Air quality modeling for accountability research: Operational, dynamic, and diagnostic evaluation

Photochemical grid models play a central role in air quality regulatory frameworks, including in air pollution accountability research, which seeks to demonstrate the extent to which regulations causally impacted emissions, air quality, and public health. There is a need, however, to develop and demonstrate appropriate practices for model application and evaluation in an accountability framework. We employ a combination of traditional and novel evaluation techniques to assess four years (2001-02, 2011-12) of simulated pollutant concentrations across a decade of major emissions reductions using the Community Multiscale Air Quality (CMAQ) model. We have grouped our assessments in three categories: Operational evaluation investigates how well CMAQ captures absolute concentrations; dynamic evaluation investigates how well CMAQ captures changes in concentrations across the decade of changing emissions; diagnostic evaluation investigates how CMAQ attributes variability in concentrations and sensitivities to emissions between meteorology and emissions, and how well this attribution compares to empirical statistical models. In this application, CMAQ captures O3 and PM2.5 concentrations and change over the decade in the Eastern United States similarly to past CMAQ applications and in line with model evaluation guidance; however, some PM2.5 species-EC, OC, and sulfate in particular-exhibit high biases in various months. CMAQ-simulated PM2.5 has a high bias in winter months and low bias in the summer, mainly due to a high bias in OC during the cold months and low bias in OC and sulfate during the summer. Simulated O3 and PM2.5 changes across the decade have normalized mean bias of less than 2.5% and 17%, respectively. Detailed comparisons suggest biased EC emissions, negative wintertime SO42− sensitivities to mobile source emissions, and incomplete capture of OC chemistry in the summer and winter. Photochemical grid model-simulated O3 and PM2.5 responses to emissions and meteorologically across the decade match results from receptor-based, statistical regression models. PM2.5 sensitivities to mobile source emissions in the summertime have decreased substantially, but wintertime mobile sensitives remain largely unchanged because decreases in negative SO42− sensitivities match decreases in positive sensitivities from other constituents. Similarly, NOX emissions have led to decreased summertime O3 and increased wintertime O3 because of opposite sensitivities. Overall, results show that emissions reductions improved air quality across the domain and remain a viable option for improving future air quality.