| 摘要: | 聯合國環境規劃署(United Nations Environment Programme, UNEP)於2019年發表《Global Mercury Assessment 2018》，估算2015年全球人為大氣汞排放量為2220噸，而東亞和東南亞為主要排放源區，貢獻全球排放量的38.6%(約859噸)。為了模擬大氣汞的濃度變化與探討影響因子，除了使用化學傳輸模型，近年來有許多研究運用廣義相加模型(Generalized Additive Model, GAM)模擬空氣污染物濃度的變化，並探討影響濃度變化的機制。
本研究使用 GAM 量化氣象因素和空氣污染物對我國低海拔測站(國立中央大學, NCU)和高海拔測站(鹿林大氣背景站, LABS)氣態元素汞(Gaseous Elemental Mercury, GEM)濃度的影響，並進一步探討各測站的主要影響機制。研究結果顯示，2019至2020年間在中央大學量測到的GEM平均濃度為2.18±3.95 ng m^(-3)，而鹿林山大氣背景測站為1.40±0.37 ng m^(-3)。春冬季較高，夏季異常升高，顯示當地排放影響。GEM日夜變化呈現日低夜高，與交通、太陽輻射和大氣擴散條件變化相關。
GAM模型結果顯示，在高海拔測站(LABS) CO是影響GEM濃度的主要因子，兩者來自相似的排放源，當CO濃度上升時，GEM濃度隨之增加，但隨著CO濃度進一步提升，GEM增長趨緩。春季GEM濃度最高，主要與中南半島生質燃燒污染物的長程傳輸有關。相對濕度與GEM濃度呈正相關，白天的谷風將污染物輸送至高山，夜間的高層氣團沈降則導致GEM濃度下降，說明此測站主要受區域性排放影響。
在低海拔測站(中央大學)，CO也是GEM變化的主要因子，與交通運輸和工業活動的排放有關。CO濃度與GEM濃度呈強正相關，且在高濃度區域出現非線性特徵。GEM濃度的日夜變化與低風速和光化學反應有關。風向對GEM濃度有顯著影響，當風向在180至225度之間時，GEM濃度上升，顯示來自南方或西南方的氣流帶來較高的汞濃度，說明此測站主要受局地性排放影響。
使用2020年觀測數據進行模型驗證，結果顯示LABS站點預測與觀測值之間有較高的相關性(R2 = 0.73)，但有高估的現象，此與2020年初中國因COVID-19實施封城措施有關，封城期間，工業與交通活動減少，污染物排放顯著下降，導致GEM濃度下降。中央大學站的R2為0.49，顯示模型能捕捉GEM濃度的整體變化趨勢，但在高濃度情況下有較大誤差，特別是在污染事件期間，模型有明顯低估的情況。模型的春夏秋三季MBE均為負值，顯示低估濃度，尤其是在夏季。這一現象與西南風攜帶污染物至觀測地區有關。;The United Nations Environment Programme (UNEP) published the Global Mercury Assessment 2018 in 2019, estimating that global anthropogenic atmospheric mercury emissions in 2015 amounted to 2,220 tons. East and Southeast Asia were identified as the major emission source regions, contributing 38.6% of global emissions (approximately 859 tons). To simulate atmospheric mercury concentration changes and investigate influencing factors, besides chemical transport models, recent studies have increasingly employed Generalized Additive Models (GAM) to model variations in air pollutant concentrations and explore the mechanisms affecting these changes.
This study applies GAM to quantify the impacts of meteorological factors and air pollutants on gaseous elemental mercury (GEM) concentrations at two monitoring stations in Taiwan: the low-altitude station at National Central University (NCU) and the high-altitude station at Lulin Atmospheric Background Station (LABS). The study further explores the dominant mechanisms influencing GEM concentrations at these sites. The results show that from 2019 to 2020, the average GEM concentration at NCU was 2.18 ± 3.95 ng m^(-3), while at LABS it was 1.40 ± 0.39 ng m^(-3). Concentrations were higher in the spring and winter, with an anomalous increase observed in the summer, likely due to local emissions. The diurnal variation of GEM concentrations exhibited lower values during the day and higher values at night, which were related to traffic, solar radiation, and atmospheric diffusion conditions.
The GAM results indicate that at the high-altitude station (LABS), carbon monoxide (CO) was the primary factor influencing GEM concentrations, with both pollutants originating from similar emission sources. As CO concentrations increased, GEM concentrations also rose, although GEM growth slowed with further increases in CO. The highest GEM concentrations occurred in the spring, primarily due to long range transport of pollutants from biomass burning in the Southeast Asian Peninsula. Relative humidity showed a positive correlation with GEM concentrations, and the diurnal wind patterns, such as valley winds transporting pollutants to the mountain during the day and subsiding air masses at night, contributed to the observed GEM variations, indicating a significant regional emission impact at this site.
At the low-altitude station (NCU), CO was also the major factor influencing GEM changes, related to emissions from traffic and industrial activities. A strong positive correlation was observed between CO and GEM concentrations, with a nonlinear relationship at higher concentration levels. Low wind speeds and photochemical reactions influenced the diurnal variation in GEM concentrations. The wind direction had a significant impact on GEM concentrations, with higher values observed when the wind direction was between 180° and 225°, suggesting that air masses from the south or southwest brought higher mercury concentrations, indicating that this station was primarily affected by local emissions.
Model validation using 2020 observational data showed a high correlation between predicted and observed GEM concentrations at LABS (R2 = 0.73), but overestimation was observed. This discrepancy was attributed to the COVID-19 lockdown measures in early 2020, which led to a significant reduction in industrial and traffic activities, consequently decreasing pollutant emissions and GEM concentrations. At NCU, the R2 value was 0.49, indicating that the model captured the overall trend of GEM concentration changes, but larger errors were observed during high-concentration periods, particularly during pollution events, where the model underestimated the concentrations. The model′s mean bias error (MBE) for spring, summer, and autumn was negative, indicating an underestimation of concentrations, especially during the summer. This phenomenon was related to the transport of pollutants by southwest winds to the observation site. |
