Investigating the Degree of Persistence, Trend and the Best Time Series Forecasting Models for Particulate Matter (PM10) Pollutant Across Malaysia

Abstract

Particulate matter is the most common atmospheric pollutant with some negative consequences on human health, environment, and the ambient air quality. In this study, the concentration of particulate matter in sixty-five air quality monitoring stations across Malaysia during January to December 2018 is analyzed. We investigated the degree of persistence and trend of the particulate matter series and developed a forecasting model using both the autoregressive integrated moving average (ARIMA) and the autoregressive fractionally integrated moving average (ARFIMA) time series methods for each monitoring station separately. Mean absolute deviation (MAD), mean absolute percentage error (MAPE) and root mean square error (RMSE) are used to determine the best fitted model for forecasting each monitoring station. Ljung-Box test of uncorrelated residuals confirmed the adequacy of each of the model. The results confirmed the evidence of transitory form of persistence in the level of particulate matter pollutant at sixty-four monitoring stations while trend increases in seventeen monitoring stations. Forecast error analysis indicates that ARFIMA models performed better than ARIMA models by producing smaller RMSE values in forty-two of the sixty-five monitoring stations. However, the overall result indicates that none of the model could be regarded as universal in forecasting particulate matter concentration, and their performance is independent of the category or location of a given monitoring station.