Air Pollution Index Determination Using ARIMA and ARFIMA Time Series Models

Air pollution is one of the major issues that affects human health, agricultural crops, forest species and ecosystems. Since 1980, Malaysia has had a series of haze episodes and the worst ever was reported in 1997. As a result, the government has established the Malaysia Air Quality Guidelines, the Air Pollution Index (API) and Haze Action Plan, to improve the air quality. The API was introduced as an index system for classifying and reporting the ambient air quality in Malaysia. The forecast of air pollution can be used for air pollution assessment and management. It can serve as information and warning to the public in cases of high air pollution levels and for policy management of many different chemical compounds. The objective of this project is to fit and illustrate the use of time series models in forecasting the API using Klang in Selangor as a case study. The time series models being considered were the Integrated Autoregressive Moving Average (ARlMA) and the Integrated Long Memory Model(ARFIMA) models. The lowest mean absolute error (MAE), root mean squared error (RMSE) and mean absolute percentage error (MAPE) values were used as the model selection criteria. Between these two models considered, the integrated ARFIMA model appeared to be the better model as it has the lowest RMSE, MAE and MAPE values. However, the difference between the forecasting abilities of these models were very small and one has to weigh the slight gain in accuracy against the complexity of the model fitting procedure for the integrated ARFIMA model as compared to the ARiMA process. As such, both of these models may be used as alternatives in forecasting the API values.The results obtained would be helpfol for any engineering design or master plan that aims to assist in improving air quality for further systematic national industrial development.