MIT developed a system that measures traffic emissions almost in real time, down to the level of an individual street
Researchers from the Massachusetts Institute of Technology have presented a new method that enables significantly more precise monitoring of road traffic emissions in a city, almost in real time and at the level of an individual road and hour of the day. It is an approach that combines existing city cameras, anonymized movement data, and computer vision to obtain a detailed picture of where, when, and because of which traffic patterns the greatest burden on air quality and the climate is created. The study was published in the journal Nature Sustainability, and the research team tested it in Manhattan, one of the most traffic-burdened urban areas in the United States. According to the authors, such a tool could become important support for city authorities when deciding on traffic regimes, transport decarbonization, and evaluating the effects of new public policies.
Unlike classic emissions inventories, which often provide a general citywide picture or rely on occasional sampling, the new model attempts to capture the actual dynamics of urban traffic. This means it does not look only at how many vehicles pass through a certain area, but also what kind of vehicles they are, how they move, how often they stop and accelerate again, and how their emissions are affected by traffic lights, congestion, and changes in travel demand during the day. This part is especially important in the centers of large cities, where short stops, queues, and so-called stop-and-go driving often produce more emissions than could be concluded from averages at the level of the entire city.
How the new model works
The team from the MIT Senseable City Lab used a network of 331 traffic cameras installed at intersections in Manhattan, along with anonymized location records from more than 1.75 million mobile phones. The cameras were used to recognize vehicle types, without collecting license plates or other personal identifiers, while the mobile data helped reconstruct broader movement patterns through the city. The researchers classified vehicles into 12 broad categories and, according to the data they cite in the paper and MIT's presentation of the research, achieved about 93 percent accuracy in assigning vehicles to the appropriate group. This gave them a sufficiently reliable basis to connect traffic flows with known emission rates for different types of vehicles and driving modes.
The special value of the model lies in the fact that it introduces signaling and traffic behavior at intersections into the calculation. In many conventional estimates, this element is precisely what remains underestimated or completely omitted, although city traffic lights significantly determine the rhythm of vehicle movement. When a car has to stop several times on a short stretch and then accelerate again, emissions can rise noticeably, and such a pattern is typical of densely built metropolitan centers. Because of this, the same traffic section does not pollute equally at every time of day, even when the number of vehicles remains similar. The new method attempts to capture precisely that difference, and on a much more detailed spatial and temporal scale than is usual in public traffic records.
The authors also emphasize that the model is conceived as a cost-effective system that does not require completely new infrastructure. Instead of expensive, separate, and limited measurement campaigns, it relies on combining data sources that cities already have or can obtain relatively easily. In this, a broader trend of modern urban management can also be seen: the use of existing digital traces and sensors to more precisely understand the consequences of traffic, noise, pollution, and other urban processes. The MIT team also emphasizes that privacy is protected because the system recognizes vehicle categories, but does not track driver identity.
Why more precise data matters for public policy
The key message of the research is not only that emissions can be measured better, but also that small methodological differences can lead to major errors in assessment. The team tested what happens when precise, local input data is replaced with coarser citywide averages. It turned out that such simplification can produce deviations from minus 49 to plus 25 percent compared with finer estimates. In other words, a city that would plan traffic or climate measures on the basis of rough averages could seriously underestimate or overestimate the real effects of its decisions.
This has direct consequences for transport planning. If a city administration wants to know whether a certain bus route, a change in traffic light timing, a new pedestrian zone, or a restriction on car entry will actually reduce emissions, a general annual inventory will not be enough. What is needed is a map that shows where the problem is concentrated, at what part of the day it reaches its peak, and which groups of vehicles participate in it the most. That is precisely why the authors argue that their approach can serve as a bridge between broad citywide estimates and extremely detailed analyses of individual vehicles, which are often too expensive and difficult to apply at the scale of an entire city.
The model was also used to simulate several scenarios of traffic changes. Among other things, the researchers observed what would happen if part of the trips shifted from private cars to buses. They also analyzed a scenario in which the morning and afternoon peaks would be relieved by slight spreading over time, that is, a situation in which fewer vehicles appear on the roads at the same time. Such simulations are especially important for cities seeking politically and socially acceptable ways to reduce emissions, because they do not necessarily start from bans, but from a different organization of demand, public transport, and traffic management.
Manhattan as a laboratory for urban mobility
The choice of Manhattan is no coincidence. It is an area with an exceptionally dense traffic network, a large number of signalized intersections, intensive taxi and delivery traffic, and strong daily oscillations between business, tourist, and residential movements. In such an environment, the differences between a calmer part of the day and the peak of the burden can be dramatic, and this is precisely what favors testing a model that attempts to map emissions down to the level of an individual block and hour. The MIT team states that such a level of detail is useful not only for understanding the existing situation but also for assessing the effects of specific interventions, from changes in traffic flows to broader climate policies.
It is also important that this is not only about carbon dioxide, but also about the broader issue of air quality in the city. Traffic is not the only source of particulate matter and other pollutants, but in dense urban environments it often plays a decisive role in the local exposure of the population. An emissions map that shows the spatial and temporal concentration of traffic can help better protect schools, hospitals, residential areas, and pedestrian corridors. In practice, this means that decisions on traffic regulation can be made less and less blindly, and more and more on the basis of measurements that show the real effect on human health.
The real test: New York congestion pricing
One of the most interesting parts of the paper concerns the analysis of a real policy measure, and not just simulation. On January 5, 2025, New York introduced a congestion pricing program for the area of Manhattan south of 60th Street, the first such system in the USA. The program, managed by the Metropolitan Transportation Authority, was introduced with the aim of reducing congestion, speeding up public transport, and securing revenue for investment in transit infrastructure. According to official MTA data published during the first year of implementation, the number of vehicles entering the zone was reduced by about 11 percent, which on average means more than 73 thousand fewer vehicles per day, or more than 27 million fewer entries into the zone during the first year.
MIT researchers observed what happened to traffic and emissions two, four, six, and eight weeks after the start of the charge. According to their results, traffic volume decreased by approximately 10 percent, but the drop in emissions was more pronounced and ranged between 16 and 22 percent. This is an important finding because it shows that reducing the number of vehicles does not act linearly on pollution. In a dense urban network, even a relatively moderate drop in traffic can produce a greater effect on emissions if it simultaneously reduces delays, the number of stops, and the need for re-acceleration. In other words, fewer cars do not mean only a lower traffic volume, but also smoother movement of the remaining vehicles.
The researchers warn, however, that the effects are not the same in every part of the network. On some main roads, the drop in emissions was more pronounced, while outside the charging zone the effects were more mixed. Such uneven spatial distribution is important for any discussion about the fairness and effectiveness of traffic measures. When city authorities introduce restrictions or financial incentives, it is not enough to say that the average is better; it is necessary to know who specifically gets cleaner air, and where traffic may simply be shifting. That is precisely why a detailed emissions map can be more useful than the figure of total reduction alone.
Comparison with other research and broader significance
MIT's findings are consistent with other recent research on the consequences of New York's congestion pricing. At the end of 2025, Cornell University published results according to which PM2.5 concentrations in the charging zone in the first six months fell by about 22 percent compared with the expected level without that policy, with declines also in other parts of the city and surrounding suburbs. Although this is a different type of measurement, the comparison is important because it shows that modeled traffic data and independent observations of air quality are moving in the same direction. This increases the credibility of the thesis that traffic measures, when carefully designed, can bring faster and larger environmental effects than skeptics often assume.
For urban planners and decision-makers, this is significant for one more reason. Debates on traffic are often conducted between two poles: on one side are general political goals such as decarbonization and a healthier city, and on the other the everyday concern of citizens that new rules will slow movement or hurt the economy. Tools like this make it possible to conduct that debate with more facts. Instead of broad assumptions, it is possible to assess what is happening by streets, neighborhoods, and parts of the day, and then compare the effects on flow, emissions, and air quality. This does not remove political disputes, but it reduces the space for decision-making without verifiable data.
Technology that can expand to other cities as well
The researchers claim that the model does not have to stop at traffic cameras. In related projects, they also experimented with data from cameras built into vehicles, including so-called dash cam systems, in order to further enrich the picture of vehicle movement through the city. This opens the possibility that future emissions assessment systems will use a broader set of sources, from city infrastructure to data generated by the vehicles themselves. In technological terms, this means that cities no longer rely only on rare fixed measurement points, but are gradually moving to a network of sensors and digital traces that enable almost continuous monitoring.
Such a direction of development is especially interesting for European and Asian cities that are already introducing low-emission zones, restrictions for older vehicles, or dynamic traffic management. If emissions can be estimated at the level of an individual street and almost in real time, then corrections to traffic policy can also be faster and more precise. Instead of waiting for years for aggregated reports, administrations could see earlier whether a certain intervention works or needs to be adjusted. At the same time, it would be easier to determine where the greatest health benefits are, and where there is a danger that the problem will simply move from one part of the city to another.
From academic model to a tool for managing the city
Although this is an academic study, its practical message is clear: cities already have a large amount of data at their disposal, but they often do not connect it in a way that allows precise management of environmental and traffic consequences. The MIT team shows that by combining cameras, anonymized location records, and existing emissions databases, it is possible to create a tool that is detailed enough to analyze an individual road, but also broad enough to cover the entire city. This is a step toward a management model in which traffic policy is not only a matter of assessment and political intuition, but also of operational analytics.
For citizens, perhaps the most important thing is that behind the technically complex methodology lies a very concrete question: where in the city is pollution the greatest and can this be changed without paralyzing everyday life. The results so far indicate that the answer may be yes, especially when traffic reduction is combined with better public transport, smarter intersection management, and precise monitoring of the consequences. At a time when urban policies are increasingly torn between climate goals, public health, and economic needs, tools that can show the real effect of an individual measure almost in real time will probably play an ever greater role in the way large cities are planned and run.
Sources:- - MIT News – presentation of the research on the traffic emissions assessment model and the main results of the study (link)
- - Nature Sustainability / Research Square – summary of the paper “Ubiquitous Data-driven Framework for Traffic Emission Estimation and Policy Evaluation” and the methodological framework of the research (link)
- - MTA – official information on the congestion pricing program in the Manhattan zone south of 60th Street and data on the drop in the number of vehicles (link)
- - MTA – overview of the results of the first year of implementation, including more than 27 million fewer vehicle entries and an average drop of about 11 percent (link)
- - Cornell University – results of research on the drop in PM2.5 after the introduction of congestion pricing in New York (link)
Find accommodation nearby
Creation time: 1 hours ago