A study on new particle formation (NPF) in the ambient atmosphere in various environments

Abstract

Atmospheric nucleation, also called new particle formation (NPF), has been studied in various atmospheric environments worldwide. Newly formed nanoparticles can grow into sub-micrometer particles, which can act as cloud condensation nuclei (CCN) and affect human health. The NPF can be affected by precursor gases, oxidizing radicals, pre-existing aerosols, and meteorological parameters (temperature, solar radiation, and relative humidity) and vary based on location and time. Field measurements of nanoparticles were conducted at various locations (Ny-Alesund, Norway (Arctic)); urban (Beijing, China and Gwangju, Korea); and agricultural (Gimje, Korea) sites), and the characteristics and major drivers of the NPF among sites were compared. The average NPF occurrence frequency at urban Gwangju, urban Beijing, agricultural (livestock) Gimje, agricultural (cropland) Gimje, and the Arctic sites was 47%, 45%, 55%, 26%, and 24%, respectively. The Greenland Sea region showed an elevated chlorophyll-α and dimethyl sulfide (DMS) production, suggesting that marine biogenic sources play an important role in the Arctic NPF. The highest formation and growth rates were observed at the agricultural site where abundant NH3 was emitted from pig and chick farms. Sulfuric acid (H2SO4), which can be formed from SO2, played a primary role in the NPF and growth, especially at urban sites. In particular, volatile organic compounds (VOCs) were correlated with the particle number concentration during the NPF event periods at agricultural (livestock) Gimje site in winter. However, further studies are needed to more clearly elucidate the NPF mechanism by site. The theoretical model parameter (LΓ) agreed well with the observed NPF occurrence at all sites. The estimated LΓ threshold values were 0.78 at the Gwangju site, 0.87 at the Beijing site, 1.02 at the Gimje (livestock) site, and 0.11 at the Arctic site. We developed single-site and multiple-sites input prediction models to predict the NPF event day using routine measurement parameters such as meteorological (e.g., temperature, pressure, relative humidity (RH), and solar radiation (SRAD)) and air quality (e.g., SO2, and PM2.5) data. The prediction results using routine measurement parameters were found to be possible to predict NPF event day or not at all sites, and the range of the classification accuracy for the singe-site input models among sites were 75–91% for the test data. Although the performance of the multiple-sites input model was lower than that of the single-site input models, the estimated trend of NPF frequency agreed well with the observed trend. The NPF predictive models can provide past or future NPF occurrence information of an environment from measured meteorological and air pollutant data without the need for aerosol size distribution data. In addition, the convolutional neural network (CNN) was used to predict NPF event days at the Arctic site. The NPF event day prediction at the Arctic site can be performed with air mass trajectory and chlorophyll-α data, verifying that chlorophyll-α and its related parameters plays an important role of the Arctic NPF prediction.