利用受體模式解析南崁溪之污染來源與臺灣地區事業廢水指紋圖譜辨識

Abstract

河川水質與國民健康息息相關，然而，由於廢水污染之流動特性及環保機關人力、物力之限制，追查污染來源實屬不易，有鑑於此，本研究旨在以質量守恆為基礎概念並透過受體模式發展水污染源鑑識方法。 研究中以桃園市南崁溪為目標河川，規劃河川水體為受體點，流域周邊事業廢水為污染源進行兩梯次採樣，在第二梯次採樣中，將安排一河川污染熱區—大檜溪橋進行30天之連續監測。本研究除了檢測該地區水質表現外，亦由52項元素之分析結果進行化學質量平衡模式 (EV-CMB) 與正矩陣因子模式 (PMF) 之解析，而針對PMF解析結果，需進一步透過相似性指標與指紋圖譜資料庫進行比對，辨識可能之污染事業別。 在兩梯次河川水樣分析結果中，全流域之「河川污染指標 (River Pollution Index, RPI)」皆屬於中度至嚴重污染，而屬嚴重污染之採樣點中，上游樂善寺主要受林口工業區之事業廢水影響，下游則由於生活污水及事業廢水雙重影響導致污染，另外，由於，該流域過去曾有Cu濃度超標之事件，桃園市政府亦於該流域進行Cu之總量管制，因此，特別針對Cu元素之污染情形與來源進行探討。在兩梯次河川水樣分析結果中，下游採樣點之Cu濃度皆高於環保署訂定之保護人體健康相關環境基準值0.03 mg/L，此外，在污染熱區之30天連續監測中，Cu之金屬指標值 (Metal Index, MI) 亦多數大於1，顯示可能造成健康之危害。 在EV-CMB解析結果中，第一梯次於河川中、下游點位皆解析出晶圓製造及半導體製造業較高之貢獻量，第二梯次則於大檜溪橋解析出未納管污水之明顯貢獻，兩梯次解析結果存在時間變異之影響。而針對下游地區之Cu污染來源，主要以印刷電路板製造業有較高之濃度占比，在大檜溪橋則以未納管污水之影響最大。在連續監測資料部分，於較多元素出現異常值之08/30、09/16及09/17三日中亦主要解析出未納管污水。 在相似性指標測試結果中，指標COD (coefficient of divergence) 與各事業別之比對正確率相較 SID (standardised identity distance) 與PD (Pearson distance) 為佳，因此，後續由COD指標輔以SID和PD進行比對。 針對污染熱區，以60筆連續監測資料進行正矩陣因子模式 (Positive Matrix Factorization, PMF) 分析，推估可能污染源為6個，此一結果與資料庫中各事業指紋資料進行比對，並且與高污染事件日中EV-CMB之解析結果交互探討後，推測6個可能之污染源分別為食品製造業 (Factor 1、5)、光電材料及元件製造業 (Factor 2)、污水處理業 (Factor 3)、未納管污水 (Factor 4) 以及電鍍業 (Factor6)，其中，Factor 4為高污染事件日之主要貢獻者，另外，在Cu之各來源貢獻比例部分，PMF與EV-CMB有相似之解析結果。 綜合上述探討，南崁溪流域之污染型態以生活污水為主，此結果與過去研究相符，而兩受體模式之解析中，較大貢獻者皆指向相同之污染源。在受體模式運用上，EV-CMB之優點在於對環境資料筆數要求不高，但若無法完整掌握該地區污染來源將影響解析之品質，亦較無法有效考量時間變異之影響。此時，透過PMF模式將有助於鑑識時序上之突發污染事件， 然而PMF需一定資料筆數才能進行解析，因此，若能於連續監測技術上有所突破，將有利於突發污染源之掌握，此一優勢也需透過事業廢水指紋資料庫之不斷擴充，方能使污染源辨識更加準確。

River water quality is closely related to population health. However, due to the fluidity of pollutants and limited resources of environmental protection agencies, it is not easy to trace the sources of pollution. Therefore, this research attempts to use receptor models which are based on the concept of conservation of mass to develop the method of source identification for water pollution. This study took Nan-kan river as an example. The sampling sites at water body were seen as receptors and the industrial wastewater along the river were sources of pollution. For the second batch of sampling, the pollution hot zone, Daguixi bridge, was continuously monitored for 30 days. Apart from testing the water quality in this area, the analysis results of 52 elements were also used in Chemical Mass Balance (EV-CMB) and Positive Matrix Factorization (PMF) modeling. The similarity comparison for PMF modeling results were further carried out to identify possible pollution sources by industrial categories. Based on the analysis results of river samples, the 'River Pollution Index (RPI)' showed that the entire river was moderately-polluted to severely-polluted. The upstream Leshan Temple was mainly affected by the industrial wastewater from the Linkou industrial park, and the downstream sites were influenced by domestic sewage and business wastewater. Besides, due to there have been cases where the concentration of copper exceeded the standard in the past, and the “Total Copper Discharge Control” by Taoyuan City Government, the sources of copper were investigated further. For samples collected from the two batches, the copper concentration at downstream sampling sites was relatively high, and was above the environmental protection standards set by the Environmental Protection Agency for protection of human health (0.03 mg/L). In addition, during the 30-day monitoring period at the hot zone, the metal index (MI) for Cu was mostly greater than 1, which was seemed to be hazardous for health. The EV-CMB modeling revealed that the wafer manufacturing and semiconductor manufacturing had a greater contribution at midstream and downstream sites. Besides, the results showed that unmanaged sewage at the Daguixi Bridge had an apparent contribution to the second batch sampling. As for the source of copper pollution in the downstream areas, the printed circuit board manufacturing industry was the main contributor. At the Daguixi bridge, the impact of unmanaged sewage was the biggest. In the part of continuous monitoring data, the main contribution of unmanaged sewage was identified on the three severely-polluted days, 08/30, 09/16, and 09/17. According to the test results of the similarity indicators, the accuracy of the coefficient of divergence (COD) in each industrial category was better than that of standardised identity distance (SID) and Pearson distance (PD). Therefore, the COD indicator was supplemented by the SID and PD indicators for comparison. For pollution hot zone, PMF modeling results based on the continuous monitoring data showed that there were six possible sources of pollution. These 6 factors were compared with the fingerprints of each industry in the database. And after interactive discussion with the analysis results of the EV-CMB on the severely-polluted day, 6 possible pollution sources were estimated to be food manufacturing (Factor 1 and Factor 5), optoelectronic materials and components manufacturing (Factor 2), wastewater treatment (Factor 3), unmanaged sewage (Factor 4) and electroplating industry (Factor 6). Among them, Factor 4 was the main contributor to the severely- polluted event. Also, the analysis results of PMF and EV-CMB on the proportion of Cu contribution were similar. Based on the above discussion, the pollution pattern in the Nan-kan River is dominated by domestic sewage and this result is consistent with previous studies. In the analysis of the two receptor models, the larger contributors all point to the same pollution source. In the application of the receptor model, the advantage of EV-CMB is that it does not require a high number of environmental data, but if it is not possible to fully acquire the source of pollution in the area, it will affect the quality of the analysis. At this time, the PMF model will help to identify sudden pollution events. However, PMF requires a certain number of data to be analyzed. Therefore, if there is a breakthrough in continuous river monitoring, it will be beneficial to understand the sudden pollution. To take this advantage, the continuous expansion of the fingerprint database can make the identification more accurate.