Chemical transport model consistency in simulating regulatory outcomes and the relationship to model performance

It is critical to evaluate an air quality model prior to regulatory applications to ensure model performance is adequate for regulatory decision making. However, no formal benchmarks currently exist in the United States (US) to judge whether a model's performance is acceptable for given purposes, and practitioners usually rely on criteria established in the past, when the extent of modeling domains and length of simulation periods were relatively limited. This study conducts modeling experiments to investigate the impact of using different modeling configurations on various policy-related modeling outcomes. Two widely used chemical transport models with different meteorological data, biogenic emissions, and aerosol modeling schemes are applied to an annual modeling period, and model performance for ozone and particulate matter (PM) constituents is evaluated for different timescales and geographical regions in the US. Results show that while both models can be considered acceptable based on criteria commonly used by modelers when evaluated on the annual basis, the model performance at finer levels can reveal differences between the two modeling configurations, which may lead to different policy outcomes. Model results for 2005 and 2014 are used to determine monitoring sites projected to violate current US standards for ozone and fine PM, sites significantly affected by emissions from two selected upwind states, and ratios of deposition to ambient concentration of oxides of sulfur and nitrogen, a.k.a. transference ratios, at each monitoring site in the eastern US. The two modeling configurations result in quite different lists of sites and transference ratios even though both show acceptable model performance in the conventional model performance evaluation. This calls for development of model performance evaluation criteria specific to various regulatory purposes.