This paper presents an artificial neural network model that is able to predict ozone concentrations as a function of meteorological conditions and precursor concentrations. The network was trained using data collected during a period of 60 days near an industrial area in Kuwait. A mobile monitoring station was used for data collection. The data were collected at the same site as the ozone measurements. The data fed to the neural network were divided into two sets: a training set and a testing set. Various architectures were tried during the training process. A network of one hidden layer of 25 neurons was found to give good predictions for both the training and testing data set. In addition, the predictions of the network were compared to measurements taken during other times of the year. The inputs to the neural network were meteorological conditions (wind speed and direction, relative humidity, temperature, and solar intensity) and the concentration of primary pollutants (methane, carbon monoxide, carbon dioxide, nitrogen oxide, nitrogen dioxide, sulfur dioxide, non-methane hydrocarbons, and dust). A backpropagation algorithm with momentum was used to prepare the neural network. A partitioning method of the connection weights of the network was used to study the relative % contribution of each of the input variables. It was found that the precursors carbon monoxide, carbon dioxide, nitrogen oxide, nitrogen dioxide, and sulfur dioxide had the most effect on the predicted ozone concentration. In addition, temperature played an important role. The performance of the neural network model was compared against linear and non-linear regression models that were prepared based on the present collected data. It was found that the neural network model consistently gives superior predictions. Based on the results of this study, artificial neural network modeling appears to be a promising technique for the prediction of pollutant concentrations.
ASJC Scopus subject areas