Abstract: In recent years, urban ozone pollution has become increasingly prominent, making the prediction of ozone concentrations crucial for early prevention and control measures. In Hefei, the number of days with ozone as the primary pollutant is increasing, and ozone has now surpassed PM_2.5 as the leading contributor to air quality degradation. In response, this study focuses on Hefei and uses the daily average ozone concentrations from January 2022 to June 2023 to create a time series. A Seasonal Autoregressive Integrated Moving Average (SARIMA) model is then developed to forecast the ozone levels for the summer of 2023 and the model's performance is evaluated. The results indicate that the model's fitted curves basically align with the actual measurements, with the relative errors of the three-day short-term forecasts remaining within 20%. The model's root-mean-square-error (RMSE) is 21.18 μg·m^-3, and the goodness-of-fit R² reaches 0.904 3, demonstrating robust fitting and predictive capabilities. These ozone forecasts provide a scientific basis for environmental protection efforts, support decision-making in urban planning, and positively impact public health.

Key words: ozone concentration, time series, SARIMA model, seasonality, short-term forecast