ASSESSMENT AND PREDICTION OF THE GROUND LEVEL OZONE CONCENTRATION IN THE EAST OF THAILAND

Abstract

To assess and predict whether the ground level ozone concentration exceeds an air quality standard in ambient, two different techniques have been applied. One is the traditional method, discriminant analysis model, and the other is an alternative scheme, neural network model. Daily ground ozone maximum concentration and other diverse variables in the air, measured from the monitoring stations in the east of Thailand for the period 2006-2010, were used to train and validate these two predictive models. The performance of the models can be evaluated by a correct classification rate (CCR). The result of performance comparison indicates the neural network model is shown to overcome the classical discriminant analysis model for both the training and the validation data set. That is, the average CCR of the neural network model is 87.22% for the training data set and 86.58% for the validation data set while the average CCR of the discriminant analysis model provides 79.77% and 78.98% for the training and the validation data set, respectively.