有限混合分佈理論運用於台灣農地重金屬污染特性之分群

Use of Finite Mixture Model to Classify Farmland Heavy Metal Pollution Characteristics in Taiwan

江莉琦(Li-Chi Chiang)

指導教授 ： 張尊國

國立臺灣大學/生物資源暨農學院/生物環境系統工程學研究所/碩士(2005年)

https://doi.org/10.6342/NTU.2005.01397

摘要

有限混合分佈理論模式(finite mixture model)在各種土壤重金屬調查結果之分類上，可達到許多目的，本研究探討五個方向之應用：
一、 以全省大樣區砷、銅金屬資料，比較Graphical Method、有母數統計
評估方法與有限混合分佈理論所得到的背景濃度值，並與環保機關
所訂定之標準進行比較，結果顯示後兩者方法較適合用來界定台灣
地區土壤重金屬背景濃度範圍。
二、 結合一般克利金法，將彰化和美地區受重金屬鉻、銅、鎳、鋅污染
的農地之調查結果，以有限混合分佈理論，依據最小誤判機率，將
推估監測區客觀劃分為兩區，其中的「高度重點監測區」代表具有
與管制區相同的污染特性，需特別加強監測。
三、 以有限混合分佈理論得到的區分值與法定管制標準作為門檻值，並結合指標克利金法，可劃分出與「污染區」具有相同特性的「可能污染潛勢區」，結果顯示Cr金屬之「可能污染潛勢區」面積為原污染區之2.24倍，Cu金屬為1.85倍，必須加強監控，較為完備。
四、 探討彰化地區受污染農地與溝渠底泥的關係，測量農地進水口與農地採樣點的距離，建立重金屬濃度與進水口距離的關係式，並以有限混合分佈理論，區分污染農地受到底泥和廢水影響的可能比例，可有效提高管理效率。
五、 以有限混合理論模式可分析彰化地區由不同實驗室所做出的樣本，有著明顯的系統性誤差存在，可提供作為數據檢核之工具。
關鍵詞：有限混合分佈理論、重金屬、土壤污染。

關鍵字

有限混合分佈理論 ； 重金屬 ； 土壤污染

並列摘要

By using the Finite Mixture Model to analyze the soil heavy metal survey data, many useful conclusions can be draw. In this study, five application cases were discussed:
1. When using the Taiwan’s large-scale As and Cu survey data, as compare the background concentrations defined by the Graphical Method, the Parametric Estimating Techniques and the Finite Mixture Model, it shows that the later two methods can properly define the range of soil heavy metal background concentration of Taiwan.
2. Combined with Ordinary Kriging and based on the investigation data of Chang-Hua polluted area, use the Finite Mixture Model to separate the estimated monitoring area. One of the parts is called the “High Monitoring Area”, which has the same characteristics of “Control Area” and needs to be monitored continuously.
3. By combined with Indicator Kringing, and using the cut-off point determined by the Finite Mixture Method and the Control Standard as the thresholds, we can separate the “Potential Pollution Area” which has the same characteristics of the “Control Area”. It shows that the “Potential Pollution Area” of Cr is 2.24 times as large as original control area, and Cu is 1.85 times.
4. To decide the pollution sources from sediments or wastewater drainages, by measuring the distances between the water inlet of the farmland and the soil concentration in the farmland, we can establish the relation formula of heavy metal concentrations and distances to the water inlet. These results can be used as reference to manage the farmland in the future.
5. By using the Finite Mixture Model, we can check the systemic error between the two data analyses by different laboratories in Chang-Hua area. It can be a useful tool for data quality control.
Keywords: Finite Mixture Model, heavy metal, soil pollution.

並列關鍵字

Finite Mixture Model ； heavy metal ； soil pollution

參考文獻

2. 行政院環境保護署，「污染農地灌溉渠道底泥及水質重金屬污染調查計劃成果報告」，2002。

8. 馬秀雯，「使用音訊與視訊資料融合方法於生物驗證」，成功大學資訊工程學系研究所碩士論文，2004。

17. Alloway, W. H., 1968, Advances in agronomy. Vol.20.p.240-243.

19. Bishop, C. M., 1995, Neural Networks for Pattern Recognition, Oxford: Oxford University Press.

21. Burgess, T.M. and R. Webster, 1980, “Optimal interpolation and isarithmic mapping of soil properties,” Journal of Soil Science, 31, pp.315-331.