EPA-99-GA103-03-A236-3

099

969 千元

2010/12/29

2011/12/28

林裕彬

土污基管會

尤衍翔

國立台灣大學

土壤及地下水

Soil pollution, Heavy metal, Stratified conditioned Latin hypercube sampling, Conditional simulation, Spatial uncertainty

張尊國;徐貴新

王鳳儀

02-33663464#

ariel730619@gmail.com

土壤採樣策略及污染調查採樣方式通常包括主觀判斷採樣（Judgmental sampling）、簡單隨機採樣（Simple random sampling）、分區採樣（Stratified sampling）、系統及網格採樣（Systematic and grid sampling）、應變叢集採樣（Adaptive cluster sampling）、混合採樣（Composite sampling）等方法，這些採樣方法依其目的與方法本身各有其優缺點，而這些方法所得採樣點又須考量同時包括多個汙染項目，且所採樣的資料影響後續汙染範圍的界定，以及其界定範圍的可靠性。因此，本研究以彰化地區為研究範圍，以該地區多種土壤種金屬為研究變數，擬發展一同時考量多項目及其空間分布與統計特性的採樣策略，以及採樣資料於汙染範圍的界定之可靠性分析方法。本採樣方法考量分區採樣及系統與網格採樣概念，結合條件拉丁超立方採樣法，發展出分區條件拉丁超立方採樣方法，此方法首先於採樣過程中先將研究區分區如採樣分區或網格和農田水利灌區，以原始土壤採樣資料或與汙染相關之變數為條件拉丁超立方採樣方法之採樣變數，以期選取的樣本於空間特性及統計特性上能更接近於原始資料或汙染相關之變數之空間特性及分佈，再以條件拉丁超立方採樣法於各分區中選取樣本所需樣本。本研究適用兩種採樣情境，一為有土壤重金屬採樣資料，另一為無土壤重金屬採樣資料，若資料為原始資料調查的土壤採樣資料，則將原始資料與不同採樣方式所得的資料以逐步指標模擬法模擬研究區內重金屬濃度空間分布情形，並比較分區條件拉丁超立方採樣與原始資料之變異圖及空間特性，且計算局部和空間不確定性，並探討其汙染範圍劃設之可靠性。若資料為汙染相關之變數(無土壤重金屬採樣資料)，則以與汙染相關變數進行分區條件拉丁超立方採樣，比較分區條件拉丁超立方採樣與原始資料之變異圖及空間特性，且計算局部和空間不確定性，並探討其汙染範圍劃設之可靠性。最後擬將研究成果及分區條件拉丁超立方採樣法採樣建置於地理資訊系統操作介面。

期末工作報告書全本.pdf

10MB

2012/3/1 上午 12:00:00

Soil sampling strategies usually include judgmental sampling, simple random sampling, stratified sampling, systematic and grid sampling, adaptive cluster sampling, composite sampling. Each one contains various objectives and the appropriateness. However, the sampling strategy should consider multivariate and then should be reliable for delineating the pollution hazard. The research aims to resample the multiple soil heavy metals at Chang-Hua County. The sampling model that can consider multivariate, statistic distribution and spatial information is developed. Moreover, the method is reliable for the analysis in delineating the pollution hazard. The method is a stratified conditional Latin hypercube sampling (scLHS) that includes the stratified sampling and grid sampling. Meanwhile, the consideration in spatial aspect for sampling sites in conditioned Latin hypercube sampling is also unignorable. First, sampling is applied based on sampling data or the other correlated data. So the incorporation of spatial data, which is regarded as the spatial cLHS, might be able to drive the data closer to their original spatial allocation. Then, the spatial distribution and uncertainty of each technique, including original data without sampling, were evaluated by the sequential indicator simulation (SIS). Furthermore, the spatial cLHS could better imitate the distribution and spatial allocation of the original data. The study considers two scenarios: with and without soil pollution data. Wherever the soil pollution occurs, the model evaluates the variogram and the spatial distribution of soil pollution based on the information offered. And then the model compares both the variogram and the spatial distribution with original data. After all the local and spatial uncertainty of the model are calculated. If there is no soil pollution at all, the model evaluates variogram and spatial distribution of soil pollution based on other information and then compares them with the original data, and calculates the local and spatial uncertainty. By the end of this project, the model and the graph presentation for related results would be showed by a built-in interface of geographic information system.

Soil pollution, Heavy metal, Stratified conditioned Latin hypercube sampling, Conditional simulation, Spatial uncertainty