Engineers from Cornell University have used artificial intelligence to simplify and strengthen models that accurately calculate the fine particulate matter (PM2.5) contained in urban air pollution. This PM2.5 consists of the soot, dust, and exhaust emitted by trucks and cars that get into human lungs.
A new study that was conducted and published in December 2022 in the journal Transportation Research Part D has made it possible for city planners and government health authorities to acquire a more accurate accounting about the health of urban residents and the air that they breathe.
Oliver Gao, the Howard Simpson Professor of Civil and Environmental Engineering at the College of Engineering at Cornell University and the senior author of the study, said that “Infrastructure impacts our living environment, our exposure.” “The effect of air pollution caused by transportation, which is released into the atmosphere in the form of exhaust by the automobiles and trucks that move on our streets, is quite difficult. Our decisions regarding infrastructure, transportation, and energy are going to have an effect on air pollution, and as a result, on public health.
In the past, determining the level of pollution in the air required laborious approaches that relied on an excessive number of data points. According to Gao, a faculty member at the Cornell Atkinson Center for Sustainability, “Older approaches to quantify particulate matter were computationally and mechanically expensive, and difficult.” However, if you create a data model that is simple to access and enlist the assistance of artificial intelligence to help fill in some of the gaps, you will be able to have a model that is correct on a smaller scale.
The purpose of the article “Developing Machine Learning Models for Hyperlocal Traffic Related Particulate Matter Concentration Mapping,” which Gao, the lead author, Salil Desai, and Mohammad Tayarani, a visiting scientist, published together with Gao, was to provide a method that is more efficient and requires less data to produce accurate models.
Pollution in the surrounding air is a major contributor to the premature mortality rate across the globe. According to a study published in the Lancet in 2015 that was mentioned in the Cornell research, air pollution was responsible for more than 4.2 million deaths worldwide per year in 2015. These deaths occurred as a result of cardiovascular disease, ischemic heart disease, stroke, and lung cancer.
In the course of this work, the team 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 people and a daily-vehicle-miles-traveled total of 55 million miles. These boroughs have a total of 55 million vehicle miles travelled each day.
In order for an artificial intelligence system to learn simulations for a large variety of traffic-related and air-pollution concentration situations, the equations employ a small number of inputs such as traffic data, topology, and meteorological.
The Convolutional Long Short-term Memory, often known as ConvLSTM, was their model that performed the best overall. This model taught the algorithm to anticipate a large number of data that were spatially associated.
According to Desai, “Our data-driven method, which is mostly based on car emission data, demands a significant reduction in the number of modelling processes.” This technique gives a high-resolution assessment of the pollution surface that is present on city streets, as opposed to concentrating on fixed places. Studies in transportation and epidemiology may benefit from higher resolution, which can assist them evaluate the consequences on public health, environmental justice, and air quality.
This study was made possible thanks to funding from the University Transportation Centers Program of the United States Department of Transportation as well as Cornell Atkinson.