Harnessing AI and advanced modeling for precision ozone control: A data-driven approach to air quality management

Ozone (O₃) is a secondary pollutant formed through photochemical reactions involving volatile organic compounds (VOCs) and nitrogen oxides (NOₓ). In the atmospheric environment, ozone further reacts with precursor pollutants to generate secondary aerosols, posing adverse effects on both environmental quality and human health. This study integrates Kriging, the Ozone Isopleth Plotting Research (OZIPR) model, photochemical loss analysis, O₃ formation potential (OFP) assessment, and machine learning techniques to develop a mathematical modeling framework for evaluating the impact of VOC and NOₓ concentration variations on ozone formation. The novelty of this model lies in quantifying the photochemical losses of ozone precursors and back-calculating the initial emission concentrations of VOCs and NOₓ, which are subsequently used as input parameters for model development.The present artificial neural network (ANN)-based precursor impact estimation model performs well with R² and RMSE values of 0.804 and 0.08 ppbv, respectively. Simulation results indicate that in southern Taiwan's air quality management region, ethylene (C₂H₄) is the most influential precursor pollutant contributing to ozone formation. A 50 % reduction in ethylene emissions is projected to decrease ozone concentration by 38.5 %. These results suggest that local environmental agencies should prioritize regulating ethylene emissions, particularly from key industrial sources such as the refining and petrochemical sectors. The proposed modeling framework provides valuable insights for policymakers in formulating effective ozone control strategies and supports environmental agencies in air quality management and regulatory enforcement.