應用多變量統計分析評估臺北水源特定區集水區水質

Watershed Water Quality Assessment in the Taipei Water Management District Using Multivariate Statistical Techniques

吳政南(Cheng-Nan Wu) ； 柳文成(Wen-Cheng Liu)

《農業工程學報》 59卷2期 (2013/06) Pp. 29-50

https://doi.org/10.29974/JTAE.201306_59(2).0002

摘要

台北水源特定區供應大台北地區近500萬人之自來水水源，係國內第一個經由都市計畫法設立之水源、水質與水量保護區，區域範圍涵蓋三個流域，包含新店溪流域及其上游之南勢溪流域及北勢溪流域。本研究以水源特定區管理局轄下的監測站之水質監測數據，配合氣象局雨量監測數據，合計12項物理與化學水質變數，包含溶氧量、生化需氧量、懸浮固體物、化學需氧量、pH值、比導電度、水溫、大腸菌類數、濁度、總磷、氯氮及兩量等。分析2008年到2010年每月的流域測站之水質資料，以量化方式探討台北水源特定區集水區複雜的水質資料結構，並應用多變量統計分析探討水質的特性。本研究利用判別分析可清楚判識三流域的特性有極大的差異，對照現況，可以驗證三個流域的水質變數各有其特性。利用主成份分析可得到台北水源特定區所包含的三個流域的水質主成份。北勢溪流域可萃取5個主成份，依序為懸浮固體成份、營養鹽成份、pH值成份、雨量成份及其他成份，可解釋的總變異量為67.260%；南勢溪流域也萃取出5個主成份，依序為懸浮固體成份、人類活動成份、總磷成份、雨量成份及氨氮成份，可解釋的總變異量為66.624% ;新店溪流域亦萃取出5個主成份，依序為懸浮固體成份、總磷成份、雨量成份、水溫成份及比導電度成份，可解釋的總變異量為73.024%。依照分析結果並對照三個流域現況，北勢溪流域為農業耕種盛行，但是污染負荷較低;南勢溪流域為溫泉及森林遊樂區，遊憩觀光盛行，人類活動造成的污染負荷影響水體甚多；新店溪流域為住宅區，人口稠密，雖有污水下水道處理，但是家庭廢水所造成的污染負荷也最為嚴重。利用群集分析進行測站分群，可將北勢溪流域測站分成4群、南勢溪流域測站分成3群、新店溪流域測站分成3群。如果在經費考量或是希望能到更好的監測效果的前提下，可以嘗試增加臨時性的測站，或是對於差異性很小的測站進行位置的改變，以評估比較水質監測的效果。

關鍵字

臺北水源特定區 ； 水質 ； 多變量統計分析 ； 判別分析 ； 主成分分析 ； 群集分析

並列摘要

The Taipei Water Management District, supplying the domestic water for 500 million people in Taipei city, is the first protection area of water sources, water quality and quantity according to the Urban Planning Law. It covers the three basins including the Pei-Shih Stream River, Nan-Shih Stream, and Hsin-Dian Stream. In the present study, the water quality monitoring data collected from the Water Management Office Stations and rainfall data collected from the Central Weather Bureau were adopted and analyzed. A total of 12 physical and chemical water quality variables, including dissolved oxygen (DO), biochemical oxygen demand (BOD), suspended solids (SS), chemical oxygen demand (COD), pH, conductivity, water temperature, coliform fecal, turbidity, total phosphorus (TP), ammonia nitrogen (NH3-N), and rainfall. The quantitative approaches were applied to analyze the long-term monitoring data from the water sampling stations located in the watershed during years 2008-2010. Also the multivariate analysis techniques were used to analyze the water quality characteristics.The three basins were identified clearly by discriminant analysis (DA), indicating that the water quality variables in the three basins have their own characteristics. The analyzed results concurred with the actual situation of the three basins. By using the principal component analysis (PCA), principal components of three basins in the Taipei Water Management District are extracted. The Pei-Shih Stream basin extracted five principal components and the percentage of explained variance is 67.26%. The first principal component is suspended solids concentration, subsequently are nutrients, pH, rainfall components, and other components. The Nan-Shih Stream basin extracted five principal components and the percentage of explained variance is 66.624%. The first principal component is suspended solids concentration, subsequently are human activities, total phosphorus, rainfall components, and ammonia nitrogen components. The Hsin-Dian Stream basin extracted five principal components and the percentage of explained variance is 73.024%. The first principal component is suspended solids concentration, subsequently are total phosphorus, rainfall, water temperature components, and conductivity components. Based on the results above the current situation of the three basins, agriculture cultivation is more prevalent in the Pei-Shih Stream basin but the pollution load is the lower. There are hot spring and recreation areas in the Nan-Shih Stream basin, and pollution load of human activities has a great impact on the water quality. In the Hsin-Dian Stream basin, there are lots of residential areas, but pollution load of domestic wastewater are the most serious. Cluster analysis (CA) was used to cluster the measured stations. The Pei-Shih Stream basin, Nan-Shih Stream basin, and Hsin-Dian Stream basin yield four clusters, three clusters, and three clusters, respectively. For financial or monitoring quality considerations, we may try to increase the temporarily measured stations or to change the measured locations for gathering similar monitoring data and then assess the effect of water quality monitoring.

並列關鍵字

Taipei Water Management District ； Water quality ； Multivariate statistical techniques ； Discriminant analysis ； Principal components analysis ； Cluster analysis