本三年期研究主要探討水文地質對地下水砷尺度變異之影響，尺度研究主要藉助地質統計方法中的多變量因子克力金方法(multivariate factorial kriging)，將導致地下水砷含量變異，依據其空間尺度分離，如此有助於探討地下水砷含量於不同尺度之釋出機制，計畫主持人曾成功利用該方法於嘉南平原砷污染研究，本三年期研究區域分別為濁水溪沖積扇雲林地區、蘭陽平原及孟加拉地區。台灣地區三大砷污染地區，嘉南平原地區、濁水溪沖積扇雲林地區及蘭陽平原地區，水文地質環境的不同，導致地下水砷污染的發生區域及深度明顯不同，於嘉南平原地區高濃度砷發生於深層(100-280 公尺)，雲林地區高濃度砷發生於淺層(0-50 公尺)，而宜蘭高濃度砷發生深度介於兩者之間(60-120 公尺)，因此尺度問題為瞭解砷受水文地質環境之重要關鍵，然而，雲林地區及蘭陽平原地區，高濃度砷因發生於淺層，探討水文地質之影響尚須需排除人為污染，因此困難度將高於嘉南平原之研究。針對方法修正部份，由於多變量因子克力金方法的核心運算為不同尺度的主成分分析(principal component analysis)，假設變數與主成分間為線性關係，導致應用受限於短梯度環境，因此將改良採用權重平均法(weighted average)，假設主成分與變數間為單峰關係曲線(unimodal curve)，而從所求得主成分軸之觀點，應可辨識各變數間之最大差異 (maximum dispersion)，修正方法結合生態權落分析中的對應分析(correspondence analysis) 及空間訊息，發展非線性多變量因子克力金方法，同時針對線性及非線性模式將自行撰寫程式並編寫使用手冊。本研究可將高濃度砷形成原因，區分為長時期的背景值效應或是短期的干擾效應，可以針對飲用高含量砷地下水致癌案例，於公共衛生致癌風險評估中，提供有關暴露強度與暴露時間更準確的推估，同時雲林地區研究，關係雲林地區民生飲用水安全外，更關係台灣地區民眾食用水產品之食品安全，蘭陽平原地區研究，因該區有龐大的公共衛生調查資料，因此未來該區域可能繼嘉南平原地區成為世界飲用水標準制訂之重要案例，此外蘭陽平原地區與孟加拉地區均屬淺層還原環境，利用相同方法分析不同地區有助於驗證方法之效能，同時可促進國際學術合作。 The main object of the proposed three-years-term program is to reveal the hydrogeological influences on the multi-scale variations of groundwater arsenic. The scaleinvestigation is achieved by the multiscale factorial kriging (MFK) in geostatistics. To filter out the spatial variation of groundwater arsenic according to the influence ranges of release mechanisms can support the interpretation of the causations of the arsenic release in groundwater at different scales. The method has been successfully applied in the Chianan plain by the program proposer. The yearly study areas of the program are the Choushui river alluvial fan in Yunlin County, Lanyang plain and Bangladesh. The three main areas of high arsenic containment in groundwater in Taiwan, the Chianan plain, the Choushui river alluvial fan in Yunlin County, and the Lanyang plain, have different hydrogeological circumstances that introduce the occurrence of high arsenic in groundwater at diverse locations and depths. The high arsenic is located in the deep aquifer (100-280 m) in the Chianan plain, while the high arsenic is marked in the shallow aquifer (0-50 m) in the Choushui river alluvial fan in Yunlin County. The high arsenic in groundwater in the Lanyang plain is located at the depth between where high arsenic occurs in the Chianan plain and the Choushui river alluvial fan in Yunlin County. Accordingly, the scale problem is essential to the hydrogeological influences on the multi-scale variations of groundwater arsenic. However, the investigation of the hydrogeological influences on groundwater arsenic in shallow aquifers should take the artificial disturbances into account that makes the analysis more complex than in deep aquifers. The core methodology of the MFK is the principal component analysis (PCA) at different scales. Due to the rectilinear assumption between the variables and the principal components in the PCA, the application of the MFK is limited to the short gradient environment. The proposed program is going to modify the assumption to the unimodal relationship between the variables and the principal components by the weighted average method to determine the maximum dispersion of the variables in the viewpoint of obtained principal components. The nonlinear modification of the MFK includes the correspondence analysis, as known in the ecological community analysis, and the spatial information of variables. The compilation of the manual for the linear and nonlinear MFK models is also included in the work projects. The MFK which can deconvolute As variations into background terms and transient terms can provide more information for predicting the exposure time of subjects to a specific As concentration in groundwater. Moreover, the study in Yunlin County can promote the edible safety of drinking water in the county and aquaculture products in Taiwan. Furthermore, the study in the Lanyang plain may affect the future revision of drinking water standards in the world due to the large amount of subjects in the sanitarian survey for the epidemiology in this area. And the similarity of the high-arsenic location, the shallow and reductive aquifer, in the Lanyang plain and Bangladesh provides the opportunity for the identification of model performance. Finally, the study in Bangladesh encourages the international science communication.
