Statistical analysis of urban air-pollution data in the Athens basin area, Greece

TLDR: In this paper, an attempt is made to predict daily and monthly concentrations by utilising a statistically based model, that is, a multiple linear regression model, using the Autoregressive Integrated Moving Average (ARIMA) stochastic model in combination with the Theil index.

Abstract: Air pollutant concentrations recorded in the Athens basin area by a network of stations are analysed and examined to estimate the typical behaviour of the air pollutants, especially with regard to daily, weekly and annual periodicities and meteorological dependencies. Also, an attempt is made to predict daily and monthly concentrations by utilising a statistically based model, that is, a multiple linear regression model. The results obtained show the presence of a very significant weekly periodicity for all the analysed air pollutants (e.g. SO2, O3, CO, NO and NO2). It also appears to be present a yearly periodicity for the primary (e.g. SO2, CO, NO and NO2) air pollutants analysed and studied. The statistical prediction by using the Autoregressive Integrated Moving Average (ARIMA) stochastic model in combination with the Theil index, shows good predicting capabilities (one day ahead) for sulphur dioxide (SO2), carbon monoxide (CO), ozone (O2) and nitrogen oxides (NO and NO2). The significance of the forecasting is controlled by the Theil index.