Forecasting Long Memory in Ozone Levels in Peninsular Malaysia using ARFIMA Model

Forecasting air pollutant concentrations is crucial for managing pollution levels that pose significant risks to human health and the environment. Among the various pollutants, ground-level ozone (O₃) is a secondary pollutant of particular concern due to its harmful health effects and complex formation dynamics involving nitrogen oxides (NOₓ)and volatile organic compounds (VOCs) under sunlight. This study focuses on predicting ozone levels using the autoregressive fractionally integrated moving average (ARFIMA) model, which effectively captures long memory behavior in time series data. The research begins by analyzing ozone concentration data from three monitoring stations across Peninsular Malaysia to identify long memory behavior. Using the Aggregated Variance (V/S) method, the presence of long memory is confirmed through the estimated differencing parameter d, which falls between 0 and 0.5 based on the Geweke and Porter-Hudak (GPH) estimator. The Augmented Dickey-Fuller (ADF) test was employed to assess the stationarity of the time series. Optimal ARFIMA models were selected based on the Akaike Information Criterion (AIC), and model performance was evaluated using the Root Mean Square Error (RMSE). The selected ARFIMA models demonstrate the lowest RMSE values, confirming their efficacy in forecasting ozone concentrations. The findings suggest the effectiveness of the ARFIMA model in forecasting ozone concentrations, which can significantly aid in air pollution assessment and management strategies. It contributes to understanding ozone pollution in Peninsular Malaysia, facilitating informed decision-making for environmental policies. Future research should explore additional time series models to strengthen the understanding of ozone dynamics and enhance forecasting capabilities in the context of long memory properties.