Time series analysis and forecasting for air pollution in small urban area: an SARIMA and factor analysis approach

Despite the existing public and government measures for monitoring and control of air quality in Bulgaria, in many regions, including typical and most numerous small towns, air quality is not satisfactory. In this paper, factor analysis and Box–Jenkins methodology are applied to examine concentrations of primary air pollutants such as NO, NO₂, NO_x, PM10, SO₂ and ground level O₃ in the town of Blagoevgrad, Bulgaria within a 1 year period from 1st September 2011 to 31st August 2012, based on hourly measurements. By using factor analysis with PCA and Promax rotation, a high multicollinearity between the six pollutants has been detected. The pollutants were grouped in three factors and the degree of contribution of the factors to the overall pollution was determined. This was interpreted as the presence of common sources of pollution. The main part of the study involves the performance of time series analysis and the development of univariate stochastic seasonal autoregressive integrated moving average (ARIMA) models with recording on a hourly basis as seasonality. The study also incorporates the Yeo–Johnson power transformation for variance stabilizing of the data and model selection by using Bayersian information criterion. The obtained SARIMA models demonstrated very good fitting performance with regard to the observed air pollutants and short-term predictions for 72 h ahead, in particular in the case of ozone and particulate matter PM10. The presented statistical approaches allow the building of non-complex models, effective for short-term air pollution forecasting and useful for advance warning purposes in urban areas.