Identifying controlling factors of ground-level ozone levels over southwestern Taiwan using a decision tree

Abstract Kaohsiung City and the suburban region of southwestern Taiwan have suffered from severe air pollution since becoming the largest center of heavy industry in Taiwan. The complex process of ozone (O 3 ) formation and its precursor compounds (the volatile organic compounds (VOCs) and nitrogen oxide (NO x ) emissions), accompanied by meteorological conditions, make controlling ozone difficult. Using a decision tree is especially appropriate for analyzing time series data that contain ozone levels and meteorological and explanatory variables for ozone formation. Results show that dominant variables such as temperature, wind speed, VOCs, and NO x can play vital roles in describing ozone variations among observations. That temperature and wind speed are highly correlated with ozone levels indicates that these meteorological conditions largely affect ozone variability. The results also demonstrate that spatial heterogeneity of ozone patterns are in coastal and inland areas caused by sea-land breeze and pollutant sources during high ozone episodes over southwestern Taiwan. This study used a decision tree to obtain quantitative insight into spatial distributions of precursor compound emissions and effects of meteorological conditions on ozone levels that are useful for refining monitoring plans and developing management strategies.