Detection of High-level PM2.5 Occurrences Applying Local Outlier Factor (LOF) Algorithm

Air quality management is significant to guarantee public health and ecosystem. For the effective control and management of atmospheric environment through different forms of approaches, analyzing of monitoring data and emission inventory should be firstly considered. Air monitoring stations and their real-time data play important roles to understand pollution circumstance around certain region. However, many of atmospheric researchers frequently face difficulties to overview large volumes of monitoring data, because it requires a lot of time and efforts. Local outlier factor (LOF) broadly applied in different research fields is useful data analysis technique to discover particular values have different pattern or/and characteristic in a data group. The authors in this paper has attempted to apply the LOF algorithm to one-year of air quality monitoring data covers the whole Korean territory for suggesting easy way to preliminarily identify the high-level of PM episodes as their occurrence causes i.e. yellow dust, non-yellow dust and domestic wildfire events. As a result, it was effective to figure out particular LOF ranges as the high-level of PM2.5 episodes, because LOF values show comparatively high numbers when concentrations of PM2.5 are rapidly increased or decreased due to LOF algorithm characteristic. The attempt of this research has found the LOF algorithm would be useful to apply as preliminary methodology for understanding which certain episode has been occurred in tremendous data group. Therefore, it would bring further consideration for effectively analyzing the environmental big data in order to establish the air quality management strategy to environmental scientists and policy makers.