SCREENING AIR POLLUTION EPISODES BY PRINCIPAL COMPONENT ANALYSIS

摘要

This investigation employed principal component analysis to examine air-quality monitoring data over Taiwan, and selected PM_(10) and ozone episodes in seven different air-quality districts. During the screening process, the component scores of the first principal component were applied as the indicators to cite the episode day. As for the suspended particles and ozone, the first principal component of varied air-quality districts can explain at least 79% and 65%, respectively, for the total variance of concentrations. As for the principal components behind the first principal component, since their contribution ratios to PM_(10) particles and ozone were, respectively, lower than 9% and 12%, they were not suitable indicators for selecting air pollution episodes. The average component scores of the first principle component and station-day numbers had significant positive correlations (greater than 0.7) for PM_(10) and ozone at most air-quality districts. While considering the PM_(10) episode, the polluted days resulted from dust storms can not be sieved as PM_(10) episode. In filtering the continuous five-day episode, several principals and two preferential episodes were cited in this study.