機率為基礎含水層脆弱度評估模式之發展-以濁水溪沖積扇為例

Developing a Probability-based Model of Aquifer Vulnerability Assessment in Choushui River alluvial fan

彭怡惠(Yi-Huei Peng)

指導教授 ： 陳世楷 ； 張誠信

國立臺北科技大學/工程學院/土木與防災研究所/碩士(2012年)

https://doi.org/10.6841/NTUT.2012.00349

摘要

地下水是極為重要的水資源之一，相較於地表水而言，地下水具有成本低廉，水溫與水量穩定、水質良好及取用方便等多項優點。目前台灣西南部地區仍大量抽取地下水供應灌溉、養殖、民生及工業用水，若地下水遭受污染，勢必嚴重影響該地區之水資源供需平衡，甚至對人體健康造成危害，故地下水資源保育業已成為極重要之議題。DRASTIC含水層脆弱度模式自發展以來經常被使用於評估地下水污染潛勢，但此模式在應用上常因相關參數資料稀少，易產生高度不確定性，進而影響模式之評估結果。因此本研究應用指標為基礎之地理統計發展以機率為基礎之DRASTIC含水層脆弱度評估模式，除估算地下水污染潛勢值外，並評估其污染潛勢預測表現。本研究將此法實際應用於濁水溪沖積扇，DRASTIC模式中之各參數分別使用三種選取方法(最大機率選取、期望值選取、傳統選取法)進行污染潛勢評估，研究結果顯示，高污染潛勢區皆位於沖積扇之扇頂區，土地利用型態以農業耕種為主，而中度污染潛勢則大多分佈於扇央地區，無污染及低度風險污染大多分佈於沖積扇西部沿海地區以及南部地區。此外，將前述三種方法之評估結果與近年來該地區地下水硝酸鹽氮實際污染情形進行相關性分析(Pearson及Spearman相關係數)，Pearson相關係數分別為0.32、0.39及0.29；Spearman相關係數則分別為0.42、0.39及0.25。再比較三種方法之污染潛勢預測能力，預測高硝酸鹽氮濃度污染(超過0.5mg/L為門檻)，其預測能力分別為91%、82%及64%。由上述結果顯示本研究發展以機率為基礎含水層脆弱度評估模式較傳統評估結果更為優異，本研究成果可提供政府劃定地下水污染潛勢範圍以及制訂污染防治與土地使用管理策略之參考。

關鍵字

DRASTIC ； 污染潛勢 ； 地下水 ； 地理統計 ； 機率 ； 相關性分析

並列摘要

Groundwater is one of the most important water resources. To compare with surface water, groundwater has the advantage of cheap cost, stable temperature and yield, excellent water quality and convenient acquisition. Nowadays, the substantial amount of groundwater has been still extracted to supply agriculture, aquaculture, household and industry needs in many southwestern Taiwan. If groundwater is polluted, the balance of regional water resources supply may be broken and polluted groundwater poses a threat to human health. Therefore, groundwater resources conservation is a very critical issue. The aquifer vulnerability model, DRASTIC, was frequently applied to assess the contamination potential of groundwater. However, because of few observation data on assessed parameters, this model typically involves high levels of prediction uncertainty. This study uses indicator-based geostatistics to develop a probability-based DRASTIC model which is adopted to determine extents of contamination potential and discuss the performance of model prediction. A case study is performed in Cho-shui River alluvial fan. The developed probability-based DRASTIC model includes three methods of parameter estimation – selecting a maximum estimation probability, calculating an expected value and using traditional parameter estimation. The study results reveal that the proximal-fan, which is an agricultural region, is the high contamination potential region, the mid-fan is the medium contamination potential region, and the western coastal and southern areas are the low or no contamination potential region. For selecting a maximum estimation probability, calculating an expected value, and using traditional parameter estimation, the Pearson correlations between the DRASTIC scores and observed nitrate-N concentrations are 0.32, 0.39, and 0.29, respectively, and the Spearman correlations between the DRASTIC scores and observed nitrate-N concentrations are 0.42, 0.39, and 0.25, respectively. To test the predicting performance of high nitrate-N pollution of more than 0.5 mg/L, the accurate prediction rates are 91%, 82%, and 64% for selecting a maximum estimation probability, calculating an expected value, and using traditional parameter estimation, respectively. The analyzed results show that the probability-based DRASTIC model is superior to the traditional one for assessing groundwater vulnerability. The results of this research can provide government administrators with establishing groundwater protection zones and land-use management strategies.

並列關鍵字

DRASTIC ； Contamination Potential ； Groundwater ； Geostatistics ； Probability ； correlation analysis

參考文獻

[63] 陳冠宇，應用地理資訊系統與因子分析方法於地下水補注潛勢評估─以濁水溪沖積扇為例，碩士論文，國立交通大學土木工程系，新竹，2010。

[49] 江介倫、鄭克聲，指標克利金空間推估應用於衛星遙測影像分類，航測及遙測學刊，第九卷，第二期，2004，第43-58頁。

[69] 廖皎妙，應用砷風險指標評估漁塭地下水使用與管理之研究，碩士論文，國 立成功大學資源工程學系，台南，2009。

[61] 黃緒瑩，高雄平原地區地下水污染潛勢之研究，碩士論文， 國立成功大學地球科學研究所，台南，2007。

[46] 方正順，嘉南平原地區地下水污染潛勢之研究，碩士論文， 國立成功大學地球科學研究所，台南，2009。