論文發表人：邱永嘉(加州大學洛杉磯校區土木研究所博士班)
http://www.agu.org/meetings/fm06/
為了要解決加州Warren子流域地下水資源管理優化的問題，一個可靠的數學模型用以預測不同的管理政策所帶來的結果是被需要的。現今，Warren子流域的部分地區已經被高濃度高的硝酸鹽所污染，因此，一個可減低硝酸鹽濃度，同時能夠持地下水位高度的計畫正被需求。美國地質調查所(USGS)根據現場探勘調查與野外記錄，為Warren子流域建構了一個概念模型，同時，利用觀測資料率定了概念模型中的參數。然而，要評估一個率定過之概念模型的可靠性仍然是件不容易的事情，因為真實世界的地質狀況往往是非常複雜，而且實際狀況通常是無法得知的。利用傳統的分析方法，通常無法得知概念模型架構上之不確定性變化對於地下水資源管理優化的衝擊。另外，判斷觀測資料的充足性與否也是另外一個難題，因為建構越複雜的模型，所需要的觀測資料越多。在這個研究中，提出一個新的概念來建構加州Warren子流域物件導向地下水模型。利用這個方法，可以藉由處理generalized inverse problem的過程，進而判斷概念模型的複雜度。概念模型的複雜度將取決於地下水資源管理模式的精確度（可靠性），在處理generalized inverse problem的同時，觀測資料的充足性與否也可以同時被決定。當建立一個可靠的概念模型當中，觀測資料不夠充足時，設計一個堅固且符合經濟效益的野外試驗將可經由處理generalized inverse problem的過程而得到。透過這個野外試驗將可獲得必要的觀測資料用以率定概念模型，最終得到一個可靠的概念模型。
In order to solve the optimal groundwater management problem for the Warren sub-basin, California, we need a reliable mathematical model to predict the results of different management policies. At the present time, certain parts of the basin are contaminated with high nitrate concentrations and a conjunctive use program is being developed to decrease the high nitrate concentration while maintaining the water table at the desired level. The USGS has constructed a conceptual model for the sub-basin based on available prior information and calibrated it with existing data. To assess the reliability of the constructed model, however, is not easy because the real structure of the groundwater basin is complex and generally unknown. The traditional methods of uncertainty analysis are unable to find the impact of the variability in model structure to the solution of the optimal management problem. To assess the sufficiency of the existing data is also difficult because more data are required to calibrate a more complex model. In this study, we use a recently developed methodology to construct objective-oriented models for the Warren sub-basin. With this methodology, the complexity of the model structure is determined by the accuracy requirement of the solution of the management problem and the sufficiency of the existing data can be judged by solving a generalized inverse problem. When the existing data are insufficient for constructing a reliable model for the specified management problem, a robust and cost-effective field experiment can be designed for collecting necessary data to make the calibrated model reliable.
