應用多方向量分析分析污染檢測數據

Abstract

鑒定化學污染源與瞭解污染團在時空上的分佈，在環境污染領域是重要且新受矚目的問題。在多重污染源地區，環境工程師面臨著如何鑒定複雜地形的污染源型態。如果污染源具有獨特的化學指紋，即可藉由多變數統計分析方法來進一步解決環境指紋問題。 多方向量分析(Polytopic Vector Analysis)是一種鑑定污染源化學指紋的多變量分析統計方法。假定某一混合系統，多方向量分析可用來計算三個重要參量：(1)來源(後端成分)個數、(2)各後端成分的化學指紋組成、以及(3)不同樣本中來自各來源的濃度相關比例。 本研究應用多方向量分析，分別針對兩個研究案例做分析(I)客雅溪流域重金屬檢測數據，以及(II)新竹科學園區地下水有機污染檢測資料。 近年來，在新竹香山海域養殖的牡蠣及沈積物的銅含量皆出現顯著增高的趨勢。綠牡蠣體內伴隨著高濃度的銅含量，不但威脅消費者健康且影響漁民每年三百多萬美金產值。香山牡蠣銅污染的來源是長久被關切且極需解決的問題。雖然在過去的研究曾指出香山海域銅污染可能源自於新竹科學園區或香山工業區，然而卻缺少強而有力的證據指出確切的污染者。於2006年檢測客雅溪及其支流的十二組數據，這十二組數據分別來自於不同兩組研究團隊的檢測成果。本研究目的在於利用統計方法-多方向量分析，鑑定綠牡蠣組織內的重金屬銅之來源。第一組及第二組數據，分別在客雅溪沿岸檢測共6個水樣及7個懸浮固體樣本，採樣時間為2006年四月，每個取樣分析18種重金屬項目(鋅、銅、鉛、鐵、鋁、錳、鎘、鍶、鉻、鋇、鎳、銀、錫、砷、釩、鎢、鎵及鉬)。第三組到第八組數據，沿客雅溪流域7個取樣點，採樣時間為2006年二月至七月；第九組到第十二組數據，沿客雅溪流域8個取樣點，採樣時間於2006年八月至十一月，每個水樣分析六個項目(懸浮固體、溶氧、生化需氧量、氨氮、銅及砷)。由多方向量分析結果顯示，新竹科學園區為綠牡蠣中銅的主要來源。 在第二個研究案例，係應用多方向量分析針對新竹科學園區地下水污染檢測資料作分析，利用計算係數以及因子負荷係數判定該工業區地下水有機污染物資料，鑒定出有六個有機污染物指紋，進而計算出各後端成分在地理上的分佈。因此，利用多方向量分析可決定後端成分的個數，並且發展出一個地下水污染物的混合模式，以模擬地下水污染源化學指紋與污染團在空間上的分佈。

Identification of chemical contaminant sources is an important environmental problem. Given multiple sources in a field area, the environmental engineer is faced with the challenge of identification and mapping multiple plumes with overlapping geographic distribution. If a contaminant source is characterized by a distinctive spectrum of chemicals, then the environmental chemical fingerprinting problem may be addressed through a multivariate statistical approach. Polytopic vector analysis (PVA) is a statistical pattern recognition technique for multivariate data used to identify fingerprints of contaminant sources. Given a mixed system, PVA is used to determine three parameters of interest in a mixed system: (1) the number of sources (end-members), (2) the composition of each source, and (3) the relative proportions of each source in each sample. In this study, polytopic vector analysis is applied to analyze two problems. In the first problem, the copper source of green oyster in Hsianshan wetland is identified while in the second problem the groundwater contamination by organic chemicals in Hsinchu science park is analyzed. In recent two decades, the copper concentrations in both the oyster organs and sediment were very high in the Hsianshan coastal area, Taiwan. The oyster with high concentration of copper poses not only a threat to human health but also results in an annual loss of about 3.1 million US dollars. What is the source of copper in the Hsianshan oyster is a question of long standing and tough problem to be solved. Previous studies indicated that the copper source originated from either the Hsianshan industrial park (HIP) or the Hsinchu Science Park (HSP) is responsible for the contamination in Hsianshan coastal area. Although several investigations had been conducted for identifying the copper source of green oyster; however, there was no clear evident or result to show who is responsible for the copper pollution. In order to search for the source of copper, the water and suspended solid samples were collected and analyzed. Water samples mainly collected from the Keya stream and some suspected source locations in 2006. Samples of suspended solid were also collected from the Keya stream and coastal area of Hsianshan. Totally 12 sets of water and suspended solid sample data were available. The Data Set 1 contains water samples taken from 6 different sites along the Keya stream, Hsinchu while the Data Set 2 has suspended solid samples taken from the same stream at 7 different sites. Both two data sets are analyzed for eighteen heavy metals including Zn, Cu, Pb, Fe, Al, Mn, Cd, Sr, Cr, Ba, Ni, Ag, Sn, As, V, W, Ga, and Mo. The Data Sets 3 to 8 were sampled at 7 locations along the Keya stream during February to July, 2006. The Data Sets 9 to 12 were collected at 8 locations, where 7 locations were the same as those taken above and one extra location was at Yuchegou creek, during August to November, 2006. Those ten data sets, Data Sets 3 to 12, were taken and analyzed for six items (SS, DO, BOD, NH3-N, Cu, As). The purpose of this study is to identify the copper source based on those sample data using a multivariable statistical analysis called polytopic vector analysis (PVA). The results of PVA indicate that a significant part of copper comes from HSP and a minor part of copper is from Yuchegou creek. In the second problem, the study applies PVA to analyze the concentration data taken from HSP, Taiwan. Based on inspecting the coefficient of determine (CD scatter) and the factor loadings index, six separate organic compound fingerprints are identified. Moreover, the geographic distribution of each end-member can be presented. Thus, this method (PVA) can develop a groundwater mixing model by searching for end-members. In addition, this mixing model can identify chemical contaminant sources and map multiple plumes with overlapping geographic distributions.