| Summary: | 碩士 === 國立高雄大學 === 土木與環境工程學系碩士班 === 102 === Due to the impact of fine particulate matters (PM2.5) on health and environment, Taiwan is scheduled to make up the control standards in 2012. The standard was 24-hour average 35ug/m3 and the annual average 15ug/m3. In this research, we used the air quality station’s monitoring data from 2007~2011. Select the three regions of northern, central and southern Taiwan include 5 different backgrounds station (general, traffic, national parks, industrial, background) total of 16 stations. The monitoring item contains Weather factors (temp, rain, humidity, wind speed) and pollutants factors (NO, NO2, NOx, CO, O3, SO2, PM10, PM2.5, CH4, THC, NMHC ). We plan to focus on the problems about understand the path of distribution of the fine particulate matters, its origin and its impacts to monitoring and control issues. To establish PM2.5 prediction models of each station. The method includes Principal Component Analysis (PCA), Multiple Regression Analysis, Time Series Analysis and Cluster Analysis. The results indicate that PCA combined with Multivariate Regression Analysis to predict with good results. The first principal component (PC1) shows the concentration of air pollutants are decreased with wind speed and temperature increased in northern and southern Taiwan, respectively. The second principal component (PC2) shows most of stations presents the high correlation between O3 and air pollutants. It appears the significant influence for air quality by photochemical reaction. In the Time Series Analysis, the model length is southern more than northern Taiwan. Show that the southern was still contaminated than other places. In the prediction part, approximately equal with the actual situation of three lags. After three lags, due to the effects of heterogeneity, the predicted value approaching the average mean. In the Cluster Analysis, we used different conditions to understand the cluster of station. The result shows that Yangming station has some of unique and representative because of geographical environment. And we used K-Means Cluster divide the station into three clusters of stations in order to understand the differences between different clusters.
|
|---|
