Abstract Description: The U.S. has seen a rapid increase in data centers over the past decade, with data centers located in every state. This growth will continue as the U.S. economy continues to digitize and integrate with artificial intelligence. From an air quality perspective, data centers are typically a minor source under the EPA’s New Source Review Program, requiring local- or state-level air quality permitting. As part of the permitting, most states require a compliance demonstration with the National Ambient Air Quality Standards (NAAQS), with the probabilistic 1-hour NO2 standard the most challenging. This air quality standard poses unique challenges for data centers, since EPA’s nearfield dispersion model, AERMOD, is not well-suited to handle intermittent source operation and has many limitations, including: out-dated treatment of building downwash, screening level NOx chemistry, and steady-state transport and dispersion. Based on experience, each of these issues can become controlling limitations on the data center’s operation and increase capital expense and permitting timeline. In contrast to AERMOD’s noted limitations, a Computational Fluid Dynamics (CFD) model explicitly solves the equations of motion in and around complex structures. This can address AERMOD’s limitations with downwash and steady-state flow assumptions. It is proposed that these science advancements can provide a more realistic estimation of potential air quality impacts from data center emission sources. This study is a continuation proposes of the effort to model a simple data center facility and compare 1-hour NO2 concentrations in the nearfield from both AERMOD and a CFD model. This study is intended to provide a proof-of-concept for this approach and a way to understand model sensitivity from a range of inputs, models, and their resultant air quality concentrations.
