Estimation of Column Aerosol Contribution in Seoul and Gangneung Using Machine Learning Clustering Technique

In this study, we estimated characteristics of the local atmospheric aerosols by the machine learning techniques with the column aerosol and meteorological data measured at Seoul and Gangneung. Based on the classified aerosol properties, contributions of aerosol inflow and outflow can be inferred and scenarios based on these results were also determined. Column aerosol observation status for two cities showed that the aerosol optical depth (ACID) observed in Seoul is 39.2% (0.18 +/- 0.16) higher than that of Gangneung (0.28 +/- 0.24), and the Angstrom Exponent (AE) was a similar range level (1.29 +/- 0.3). Although aerosol loads are differences between the two regions, the particle size distribution of regions is similar. For the machine learning clustering analysis, all data samples were classified as the best number of classes and optimized scenarios for Seoul and Gangneung were determined. In order to verify the created scenarios, a case was selected from the GOCI RGB image and MODIS L1B, and the scenario algorithm was performed using the actual ground observation data. This methodology is useful for monitoring and predicting fine dust through the characterization and contribution calculation of atmospheric column aerosols.