Emissions do not emerge uniformly across the city.

They form at moments of interruption, acceleration, and imbalance.

Understanding them requires the right lens.

Cameras are everywhere. Powered by AI, we can identify each vehicle model and their precise motion.

Car make inventory
The Anatomy of Traffic
Fleet Density
Signal Rhythm
Acceleration Roughness
Fleet Density
Not all vehicles weigh the same inside system.
A small number of heavy-duty vehicles can dominate local emissions. Who moves matters as much as how many.
Signal Rhythm
Congestion is not slow. It is rhythmic.
Traffic signals impose stop–start cycles that concentrate idling and acceleration. These repeated interruptions generate emission spikes invisible to average-speed maps.
Acceleration Roughness
Acceleration is invisible in inventories, but dominant in emissions.
Emissions respond to how unevenly traffic accelerates and decelerates, not to speed alone. Streets moving equally fast can pollute very differently.

Policy as Treatment

When cities can see emissions,
policy becomes treatment.
Diagnosis enables targeted, timely intervention.

Congestion pricing in New York City shows how structural visibility can change outcomes: emissions display strong spatiotemporal variation, and congestion pricing led to a significant reduction in emissions.

16–22%
reduction in emissions
You can't treat what you can't see.
Fleeting Emissions
Visualizing the dynamics of urban emissions.
Description
Urban traffic emissions change constantly, shaped by vehicles’ speed, acceleration, interruption—what fixed sensors cannot capture. In Fleeting Emissions we integrate visual data from widely present traffic cameras and dashboard cameras with AI-driven vehicle recognition and network-scale modeling, to capture fine-grained variations in fleet composition, vehicular rhythms, and acceleration behavior. We tested the Fleeting Emissions model using dashcams in Amsterdam and Abu Dhabi, and grew it up to city scale in New York using extensive traffic cameras and data fusion to uncover hidden emission hotspots and temporal spikes that conventional models overlook. As demonstrated in Manhattan, such high-resolution insight enables real-time evaluation of policies—from congestion pricing to demand shifts—transforming emissions from an abstract metric into a visible, actionable layer of the city.
Publication
Hu, S., Santi, P., Benson, T., Zhou, X., Wang, A., Kumar, A., Ratti, C. (2026). Ubiquitous Data-Driven Framework for Traffic Emission Estimation and Policy Evaluation. Nature Sustainability.

Press
Download the Press Material
The material on this website can be used freely in any publication provided that
1. it is duly credited as a project by the MIT Senseable City Lab.
2. a PDF copy of the publication is sent to senseable-press@mit.edu
Team
Carlo Ratti, Director
Paolo Santi and Fábio Duarte, Principal Investigators
Research: Songhua Hu (lead), Tom Benson, Xuesong Zhou, An Wang, Ashutosh Kumar
Visualization: Jingrong Zhang (lead), Martina Mazzarello
Collaboration
Department of Municipalities and Transport, Abu Dhabi AMS Institute
Accessibility