結合經驗正交函數法與小波相關法建立屏東平原降雨地下水位時空變動系統之貝氏網路風險評估架構

近年來，極端天氣事件頻傳，包括極端降雨、乾旱等發生的頻率與嚴重度都大幅增加。例如莫拉克颱風、凡那比颱風以及梅姬颱風等都在台灣各地造成了破紀錄的降雨也導致了嚴重的災情。然而，研究顯示這些極端降雨事件並沒有造成總年度雨量的上升，而是受到了更多極端乾旱事件所影響，這使得水資源的控管以及災害的防治需要有效的調適策略變成了政府以及學界非常重要的課題。而地下水一直都是較穩定且可信賴的水資源來源之一。在台灣，超過40%的水資源來源仍是使用地下水，並廣泛使用在民生、農業及工業用水。貝氏網路圖形為一種有向非循環機率圖形架構，能夠將變數之間的因果關係系統性地建立，並透過節點與連結箭頭及條件機率分配將整個系統條理化在條件機率關係之下。本研究以屏東平原為例，首先透過經驗正交函數法(EOF)找到降雨以及地下水位變動之主要時間空間型態並對應其補注區與水文地質特性，接著透過交叉小波分析法找出降雨與地下水位變動之非穩態時間頻率的關係以瞭解其時間序列及其相變化。瞭解降雨與地下水變動的時頻與空間關係後，整合EOF與小波分析結果資訊，評估空間上各地區之人為與自然補注量與出水量，並建立降雨地下水系統之貝氏網路圖形模式條件機率架構，在此架構下評估當乾旱情況時，整個空間上地下水位洩降至極端低水位的風險。結果顯示，高屏溪流域附近之地下水位具有較高的低地下水位風險，而平原東邊沖積扇則為主要補注區，有較多的山區側向補注具有較低的低地下水位風險。

關鍵字

地下水位；經驗正交函數法；交叉小波分析；貝氏網路圖形

並列摘要

The increasing frequency and intensity of extreme rainfall events has been observed recently in Taiwan. Particularly, Typhoon Morakot, Typhoon Fanapi, and Typhoon Megi consecutively brought record-breaking intensity and magnitude of rainfalls to different locations of Taiwan in these two years. However, records show the extreme rainfall events did not elevate the amount of annual rainfall accordingly. Conversely, the increasing frequency of droughts has also been occurring in Taiwan. The challenges have been confronted by governmental agencies and scientific communities to come up with effective adaptation strategies for natural disaster reduction and sustainable environment establishment. Groundwater has long been a reliable water source for a variety of domestic, agricultural, and industrial uses because of its stable quantity and quality. In Taiwan, groundwater accounts for the largest proportion of all water resources for about 40%. Bayesian networks (BN) is one of the probabilistic graphical models that represents a set of random variables and their conditional dependencies via a directed acyclic graph (DAG), which can help us to connect the probabilistic causal relationship between random variables with those probability distributions evaluated from different methods. In this study, first, the Empirical Orthogonal Function (EOF) method is used for investigate the spatial relationship of groundwater, the area of confined and unconfined aquifer, and the most important recharge zone can be identified. Second, investigate the time-frequency relationship between rainfall and groundwater level signals by using wavelet coherence method in order to figure out the latent connections. Estimating the human and natural effect on groundwater recharge and discharge amount spatially. Further, the information of EOF and wavelet analysis is integrated into the framework of Bayesian Network. The risk assessment of low groundwater level based on the framework not only has a good performance of cross validation but also provides conditional probability map in Pingtung Plain indicating the possible direction of groundwater flow. The lateral groundwater flow primary from alluvial fans in the east side of mountains and flow into plain area then tends to flow along with the main rivers in Pingtung Plain.

並列關鍵字

Groundwater level ； Empirical Orthogonal Function ； Cross Wavelet Analysis ； Bayesian Network

參考文獻

193. 徐年盛 et al., 2011. 地下水系統水平衡分析與補注源水量推估之研究. Journal of the Chinese Institute of Civil and Hydraulic Engineering, 23(4): 347-357.

194. 徐年盛 et al., 2012. 多類灌溉型式下地下水系統抽水量與補注量之估算. 農業工程學報, 58(1): 69-90.

199. 黃偉誠, 2011. 屏東平原海水入侵之整治探討及氣候變遷引起海平面上升對海水入侵之影響評估, 國立成功大學, 台南市, 83 pp.

203. 戴宏育, 2013. 沿海地區海水入侵之數值研究－以屏東平原為例, 國立成功大學, 台南市, 65 pp.

1. Adamowski, J., Chan, H.F., 2011. A wavelet neural network conjunction model for groundwater level forecasting. Journal of hydrology, 407(1-4): 28-40.

主題瀏覽