以土地利用迴歸模型結合貝氏最大熵法回推台北地區細粒懸浮粒子時空分佈之研究

Abstract

目前有很多的研究指出，漂浮在空氣中的細懸浮微粒（PM2.5）對人體所造成的健康危害要比粗懸浮微粒（PM10）還要來的嚴重。然而，直到2005年的八月，台北地區才完成PM2.5觀測網。由於土地利用迴歸（Land Use Regression，LUR）包含地理資訊系統（Geographic Information System，GIS）資料已經在模式中被成功地應用在估計戶外空氣污染的空間變異，並且研究結果顯示其表現類似於其它地理統計方法如克利金（kriging）或更好。故本研究利用土地利用迴歸模型結合貝氏最大熵法（Bayesian Maximum Entropy method，BME）去驗證2005-2009年新莊超級測站、1998-2004年古亭與三重測站以及2003-2004年新莊、中山、萬華測站的PM2.5。其推估的平均誤差介於2.72ug/m3至3.63ug/m3之間。從時空推估圖來看，發生PM2.5高濃度的地方，主要在士林區偏南、大同區、三重區、信義區、永和區。

Many researches have pointed out that the fine particulate matter (PM2.5) influences people health more seriously than coarse particulate matter (PM10). The network of air quality monitoring stations of PM2.5 in Taipei, however, has not been set up completely until 2005 August. Model combined LUR (Land Use Regression) and GIS (Geographic Information System) has already been used successfully in estimating the spatial variation of outdoor air pollution, and the performance is also comparable with, even better than, other geostatistical methods (e.g. kriging). This research combines LUR and BME (Bayesian Maximum Entropy) method for spatiotemporal PM2.5 retrospective estimation. The validation is performed at Hsin-Chuang supersite during 2005-2009, Ku-Ting and San-Chung station during 1998-2004, and Hsin-Chung, Chung-Shan, Wan-Hua station during 2003-2004. The estimated mean errors lie between 2.72 μg/m3 and 3.63μg/m3. Judging from the spatiotemporal estimated maps, it can be found that higher concentration of PM2.5 happens in southern Shih-Lin district, Da-Tung district, San-Chung district, Hsin-Yi district, and Yung-Ho district