The Prediction of Surface Layer Ozone Concentration Using an Improved AR Model

In order to forecast the surface layer ozone concentration in the eastern coastal cities of China, an improved autoregressive method is used to dispose the ozone concentration data observed in November, 2008 in Tianjin, China in this paper. First the data are disposed by traditional auto-regressive model, then the real observed data are subtracted by the initial prediction value, through which the error terms are obtained. Next the error terms are disposed by filtering. The obtained filtered error terms are used in the AR model again and the new error terms are obtained, finally they are used to predict the concentration data a few hours ahead. Empirical results show that the proposed model is better than the traditional AR model, furthermore, the shorter the prediction time is, the better the model's prediction result is. So it is concluded that the proposed method is a nice method in predicting short term ozone concentration.