Modelling air quality during the EXPLORE-YRD campaign – Part I. Model performance evaluation and impacts of meteorological inputs and grid resolutions

The EXPeriment on the eLucidation of the atmospheric Oxidation capacity and aerosol foRmation and their Effects in the Yangtze River Delta (EXPLORE-YRD) campaign was carried out between May and June 2018 at a regional site in Taizhou, China. The EXPLORE-YRD campaign helped construct a detailed air quality model to understand the formation of O3 and PM2.5 further, identify the key sources of elevated air pollution events, and design efficient emission control strategies to reduce O3 and PM2.5 pollution in YRD. In this study, we predicted the air quality during the EXPLORE-YRD campaign using the Weather Research and Forecasting/Community Multiscale Air Quality modelling system (WRF/CMAQ) and evaluated model performance on O3 and PM2.5 concentrations and compositions. Air quality was predicted using two sets of reanalysis data—NCEP Final (FNL) Operational Global Analysis and ECMWF Reanalysis v5.0 (ERA5)—and three horizontal resolutions of 36, 12, and 4 km. The results showed that PM2.5 concentration was generally under-predicted using both the FNL and ERA5 data. ERA5 yielded slightly higher PM2.5 predictions during the EXPLORE-YRD campaign. Both reanalysis data sets under-predicted the high PM2.5 pollution processes on 29–30 May 2018, indicating that reanalysis data is not essential for under-predicting extreme PM2.5 pollution processes. The performance of O3 was similar in both the reanalysis data sets, because O3 is mostly sensitive to temperature predictions and FNL and ERA5 yielded similar temperature results. Although the average performance of PM2.5 and O3 predictions yielded by FNL and ERA5 was similar, large differences were observed in certain locations on specific days (e.g. in Hangzhou between 29 May and June 6, 2018 and in Hefei on 1–3 June 2018). Therefore, the choice of reanalysis data could be an important factor affecting the predictions of PM2.5 and O3, depending on locations and episodes. Comparable results were obtained using predictions with different horizontal resolutions, indicating that grid resolution was not crucial for determining the model performance of both PM2.5 and O3 during the campaign.