本文提出一個結合指數疊加法與物理基底數值模型的概念，以評估台灣南部屏東地下水盆地的地下水脆弱性（GV）和可持續性。研究一開始使用經土地利用因素的層次分析法（AHP）修改後的DRASTIC法，量化通過現地驗證的地下水脆弱性和地下水污染風險的時空變化，並藉此創建DRASTIC的參數空間分布。研究中使用MODFLOW模型以預測未來在不同氣候條件下的地下水脆弱性與評估地下水資源可持續性的現狀。本研究更另外採用逐步高斯模擬（SGS）的隨機方法獲得不同的隨機水力傳導係數空間分布以量化其不確定性。經量化後的水頭和水力傳導係數不確定性將添加到改進後AHP-DRASTIC模型的水深和水力傳導係數參數之中，從而量化地下水脆弱性的不確定性。 研究結果顯示在過去的20年（1995-2017）內，由於土地利用變化導致農業面積減少造成GV略有下降。在DRASTIC參數中使用依年度或更短觀測期的土地利用圖，將可以在特定地點條件下獲得更好的GV分布。同時年度污染風險圖表示，九如鄉和里港鄉自2016年以來便面臨著較高的硝酸鹽污染風險。屏東地下水盆地的鹽埔鄉，長治鄉和高樹鄉等其他農業區，氣候條件對於地下水污染風險的時間變化造成的影響較小。 此外，從地下水模型獲得的水深和淨補注量可以提高地下水脆弱性的預測準確性。研究結果表明屏東平原地下水盆地內的密集河網控制了淺層地下水位，因此氣候條件不會顯著影響GV的變化，其對地下水污染風險的影響也相對較低。另外通過對可持續性指標的分析，我們發現屏東平原地下水盆地的地下水資源系統處於高脆弱性的臨界狀態。 對於所有的所有GV類型而言，輸入參數的不確定性會使 GV值的結果在空間分佈和強度上出現很大的差異。輸入參數的不確定性對於GV的變化影響取決於參數值在其類型中的變化。隨機的GV能較原始GV更能驗證現場的硝酸鹽濃度，較高的 GV不確定性主要分佈在河流附近，主要是因為受到了地下水位和水力傳導係數變化的控制。 ;This study aims to propose a concept that integrates the index-overlay method and a physical-based numerical model to assess the dynamic of groundwater vulnerability (GV) and sustainability in the Pingtung groundwater basin in southern Taiwan. In the current study, the conventional DRASTIC was initially modified using the analytical hierarchy process (AHP) incorporated with land-use factors to quantify the spatial-temporal variation of groundwater vulnerability and groundwater contamination risk validated by field measurement data. The relevant data were collected to create the DRASTIC parameter maps. The physical-based MODFLOW model was also used to predict future groundwater vulnerability under different climate conditions and evaluate the current state of groundwater resource sustainability. A stochastic approach using Sequential Gaussian Simulation (SGS) is then employed to obtain the limited sets of realization of hydraulic conductivity and simulate hydraulic head realization and quantify their uncertainties. The information on hydraulic head and conductivities’ uncertainties will be added to the depth of water and hydraulic conductivity of the modified-AHP-DRASTIC model, whereby quantifying the uncertainty of groundwater vulnerability. Results showed that the GV has slightly decreased due to decreased agricultural areas under land-use change over two decades (1995-2017). The yearly changes or a shorter period of observations incorporated with the accurate land-use map in DRASTIC parameters can improve GV maps to obtain a better representation of site-specific conditions. Meanwhile, the maps of yearly contamination risk indicated that the counties of Jiuru and Ligang were at high risk of nitrate pollution since 2016. In other agriculture dominated regions such as Yanpu, Changzhi, and Gaoshu in the Pingtung groundwater basin, the climate conditions influence less the temporal variations of groundwater contamination risk. In addition, the depth of water and net recharge obtained from the groundwater model can improve the accuracy of the groundwater vulnerability prediction. Climate conditions do not significantly affect GV variations because of the dense river network that controls the shallow groundwater levels in the Pingtung plain groundwater basin. Therefore, the influence of climate conditions on the risk of groundwater contamination is also relatively low. Based on the analysis of the sustainability indicators, we found that the groundwater resource system in the Pingtung plain groundwater basin is in a critical condition of high vulnerability. Accordingly, the large discrepancies of GV values occurred in both the spatial distribution and intensity in all GV classes when the uncertainty information of input parameters was added to GV mapping. The uncertainty information of the input parameters may affect the variation of GV slightly that depends on the variation of parameter values in their value classes. The stochastic-based GV performs a better match of nitrate concentration than the original GV. High GV uncertainty is mostly distributed near the rivers, which significantly controlled the variability of groundwater levels and hydraulic conductivity.
