Time-series analysis of air pollution data

Time-series analysis of air pollution environmental levels involves the identification of long-term variation in the mean (trend) and of cyclical or periodic components. A model based on a stepwise approach to time-series analysis was applied to the daily average concentrations of strong acidity (SA) and black smoke (BS) in the Oporto area, using an available computer program. Each step is completed by a correlation analysis of the residuals, allowing the identification of an optimal structure with a residual white noise. A periodic component with harmonics defined through "peaks" of concentration on week middle days and "troughs" on weekends was observed. SA concentration behaviour can be related with industrial activities, mainly through fossil-fuel burning in discontinuous working cycles. The observed evolution for BS is most probably related with weekly patterns of motor traffic, with observed minimum values during weekends. The periodic components represent, on the average, about 5% of the total variance for the SA series and 15% for the BS series. However, the weekly cycles are predominant in the SA series, representing on the average 75% of the periodic variance, against 46% for the BS series. Statistically significant higher frequency (approximate to 2-4 day) periodic components were observed for both pollutant indicators and for all collection sites analysed. This may be due to synoptic weather variations of minimum and maximum daily temperature and precipitation, which show similar periods in the study area. (C) 1999 Elsevier Science Ltd. All rights reserved.