|摘要: ||大氣背景值之定義為局地自然源、外來自然源與背景測站偶受之少量外來人為污染物貢獻量之總和，即人為活動無法改變之污染物濃度，透過大氣背景值之推估方能定量來自當地污染或長程傳輸之貢獻。自然存在之氣體背景值常使用背景站之直接觀測分析，然而臺灣位處東亞中心且人口稠密，無法找到嚴格定義之地面背景測站，且PM2.5易受天氣與其他環境因素影響，導致臺灣PM2.5之大氣背景值不易估量。本篇研究嘗試定量臺灣大氣背景PM¬2.5濃度，以應用於區分其他來源之貢獻。本研究使用2005年冬季至2016年秋季臺灣之萬里、臺東與恆春等背景測站資料，應用調整之AGAGE (Advanced Global Atmospheric Gases Experiment) 方法標記屬大氣背景情境之資料以求得大氣背景值。除AGAGE方法外，另使用隱馬可夫模式（Hidden Markov Model, HMM）分析PM2.5觀測資料作為比對參考，以求得更客觀之定量結果，並與過往文獻相互驗證。|
研究結果顯示使用調整之AGAGE方法求得臺灣地面之大氣背景PM2.5質量濃度約為4.4 μg m-3，與HMM及過往文獻之結果相近。使用標記為背景情境之資料與氣象參數進行相關性分析可發現相關係數低（r < 0.1），且具高可信度，顯示此大氣背景值非因單一氣象參數影響所導致之低值。分析影響大氣背景值之潛在因素可發現其主因為季節變化，可造成背景值約50 %之變化量，降水與日夜變化造成之差異則約10 %。透過趨勢分析之結果可發現近11年背景PM2.5之下降率約為0.1 – 0.2 μg m-3 yr-1，與文獻記載東亞地區PM2.5前驅物同樣呈現下降趨勢，顯示背景值變化趨勢可能與大環境之趨勢變化有關，或是背景值仍然不可避免受到外來源影響。
基於前述之大氣背景值，本篇研究更進一步探討臺灣本地PM2.5之污染特性，假設臺灣三個背景測站（萬里、臺東與恆春）之PM2.5小時濃度均小於平均大氣背景值（4.4 μg m-3）時，可視作當時不受境外污染物影響之情境，此情境下之PM2.5質量濃度自北而南逐漸增加、平均值約5 – 35 μg m-3。相比於平均大氣背景值，最大變化率發生於高屏地區，該區人為活動造成之PM2.5排放、地形與天氣系統之潛在效應可使其增為背景值之5倍。透過東北、西南風情境分析之分析結果，可發現不同季風盛行條件下，跨縣市傳送量約5 μg m-3。由情境假設結果顯示，假設PM2.5排放量不變、且不受境外污染物影響下，非因當地污染導致之PM2.5貢獻量約為10 μg m-3（即背景值與跨縣市傳送量之總和），其中，大氣背景值即為潛在之PM2.5減排極限，未來考量相關政策修訂或定義高污染事件時，或可將此貢獻量納入考量，以制定合理並且可達成之環保規範。
;The definition of background air quality (AQ) concentration is the sum of local natural emissions, foreign natural and trace anthropogenic pollutants advected into an area. In other words, those AQ level cannot be affected by local anthropogenic activities. Through the estimation of background level of AQ concentration, one can quantify the contributions originate from other sources (e.g. local emission or long-range transport). In the literatures, background concentration of gas pollutants is often obtained by direct observations at natural background sites. However, Taiwan is located in the center of East Asia and populated region. It is scarce to find a standard background site that meets the strict limitations. To quantify the background PM2.5 concentration in Taiwan seems to be even difficult due to PM2.5 concentration strongly varied with weather, long range transport, chemical reactions, and so on.
In this study, we attempt to quantify PM2.5 background concentration level in Taiwan by using 11-year PM2.5 monitoring data obtained from EPA air quality stations in Wanli, Taitung and Hengchun. These sites are chosen as background sites since there are no large PM2.5 emission source adjacently. Modified AGAGE (Advanced Global Atmospheric Gases Experiment) method is used to flag the background condition for calculating PM2.5 background concentration level. In addition to the modified AGAGE method, the HMM (Hidden Markov Model) statistical method and literature reviews are also performed in this study to confirm the reliability of background PM¬2.5 concentrations.
The results show that the PM2.5 background concentration in Taiwan is about 4.4 μg m-3. Low correlation coefficients (r < 0.1) with at least 95 % significant level between background-flagged data and meteorological parameters indicates that the low concentration of background value is weakly related to the influences resulting from meteorological condition. The main factor of affecting background concentration is seasonal variation. It could result in a change in background value of around 50 %. Precipitation and diurnal pattern could contribute for 10 % variation of background PM2.5 value. According to the linear trend analysis, it shows that the decreasing trend of background concentration is -0.1 to -0.2 μg m-3 yr-1 in the past 11 years in corresponding to the decreasing trend of air pollution emissions in East Asia in recent years. This study suggested that the background trend may be related to the trend of surrounding environment or the influences caused by Asian continental sources.
Based on the result of background PM2.5 concentration level, we attempt to further investigate the local PM2.5 characteristics in Taiwan. Assuming that when all of hourly PM2.5 concentration measured at background sites (i.e., Wanli, Taitung and Hengchun) less than averaged background PM2.5 concentration (i.e., 4.4 μg m-3), the situation could be considered as no influence of long-range transport. Under this scenario, the PM2.5 concentrations gradually increase from North to South, and averaged PM2.5 concentrations are 5 – 35 μg m-3 due to local emission and topographic effect. The highest increment appears happened at Southwestern region, and it could be 5 times larger than background concentration level. Following the aforementioned dataset, we further analyze the PM2.5 spatial distribution under Northeast wind and Southwest wind, respectively. We found that the transport amount of PM2.5 across counties could reach 5 μg m-3 and it shows the accumulation effect at specific sites. Results of scenario assumptions show that there are about 10 μg m-3 of PM2.5 due to non-local pollutions (i.e., sum of background value and transported PM2.5 across counties). The 10 μg m-3 represents the lower bound of PM2.5 reduction limit as current emission condition. The results from this study provide a useful information to policy makers on making an achievable and reasonable PM2.5 deduction goal.