Extreme Value Distribution for Prediction of Future PM10 Exceedences

Ahmad Shukri Yahaya; Nor Azam Ramli Ramli; Noor Faizah Fitri Md Yusof
Central fitting distribution (CFD) such as Weibull, gamma and lognormal distribution can give a good result for fitting the mean concentration of air pollutants data. However, it cannot precisely fit the high concentration region. Therefore, extreme value distributions (EVD) that are Gumbel and Frechet distributions were used in this research to fit the high particulate event in Seberang Perai, Penang from 2002 to 2006 to reduce the predicting error. The cfd (Weibull, gamma and lognormal distributions) and evd (Frechet and Gumbel distributions) were used to fit the daily maximum concentration. The best distribution that can fit the data was selected based on performance indicators. Furthermore, the exceedences of a critical PM10 concentration over the Malaysian Ambient Air Quality Guidelines were estimated using the best distributions. The results of performance indicators show that the extreme value distribution gives better fit to the actual high PM10 I. INTRODUCTION concentration than the central fitting distribution. The exceedences over a high particulate event were successfully predicted. In 2002, the exceedences is 291 days, 224 days in 2003, 151 days in 2004, 156 days in 2005 and 9 days in 2006.
central fitting distribution; Gumbel distribution; Frechet distribution; method of moments; daily maximum concentration
Download | Back to Issue| Archive