堰塞湖的形成、穩定性、影響及防制對策---堰塞湖的形成、穩定性、影響及防制對策

Abstract

由防災之觀點與需求，堰塞湖最值得關心的三項議題分別是：（1）何處容易形成堰塞湖；（2）天然壩會不會潰決、何時會潰決；以及（3）天然壩萬一潰決了會有什麼影響。基於堰塞湖基礎研究、防災規劃、及救災決策之實用性與必要性，此單一整合性計畫擬鎖定這三項重要議題為研究重點，擬由總計畫及六個子計畫協力達成，以解決上述與堰塞湖防災相關之重要課題。總計畫將負責資料蒐集資訊平台建立與建置、研究整合及會議召開、堰塞湖案例選定及案例分析彙整、堰塞湖災害處理及防治對策-防災作業手冊檢討修正等。整體計畫完成後，希望能提升未來對堰塞湖災害之整合預測能力。 堰塞湖形成與存在性之預測需將地文及水文因素納入考量，透過適當之規模門檻進行分析，才能完整評估某一塊體動是否能堵塞河道而形成堰塞湖。子計畫一旨在進行堰塞湖形成潛感分析，將由本計畫建立的全球堰塞湖資料庫篩選可供本項目使用的案例，以定量方法（鑑別分析及邏輯斯迴歸）進行潛感分析方法的建立，並以莫拉克颱風災區的選定流域為案例進行案例分析。 天然壩穩定性快速評估為堰塞湖防減災相關決策之關鑑步驟之一，子計畫二擬以邏輯斯迴歸建立天然壩破壞機率預測模式，並將以非線性迴歸模式評估天然壩壽命。研究成果預期將可提供作為台灣堰塞湖防減災緊急處置決策參考。 堰塞湖若發生壩體潰決，其對下游河道影響與沿岸居民生命財產恐造成莫大損失。子計畫三擬利用有限體積法建立一維堰塞湖溢流沖刷潰壩之數學模型、進行物理模型試驗，依不同條件設計多組實驗室案例，觀察壩體溢流沖刷、潰決之過程，供數學模型檢定驗證所需。並完成數學模型之各項測試。另選定流域堰塞湖案例進行現況模擬，分析堰塞湖下游河道可能之沖淤與河防安全影響。 天然壩壩體材料可視為崩積土，通常甚不規則，其母岩可影響崩積土之材料特性，子計畫四擬選定具代表性的試驗區位以探討不同之崩積土材料特性與性質，以用於天然壩壩體破壞模擬時所須要的材料參數。堰塞湖天然壩體破壞主要由溢流、壩體邊坡不穩定、溯源沖刷等三種不同的機制所造成。堰塞湖多形成於陡峭狹窄之河谷，三維效應特別顯著，子計畫四亦運用數值模擬，以選定流域堰塞湖天然壩體為案例，來模擬天然壩可能因溢流、壩體邊坡不穩定、溯源沖刷等破壞之潰決過程，以充分探討與了解整體堰塞湖之天然壩體之破壞機制。 既有穩定監測主要採用地表與地中位移監測及水位與水壓監測，並不能直接反應沖刷與管湧情形；且相較於一般水庫壩體穩定性之監測，天然壩之監測設備常需有耗損的準備，因此如何建構能兼顧大範圍涵蓋與經濟性的監測系統仍有發展空間。子計畫五將以時域反射技術為基礎，發展適用於天然壩破壞機制且同時兼顧大範圍涵蓋與經濟性的分佈式溢流沖刷管湧監測系統。 除了運用監測技術，對於堰塞湖的地形監視，可以採用視覺式監視技術以達到即時整治與防範的效果，達到對堰塞湖地形潰堤的預測與事後的分析。子計畫六將探討未來運用視覺式監視技術時所需克服之影像品質技術問題。

The major concerns for disaster prevention of landslide dams may include: (1) Where the landslide may happen? (2) If it happens, will it breach and when? (3) If it does breach, what will happen? A collaborative research project (consisting of six subsidiary projects) is teamed up to explore these problems. The main project is responsible for establishing a platform for research communication and data sharing. Common landslide-dam cases will be adopted for the verification of models to be established by different subsidiary projects. In addition, an earlier version of SOP of disaster-prevention management for landslide dam will be revised according to the new findings from this project. It is anticipated that the results of this project should be able to enhance the prediction ability for future landslide-dam disaster. Besides hydrological and seismic effects, geomorphology conditions are often the dominant factors that determine whether a landslide dam will happen. The subsidiary project #1 aims to propose assessment models for the susceptibility of the landslide dam formation. Discriminant analysis and logistic regression are adopted for developing the models. A catchment in the disaster area caused by the typhoon Morakot will be taken as an example for performing the susceptibility assessment of landslide dam formation using the proposed approaches. Rapid assessment of the landslide-dam stability is one of the crucial steps for decision-making to reduce the related disasters. The subsidiary project #2 aims to build a logistic regression models for predicting the failure probability of a landslide dam and its longevity on the basis of worldwide landslide inventory. It is anticipated that the proposed statistical models can be used as an evaluation tool for decision making on hazard mitigation actions. A breached landslide dam may turn into a serious disaster in its downstream areas. The subsidiary project #3 aims to develop (1) a 1-D model for modeling the overtopping erosion of landslide lakes by finite volume method, and (2) a test set-up for physical model test. The numerical model will be calibrated and verified by using the results of the physical model tests as well as field cases. The material of a landslide dam is a composition of colluvium depending on parent rocks. The subsidiary project #4 aims to explore the mechanical properties of various colluvium and the role of its parent rocks. The failures of a landslide dam may be the causes of overflow, slope instability, and the progressive head-cutting. To model the entire failure mechanisms of landslide dams, the subsidiary project #4 will also make use of numerical simulation to model the progressive failure processes of landslide dams. Existing instrumentation in landslide dams consisting mainly of water level and ground movement sensors does not directly reflect the major failure mechanisms of overflow scouring and seepage piping. Unlike man-made earth dams, instrumentation in landslide dam is prone to sacrifice in a short period of time. The subsidiary project #5 will utilize the principle of time domain reflectometry to develop economical distributed sensing rod for monitoring of overflow, scouring, and piping phenomena in landslide dams. It is also possible to use visual type (noncontiguous) technique to provide real-time image information for disaster prevention and mitigation. The subsidiary project #6 proposes to arrange video cameras in mountain regions to monitor potential landslide dams in long terms. The required technique for image refinement will be explored.