Comparison of Urban Air Quality Simulations During the KORUS‐AQ Campaign With Regionally Refined Versus Global Uniform Grids in the Multi‐Scale Infrastructure for Chemistry and Aerosols (MUSICA) Version 0

Model intercomparison studies often report a large spread in simulation results, but quantifying the causes of these differences is hindered by the fact that several processes contribute to the model spread simultaneously. Here we use the Multi‐Scale Infrastructure for Chemistry and Aerosols (MUSICA) version 0 to investigate the model resolution dependencies of simulated chemical species, with a focus on the differences between global uniform grid and regional refinement grid simulations with the same modeling framework. We construct two global (ne30 [∼112 km] and ne60 [∼56 km]) and two regional refinement grids over Korea (ne30x8 [∼14 km] and ne30x16 [∼7 km]). The grid resolution can change chemical concentrations by an order of magnitude in the boundary layer, and the importance increases as the species' reactivity increases (e.g., up to 50% and 1,000% changes for ethane and xylenes, respectively). The diurnal cycle of oxidants (OH, O3, and NO3) also varies with the grid resolution, which leads to different oxidation pathways of volatile organic compounds (e.g., the fraction of monoterpenes reacting with NO3 in Seoul around midnight is 90% for ne30, but 65% for ne30x16). The models with high‐resolution grids usually do a better job at reproducing aircraft observations during the KORUS‐AQ campaign, but not always, implying compensating errors in the coarse grid simulations. For example, ozone is better reproduced by the coarse grid due to the artificial mixing of NOx. When developing new chemical mechanisms and evaluating models over urban areas, the uncertainties associated with model resolution should be considered. Plain Language Summary A new model framework, the Multi‐Scale Infrastructure for Chemistry and Aerosols version 0 (MUSICAv0), has been developed at NCAR to enable a computationally feasible global modeling framework while still resolving chemistry at urban scales. Using the MUSICAv0 framework with different horizontal grid resolutions (∼112, ∼56, ∼14, and ∼7 km), this work examines how much horizontal grid resolution can affect simulated chemical concentrations in 3D chemistry models. Model concentrations can vary up to 10 times between ∼112 and ∼7 km grids over urban areas at the surface. On the other hand, a region‐specific emission inventory with detailed local information is essential for some chemical species, although it is generally less important than the grid resolution for many chemical species. The model with a high‐resolution gri