Traffic related PM predictor for besiktas, Turkey

Abstract

The main objective of this study was to develop an Artificial Neural Networks (ANN) based model, which could be used as a tool for the prediction of traffic related PM2.5 and PM10 emissions. In this purpose, about 70 pairs of daily PM2.5 and PM2.5-10 samples were collected near to a main artery in Besiktas, Istanbul, Turkey. In addition to the PM data, hourly meteorological data, air quality data (CO, SO2, NO, NO2, NOx) and traffic data (traffic counts, speed, and density) were employed in the model. The results obtained from two different Neural Networks namely Forward NN (FFNN) and Radial Basis Function NN (RBFNN) were compared. While FFNN did not give good results due to limited number of data (60% of 70 data points) in high dimensional space (i.e., 14 dimensional space), more robust results were obtained with RBFNN with 72% prediction performance. © Springer-Verlag Berlin Heidelberg 2009.