By: Edouard Philippe Martin
Ground–level ozone (O3) and fine particulate matter (PM2.5) are two air pollutants known to reduce visibility, to have damaging effects on building materials and adverse impacts on human health. O3 is the result of a series of complex chemical reactions between nitrogen oxides (NOx) and volatile organic compounds (VOCs) in the presence of solar radiation. PM is a class of airborne contaminants composed of sulphate, nitrate, ammonium, crustal components and trace amounts of microorganisms. PM2.5 is the respirable subgroup of PM having an aerodynamic diameter of less than 2.5 μm. Development of effective forecasting models for ground-level O3 and PM2.5 is important to warn the public about potentially harmful or unhealthy concentration levels.
The objectives of this study is to investigate the applicability of Multiple Linear Regression (MLR), Principle Component Regression (PCR), Multivariate Adaptive Regression Splines (MARS), feed-forward Artificial Neural Networks (ANN) and hybrid Principal Component – Artificial Neural Networks (PC-ANN) models to predict concentrations of O3 and PM2.5 in Montréal (Canada). Air quality and meteorological data is obtained from the Réseau de surveillance de la qualité de l’air (RSQA) for the Airport Station (45°28′N, 73°44′W) and the Maisonneuve Station (45°30′N, 73°34′W) for the period January 2004 to December 2007. Air pollution data include concentration values for nitrogen monoxide (NO), nitrogen dioxide (NO2), carbon monoxide (CO) and 142 different volatile organic compounds. Meteorological data include solar irradiation (SR), temperature (Temp), pressure (Press), dew point (DP), precipitation (Precip), wind speed (WS) and wind direction (WD).
Analysis of the available volatile organic compound data expressed on a propylene–equivalent concentration indicated that m/p–xylene, toluene, propylene and (1,2,4)–trimethylbenzene were species with the most significant ozone forming potential in the study area.
Different models and architectures have been investigated through five case studies. Predictive performances of each model have been measured by means of performance metrics and forecast success rates. Overall, MARS models allowing second order interaction of independent basis functions yielded lower error, higher correlation and higher forecast success rates. This study indicates that models based on statistical methods can be cost-effective tools to forecast ground-level O3 and PM2.5 in Montréal and to provide support for decision makers in protecting human health.
The complete paper can be viewed online at http://spectrum.library.concordia.ca/35960/