Linda Poon is an assistant editor at CityLab covering science and urban technology, including smart cities and climate change. She previously covered global health and development for NPR’s Goats and Soda blog.
The age-old strategy is “see a rat, kill a rat.” The new plan is to end an infestation before it ever begins.
It used to be said that there was one rat for every person in New York City. If that were true, there would be as many as 8.6 million beady-eyed rodents lurking in the streets, alleys, and tunnels of Gotham. That claim has been debunked, but good luck finding comfort in the more modern estimate that New York is a city of a mere 2 million rats.
The thing is, no one really knows how many rats there are. Not in New York City, nor Washington, D.C., nor Chicago—all three of which rank among the most rodent-infested cities in the U.S. Even the best numbers available are educated guesses, often based on surveys about calls to 311. It’s even more challenging to understand how rats move or anticipate where they will wreak havoc next. But cities have entire teams dedicated to controlling the rat population, and for good reason. The rodents sneak into buildings and subways, nibbling on trash, chewing through cables, and spreading nasty diseases, often while remaining out of sight. Despite the best efforts to deploy traps, poison, and—yes—even cats, the rodents persist.
“Rats generally are thought of as pests, so the [strategy] is usually very straightforward: There's a rat, let's get rid of it,” says Michael Blum, an ecologist who heads the New Orleans Rat Project at Tulane University.
But some cities are shifting gears to get one step ahead of the rats. They’re teaming up with researchers and computer scientists not just to eliminate the pests, but also to understand, as Blum puts it, “the subtleties of what sustains the populations.” And their number one weapon is big data.
311 smells a party of rats
For the past 12 years, Gerard Brown has been trying to figure out how to keep rats away from the nation’s capital. As program manager for Washington, D.C.’s Rodent Control Division, he’s considered a local rat guru—but the answer to his challenge may just come from someone who didn’t know anything about rats before this year.
In January, Brown partnered up with Peter Casey, a senior data scientist in D.C.’s Office of the Chief Technology Officer. Casey is working with the the city’s new applied research team to develop a machine learning model to predict rat infestations. Their goal is to build a model that can tell Brown’s team which neighborhoods will see a surge of rats days, even weeks, before they happen, giving him a head start on making the areas as unwelcoming to the vermin as possible.
To start, the team turns to the city’s records on 311 calls about rodent sightings—for now, that’s the most reliable insight for finding rodent hotspots. Then they compare those calls with other city data, including the number of registered businesses (particularly food businesses), apartments (an indication of human density), and the breakdown of the area’s landscape (concrete versus “penetrable Earth” like parks). They’re looking for patterns that signal favorable conditions for rats, highlighting places where Brown’s team could soon find an infestation.
That still doesn’t reveal the full picture: Some people may be more likely to call 311 when they see rats, multiple people could call about the same animals, and others might not even think to pick up the phone. So the data team zeroes in on places where Brown has already inspected and actually found rats, using that intersection of data to fuel the computer models that could soon take much of the guesswork out of predicting the rat infestations of the future.
It’s still early in the project, and the effectiveness of 311-based predictions remains up in the air. But early results from other cities are encouraging.
In 2011, Chicago faced a rampant rodent problem. Records show that the 311 call center received more than 34,000 related complaints that year alone. Desperate for a more effective strategy, the city enlisted the help of Daniel Neill, a computer scientist at Carnegie Mellon University’s Event and Pattern Detection Lab. Neill had been working with the city’s innovation team on software to predict hotspots of crime. The city asked if he could tweak the program to help predict future rat complaints.
Neill had three years’ worth of rat data to work with, plus a trove of other complaints, relating to things like overflowing trash bins, tree debris, food poisoning at restaurants, and building vacancies. In his search to find the signals that precede clusters of rat complaints, he fed all that data into his program to determine how accurate it was in predicting where and when populations would spike next.
Perhaps not surprisingly, the data team found that 311 calls related to food and shelter were the strongest predictors of a rat infestation. So areas where residents called about sanitation violations or tree debris were most likely to be see a spike in rat complaints. “Any one call type in isolation is a weak predictor of rats,” Neill says. “The real bang for our buck comes when all of those call types are integrated.”
The result was impressive: Neill’s model could predict a spike in rodent complaints a full week before it happened.
He eventually showcased this method to officials in Baltimore and Pittsburgh, while Chicago’s sanitation department put the predictive model to the test around 2013. After an error in the trial period, Neill says he still lacks some crucial quantitative evidence about the accuracy of his model, but he’s confident in the results based on historical data. In fact, the city of Chicago is still running Neill’s predictive analytics approach and has touted that it’s 20 percent more effective than the traditional method of baiting rats after they’ve been discovered.
Can curiosity kill the rat?
While Neill’s model is based purely on data, Casey in D.C. says his team depends heavily on Brown’s expertise in rodent control. Not all abatement methods are created equal. Using rodenticide may be more effective in one part of the city while deploying smart, heavy-duty trash bins may be the key in another. Brown provides the context, revealing the subtle observations about rat movements that help Casey’s data team narrow the scope of their analysis.
“One of the big things was getting in touch with Gerard and [making sure] that I'm not just creating a predictive model in a vacuum,” says Casey. “But we wanted to make sure that this is something that the team could use in the field.”
The fact is, numbers alone say very little about rats. We know there are a lot of them, and we know they live close to humans, but they remain a mystery even to leading experts like Jason Munshi-South at Fordham University.
“Cities are such complex environments, especially with three dimensional and subterranean micro habitats that rats exploit, and some of them we can't even really access like sewers,” he says. “We don't always know what they're doing when they're out of sight, and they're notoriously difficult to count, so even understanding the basic question of how many rats there are is extremely difficult.” It doesn’t help that humans are constantly manipulating their environment in an effort to get rid of them.
So far, research suggests that rats stay close to their burrows. In fact, the average rat might only venture a few dozen feet in its entire lifetime. They also stick to the same paths, creating “runways” that connect their homes straight to their food source. These trails sometimes appear as streaks of grease left behind from their fur, Brown says, or footpaths embedded in the dirt.
It’s still a mystery, though, as to why rat populations can suddenly rebound after being exterminated in an area. “One possibility is that [when] you knock the population down, the survivors have more resources, so they breed faster,” Munshi-South says. “Or it could be that there are rats nearby that weren't affected, so they started dispersing into the now empty territory.” We just can’t say for sure.
But researchers are right on the rats’ tails. Munshi-South has been studying the genetics of rats across Manhattan to determine how closely one colony is related to another elsewhere in the city. His lab has captured more than 500 rats from almost all of the borough’s 40-some ZIP codes, and sequenced the genome of around 250 of them, looking to see how the rats relate to one another.
So far, his lab has found more evidence that rats really don’t travel too far. Rats in northern Manhattan are more closely related to one another than to rats farther south, and vice-versa.
Then there are the more curious patterns. “Rats start to split into two major groups right around Midtown, where you have kind of an uptown set of rats and a downtown set of rats,” Munshi-South tells CityLab. He can’t say why for sure, though. For one thing, the downtown rats are more diverse, possibly because there’s more of them overall.
The urban makeup of Midtown itself could be what’s separating the two populations. Historically, Midtown just isn’t as popular for rats as the areas north and south of it, perhaps because of the abundance of office buildings and better maintenance. That’s created a relative “no-go zone” for rats, he says.
For now, Munshi-South can’t tell if this pattern holds true for other cities, but he hopes to find out. He’s teamed up with a slew of other rodentologists from cities including Vancouver and New Orleans to get at that very question. And one of his partners is Tulane University’s Michael Blum, who for the past four years has been in New Orleans studying rats in the aftermath of Hurricane Katrina.
Much of Blum’s study focuses on the Lower Ninth Ward, where vacant and redeveloped areas offer a mix of landscapes to study. He and his students have been painstakingly trapping as many rats as they can to study their movement and create their own predictive model. Trapping rats is a feat in itself: rats are clever and don’t easily fall for new traps. “The lack of curiosity probably saved the rodents,” Blum says. But for him, 311 data just isn’t good enough.
So he’s created a model of “nodes” and “pathways” to show how rats repopulate parts of the city. Think of nodes as city blocks or parks— the places that sustain a local population of rats—and pathways as the alleyways and streets by which they travel.
“We're running different hypothetical scenarios where we can literally take out [rat populations from] particular blocks or neighborhoods, and simulate what the likelihood will be that those areas will be recolonized from different parts of the city,” he tells CityLab. “We can estimate the time it takes, the number of individuals that move from different parts of the city, so basically how the whole city landscape readjusts based on a particular control effort or, say, a deficit where you have a very low population density.”
Blum says the model isn’t just based on information collected about the animals, but also on a mountain of archival data provided by the city. “Not only are we very carefully estimating population density across the city, we also have very rich GIS and environmental archives of have land cover data, municipal coverages that tell us about the structure of the streetscape and the block sizes, as well as the socioeconomic landscape across the city.”
Still, it’s a work in progress. Blum’s team recently secured funding for another six years of research, and he hopes they can eventually hand over the model to New Orleans officials and other cities so they can see the benefits. After all, rat traps and birth control initiatives barely scratch the surface of a much deeper problem: The rise of rats isn’t just about the animals themselves. Humans inevitably play a part.
Back in D.C., Brown is also tasked with educating the community about why a city block is infested by rats and make recommendations about what businesses and residents should do to address it.
“There are a few things that have to happen to have people show up to a rat meeting: You have to have a trash problem, you have to have rats, and you have to have people who are sick and tired of it,” he says. “One thing that I tell people at the beginning of these meetings is that they have to follow my recommendations or we’ll be back here next year talking about the same things.”
So, sure, rat poison can be effective, and so can big data. But, perhaps the key to eliminating rats lies in something else, says Brown: “It’s about human behavior.”