Sub-micron aerosol and CCN characteristics in Seoul measured during 2019–2021 and CCN prediction using machine learning

Abstract

The seasonal characteristics of aerosol number concentration (N_CN) and cloud condensation nuclei number concentration (N_CCN) in Seoul from 2019 to 2021 were analyzed. The average N_CN and N_CCN at 0.6 % supersaturation (SS) were 12,612 cm⁻³ and 2,829 cm⁻³, respectively, representing significant decreases of approximately 30 % and 50 % compared to the values from 2004 to 2010. Despite these reductions, diurnal and seasonal patterns remained consistent over time. However, distinct seasonal differences were noted, particularly in the relationship between N_CN and N_CCN: in winter, N_CN and N_CCN exhibited similar trends in both diurnal variation and back trajectory analysis, whereas in summer, their trends frequently diverged. To account for these pronounced seasonal differences, predictions of N_CN and N_CCN were conducted separately by season using the Random Forest Regression Model. The seasonally tailored predictions provided improved performance compared to the Multiple Linear Regression Model used as a baseline. Analysis of importance of each predictor revealed that N_CN was the most significant predictor of N_CCN at 0.4–1.0 % SS in winter, while particulate matter (PM₁₀) played a more critical role across all supersaturation levels in summer and at 0.2 % SS in winter. These results suggest seasonally distinct mechanisms driving CCN formation in Seoul. Overall, this study highlighted recent trends in aerosols and CCN reflecting the rapid changes of air quality at a representative urban region in East Asia, and the data presented in this study can be used as representative values of aerosol and CCN in East Asian urban region for aerosol–cloud interaction studies in climate models.