AI improves detail, urban air pollution estimate – Zoo House News
Using artificial intelligence, engineers at Cornell University have simplified and reinforced models that accurately calculate the particulate matter (PM2.5) — soot, dust and exhaust fumes emitted by trucks and cars that enter human lungs — that are responsible for urban air pollution are included.
Now city planners and state health officials can take a more accurate look at the well-being of city dwellers and the air they breathe from new research published in December 2022 in the journal Transportation Research Part D.
“Infrastructure determines our living environment, our exposure,” said senior author Oliver Gao, Howard Simpson Professor of Civil and Environmental Engineering in Cornell University’s College of Engineering. “The effects of air pollution from transport – caused by the exhaust fumes from the cars and trucks that drive on our roads – are very complicated. Our infrastructure, transport and energy policies will affect air pollution and therefore public health.”
Previous methods of measuring air pollution were cumbersome and relied on inordinate amounts of data points. “Previous models used to calculate particulate matter were computationally intensive and mechanically expensive and complex,” said Gao, a faculty member at the Cornell Atkinson Center for Sustainability. “But if you develop an easily accessible data model and use artificial intelligence to fill in some of the gaps, you can have an accurate model at the local level.”
Lead author Salil Desai and visiting scholar Mohammad Tayarani co-published “Developing Machine Learning Models for Hyperlocal Traffic Related Particulate Matter Concentration Mapping” with Gao to offer a lighter, less data-intensive way to create accurate models.
Air pollution is one of the leading causes of premature deaths worldwide. According to a Lancet study cited in Cornell Research, in 2015 more than 4.2 million deaths a year worldwide — from cardiovascular disease, ischemic heart disease, stroke and lung cancer — were attributed to air pollution.
In this work, the group developed four machine learning models for traffic-related particulate matter concentrations in data collected in the five boroughs of New York City, which have a combined population of 8.2 million residents and daily vehicle mileage of 55 million.
The equations use few inputs such as traffic data, topology, and meteorology in an AI algorithm to learn simulations for a variety of traffic-related air pollutant concentration scenarios.
Their most powerful model was the Convolutional Long Short-Term Memory, or ConvLSTM, which trained the algorithm to predict many spatially correlated observations.
“Our data-driven approach – based primarily on vehicle emissions data – requires significantly fewer modeling steps,” said Desai. Rather than focusing on stationary locations, the method provides a high-resolution estimate of the pollution surface of city streets. Higher resolution can help transport and epidemiology studies assess impacts on health, environmental justice, and air quality.
This research was funded by the US Department of Transportation’s University Transportation Centers Program and Cornell Atkinson.