Is EPA undermonitoring air quality in non-white areas?

Using a novel methodology, researchers have uncovered gaps in block-level monitoring coverage that has often been assumed to be uniform.  

By Jed Thomas

A new University of Utah study reveals that the U.S. Environmental Protection Agency’s (EPA) air quality monitoring network is disproportionately located in predominantly white neighbourhoods—leaving many communities of colour underrepresented in national pollution data. 

Researchers found significant disparities across six key pollutants, particularly for lead and sulphur dioxide, followed by ozone and carbon monoxide. These regulatory monitors are the backbone of pollution control efforts, informing decisions in public health, environmental regulation, and urban planning. 

“If we’re relying on this data to make decisions, we have to ask—whose air is actually being measured?” said lead author Brenna Kelly, a doctoral student in Population Health Sciences at the U. “Even high-quality data can still fall short if it doesn’t reflect all communities equally.” 

Challenging assumptions

While it's well-established that marginalized populations often face higher exposure to pollution, past research has assumed that monitoring coverage is evenly distributed.

This study is the first to examine those assumptions using neighbourhood-scale data across all U.S. Census groups.

The most pronounced monitoring gaps were found for Native Hawaiian and Pacific Islander communities, followed by American Indian and Alaska Native populations. 

The findings raise serious concerns for environmental justice and big-data research.

“Bias in AI algorithms gets a lot of attention, but here we’re seeing bias baked into the datasets themselves,” said co-author Simon Brewer, associate professor of geography and a member of the U’s ONE-U Responsible AI Initiative. “The fact that these disparities exist across all major pollutants points to deeper systemic issues in how monitors are sited.” 

How was this discovered?

Published in JAMA Network Open on December 4, 2024, the study mapped EPA monitor locations against census-block-level demographic data—one of the most granular ways to analyse residential patterns.

Using EPA data for six harmful pollutants (lead, ozone, nitrogen dioxide, sulphur dioxide, carbon monoxide, and particulate matter) and 2022 U.S. Census figures, researchers found that every racial and ethnic group, when compared to the non-Hispanic white population, had less monitoring coverage—particularly for lead, ozone, nitrogen dioxide, and particulate matter. 

Kelly’s interest in monitor placement emerged during her research on air pollution’s effects on pregnant people. Epidemiological studies often depend on accurate, comprehensive environmental data—something this study suggests is lacking. 

Next steps for monitoring professionals

“It’s not just one pollutant or one group being overlooked—we’re missing the bigger picture for entire populations,” she said. “That undermines the quality of research and leads to biased conclusions.” 

The study underscores a broader challenge for data-driven science: ensuring that the tools and datasets used are equitable from the ground up. The University of Utah’s ONE-U Responsible AI Initiative brings together researchers to address exactly these kinds of concerns. 

“In a data-driven world, we need to be just as vigilant about the quality and fairness of the data as we are about the models we apply to it,” Brewer said. 

To read the study in full, click here.