Improving understanding of Korea's air quality using chemical transport modeling and machine learning data fusion

Air pollution, the invisible murderer, causes four million deaths globally per year. With the rapid economic growth in East Asia, more focus has been paid to improving air quality. During the last decades, China and Korea have published regulations on air quality control. As a result, air pollution problems have become less severe than decades ago. However, there remain problems on further lowering near-surface air pollution concentrations, and who is responsible for it? In this dissertation, we applied two methods to quantify at various locations in Korea the impact of local, regional, and remote emission sources. It was found that contributions depend on meteorological phases and location of interest and vary between different constituents of air pollution. Another thesis focus is model optimization aiming at higher spatial resolution, better performance, and faster run times. To achieve this objective, a machine learning algorithm was applied with multiple datasets as inputs, including air quality model predictions, satellite retrievals, emission inventories, elevation, population, land use cover, and ground observations. Machine learning model performance using different air quality model inputs, satellite AOD inputs, and observational densities are explored. The findings in this thesis guide the development of future mitigation strategies in Korea and the establishment of next-generation air quality forecasting system.