Research on driving factors of ozone in Henan province based on explainable machine learning

Abstract Near-surface ozone pollution poses a severe threat to human health and the ecological environment. To investigate the characteristics and driving factors of O 3 concentrations in Henan province in recent years, this study combined the extreme gradient boosting model with the SHapley Additive exPlanations interpretability framework, using air quality monitoring data and meteorological reanalysis datasets from 2022 to 2024. The impacts of meteorological variables and atmospheric pollutant emissions on O 3 were systematically examined. Results showed significant O 3 pollution in Henan Province, with ρ (O 3 − 8 h) of 170.6, 169.8, and 183.2 μ g m −3 for 2022, 2023, and 2024, respectively. Spatially, ρ (O 3 − 8 h) in central and northern Henan exceeded that in southern regions, with Jiaozuo, Xinxiang, and Zhengzhou exhibiting the highest values among 17 cities. Temporally, monthly variations in ρ (O 3 − 8 h) followed an inverted ‘V’ pattern, peaking in summer (summer > spring > autumn > winter). Meteorological factors accounted for 83.3% of the variance in ρ (O 3 − 8 h), while pollutant emissions accounted for 16.7%. Key meteorological factors exhibited complex nonlinear relationships with O 3 , including 2 m surface temperature, surface solar radiation downwards, surface thermal radiation downwards, and 2 m dew point temperature. Although city-specific driving factors varied, T2M and SSRD remained dominant across all regions, with cumulative contributions ranging from 52.7% to 57.0%. Seasonally, T2M was the primary driver in spring (31.4%), while T2M dominated in summer, autumn, and winter, reaching maximum influence in winter (41.5%). The prevention and control of ozone pollution require the adoption of targeted measures based on regional and seasonal differences.