| 摘要: | 氣膠為法規污染物中最常造成造成空氣污染指數(PSI)不良(unhealthy)的原因之一，氣膠不但會對人體健康造成危害，降低環境能見度，亦會對氣候變遷造成影響；國內外量測氣膠的大型研究計劃已積極進行中，希望進一步求得時間、空間解析度上更縝密的資料，對氣膠物理、化學更深入的了解，獲得資料亦可支援其它相關研究，或提供訂定空氣污染管制策略時參考。 本研究以蜂巢式套管採樣器採集粒徑小於2.5μm的細粒氣膠，採樣站選定環保署台北新莊站及高雄小港站，以比較典型工業都會區及非工業都會區氣膠污染特性，所得結果亦可直接與附近環保署測站的連續監測值進行比對；採得樣品主要以電腦控制掃描式電子顯微鏡(CCSEM)分析單顆粒氣膠的元素組成，並選取代表性氣膠獲取高解析度影像，元素組成數據以聚類分析、因子分析推估可能污染來源，顯微影像以碎形維度進行量化計算，並以鑑別分析、類神經網路評估氣膠元素組成資料庫進一步利用的成效；混合氣膠樣品則分析氣膠的質量濃度、水溶性離子、OC/EC及金屬離子，所得結果轉換成元素濃度百分比後與CCSEM分析結果進行比對，或直接以絕對主成份進行污染源貢獻量推估。 聚類分析結果可將台北站分成18類，其中氣膠個數較多的主要類別有6個；高雄站則可分成16類，主要類別有5類，其中柴油引擎廢氣類別氣膠數佔所有氣膠數台北站為39%，高雄站則佔53%；因子分析結果兩站都有6類主要污染來源，台北站包含：木材農廢燃燒及二次氣膠、工業污染、塵土及鍋爐燃燒，肥料、水泥及生物氣膠、鋼鐵工業等六類；高雄站6類中有五類與台北站相當，另一類則為海鹽及工業冶煉程序；顯微影像以碎形維度量化計算後，可依氣膠邊界破碎程度、投影面積密實度及表面粗糙度將不同氣膠表現於三個碎形維度所形成的空間座標中的不同位置；鑑別分析與類神經網路兩種方法可對其它未分類氣膠進行鑑定，或對已分類氣膠進行評估。 混合相分析結果兩站未分析約佔21%、碳含量37%、水溶性離子則佔39%，其餘金屬約為3%；以受體模式(絕對主成份分析)推估台北站及高雄站主要有4個污染來源，其中與人為及工業排放有關的NO3-、NH4+及SO42-，出現在兩站的主要二個類別，佔PM2.5質量濃度達78%，台北站則佔59%。 整體來說CCSEM分析結果與混合相分析結果各主要元素所佔的百分比具有類似的趨勢；因子分析及絕對主成份分析所推估的污染源類別多以混合來源為主，而聚類分析所得類別則具有更佳的解析度，可將混合污染源解析至更細的類別。但以單顆粒氣膠的元素組成資訊進行污染源推估對於某些元素組成相近但來源不同的氣膠仍無法分離，必須輔以其它資訊才能合理地將不同類別分離。 Particulate matter (PM) is one of the regulated pollutants that seriously deteriorates air quality. PM (also called aerosol) not only causes visibility impairment, global climate change but also increases the risks of adverse health effects. Consequently, a comprehensive investigation on ambient aerosol to understand their physical and chemical properties is urgently needed. In order to compare the characteristics of fine particles between typical industrial urban and non-industrial urban, we collected PM2.5 by using Honeycomb denuders from Sin-chung and Shao-gawn sites situated in Taipei metropolitan area and Kao-hsiung city, respectively. Particles were collected on PC (polycarbonate) filters and analyzed with CCSEM (computer controlled scanning electron microscopy) to characterize elemental composition and morphology of individual particles. Source apportionment of particles were conducted on elemental composition using statistical cluster and factor analyses. The discriminate analysis and neural network were applied to classify unsorted particle into target category based on known data set or to assess the classified data. Morphology of selected particles was quantified using fractal algorithm on their boundary, projected area, and surface roughness, respectively. Meanwhile, aerosol bulk properties like mass, water-soluble ions, carbonaceous contents, and metal concentrations were obtained from collocated filter samples. The APCA (absolute principal component analysis) was applied for source apportionment of particles to compare that from CCSEM data. Cluster analysis classifies Taipei aerosols into 18 categories with 6 are major categories. In contrast, Kao-hsiung aerosols are classified into 16 categories and with 5 major ones. In both site C, O rich particles are predominant, 39% and 53% particles can be attributed to diesel vehicle exhausts from Taipei and Kao-hsiung, respectively. The results were further confirmed by the discriminate analysis and neural network algorithm. Simultaneously, factor analysis shows 6 source types both in Taipei (wood burning and secondary, industry, soil and boiler, fertilizer, cement and bioaerosol, and ferrous furnace) and in Kao-hsiung (sea-salt and industry, industry, soil and boiler, fertilizer, cement and bioaerosol, ferrous furnace). A Cartesian coordinate system for fractal dimensions on boundary, projected area, and surface roughness from single particles is established to identify particles from different sources. Aerosol bulk analysis reveals the averaged fraction of carbon in PM2.5 is 37%, that of water-soluble ion is 39%, that of metal is 3%, and the remaining 21% is unknown. The receptor model(APCA) estimates 4 different source types contributing to both sites, among them industrial sources containing precursors of NO3-, NH4+and SO42- is the major one. This source type accounts for 59% and 78% of PM2.5 in Taipei and Kao-hsiung area, respectively. Finally, single particle and bulk analysis are agreed in reconstructed elemental compositions in this study. The source types apportioned from factor analysis and APCA are mixed compared to more resolved ones from cluster analysis. However, supplemental information is needed to resolve source contributions for particles with similar elements. |
