Influence of Emission Inventories and Secondary Organic Aerosol Chemistry on Simulated Fine Particulate Matter Over the U.S. in a Chemical Transport Model

Abstract Accurately modeling ambient fine particulate matter (diameter <2.5 μm, PM 2.5 ) is crucial to air quality management. We evaluate how uncertainties in biomass burning (BB) emissions, anthropogenic emissions, and secondary organic aerosol (SOA) production mechanisms impact simulated PM 2.5 mass, composition, and statistical accuracy of modeled PM 2.5 over the continental U.S. using the GEOS‐Chem chemical transport model. We compare simulations driven by different BB inventories (Global Fire Emissions Database, GFED; Global Fire Assimilation System, GFAS) and anthropogenic emissions inventories (Community Emissions Data System, CEDS; U.S. Environmental Protection Agency National Emissions Inventory, NEI), and SOA chemistry schemes (simple, S; complex, C).Six scenario combinations (S‐GFED‐CEDS, C‐GFED‐CEDS, C‐GFED‐NEI, S‐GFED‐NEI, S‐GFAS‐CEDS, and C‐GFAS‐NEI) were analyzed for August and December 2021, representing wildfire and non‐wildfire periods. In August, GFED‐driven scenarios produce PM 2.5 concentrations >38% higher than the GFAS‐driven simulation (C‐GFAS‐NEI) change. GFED leads to increased organic aerosols (OA) in August, while in December the SOA chemistry scheme drives OA differences between simulations. Compared to ground‐based observations, the C‐GFAS‐NEI simulation has the highest spatial correlation and lowest bias in August, when GFED‐based simulations overestimate PM 2.5 by more than 80% on average. In December, performance was similar across scenarios, with the simple SOA scheme yielding a lower bias than the complex scheme. These findings demonstrate that BB inventories dominate uncertainty in simulated PM 2.5 during wildfire‐impacted periods, whereas SOA chemistry exerts greater influence when wildfire activity is limited. Accounting for these uncertainties can guide emission inventory selection, choice of SOA treatment, and inform future model developments to improve policy or source attribution applications.