Simple Forecasting of Surface Ozone through a Statistical Approach

Objectives: Ozone ($O_3$) advisories are issued by provincial/prefectural and city governments in Korea and Japan when oxidant concentrations exceed the criteria of the related country. Advisories issued only after exposure to high $O_3$ concentrations cannot be considered ideal measures. Forecasts of $O_3$ would be more beneficial to citizens' health and daily life than real-time advisories. The present study was undertaken to present a simplified forecasting model that can predict surface $O_3$ concentrations for the afternoon of the day of the forecast. Methods: For the construction of a simple and practical model, a multivariate regression model was applied. The monitored data on gases and climate variables from Japan's air quality networks that were recorded over nearly one year starting from April 2016 were applied as the subject for our model. Results: A well-known inverse correlation between $NO_2$ and $O_3$ was confirmed by the monitored data for Iksan, Korea and Fukuoka, Japan. Typical time fluctuations for $O_3$ and $NO_x$ were also found. Our model suggests that insolation is the most influential factor in determining the concentration of $O_3$. $CH_4$ also plays a major role in our model. It was possible to visually check for the fit of a theoretical distribution to the observed data by examining the probability-probability (P-P) scatter plot. The goodness of fit of the model in this study was also successfully validated through a comparison (r=0.8, p<0.05) of the measured and predicted $O_3$ concentrations. Conclusions: The advantage of our model is that it is capable of immediate forecasting of surface $O_3$ for the afternoon of the day from the routinely measured values of the precursor and meteorological parameters. Although a comparison to other approaches for $O_3$ forecasting was not carried out, the model suggested in this study would be very helpful for the citizens of Korea and Japan, especially during the $O_3$ season from May to June.