應用遺傳規劃法與二維水理模式評估河川魚類棲地-以大屯溪日本禿頭鯊為例

Abstract

河川水利工程規劃對於生態保育逐漸重視，許多研究將魚類與棲地因子(如流速、水深)間之關係量化，藉此評估工程的生態效益。魚類棲地喜好性反應於多項棲地因子非線性、複雜的交互關係，傳統使用單一因子建立之棲地適合度曲線被認為不再適用，因此找尋改良的新方法顯得相當重要。 本研究針對大屯溪下游河段，調查日本禿頭鯊出現頻度與物理棲地因子發生頻度，計算複合因子之第三型棲地適合度指數，並利用遺傳規劃法建立魚類出現頻度與棲地因子之最佳關係方程式，同時與傳統的棲地適合度曲線進行比較，驗證兩種方法預測魚類出現頻度，以及分析其對於各棲地因子的敏感度。應用二維水理棲地模式River2D，模擬流況與魚類權重可用棲地面積之空間分布，並進一步將流況分類，探討魚類對於流況的偏好，將分析結果比對文獻論述，判斷兩種方法於評估魚類棲地之適用性。 研究結果顯示，傳統與遺傳規劃複合棲地適合度曲線所反應之魚類最佳棲地狀態大不相同，後者於預測春夏兩季魚類出現頻度時，效率係數與判定係數皆達到0.87與0.61以上，具有較高的準確度。敏感度分析結果指出，兩種方法所反應之顯著敏感因子不盡相同。此外，於最佳棲地狀態時，傳統棲地適合度曲線對於流速因子之敏感程度相當大，應用於工程規劃需更審慎考量。將兩種適合度曲線輸入River2D模擬魚類棲地空間分布的結果有明顯差異，進一步加入分類流況探討魚類之偏好，遺傳規劃複合棲地適合度曲線所反應之結果與文獻描述較為符合。本研究證明將物理棲地因子發生頻度納入量化魚類棲地喜好性考量之必要，以及遺傳規劃法於建立環境變數與魚類之間關係的可行性，並且提供改良傳統量化魚類棲地喜好性的方法，作為未來河川生態工程規劃更具價值的考量準則。

The river ecological conservation concept became important, the living right and diversity maintenance were main topics on river engineering and water resources development. The relationship between fishes and environmental factors had been quantified to assess the efficiency of the conservation projects. Some studies concluded that habitat suitability curves (HSC) can’t describe fish occurrence appropriately. This failure would be ascribed to non-linear and complex interactions between environmental factors affecting the habitat selection and to uncertainty in ecology. Therefore, it was important to find a new method to improved. The study validated the HSC built by linear regression and genetic programming (GP) respectively, and then the accuracy of Sicyopterus Japonicus habitat preference in Datuan stream was predicted. Sensitivity analysis was applied for the relationship between the frequency of fish occurrence, current velocity and depth. River2D was considered as an efficient tool for simulating the flow condition of constant discharge here and then used to calculate the weighted usable area (WUA) for assessing the fish habitat distributions. Finally, discussed with flow conditions and compared the result with various of literatures. The results showed that the best fish habitat condition described by the GP HSC and the traditional one were different. The traditional HSC was accurate and reliable to predict the frequency of occurrence of fish. Two type HSC reacted differently in sensitive analysis. In the best state, current velocity was so sensitive for traditional HSC as to be unsuitable for ecological engineering planning. For the result obtained from River2D, the WUA of downstream habitat from two type HSC were divergent. Furthermore, discussing with flow conditions and found out the results of GP habitat suitability curve were more in line with literatures. This study showed up an improved method for the traditional fish habitat suitability estimation and also makes the prediction of Sicyopterus Japonicus more convincible. This confirms the fact that suitability could not be constructed through a linear process, which means the fish habitat preference could not be described through a simple linear trend. The nonlinear genetic programming could be a better tool for searching the relationship for suitability curve construction. All the consequences showd that the consideration for habitat factors frequency distribution was essential in a study of habitat suitability for fishes. This study aimed to apply the hydraulic simulation and habitat assessment measurement in reality and could provide information for ecological engineering design in the future.