A Performance Analysis of Air Quality Prediction using the Korean Air Chemistry Modeling System version 2.0 (K_ACheMS v2.0)

Accurate air quality forecasting is essential for issuing early warnings of high-concentration air pollution episodes, mitigating their adverse health impacts. This study evaluates the performance of the Korean Air Chemistry Modeling System version 2.0 (K_ACheMS v2.0), an integrated system designed to enhance the predictability of air pollutant concentrations in South Korea. The K_ACheMS v2.0 incorporates (i) a modified Weather Research and Forecasting (WRF) version 4.1.5 with machine learning (ML) based wind-speed corrections; (ii) GIST Multiscale Air Quality (GMAQ) v1.0, utilizing an updated Statewide Air Pollution Research Center 07 (SAPRC07TC) chemical mechanism; and (iii) a three-dimensional variational (3D-VAR) data assimilation method to optimize the chemical initial conditions. The 5-day PM_2.5 predictability of the K_ACheMS v2.0 was evaluated against ground-based PM_2.5 observations in South Korea for 2024. Furthermore, we conducted an intercomparison of the K_ACheMS v2.0 against two global real-time air quality prediction systems, the Copernicus Atmosphere Monitoring Service (CAMS) from ECMWF and the Goddard Earth Observing System Composition Forecast (GEOS-CF) from NASA GMAO. We further analyzed the chemical composition of PM_2.5 to identify the key drivers of performance variability among these systems, using observations measured at two supersites in South Korea during the Airborne and Satellite Investigation of Asian Air Quality (ASIA-AQ) campaign. Our findings demonstrate that K_ACheMS v2.0 exhibits robust predictive performance for PM_2.5 and its chemical components compared to the global models, achieving an Index of Agreement (IOA) of 0.71, which outperforms CAMS (0.66) and GEOS-CF (0.46).