Using traffic cameras and mobile phone data, MIT researchers built a system that maps car emissions in Manhattan block by block and hour by hour. The approach could help cities design smarter, fairer climate and transportation policies.
Imagine being able to see, almost in real time, which city blocks are choking on exhaust and which streets are suddenly cleaner after a new traffic rule kicks in.
MIT researchers say cities can now do just that.
In a new study focused on New York City, a team from MIT’s Senseable City Lab has shown that existing traffic cameras and anonymized mobile phone data can be combined to generate a highly detailed, near real-time picture of vehicle emissions. The method produces block-by-block, hour-by-hour maps of pollution that could help city leaders design and test transportation and climate policies with far greater precision.
The work, published in the journal Nature Sustainability, aims to fill a major data gap. Traditional citywide emissions inventories average pollution over large areas and long time periods, while highly detailed studies often focus on a small number of vehicles and are hard to scale. The MIT framework sits in between, offering citywide coverage with fine-grained detail, using data that many cities already collect.
Co-author Paolo Santi, a principal research scientist in the Senseable City Lab, noted the approach hinges on combining multiple streams of information.
“Our model, by combining real-time traffic cameras with multiple data sources, allows extrapolating very detailed emission maps, down to a single road and hour of the day,” he said in a news release. “Such detailed information can prove very helpful to support decision-making and understand effects of traffic and mobility interventions.”
The study was conducted by former MIT postdoctoral researcher Songhua Hu, now an assistant professor at City University of Hong Kong, Santi, MIT Senseable City Lab researcher Tom Benson, MIT Senseable City Lab visiting doctoral student Ashutosh Kumar, MIT Senseable City Lab director Carlo Ratti, and collaborators at Arizona State University and Hong Kong Polytechnic University.
Hyperlocal view of Manhattan’s traffic pollution
To test their method, the researchers turned Manhattan into a living laboratory.
They drew on images from 331 traffic cameras already installed at intersections across the borough. Using computer vision, they classified vehicles into 12 broad categories, such as cars, buses and trucks. The system correctly identified about 93% of vehicles, giving the team a detailed snapshot of what types of vehicles were on which streets, and when.
Those cameras also captured how traffic lights shape driving behavior. Stop-and-go patterns caused by signals are a major driver of urban emissions, but they are often glossed over in conventional inventories that assume smoother traffic flow than many city streets actually experience.
The team then layered in anonymized location records from more than 1.75 million mobile phones. Those data revealed how vehicles move through the city over time, not just at camera locations. By combining the traffic patterns from phones with the vehicle types seen on camera, and linking both to known emissions rates, the researchers could estimate how much pollution different streets were generating at different hours.
Hu noted the power of the approach lies in reusing data that cities and telecom providers already have.
“The very basic idea is just to estimate traffic emissions using existing data sources in a cost-effective way,” he said in the news release. “We just need to input all emission-related information based on existing urban data sources, and we can estimate the traffic emissions.”
Testing “what if” policies before they roll out
Once the model was built, the team used it to explore how different policy choices might change emissions.
In one scenario, they simulated what would happen if a share of trips shifted from private cars to buses. In another, they examined the impact of spreading out rush hour, so that the same number of trips occur over a longer period, reducing peak congestion.
They also tested what happens when cities rely on rougher, citywide averages instead of detailed, street-level inputs. The researchers found that using simplified averages could lead to emissions estimates that were off by as much as nearly half in either direction compared with their fine-grained model. That suggests that small shortcuts in data can translate into big errors when evaluating policies.
Real-world test: New York’s congestion pricing
The most striking test came from a real policy change. In January 2025, New York City implemented congestion pricing for vehicles entering Manhattan south of 60th Street, charging drivers a fee to reduce traffic in the busiest part of the city.
The MIT team examined traffic and emissions at two, four, six and eight weeks after the program began. They found that while traffic volume in the zone dropped by about 10%, emissions fell by roughly 16-22%, a much larger reduction.
Those findings line up with earlier work by Cornell University researchers, who reported a 22% drop in fine particulate pollution (PM2.5) inside the pricing zone. The MIT analysis also showed that the benefits were not evenly spread: some major streets saw much larger improvements, while areas outside the zone experienced more mixed effects.
By pinpointing where and when emissions fall, the model could help city officials refine congestion pricing or other measures to make sure they are effective and equitable.
A new toolkit for cleaner, fairer cities
The Senseable City Lab’s work is part of a broader push to measure environmental conditions at the “hyperlocal” level — down to individual blocks and hours.
Ratti has described the emissions framework as one piece of that larger effort, which “is part of our lab’s ongoing quest into hyperlocal measurements of air quality and other environmental factors. By integrating multiple streams of data, we can reach a level of precision that was unthinkable just a few years ago — giving policymakers powerful new tools to understand and protect human health.”
The researchers emphasize that their method also protects privacy. The computer vision system recognizes vehicle types, not license plates, and the mobile phone data are anonymized and aggregated.
Looking ahead, the team sees opportunities to plug in even more data sources. In related work in Amsterdam, for example, they have used dashboard cameras from vehicles to capture detailed information about how cars move through the city. That kind of input could further sharpen emissions estimates in places where traffic cameras are sparse.
For students and young researchers, the project highlights how data science, urban planning and environmental engineering can come together to tackle climate and public health challenges. By turning everyday technologies — traffic cameras, smartphones, even dash cams — into tools for environmental monitoring, the MIT team is offering cities a way to see the invisible and act on it.
If widely adopted, their framework could help urban leaders test ideas on a computer before rolling them out on the street, track whether policies are working in near real time and ensure that the benefits of cleaner air reach every neighborhood, not just a lucky few.