A Systematic Approach to Mitigate Ozone Pollution in Northern Taiwan: Evidence from De-pollutant Analysis

Abstract

Strict air pollution control regulations in Taiwan have led to a gradual decrease in PM_2.5 and most gaseous pollutants from 2012 to 2022, except for ozone (O₃). With annual average concentrations frequently surpassing the Taiwan Ambient Air Quality Standards (AAQS) of 60 ppb (95^th percentile of daily maximum 8-hour averages), O₃ remains a major air quality concern in northern Taiwan. The present study applied machine learning (ML) models, including positive matrix factorization-eXtreme Gradient Boost-SHapely Additive Explanation (PMF-XGB-SHAP), on three years of hourly data to investigate the influence of meteorological parameters, emission sources and other pollutants on O₃ formation at an urban site in Taipei. Then, novel de-pollutant models were developed by controlling the anthropogenic emission factors in the model to quantify the impact of reduction on ambient O₃ levels, and de-weather was applied to assess the impact of meteorological parameters. Findings showed that meteorology contributed 46.7–54.8% and 44.9–54.0% to daytime and nighttime O₃ levels, respectively, with relative humidity (RH) and boundary layer height (BLH) as dominant influencing factor. Among pollutants, NO_X displayed a consistent negative association, while PM_2.5 showed a positive relationship with daytime O₃ levels. The association between vehicular VOCs and O₃ varied across years, reflecting changes in traffic patterns. Furthermore, de-pollutant analysis demonstrated that simultaneous 50% reductions in CO, SO₂, and VOCs from industrial emissions could lower O₃ concentrations by 13.4–22.6% during pollution episode days. By providing a quantitative, source-specific pathway for precursor control, the de-pollutant modelling approach establishes a framework for air quality management in other regions grappling with complex, multi-source ozone pollution.