Systematic bias of WRF-CMAQ PM10 simulations for Seoul, Korea

Abstract

For evaluating the performance of the Korean operational air quality model (Weather Research and Forecasting-Community Multiscale Air Quality), we compared the simulated concentrations of particulate matter with diameters ≤ 10 μm (PM10) against observations in Seoul, Korea for the cold seasons (October through March) of 2015–2018. The simulations yield about 70% hit rate for PM10 concentrations within 10 μg m−3 from the observations. For the 50–90 μg m−3 range, the simulations systematically overestimate the PM10 frequency mainly due to overestimation/underestimation of PM10 concentrations in the moderate-PM10 (30 < PM10 ≤ 60 μg m−3)/high-PM10 (60 < PM10 ≤ 100 μg m−3) range. Considering that the model simulates PM10 transports using input meteorological fields and emissions amounts, 72-h back-trajectories from Seoul (37.51°N, 126.99°E) at the 500 m above ground level are analyzed to investigate the cause of the systematic bias. The 410 days when PM10 concentrations are within the 30–100 μg m−3 range are grouped into three major wind flows: westerly (163 days), northerly (99 days), and northwesterly (148 days). Overestimations of moderate-PM10 days frequently occur for the westerly group associated with a synoptic pattern characterized by calm anticyclonic systems over China and Korea, especially when the emission amounts were overestimated in eastern China, the main PM10 source region for Seoul for the westerly group. Underestimations of high-PM10 days frequently occur for the northerly and northwesterly groups associated with climatological patterns or strong windy cyclonic systems along with some small inflow from underestimated emission in northeastern China. To improve the model performance, statistical models trained for each back-trajectory group along with improved estimations of emission amounts may prove helpful.