ANALYSIS OF PM10 USING EXTREME VALUE THEORY

The literature review had identified that the extreme value theory is widely used in hydrological studies. However, its contribution in air pollution is indisputably important. This paper assesses the use of extreme value distributions of the two-parameter Gumbel, two and three-parameter Weibull, Generalized Extreme Value (GEV) and two and three-parameter Generalized Pareto Distribution (GPD) on the maximum concentration of daily PM10 data recorded in the year 2005 in Shah Alam, Selangor. Parameters estimations for all distributions were evaluated using the method of Maximum Likelihood Estimator (MLE). The goodness-of-fit of the distribution was determined using six performance indicators namely; the accuracy measures which include Predictive Accuracy (PA), Coefficient of Determination (R2), Index of Agreement (IA) and error measures that consist of Root Mean Square Error (RMSE), Mean Absolute Error (MAE) and Normalized Absolute Error (NAE). The best distribution was selected based on the highest accuracy measures and the smallest error measures. This study reveals that the three-parameter Weibull was the best fit for daily maximum concentration for PM10. The analysis also demonstrates that the number of days in which the concentration of PM10 exceeded the Malaysia Ambient Air Quality Guidelines (MAAQG) of 150 mg/m^3 for 2005 was 25 days as compared to the actual 15 days.