MAPPING THE INVISIBLE CITY

Data science is uncovering some of the most pressing issues of urban living.

By Christian Elliott

Imagine an October day in Brooklyn. It’s 2036. A week ago, a Category 5 hurricane, intensified rapidly by warm ocean waters, spun up in the Caribbean Sea. Now, it bears down on the city. Sea levels, rising every year, mean the coastal storm surge on its way will soon cause billions of dollars in damage. Or, instead, imagine a July day earlier that same year. A heat dome has settled over the city. Temperatures are well into triple digits Fahrenheit. Those without air conditioning head to the city’s parks, fleeing the sweltering pavement.

Climate change is already reshaping New York City and its residents — a flooded subway station can send cascading effects through the power grid and daily commutes; an especially hot block shapes who lingers where on a summer afternoon. To live with the uncertainty of a future shaped by climate change, cities like New York need new ways of seeing and understanding themselves. At NYU Tandon’s Center for Urban Science + Progress, three researchers in neighboring offices are starting on that difficult task now — reinventing what it means to map a city.

To Yuki Miura, Anamika Shreevastava and Takahiro Yabe, the urban environment is more than a grid of blocks; it’s a living, organic system of water, heat and urban movement. One that, with the right tools, we can study, anticipate and prepare for disasters. Though they focus on different facets of urban risk — modeling floodwaters, dangerous heat and how people move through cities during disasters, respectively, Miura, Shreevastava and Yabe share a common goal — collecting data in new and innovative ways and turning that data into information decision-makers can use to keep New Yorkers safe. And with disasters increasingly overlapping, all three are collaborating across traditional disciplinary lines — exploring how flood simulations might plug into mobility models, and how to depict urban heat disparities in 3D like floodwaters.

A Flood of Data

Yuki Miura, Assistant Professor at the Department of Mechanical and Aerospace Engineering and the Center for Urban Science + Progress, grew up conscious of natural hazards. She is from Japan, where earthquakes and tsunamis have shaped public awareness and disaster preparedness. The devastating Tohoku earthquake and tsunami struck in 2011, just before she entered college; her high school graduation ceremony was postponed. The disaster brought home both the importance of preparation and the challenges that remain even in communities accustomed to natural hazards. When she came to Columbia University for her master’s and Ph.D. in civil engineering and engineering mechanics, specializing in flood risk and hurricanes, she encountered a different set of hazards and urban vulnerabilities. Those experiences shaped her interest in research that helps communities understand their risks and make decisions about how to reduce them.

Those research questions? How to uncover climate risks that conventional maps miss, and use that information to guide resilience investments. After a stint at Morgan Stanley, where she modeled risks to businesses, real estate and the energy sector from drought, wildfire and other climate hazards, Miura joined NYU Tandon to run the Climate, Energy, and Risk Analytics Lab. Her lab takes a three-part approach: identify risks, measure them and manage them. Through RoofNet, the team is developing a global roof-material dataset and AI methods to extract building characteristics from imagery, helping reveal vulnerabilities that existing building records often leave out. In New York City, a major focus is flooding and its consequences for interconnected infrastructure.

Different storms can flood the city in a wide variety of ways. In 2012, Hurricane Sandy did so through storm surge — mostly affecting parts of New York City close to the coast. Hurricane

Ida in 2021 did so through heavy rain — inundating low-lying areas across the city, especially wherever drainage systems were overwhelmed. And the city’s vulnerability is constantly changing — new seawalls have sprung up in places, for example, which will help with future storm surge events, but not with flash flooding from heavy rain. So, the city needs good maps of where flooding will happen in different disaster scenarios.

Street with rubble and a boat in the road. — *Aftermath of the Tohoku earthquake and tsunami. Editorial credit: Fly_and_Dive / Shutterstock.com*

*Aftermath of the Tohoku earthquake and tsunami. Editorial credit: Fly_and_Dive / Shutterstock.com*

Mapping urban flood risk in cities like New York isn’t a new concept. Since 1968, the Federal Emergency Management Agency, or FEMA, has mapped flood risk, identifying 100-year flood plains (which have a 1% annual risk of flooding) for the insurance industry and for urban planners. But there’s a problem with those maps. “It’s not very intuitive,” says Miura. “I feel like it’s hard to understand what it actually means to you. You might be in that area, but it’s hard for someone who doesn’t have a background in probability statistics to wrap your head around.”

So, Miura set out to create a better tool for community members. She started out by interviewing stakeholders, trying to figure out their understanding of various flood scenarios. Miura recalls at one event someone asked her, “Is it the kind of flood that I can wear rain boots in, or is it the kind of flood that I need a canoe for?” Her first reaction as a scientist? “Well, either way, you probably shouldn’t be going out!” But the question stuck with her — she imagined a 3D map that showed flood depths compared to the silhouetted shapes of an adult and child.

“I think looking at water height changing on a human scale and seeing your child is going to be completely underwater, that’s very alarming for anybody,” she says. “You don’t need to know statistics, you don’t need to know fluid dynamics or anything.”

Thus, GeoFlood Studio was born: an interactive 3D flood visualization platform that lets users zoom in on their neighborhood and explore modeled flood depths during storms like Sandy, making the consequences easier to understand than a flood- zone designation alone. The tool also shows evacuation routes. Miura’s lab is currently working with the New York City Department of Environmental Protection (DEP) to adapt GeoFlood Studio into a DEP platform for communicating the East Side Coastal Resiliency (ESCR) project. The platform will help residents explore the project’s flood protections and understand how they would change flooding in their neighborhoods.

“I think looking at water height changing on a human scale and seeing your child is going to be completely underwater, that’s very alarming for anybody. You don’t need to know statistics, you don’t need to know fluid dynamics or anything.”

City decision-makers can also use GeoFlood Studio to explore how protective measures, such as seawalls, change flooding across a neighborhood. The broader research goal is to connect those changes to impacts on buildings and infrastructure, helping decision-makers compare where resilience investments could do the most good. Miura also plans to add animations showing how flooding evolves over time, beyond a static map of maximum flood depths. Projects are also in the works to expand GeoFlood Studio to flood-prone areas of Italy and Miami.

And the city has taken notice — the MTA asked Miura to create a computer model to test the effectiveness of various interventions to protect stations and tunnels from flooding by sealing ventilation openings and building seawalls. And Con Ed, interested in how subway flooding affects the power grid, is now a client too.

These consequences can extend far beyond the flooded area. A hospital can lose power or communications even if floodwater never reaches its doors. Miura’s lab models the dependencies between infrastructure systems to trace how an initial failure spreads across power, transportation, communications, and other essential services. The team is developing graph neural networks, AI models that learn relationships within connected systems, to estimate these cascading impacts rapidly and help identify where protective investments could prevent wider disruption.

The work has also opened conversations about making another urban climate hazard more visible: extreme heat. To explore that challenge, Miura teamed up with fellow NYU Tandon researcher Anamika Shreevastava.

Flood zones in New York City

It’s Getting
Hot in Here

Shreevastava likes to explain her career path as a journey of zooming out. She started at the building scale, studying civil engineering as an undergraduate, where she learned how to design buildings. For her master’s, she switched to architectural engineering and learned how to keep buildings comfortable for their occupants through ventilation, mostly through air conditioning. In such cases, excess heat leaving the building’s interior simply goes outside. Zooming out to city scale, she learned that a city full of buildings with air conditioners pumping out heat measurably warms up the entire city, by a process called anthropogenic heating.

Studying that phenomenon, Shreevastava learned about the urban heat island effect–due to dense human activity, cities are warmer than their rural surroundings. For her Ph.D., she studied heat at the city scale and the feasibility of building thermally equitable cities. For her postdoc at the NASA Jet Propulsion Laboratory she zoomed out even further, studying heat in diverse cities across the world using satellite data, and learned that despite the different cultural and architectural contexts, the issues that arise from unequal extreme heat impacts are similar for us all. Now an Assistant Professor in the Department of Mechanical and Aerospace Engineering and CUSP, Shreevastava brings a multi-scale perspective on heat at all scales through her Urban Planet Lab.

Busy street with smoke; inset of heat map.

“So now I’ve managed to zoom out all the way to view the planet as a whole,” she says. “It’s an urban planet — the signature that cities leave extends beyond the city’s footprint. You can see the impact of urbanization from across the planet, and we need to figure out a sustainable pathway for urbanization. That’s where I’m coming from.”

Shreevastava’s major contribution to the field of urban heat is a concept called fractal intra-urban heat islands. Diagrams of urban heat islands typically show cities as “big domes” of heat. But really, temperatures within a city are highly heterogeneous, Shreevastava says. “Some parts are cold, some parts are hot, and it’s a cluster of many, many islets that form this complex system as a whole.” For her Ph.D., Shreevastava studied those islets and how they behave and interact.

“It’s an urban planet — the signature that cities leave extends beyond the city’s footprint. You can see the impact of urbanization from across the planet, and we need to figure out a sustainable pathway for urbanization. That’s where I’m coming from.”

Shreevastava is new to New York — she conducted most of her research on Los Angeles from her time at NASA JPL. But the key processes hold true for most coastal cities: temperatures are governed largely by the geographic setting and the background climate, such as the land-sea breeze interaction. But the local thermal heterogeneity can indeed be engineered. On a smaller scale, for instance, local parks and other green spaces provide cooling. “What I’ve found personally very remarkable is that New York, despite being a dense metropolis, is quite good at having parks every here and there,” she says. “Compact living is indeed an effective pathway for minimizing the urban footprint on the planet.”

Mapping those heat islets matters because, while the impacts of heat might not be as dramatic as hurricane-induced flooding, heat is still a deadly threat in cities. It doesn’t typically destroy infrastructure, but it’s an invisible, slow killer of the city’s residents, often the most vulnerable and isolated. “Heat is not a shock event. Heat is a creep,” says Shreevastava. “It doesn’t capture media attention in the way a remarkably disastrous flood photograph would, but it kills.” Unlike with a hurricane, there’s often a time lag of 3-7 days before deaths are reported — patients come into hospitals in cardiac arrest or other symptoms. “Heat exaggerates other underlying diseases and health conditions in people, and that’s often reported as the cause of death, so it is easy to miss the true scale of heat impacts. It’s a very insidious problem to track, but we know it’s among the leading causes of death in the United States.” What’s more, heat discriminates, unlike, say, a flood, based on income — those who can’t afford air conditioning are most affected.

Heat map of the New York City metro area.

In a recent study, Shreevastava looked into the degree to which historic redlining from the 1930s to ‘60s — the process of systematically denying mortgages to predominantly minority neighborhoods, thus further segregating U.S. cities — impacts disparities in urban heat today. In other words, are neighborhoods that were redlined hotter than ones that were not? That is largely the case in Los Angeles, Shreevastava’s mapping work showed, but a stronger predictor now is the present-day income. Once-redlined hilly and coastal areas with cool sea breezes are now affluent neighborhoods — like Beverly Hills and Santa Monica. Poorer neighborhoods are, by and large, hotter, with less access to green space.

The city “can’t just rip up its fabric and rebuild,” says Shreevastava, but decision-makers should be aware of the long legacy of past decisions that created inequality. “The decisions they make today determine the thermal landscape of tomorrow,” she says, “because climate change is happening now, it’s warming now.” The equitable solution isn’t as simple as investing in shade to make neighborhoods cooler — doing so could force out lower-income residents when property values inevitably rise.

In New York, the city wants a similar analysis, and they want a 3D map like Miura’s flood map for urban heat. Heat varies neighborhood to neighborhood, as Shreevastava’s past work shows, but it also varies on the street level as it moves around buildings and through the city. But there’s an unsolved challenge that makes this work tricky — and it comes down to air temperature versus surface temperature. For her Los Angeles research, Shreevastava used data from ECOSTRESS, a thermal radiometer mounted on the International Space Station. But from remote sensing we can only measure the surface temperature — the temperature of the ground. Imagine you’re standing on a very hot parking lot. The air could still be tolerable — air temperature is what we actually feel. To create a 3D map, Shreevastava would need a network of sensors throughout the city measuring air temperatures, which isn’t yet in place.

The Human
Element

In the meantime, there’s one thing flooding and extreme heat have in common — they make people move. Studying that movement is Takahiro Yabe’s forte. He’s a civil engineer by training, but uses large-scale human behavior data collected from mobile devices to study how disasters of all kinds — floods, pandemics, earthquakes — affect the way people move through cities. His ultimate goal? To predict how people evacuate, and when they return.

Yabe, Assistant Professor of Technology Management and Innovation and CUSP, got into this line of work, like Miura, thanks to the 2011 earthquake and tsunami in Japan. It struck during his first year of college. “I had to walk five hours to get home that day because the transportation collapsed,” he says. “So as a college kid who’s ambitious, I thought, I have to do something about this, right?” He saw a lot of discussion in the aftermath about building seawalls and strengthening buildings, but comparatively little about increasing preparedness and resiliency through understanding human behavior. That, he decided, would be his niche.

For Yabe’s Ph.D. at Purdue University in the U.S., he studied how infrastructure affects recovery after a disaster — how long it takes businesses to reopen and when residents return. That requires “thinking of cities not just as networks of hard, grey infrastructure, but also as dynamic and complex systems where people are living and interacting with that infrastructure,” he says. “People’s behavior is determined by the physical structure of the city, but that behavior also shapes how cities are being built.”

Early on, mobility data was pretty crude. In an attempt to predict whether people would evacuate from Hurricane Sandy in 2012, Yabe relied on geotagged Twitter posts — users who tweeted lots of disaster-related words were more likely to evacuate. Cell phone mobility data made things much easier — access to anonymized data that included where people went throughout their days and what they did and liked, and who else they associated with. That data allowed him to ask more complex questions, like when people come back after disasters. That’s his novel contribution to the field — for all disasters, there’s a recovery “curve” — populations return in an exponential decay pattern. The steepness of that curve can vary, but with a few variables — the initial evacuation rate, the speed of return in the first couple of days — Yabe can predict how many people will end up returning longer term. “It was like, ‘Wow, this is crazy,’” he says. “And it was one of the papers that I’ve published I’m most proud of.”

While he was defending his Ph.D., another disaster struck — the COVID-19 pandemic. He watched, as a new postdoc at MIT’s Media Lab, as his favorite seafood restaurant next door to campus shut down. It depended on faculty as customers, who were now working remotely. That realization shifted his work in the direction of economic dependencies in cities. After any disaster, “there’s going to be knock-on effects, or spillover effects, in economists’ words, on neighboring businesses, and that’s been under- studied,” he says. Traditionally, urban planning just used distance — assuming two businesses were dependent on one another as long as they were close. Yabe’s innovation is using cell phone movement data, which provides insight into the daily habits, routines and preferences of New Yorkers.

That pandemic-era research was a “big hit,” he says. It revealed hidden patterns in the city and problems with urban design. “You might have no correlation between two different nearby places, like a Five Guys and an Opera House,” which is the case in the Back Bay in Boston, he says. Those two businesses might have zero customer overlap. “It’s not just about where these places are, but it’s actually behavior that links places,” he says. “That’s been the really cool insight.” Now, organizations like the Downtown Brooklyn Partnership have reached out for help selecting retail tenants for neighborhoods and even festivals for urban parks that will attract the right demographics of customers for nearby businesses.

That’s not to say he isn’t still very interested in disasters. Yabe’s been talking with Miura about studying the longer-term impacts of flooding on city infrastructure. “You can do these flood simulations and say, ‘This building can be flooded by X percent within the next Y years,’ but can you say what is the secondary impact of that?” he says. “Because human behavior is going to change.”

Yabe’s also planning collaborations with Shreevastava, whom he studied with at Purdue. In collaboration with the NYC Department of Parks and Recreation, his students recently studied city park usage in hot weather using mobile phone data to track visitation. During hot days, they found, fewer people visit the city’s 1,200 parks — but those who do visit are disproportionately lower income. That validates her findings in Los Angeles, Yabe says. “The human mobility analysis and built environment characteristics, whether it’s heat, flooding, pollution, anything, it’s really easy to overlay and combine.”

New York City’s future under climate change is far from certain. But one thing is — Yabe, Shreevastava and Miura, these three colleagues and friends, they will continue to overlay, combine and collaborate to map changing neighborhoods and help people adapt as best they can.