Does Downscaling Improve the Performance of Urban Ozone Modeling?

Increasing the model resolution is expected to be one way for improving air quality forecasts in urban areas. In this study, we evaluate the model performance in a large city at various resolutions to examine the best resolution for air pollution simulations. The comparison with measurements at a station near the traffic emissions shows the advantage of using high resolutions for capturing the extreme values. The statistical evaluation indicates that the highest model resolution (33 m) provides the best results for NOX concentration distributions near the traffic roads, while the improvement for roadside O3 with decreasing grid spacing stops at a certain point. The best model performance for the areas with a distance to the pollution sources is with the resolution of 100–300 m, at which the transport errors are equivalent to the emission biases. Plain Language Summary: As the increasing needs in the air quality forecasting in large cities, there is a trend in decreasing the model grid spacing to obtain more detailed pollutants distributions between neighborhoods or at street levels. To determine at which resolution the model can obtain the best representation of the pollutants' concentrations, we evaluate the model performance at different resolutions taking Hong Kong as an illustration. The analysis shows that the improvement with increasing model resolution is not monotonic for the areas far away from the intense emissions; however, the model with the highest resolution (33 m) reproduces the best results for the short‐lived species near the pollution sources. Key Points: Increasing horizontal resolution to 33 m improves the prediction of NOX near the traffic emissionsThe threshold of the model resolution is around 300 m for areas with a distance to the pollution sourcesThe changes of model performances with varied resolutions are different for NOX and O3 [ABSTRACT FROM AUTHOR]