An analysis by a North Carolina State University professor found that a proposed data center in Richmond County, North Carolina may worsen local air quality, potentially raising the risk of respiratory, cardiovascular and neurological health effects for residents. 

Earlier this year, Amazon Web Services and Duke Energy Progress submitted air quality permit applications to the North Carolina Department of Environmental Quality for the installation of on-site diesel generators that would help power a $10 billion, 21-building data center. 

The companies split their requests into two permit applications, and by doing so, the estimated emissions associated with the generators “appear much lower than they actually are,” said Jennifer Richmond-Bryant, an associate professor in the Department of Forestry and Environmental Resources at NC State.

Richmond-Bryant, whose research focuses on human exposure to air pollution, reviewed both companies’ permit applications to assess whether their analyses were complete and fully accounted for the project’s combined emissions.

By comparing projected emissions from the data center’s more than 650 diesel generators with 2024 emissions from Duke’s natural gas turbine in Hamlet, Richmond-Bryant found that the individual permit applications underestimated the project’s total potential pollution.

Her analysis found that the total amount of Hazardous Air Pollutants (HAPs) will increase by 28% compared to 2024 levels of HAPs emitted from Duke’s gas turbine and that the level of volatile organic compound emissions may increase by over 400%. 

Among the pollutants expected to be emitted by the data center, Richmond-Bryant identified fine particulate matter, or PM2.5, as one of the most concerning. She said the pollutant — a mixture of tiny airborne particles that can travel deep into the respiratory system — is “well known to cause cardiovascular disease and increased mortality,” with evidence linking exposure to respiratory and neurological effects.

“… what matters from an individual health perspective is the total pollution one is exposed to.”

Richmond-Bryant said the key issue is that the two projects are located on the same site and should therefore be evaluated together. Combined, their emissions would exceed the federal threshold that triggers a more rigorous review under the Clean Air Act’s Prevention of Significant Deterioration (PSD) program, which is designed to ensure that new sources of pollution do not worsen air quality in areas that already meet federal standards.

According to Richmond-Bryant, the North Carolina Department of Environmental Quality should consider the cumulative impact of nearby pollution sources when reviewing projects like the data center to ensure they do not significantly worsen local air quality.

In addition to Duke’s natural gas plant, Richmond County already has other sources of air pollution, including an Enviva wood pellet facility, a sand mine, smaller industrial sources, nearby farms and vehicle emissions from major roadways such as I-74 and US-1.

“When multiple sources are emitting at once, the exposed respiratory system does not differentiate between emissions from one site or another. So, what matters from an individual health perspective is the total pollution one is exposed to,” Richmond-Bryant said. 

She added, “What matters from a public health perspective is whether and to what extent the North Carolina Department of Environmental Quality accounts for the multitude of sources when estimating total allowable air pollutant emissions of each type of pollutant.” 

The state maintains that Duke and Amazon’s permit applications should be reviewed separately since the diesel generators are located on different properties, despite Duke operating 57 temporary diesel generators that would run continuously for up to a year to power the site until a permanent grid connection is built.

“Duke Energy Progress and Amazon Web Services are both serving the same operation. It is not the same situation as two neighboring facilities,” she said. 

More Testing is Needed to Uncover Pollution Risks

Richmond-Bryant said additional air quality analysis is needed before the state approves Duke and Amazon’s permit applications for the diesel generators, largely because the companies’ dispersion models are insufficient and there are no air quality monitors nearby to assess local conditions.

Duke and Amazon used computer models to predict how pollutants from their diesel generators would spread through the air. However, the modeling did not fully evaluate the combined impact of emissions from both companies’ generators and only assessed certain pollutants, not all criteria air pollutants.

“It is not clear why dispersion modeling was not performed for criteria air pollutants in order to predict potential exceedances in different parts of Richmond County,” Richmond-Bryant said. 

Criteria air pollutants include nitrogen oxides, carbon monoxide, fine particulate matter and other common pollutants that can harm human health and the environment. The Clean Air Act sets national limits for these pollutants through the National Ambient Air Quality Standards (NAAQS).

Amazon also did not provide enough information to support its emission estimates, according to Richmond-Bryant. The company claimed its pollution controls would reduce carbon monoxide and nitrogen oxide emissions below regulatory limits, but it did not clearly explain how it calculated those reductions.

Richmond-Bryant recommended that the state install a Federal Equivalent Method monitor, an automated device used to measure air pollution, in or near Hamlet “to ensure concentrations from localized sources are captured near where people live.” 

The monitors can collect data quickly and frequently, capturing short-term changes in pollution levels. This level of detail can help researchers and regulators better understand pollution patterns and identify changes that might be missed with less frequent monitoring.

Currently, the nearest air quality monitor to Hamlet is roughly 30 miles away in Candor, which is enough distance “that pollutant concentrations emanating from sources in Hamlet might be substantially diluted so that they would not register substantial change on the monitor,” Richmond Bryant said. 

Richmond-Bryant said the state should consider installing a bank of Federal Equivalent Method monitors in the town to measure nitrogen oxides, ozone, fine particulate matter and other air pollutants. The monitoring data would help community members understand their exposure risk and evaluate pollution impacts under the PSD program.

Richmond-Bryant has assisted residents in installing low-cost air quality monitors in Richmond County to track fine particulate matter concentrations. After correcting for relative humidity and temperature, which are known to bias air quality data, she found that average concentrations already exceed the level of the annual average NAAQS for fine particulate matter.

“If nothing else, this indicates that the state should add a Federal Equivalent Method monitor here to confirm or refute the low-cost sensor findings,” Richmond-Bryant said.