Characteristics and simulation of ozone pollution in Haikou City under combined effects of meteorological factors and pollutants

Ozone (O 3 ) pollution has become a central concern in regional air-quality management. Using environmental monitoring and co-located surface meteorological observations for Haikou from 2016 to 2024, we examine nine-year variations in the daily maximum 8-h average O 3 concentration (O 3 -8 h) Twelve routinely measured air pollutants and meteorological variables were used as inputs to three models—random forest (RF), XGBoost, and LightGBM—to simulate and verify O 3 -8 h in 2024. Over the study period, O 3 -8 h increased at 1.61 µg (m 3 a) −1 . Haikou averaged 9.1 days of light O 3 pollution per year, with an overall exceedance rate of 2.49%. Seasonally, O 3 -8 h was highest in autumn, followed by winter and spring, and lowest in summer. The monthly cycle was bimodal, peaking at 78.7 µg m −3 in April and 101.3 µg m −3 in October. O 3 -8 h correlated positively with PM 10 and PM 2.5 ( r = 0.67 and 0.64). All three models reproduced the 2024 temporal pattern but underperformed for extreme values. RF performed best (RMSE = 21.35 µg m −3 , MB = − 5.29 µg m −3 , MNB = − 0.40%, and R 2 = 0.566), indicating stronger applicability for dynamic O 3 simulation in Haikou. The results provide a basis for ozone early warning and air-quality management in tropical coastal cities.