蕭朱杏臺灣大學:流行病學研究所王偉銘Wang, Wei-MingWei-MingWang2007-11-272018-06-292007-11-272018-06-292004http://ntur.lib.ntu.edu.tw//handle/246246/56181在現行的土壤整治辦法之決策過程中,只要有採樣到超過污染濃度管制值之樣本,便進行整塊土地的管制及整治。這種作法有兩個問題,第一是僅靠少數幾個採樣值便決定了整塊待整治決策的土地污染濃度,沒有考慮到估計的誤差。第二,因為樣本數小,即使誤差被考慮了,其值也可能很大。對於第二個困難,一般來說,因為受限於經費與執行困難,增加土壤採樣的個數較不可行;因此,可以利用機率的概念將統計誤差納入考量,來處理這種依賴稀少樣本推估整治單位之平均污染濃度的情況,。本研究依據地理資料的特性,考慮隨機效應模式;並引入內生之先驗分佈,來描述不同的整治單位之平均污染濃度之間的相關性,藉以評估土地的污染濃度,進而利用事後分配找出最適的整治決策。本研究並與現行之整治決策做比較,此整治決策較現行辦法穩健,不易受過高或過低的污染濃度影響,但亦可能會有小區域隱含高度污染之風險。In the procedure of the soil remediation practice in effect, the land should be under control and remediation as long as there is one sample over polluted concentration threshold. There is two problem of this strategy. First, to decide the polluted concentration of the whole unit for remediation by a few samples will ignore the error of estimate. Because of the small amount of samples, even though the error is taken into consideration, there may be big error. Furthermore, generally, due to the limitation of budget and the difficulty of practice, it is not feasible to increase the sample size. Using the concept of probability to take statistic error in to consideration the problem of the soil remediation practice in effect can be solved. The research, in light of the characteristics of geographic information, adapt random effect model and use intrinsic prior to formulate the correlation of the average polluted concentration in different remediation units. The research can estimate effectively the average soil polluted concentration and use posterior to find the best remediation strategy. To compare the remediation strategy of this research with the recent one, the former is much more robust, because it is not affected by the polluted concentration under or over. However, there is a risk that the small area in a unit may contain high polluted concentration.第一章 前言 1 1.1 研究背景 1 1.2 土壤污染調查實例 4 第二章 貝氏模式 7 2.1 隨機效應模型 7 2.2 事前分配的設定 8 2.2.1 共軛事前分配 9 2.2.2 內生事前分配 9 2.3 計算方法 10 2.4 實際土壤污染調查之分析 12 第三章 與現有統計方法比較 20 3.1 現行整治辦法 20 3.2 克利金法 21 3.2.1 等污染線圖 22 3.3 地理統計的結果與討論 25 第四章 模擬 -26 4.1 較少RU資料的模擬 26 4.1.1 與實際污染平均值比較 28 4.1.2 與現行整治決策比較 31 4.1.3 最佳決策線 34 4.2 較多RU資料的模擬 37 4.3 模擬的結果與討論 42 第五章 討論 44 參考文獻 46 附錄一:試驗半變異圖 49 附錄二:模擬土壤污染濃度程式 51 附錄三:WINBUGS之Gibbs Samplers程式 52 附錄四:WINBUGS之script程式 54 附錄五:模擬資料之事後分配以及其期望值、眾數之程式 55893383 bytesapplication/pdfen-US克利金法內生事前分配地理統計隨機效應模式Intrinsic priorGeostatisticsKrigingRandom effect model內生隨機效應之貝氏模式於土壤污染之應用Application of Bayesian Intrinsic Random Effect Model to Contaminated Soilthesishttp://ntur.lib.ntu.edu.tw/bitstream/246246/56181/1/ntu-93-R91842019-1.pdf