Document Type : علمی - پژوهشی


Department of Power and Control Engineering, School of Electrical and Computer Engineering, Shiraz University, Shiraz


Introduction: Transportation and traffic flows are the main factors of air pollution in Iran. Therefore it is essential to investigate their effects on the air quality of urban areas to forecast more accurately and manage better traffic and pollution. Unfortunately, not much research has been conducted on this issue in Iran. Identification of pollutant sources is the most important and time-consuming stage of air pollution modelling. We cannot consider only one variable for air pollution modelling in a single region. Hence, different variables should be taken into account, studied and planned. Some measures make significant changes in the air pollution of a metropolis. Hence, undertaking a series of measures can reduce the air pollution, and adopting new methods to evaluate the air pollution is one of these measures. The main goal of this research is to offer a smart model by which concentration of pollutants such as CO can be estimated with the appropriate accuracy and, by examining the causes of these pollutants and predicting the air pollution, the necessary actions to manage and control the air pollution can be planned (Hassan and Croether, 1998). Materials and methods: In this paper, a neural network model and a nonlinear state space model were designed based on urban traffic in Shiraz. In these models the concentrations of CO, NO2 and SO2 pollutants were analyzed and also estimated using a Kalman Filter for a 24 hour cycle. The models are based on the correlation between the volume of pollution, traffic, initial pollution and meteorological information. The extended Kalman Filter algorithm was used to analyse and predict the air pollution in Shiraz over a 24 hour period. A key factor of the proposed system is its adaptation with the short time pollution changes (Safavi, 2008; Brown et al., 2007). Result and discussion: In this research, traffic and pollution data caused by pollutant concentrations has been studied, then an attempt was made to match these air pollution data with significant parts in Shiraz city and many traffic and pollution data were excluded due to a mismatch in terms of location. Finally, modelling was updated based on this data and the result was adapted to real data. This nonlinear model structure offers the advantage of being evolutionary and sufficiently flexible, in the sense that the overall evaluation of the model performance can easily be undertaken by excluding or adding input variables. On the other hand, if the corresponding data of each new station is available, the study can be extended to other parts of Shiraz city. So, if traffic data is available in some parts of the city, the pollution can be extended to some other parts and be reduced in critical areas using certain traffic strategies. Conclusion: The neural network method and the Kalman Filter were tested on Shiraz pollution data which revealed that the models, specially the Kalman Filter, work reasonably well.


  1. Hassan A.A, Crowther J.M. Modelling of fluid flow and pollutant dispersion in a street canyon. Environmental Monitoring and Assessment Journal; 1998; 52:281-297.
  2. Branis M. Air quality of Prague: traffic as a main pollution source. Environmental monitoring and assessment Journal; 2009; 156.1-4:377-390.
  3. Kim Y, Guldmann J.M. Impact of traffic flows and wind directions on air pollution concentrations in Seoul, Korea. Atmospheric Environment Journal; 2011; 45.16:2803-2810.
  4. Gokhale S. Traffic flow pattern and meteorology at two distinct urban junctions with impacts on air quality. Atmospheric Environment Journal; 2011; 45.10:1830-1840.
  5. Keuken M.P. Elemental carbon as an indicator for evaluating the impact of traffic measures on air quality and health. Atmospheric Environment Journal; 2012; 61:1-8.
  6. Marsik T, Johnson R. Model for Estimation of Traffic Pollutant Levels in Northern Communities. Journal of the Air & Waste Management Association; 2010; 60 (11):1335-1340.
  7. Zolghadri A, Cazaurang F. Adaptive nonlinear state-space modelling for the prediction of daily mean PM10 concentrations. Environmental Modelling & Software Journal; 2006;21, (6):885-894.
  8. Safavi A.A. Wavelet-based neural network and multiresolution analysis with applications to process systems engineering. Ph.D.: Dept. of Chemical Eng., the University of Sydney, Australia; 1996.
  9. Haykin S. Neural networks-a comprehensive foundation. 2nd Ed, Prentic-Hall; 1999.
  10. Ding X, Canu S, Denceux T. “Neural network model for forecasting.” Neural networks and their applications, In J. G. Taylor, Ed., John Wiley and Sons; 1996. pp 153-167.
  11. Winer N. Extrapolation, Introduction, and Smoothing of Stationary Time Series. New York. Wiley; 1949.
  12. Kalman R. E. A New Approach to Linear Filtering and Prediction Problems.Trans. ASME-J. Basic Eng., 35-45; 1960.
  13. Brown R. G, Hwang P. Y. C. Introduction to Random signals and applied Kalman Filtering. Wiley, 4th Ed; 2007.
  14. Young P.C, Ng C.N, Lane K, Parker D. Recursive forecasting, smoothing and seasonal adjustment of non-stationary environmental data. Journal of Forecasting; 1992; (10), pp 57-89.