|摘要: ||民眾長時間處於室內環境，室內生物性空氣污染物可能危害人體健康，因此，除了藉由良好的室內空氣品質管理，來建立健康安全的室內空間外，發展便捷評估空氣中微生物汙染程度之技術，才可適時提供正確資訊及早回應。目前廣泛使用來評估室內空氣微生物污染程度之衝擊式生物氣膠採樣器與傳統微生物培養法不但無法長時間連續監測，且耗費時間及成本。而目前市面上亦無生物氣膠量測之直讀式儀器可使用，因此，本研究採集多種類型室內場所生物氣膠及使用直讀式儀器可即時量測之空氣參數，來建立預測室內生物氣膠濃度之多元線性及非線性迴歸模型，以期能運用於室內空氣品質維護管理。本研究選擇台灣南部不同類型開啟空調系統之室內場所，分別使用Grimm粉塵粒徑分析儀來量測室內外微粒質量濃度及粒數濃度、MAS-100採樣器採集室內外空氣細菌及真菌、IAQ-Calc室內空氣品質監測儀量測室內外二氧化碳、溫度及相對濕度，空氣採樣點數共計室內83點及室外43點。再將量測之室內生物氣膠濃度與室內及室外各空氣參數值(分徑微粒質量濃度、分徑微粒粒數濃度、二氧化碳、溫度及相對濕度)以SPSS統計軟體進行相關性分析及建立四種情況之多元線性與非線性迴歸預測模型，最後再以平均絕對值誤差率( Mean Absolute Percentage Error , MAPE )來評估迴歸模型之預測準確度。經由Pearson相關性分析顯示室內空氣細菌濃度與室內的二氧化碳、溫度、分徑質量濃度、室內及室外之分徑微粒粒數濃度具有顯著相關性;室內空氣真菌濃度則與室內外相對溼度、室內分徑微粒質量濃度及室內外之分徑微粒粒數濃度有顯著相關。單一自變數之冪函數非線性迴歸分析，顯示室內空氣細菌濃度與室內二氧化碳、分徑微粒質量濃度及分徑微粒粒數濃度具顯著相關性；室內空氣真菌濃度則與室內外相對溼度、室內分徑微粒質量濃度及分徑微粒粒數濃度具顯著相關性。多元線性迴歸得到之預測模型皆是以室內外之分徑微粒粒數濃度來預測室內細菌與真菌生物氣膠濃度為最佳(R2 =0.860、0.677)，顯示在監測之空氣參數(自變數)中，微粒粒數濃度較微粒質量濃度能得到更好的預測生物氣膠濃度。多元非線性預測模型之判斷係數R2太低而不適合用來預測室內細、真菌生物氣膠濃度。經MAPE評估多元線性模型之預測能力，僅以室內外分徑微粒粒數濃度為自變數直接迴歸得到之模型，可合理預測室內空氣細菌濃度；而室內空氣真菌濃度皆無法以建立之迴歸模型做合理預測，顯示需尋找更好的空氣或環境預測參數來加入迴歸，或者選擇更適合的迴歸數學模式來分析。本研究建立之多元線性迴歸預測室內細菌生物氣膠模型，可輔助公共場所簡單量測室內外微粒粒數濃度來快速評估室內空氣細菌污染潛勢，有助於提升室內空氣品質維護與管理能力。|
People spent most of their time indoors. Indoor airborne biological contaminants could have adverse health effects on people. In addition to performing good indoor air quality management to establish safe and healthy indoor environments, it is necessary to develop convenient techniques to assess airborne biological contaminants and obtain correct information timely in response to contamination early. The widely used methods to assess indoor airborne microbial contamination are the bioaerosol impaction sampler and traditional microbiological culture technique. They cannot continuously monitor bioaerosols and were time-consuming and expensive. This study used the simultaneous detected data, including indoor bioaerosols collected by an air sampler and air parameters monitored in real time by portable direct reading instruments in public places to establish multiple linear and nonlinear regression models to predict indoor bioaerosol concentrations. This study explored the feasibility of applying the established regression models for managing indoor air quality under lacking of direct measuring instrument for bioaerosols currently on the market.This study selected various categories of public places located in the south of Taiwan during the periods of operation of air-conditioning to measure indoor and outdoor bioaerosol concentrations and air parameters, including particulate mass and number concentrations by a Grimm dust analyzer, bacteria and fungi by an MAS-100 sampler, carbon dioxide, temperature, and relative humidity by an IAQ-Calc indoor air monitor. The 83 sampling sites indoors and 43 sampling sites outdoors were measured. The measured indoor bioaerosol concentrations and indoor and outdoor air parameters, including particulate size-segregated mass and number concentrations, carbon dioxide, temperature, and relative humidity were used to perform correlation analysis and established multiple linear and nonlinear regression models of four cases for prediction of indoor bioaerosols using SPSS statistical software. Finally, the prediction accuracies of the regression models were evaluated by calculating the mean absolute percentage error (MAPE).Pearson correlation analysis showed indoor air bacteria levels significantly correlated with indoor carbon dioxide, temperature, particulate size-segregated mass concentrations, and indoor and outdoor particulate size-segregated number concentrations. Indoor air fungi levels significantly correlated with indoor and outdoor relative humidity, indoor particulate size-segregated mass concentrations, and indoor and outdoor particulate size-segregated number concentrations. The power function nonlinear regression of a single independent variable showed indoor air bacteria levels significantly correlated with indoor carbon dioxide, particulate size-segregated mass concentrations, and particulate size-segregated number concentrations. Indoor air fungi levels significantly correlated with indoor and outdoor relative humidity, indoor particulate size-segregated mass concentrations, and indoor particulate size-segregated number concentrations.The models for prediction of indoor air bacteria (R2 =0.860) and fungi (R2 =0.677) levels using indoor and outdoor particulate size-segregated number concentrations were the best among the established multiple linear regression models. Among the monitored air parameters (independent variables), the indoor bioaerosol levels predicted using particulate number concentrations were better than those using particulate mass concentrations. The coefficients of determinant R2 of the established multiple nonlinear regression models were too low to suitable for prediction of indoor bacterial and fungal bioaerosol concentrations. Only the model obtained using the indoor and outdoor particulate size-segregated number concentrations as independent variables can reasonably predict indoor air bacteria levels after evaluation by MAPE. However, all the established multiple linear and nonlinear regression models cannot reasonably predict indoor air fungi levels. This result suggested more air or environmental parameters related to fungi or more appropriate mathematical models should be selected to perform regression.The established multiple nonlinear regression model for prediction of indoor bacterial bioaerosols can help public places to assess quickly the potential airborne bacterial contamination by simply measuring indoor and outdoor particulate number concentrations. The regression model contributed to promote management ability of indoor air quality.