Stochastic modelling of daily air pollution in Burgas, Bulgaria

Abstract

Exceeding the norms and limits of atmospheric air pollution causes enormous damage to the population’s health and the environment. Determining the factors affecting air quality is a current task in a local, regional, and global scale. In this study, we use daily time series data for the main air pollutants in Burgas, Bulgaria – O3, NO, NO2, CO, SO2, and PM10, to analyze, model, and forecast these levels depending on meteorological factors. For this purpose, the stochastic ARIMA method and ARIMA with transfer functions are applied. Results are obtained for univariate and multivariate time series. Particular attention is paid to the concentrations of the secondary pollutant ground-level ozone (O3), which are modelled as a function of all variables considered. Results were evaluated using root mean square error, mean absolute percentage errors, and the coefficient of determination. Short-term forecasts have been obtained for seven days ahead. Model accuracy up to 84% has been established.