Edit Module

The Doomsday Squad

A team of computer modelers at Argonne National Laboratory is preparing for the urban calamities most Chicagoans would rather not think about.

Photo: Getty Images/istockphoto

The suburban campus, as tightly screwed down as a bank vault, unfurls across a manicured landscape of low-slung white glass-and-steel structures like something out of a sci-fi writer’s imaginings. It all feels disconnected from the hive of city life 25 miles away: the L trains clattering, the ambulances whoop-whooping, the millions of people going about their days without giving a passing thought to earthquakes, terrorist attacks, plagues, and other cataclysms. But inside the guarded gates of this cloistered realm in Lemont, Illinois, thrums a group of researchers who think of little else.

The leaders of the Global Security Sciences division at Argonne National Laboratory don’t look the part of superheroes—more like PTA members. There’s Nate Evans, the head of cyberthreats; Megan Clifford, who handles infrastructure, such as power grids, telecommunications, and highways; and Charles Macal, the leader of the Social and Behavioral Systems Group.


But much of what they and their 270-plus full-time colleagues do is the stuff of a Michael Bay disaster flick. Harnessing Argonne’s massive computing power, they think about precisely what most of us would prefer not to. That is, they imagine the disastrous, envision the catastrophic, and intricately model the apocalypse in its various and horrific guises—crippling cyberattack, dirty bomb, Ebola outbreak, and so on. Their goal isn’t to come up with a screenplay, but they are creating narratives nonetheless: building minutely detailed predictive models that show how a given disaster might unfold and, not for nothing, how we might alter the story before the plot turns ugly.

“We want to be able to say, ‘Based on a lot of hardcore science and math, here’s what we believe could transpire if this were to happen,’ ” says Clifford. “Then people can start to figure out those resilience enhancements or intervention strategies.” In other words, better ways to be prepared when the curtain goes up on the shit show.

In the unsettling story lines developed by the GSS, which was created in 2014, Chicago is often cast in the leading role, serving as a stand-in for other cities. With soaring skyscrapers, one of the world’s busiest airports, traffic-snarled highways, packed commuter trains, sports stadiums galore, and (fun fact) one of the largest internet hubs on the planet, Chicago is, in the estimation of disaster modelers, a “huge asset,” says Evans.

Lucky us.

GSS researchers Megan Clifford, Charles Macal, and Nate Evans
GSS researchers Megan Clifford, Charles Macal, and Nate Evans. Their predictive models can foretell the course of a disaster in granular detail. Photo: Jeremy Bolen

Last fall, looking for a unique way to draw attention to their work in modeling the spread of highly communicable diseases like MRSA and Ebola, some of the GSS team members posed a question: What would happen if some zombies—oh, say, 25 or so—were turned loose on the city and started biting folks? How long would it take before the entire city was a bunch of undead brain eaters? “We took an existing computer model we’ve been applying to infectious disease transmission and propagation,” says Evans, “and Chicago was our test case.”

Referring to it as a “computer model” is a little like calling a Mars mission a trip to the Adler Planetarium. The researchers at the GSS boast one of the fastest supercomputers in the world. Mira, Argonne’s mammoth IBM Blue Gene/Q system, consists of 48 server racks capable of 10 quadrillion calculations per second—allowing it to do in one day what would take 20 years on a desktop Mac.

This power enabled GSS scientists to predict the progress of a zombie contagion in ticktock detail by using what’s known as agent-based simulation. The particular simulation the GSS team used for the zombie exercise is called ChiSIM (Chicago Social Interaction Model). Based on detailed demographic data (population density by ZIP code, socioeconomic status, gender, household structure) and surveys revealing how city dwellers spend their time (work, play, dining out, taking the kids to school, hailing a cab), ChiSIM is, in essence, a fully actualized virtual population: millions of computer “scripts” that mirror the ages, locations, activities, work and leisure habits, and social proclivities of Chicago’s real 2.7 million souls. Evans describes it as a “unique set of data that allows us to know a type of schedule for everyone in Chicago.” They’ve even modeled the movements of students going to school and playing sports, the better to accurately predict disease transmission pathways.

“We let the computer do what it’s good at,” Evans says. “Just crank, working all night, you know, finding these scenarios. Then the computer can tell us, ‘Look, here are the most vulnerable places, here are other areas that need some further study.’ ”

A video posted on YouTube by the GSS just before Halloween, complete with a fog-in-the-cemetery soundtrack, showed the day-by-day spread of the zombie contagion in its full glory: rapidly multiplying red dots marching unchecked across the city of Chicago. By day 7, the city map looked as if it had contracted a bad case of measles. By day 30, huge splotches engulfed entire neighborhoods. By day 60, the city was full-blown Dawn of the Dead—no surprise to anyone who’s seen a zombie movie or a medical disaster thriller like Contagion. One zombie infects five people, all of whom bite five more people, who each bite five more, and so on.

The video had the intended effect, generating lots of hits and going, ahem, viral worldwide. But it didn’t even touch on the truly fascinating stuff revealed by the exercise, which was meant to mimic the propagation of an ultracontagious disease like Ebola or some yet-to-be-discovered strain of influenza.

Consider, for example, what the model shows about hospitals. They are obviously the most critical hub in the outbreak of a contagion—in terms of treatment, certainly, but also for establishing protocols for quarantining people—and as soon as they become anything less than fully operational, the effect of the outbreak grows exponentially. But doctors and nurses are people, too, Macal suggests, meaning they are as frightened to catch a deadly disease as anyone. Accordingly, in scenarios like the ones the GSS modeled, numerous medical workers refuse to come in. Hordes of new cases, in turn, overwhelm the few facilities that are staffed, infecting the brave souls who report for work. More infections lead to more people seeking treatment that isn’t there, which leads to more hordes in the streets biting more people.

The models also revealed that transportation systems, grocery stores, water purification and treatment facilities, power grids, and other businesses and services would stop functioning for lack of people to run them. Airports: down. Trains: down.


The GSS scientists introduced different variables to understand how they might accelerate or slow the pace of the outbreak. One interesting thing they found was that even small “wins”—killing a zombie here or there, providing information more quickly on where to take shelter—yielded big results, cutting the number of infections by hundreds of thousands. In the event of a real contagion, like Ebola or MRSA, that kind of information could allow hospitals to activate protocols—how and where to set up quarantines, what kind of protective garb to wear, how best to determine the pathways the disease would travel—that would help medical workers feel safe enough to come in. That would slash the numbers of those turned away untreated, which could be the key factor in turning the tide of an epidemic.

The GSS scientists ultimately share their scenarios and recommendations for new preparedness protocols with local, state, and federal authorities, identifying areas of vulnerability. Alas, they can’t make anyone implement them. “That’s not our role,” says Clifford. “Our role really is to provide the best available science and engineering and mathematics to help inform the decisions that need to be made.”

Let’s hope the people in charge are listening.


On December 16, 1811, an 8-magnitude earthquake rocked northeast Arkansas. Sparsely populated at the time, the region was spared a massive death toll, but by all measures the event was catastrophic. Huge chasms opened, rivers altered course, sand and water shot up through fissures in the earth, structures hundreds of miles away collapsed. Aftershocks nearly as powerful as the original temblor shook the country’s midsection, from Arkansas to Illinois, for two months. The fault line that generated that earthquake is known today as the New Madrid seismic zone. It runs as far north as southern Illinois.

If the image of toppling skyscrapers has you pricing out-of-town real estate, rest easy: Though the fault poses a threat downstate, and while it has triggered small earthquakes in Chicago in the past, seismologists say the likelihood of a major seismic event hitting the city itself is remote.

In the worst-case mindset of the GSS scientists, however, that’s plenty likely enough to justify detailed modeling of urban earthquake scenarios. Those scenarios, it turns out, yield useful information for cities well outside of seismic zones. Clifford and her colleagues point to some responses to recent natural disasters where the kind of modeling the GSS does would have been immensely valuable. For example, during Hurricane Sandy, officials at one of New York City’s major telecommunications companies stored the fuel tanks for their backup generators in the basement of their Lower Manhattan facility. When the basement flooded, the fuel tanks became inaccessible, rendering the backup generators useless. Cellular communications went dark for many thousands of customers. In the aftermath of the storm, Clifford says, the company spent enormous sums to replace and upgrade damaged equipment—expense and disruptions that could have been avoided if officials had had models showing their facility’s vulnerabilities in more detail.

What Clifford and her group focus on most closely in their models are the “lifeline infrastructures”: energy, transportation, water and wastewater, and telecommunications. “Those are all the things that people are relying on. Every day we pick up our cell phones. Every day we get on a road and probably drive somewhere or ride somewhere. We turn on our light switch. We expect the tap to run. Those are the services that are essential to our livelihoods.”

The earthquake models also reveal in chilling detail the cascading nature of infrastructure failures. Water-pumping stations, for example, “are dependent on power,” says Clifford. “So if [the power goes out] and water can’t be pumped or, let’s say, you can’t get transportation in with the chemicals to treat the water, then you’re not going to have purified water.” Such vital facilities have backup generators, of course, “but that backup generation is dependent on fuel,” Clifford points out. “Then how do you get the fuel in? Well, you need roads and trucks to get the fuel in. Are the roads accessible? Are rail signals or street signals working? It’s like a full circle.”

The same goes for the food supply: How do you get goods into the city if the roads are out? Clifford says her team’s modeling accounts for rerouting deliveries based on different road-failure scenarios. But delivering food is just one problem; paying for it is another. One of the greatest vulnerabilities the GSS earthquake models have revealed is our reliance on credit cards. “If the power is out, you can’t use a card at a gas station, you can’t use a card to get food.” The city reverts to an all-cash economy.

In light of such scenarios, it’s not surprising that many of the GSS’s recommendations for disaster preparedness have to do with restoring and conserving power or finding alternate sources. The GSS has been collaborating with a division at Argonne that’s working on a new generation of long-life batteries—ones that may be able to push electric cars well beyond the 400-mile threshold or extend cell phone life to weeks instead of hours.


One truth Clifford stresses is that, in our heavily interconnected era, a disruptive event like an earthquake wouldn’t have to be local to have drastic local effects. She describes a scenario the team has been modeling. “We went into North Dakota—hypothetically—and we looked at the natural gas and electric power systems,” she says. “Then we said, ‘What if these two substations failed? What does that mean for North Dakota?’ We were thinking that’s the only area it would impact.

“Well,” she continues, “they distribute natural gas downward to four states, Illinois being one of them. So when the fuel is curtailed going down, all of a sudden you see the service areas now that are going out.” The upshot: Two failed substations in the remote Great Plains can knock out power to a whole swath of the country, including Chicago.

The good news, Clifford says, is that by running such a scenario, “we can look at restoration models that say, ‘This is what you would want to get back online first so that the country’s back up and running.’ ”


The students seated in front of computers in Argonne’s Theory and Computer Sciences Building were frantic. Each group had been given a virtual power company, complete with its own website and imaginary members of the public trying to pay bills. Their task: to shield their companies from hackers. Playing the part of the cyberintruders were Argonne’s cyberexperts, as well as other computer science students and volunteers from cybersecurity companies such as Chicago-based Juniper Networks.

The students on defense—some 75 in all, divided into eight teams and drawn from seven universities in Illinois and Iowa—weren’t faring well. When the power stopped flowing at their station, a light bulb next to them went dark. And each time a power company was compromised in any way—its website defaced, its power grid attacked—a message would flash on a nearby wall screen, triggering a lot of groaning and laughing.

“There were actually a couple teams that were completely wiped out,” Evans told me later, “yet they still sort of came back and tried to stand stuff up quickly. This is what you would do in the real world: ‘My power grid’s down. Well, I need to get that working again. What can I do, really quickly, to set something up and make it work?’ ”

As the 2016 presidential election demonstrated, cybersecurity is a major national issue. There’s a special urgency in Chicago, which is home to one of the biggest internet network hubs in the country: a massive “carrier hotel” on the Near South Side called the Lakeside Technology Center. Last fall, a single glitch at that data center caused major disruptions of service, from the Midwest to the East Coast. “There was a configuration mess-up that prevented the interconnects from actually occurring properly,” says Evans, “and it made a considerable portion of the internet inaccessible for 10 or 12 hours.” Weeks earlier, malware had invaded a similar connector on the East Coast, shutting down websites such as Spotify and Amazon and crippling Twitter.

Among the concerns the GSS is addressing is the ability of hackers to take control not just of the internet’s hubs but of its spokes—specifically, things like routers, webcams, and, increasingly, web-enabled home amenities like baby monitors and TVs. The malware used in the East Coast attack, in fact, was specially designed to scour the internet for devices to manipulate remotely. Among them: onboard Wi-Fi systems in cars and trucks. The newest models have web access, and this opens the door to any number of scary scenarios—particularly ones in which hackers commandeer a vehicle’s steering, brakes, and other computer-reliant features. The advent of driverless cars conjures a whole new set of nightmares. Imagine entire fleets of trucks under the control of a bad guy.


Evans and his cybersecurity team have been preparing for just such scenarios. “We actually have people that hack into vehicles and create defenses to stop that from happening,” he says. His team is partnering with Argonne’s Advanced Powertrain Research Facility to prevent automotive hacking, as well as to create traffic prediction models to see just how much chaos could be unleashed in various hacking scenarios. What their models show, among other things, is that gridlock can be achieved on a massive scale in a relatively short amount of time by causing accidents in just a few key locations, cutting off the entire urban area from any incoming or outgoing transit.


Do the men and women of the GSS sleep well at night? To hear Clifford tell it, absolutely. For one thing, she says, they come home each day with the satisfaction of having done work they believe in. “For me, getting into this business was about making a difference and having a positive impact,” she says. “Did it mean that I had to become more aware of some of the risks that we’re faced with? Absolutely.”

Indeed, given the nature of their work, it might be expected that members of the GSS would take extreme protective measures in their own lives—hoarding batteries and bottled water, say, or converting cash to gold bars. For Clifford, at least, that’s not the case: “I think it is good to change our behaviors in a positive way, whether it’s because we’ve experienced a disaster or we know someone who has. But we all live with some level of risk. I also am going to live my life.”


Edit Module
Edit Module
Submit your comment

Comments are moderated. We review them in an effort to remove foul language, commercial messages, abuse, and irrelevancies.