河川污染源鑑識的方法建立與指紋圖譜的應用

Abstract

隨著科技的進步，工業化的比例提高，使工業廢水污染對環境和人類健康的危害變得更大。因此，能夠精準的識別工業廢水的污染源有助於河川水質的管理。近年來受體模式之正矩陣因子法（Positive Matrix Factorization，PMF）已逐漸應用於河川水體的污染源鑑識分析，然而作法上會遇到一些常見的限制，這些限制也影響了模式結果後續在對應污染的工業類別的應用。為改善上述限制，在本論文中以三種方法與PMF 模式結合以改善河川的來源解析。 第一個研究為從污染物的資料前處理上精進PMF模式，建立了一種用於識別河川中特定河段污染源的方法。本研究選擇在一條污染嚴重的河段上游和下游分別設置採樣點，每天採集兩次河水樣本，共連續監測30天，監測期間同時也採集了潛在污染源的工業廢水樣本。河段的污染物濃度資料是根據上下游兩個採樣點所量測到的污染源質量流量除以體積流量計算所獲得，並針對計算過後的 31 種元素濃度進行正矩陣分析，以解析污染源可能的工業類別與其指紋圖譜及貢獻。模式分析結果解析出河段的污染工業類別。透過比較單一受體點分析和多個受體點分析的結果，顯示出河段的污染源解析可以排除上游污染物所造成之影響，並降低獲得混合指紋圖譜而導致污染源識別的難度。因此，從結果可以證實在此階段的研究中所提出的兩個受體點間污染物濃度計算與預處理方法的適用性和有效性。 第二個研究為從輸入資料中所包含元素的角度完善PMF模式，透過鑭系元素的測試結果探討低貢獻的元素在PMF分析中的重要性。本研究選擇在工業區附近一條受污染的河川進行連續45天的監測、每天採集兩次河水樣本，並對測量的 31 種元素進行正矩陣分析。在監測期間，同時對潛在污染源的廢水樣本進行採集，以建立污染源指紋圖譜。透過鑭系元素的敏感性測試結果得知，排除部分或全部稀土元素將無法將污染源歸入正確的工業類別。此外，部分鑭系元素是工業類別的關鍵元素，使它們成為區別部分工業類別的污染源指紋圖譜的因素之一。因此，低濃度元素的指紋圖譜是有助於識別污染源的工業類別，並且可增加其指紋圖譜間的差異。 第三個研究為建立各個工業廢水指紋圖譜與其重要元素，增加PMF模式對於河川污染源鑑識的可用性。透過彙整來自不同河川附近所採集的潛在污染源的廢水樣本，採用聚類分析的方法計算產業間的相似度後進行分類，並進一步透過皮爾森相關係數確認分類結果的合理性，建立各個工業類別的複合指紋圖譜。進一步歸納各行業別中顯著貢獻的元素來識別污染源，以提高複合指紋圖譜的可用性，最終應用於河川的污染源鑑識分析。透過所建立的工業類別的複合指紋圖譜資料庫應用於阿公店溪的污染源解析，在沒有當地的工業廢水指紋圖譜的情況下，儘管模式分析結果無法比對出正確的污染源，但仍可以透過複合指紋圖譜獲得可能的潛在污染源。進一步透過複合指紋圖譜中主要的元素判斷工業類別，再縮小可能的行業別範圍。結果證實複合指紋圖譜的資料庫建立對於河川污染管制有其效益之存在。 總結來說，透過受體模式方法上的修正進行污染源鑑識是有助於河川污染的管理。識別河川河段污染源的方法建立可以減少混合來源的產生，使其更有效的確認高污染區域的污染源。此外，除了常被關注的重金屬元素之外，低濃度元素濃度的監測亦不可或缺，完整的元素濃度資料有助於模式分析。對於環境管理的政策推動上，建立完整的工業類別指紋圖譜供各個的河川污染防治所使用，可以提升污染源鑑識的實際應用效能。

With advancements in science and technology, the degree of industrialization has increased, exacerbating the impact of industrial wastewater pollution on both the environment and human health. Therefore, accurately identifying the sources of industrial wastewater pollution is essential for effective river quality management. In recent years, the Positive Matrix Factorization (PMF) model, a receptor model, has been increasingly applied in the identification and analysis of pollution sources in river water. However, there are some common limitations associated with this approach, which can affect the subsequent application of the model results in identifying the corresponding industry categories of pollution. To address these limitations, this dissertation proposes the integration of three studies with the PMF model to improve the source apportionment of river water pollution. The first study aims to improve the PMF model through the pre-processing of pollutant data, establishing a method for identifying pollution sources in a river reach. River water sample collection was conducted twice a day for 30 days in a polluted river in southern Taiwan at sampling sites upstream and downstream of the river reach. Wastewater samples from potential pollution sources were also taken during the monitoring period. The pollutant concentration data for the river reach was calculated based on the mass flow rate at two sites divided by the volume flow rate. Positive matrix factorization was applied to the 31 elements measured in the river to resolve source profiles and contributions. The source profiles of potential pollution were identified by the receptor models and then the corresponding industry categories were determined. By comparing the results of the single-site analysis and multiple-site analysis, it was demonstrated that source apportionment for the river reach effectively excluded the influence of upstream pollutants, thereby reducing the difficulty of pollution source identification caused by the mixed-source profiles. Consequently, the applicability and effectiveness of the proposed pretreatment method for calculating pollutant concentration calculations from two sites were demonstrated. The second study refines the PMF model by considering the elements present in the input data, using the results of lanthanides to assess the importance of low-concentration elements in PMF model analysis. Over a period of 45 days, water samples were obtained twice daily from a polluted river in southern Taiwan near industrial areas. During the monitoring period, wastewater samples from potential contamination sources were conducted to establish the source profiles. Positive matrix factorization was used to determine the pollution sources from the 31 elements measured in the stream. The sensitivity test results for lanthanides show that excluding some or all of them renders it impossible to classify pollution sources into the correct industry categories. Because lanthanides are key elements in certain industry categories, they are distinguishing factors in profiles among industries. Therefore, the source profiles of low-concentration lanthanides aid in identifying pollution sources. The third study aims to enhance the usability of the PMF model for source identification of river pollution by developing industrial composited profiles and their key elements. After collecting and integrating wastewater samples from potential pollution sources near different rivers, cluster analysis was used to calculate the similarity between source profiles of industries. Then the source profiles of industries were classified. The rationality of the results was further confirmed through the Pearson correlation to establish a composite profile of each industry category. Moreover, the source profiles of industries were further identified through major elements to improve the usability of composite profiles, which were ultimately applied to different rivers. According to the application results, it was indicated that although a specific industry category of pollution source cannot be directly pointed out through database comparison, the proposed approach can narrow down the range of possible polluting industry categories. Therefore, the database of composite profiles had benefits for river pollution control. In conclusion, improving pollution source identification through modifications to receptor models is helpful for more effective river pollution management. The establishment of methods for identifying pollution sources in river reaches reduces the occurrence of mixed sources, making it more effective in pinpointing pollution sources in heavily polluted areas. Additionally, monitoring low-concentration elements, alongside the commonly monitored heavy metals, is essential, as comprehensive elemental concentration data is crucial for effective model analysis. In terms of environmental policy implementation, establishing complete composited profiles of industry categories for use in river pollution control can enhance the practical application effectiveness of pollution source identification.