106, with more than 300 total detections. “We were activated to try and figure out, ‘Well, what’s going on? Where did this come from?’” says Glascoe. 

As NARAC started its analysis, Glascoe remembered an internal research project involving the use of measurement data, atmospheric transport models, statistical methods, and machine learning that he thought might be helpful in tracing the radioactivity backward, rather than making the more standard forward prediction. “As the data comes in, the modeling gets adjusted to try and identify where likely sources are,” says Glascoe. 

Like the prototype that DOE had called up for use with Three Mile Island, this one wasn’t quite ready, but Glascoe called headquarters for permission anyway. “I said, ‘Hey, I know we haven’t really kicked the tires too much on this thing, except they did conclude this project and it looks like it works.’” They agreed to let him try it. 

Four days and many supercomputer cycles later, the team produced a map of probable release regions. The bull’s-eye was on a region with an industrial center. “And sure enough, a release from that location would do the trick,” says Glascoe. 

The suspect spot was in Russia, and many now believe the radioactivity came from the Mayak nuclear facility, which processes spent nuclear fuel. Mayak is located in a “closed city,” one that tightly controls who goes in and out. 

Ultimately, no one can stop the atmosphere’s churn, or its tendency to push particles around. The winds don’t care about borders or permits. And NARAC is there to scrutinize, even if it can’t stop, that movement.

Read more PopSci+ stories.