堰塞湖形成潛感分析

Susceptibility Analysis for Landslide Dam Development

鄭伊婷(Yi-Ting Cheng)

指導教授 ： 廖志中

國立交通大學/工學院/土木工程系所/碩士(2013年)

https://doi.org/10.6842/NCTU.2013.00663

摘要

國內近年來由於地震或颱風豪雨於山區河流形成許多堰塞湖，堰塞湖的天然壩若是潰決，壩體土石與堰塞湖湖水一起往下游流動會對下游造成嚴重的災害。以防災角度而言，若能了解堰塞湖形成之潛感(機率)與形成之位置將有助於未來可能形成堰塞湖地區加強相關的防災工作。本研究目的即在於蒐集及建立較為完整之堰塞湖目錄，建立適當方法求取量化之機率代表堰塞湖形成潛感(機率)並且繪製堰塞湖形成潛勢圖詮釋堰塞湖形成之位置。
本研究蒐集各種山崩與堰塞湖相關文獻與圖資，進入地理資訊系統整合資料並建立山崩與堰塞湖目錄，根據目錄挑選適當山崩與堰塞湖案例建立山崩後堰塞湖形成之相關因子，包含山崩因子、運移因子及水文因子。由相關因子中透過相關性分析選出影響因子，並將因子資料輸入統計分析軟體進行羅吉斯迴歸，算出山崩後堰塞湖形成潛感模式。此山崩後堰塞湖形成潛感模式驗證的結果，透過分類誤差矩陣驗證整體正確率約80.6%，ROC曲線之AUC為0.893，表示本研究之山崩後堰塞湖形成潛感模式可有效代表山崩後堰塞湖形成與否的預測。後續亦根據條件式機率精神，將山崩潛感模式與山崩後堰塞湖形成潛感模式互相搭配以得到堰塞湖形成潛感模式。
本研究也選擇適當的堰塞湖潛感分析單元，作為堰塞湖形成潛感模式運算及繪製堰塞湖形成潛勢圖之基礎，引用蔡雨澄(2012)之山崩潛感模式及經濟部中央地質調查所(2009)之土石流、岩屑崩滑及岩體滑動潛感模式，代入本研究所建立之山崩因子圖層與山崩後堰塞湖因子圖層計算其堰塞湖潛感分析單元內各類型山崩潛感與山崩後堰塞湖形成潛感，兩者套疊運算後，即可得到堰塞湖形成潛勢圖。目前本研究以高屏溪流域為例進行堰塞湖形成潛感運算及堰塞湖形成潛勢圖繪製，可提出適當的三種山崩型態之堰塞湖形成潛感模式並且繪製出合理的堰塞湖形成潛勢圖。

關鍵字

堰塞湖 ； 天然壩 ； 堰塞湖目錄 ； 山崩後堰塞湖形成潛感 ； 堰塞湖形成潛感 ； 堰塞湖形成潛勢圖 ； 羅吉斯迴歸

並列摘要

In Taiwan, earthquakes or rainfalls caused many natural dam lakes. If these dams break, the debris will flow to downstream and become terrible disasters. Susceptibility analysis for landslide dam formation may be helpful for the disaster prevention. This study aims to collect and establish a more complete inventory of landslide dam, establish an appropriate method to evaluate the probability of forming landslide dam and prepare the potential landslide dam forming map.
In this study, literatures related to landslide dams were reviewed and a landslide dam inventory was created from the data retrieved from the cases in the literatures. Based on the inventory, a susceptibility model of landslide dam formation after slope failure was proposed by logistic regression. 31 appropriate landslide cases with landslide dam formed and 31 corresponding landslide cases without landslide dam formed were selected from the inventory for the model regression. The parameters adopted for the model include landslide types, transporting factors and hydrological factors. The proposed model was verified by the other 10 cases in the inventory. The results show that the overall accuracy rate is 80.6% and the AUC of the ROC curves is 0.893; it reflects the facts that the proposed model is good enough for evaluating the susceptibility of landslide dam formation after slope failure.
Based on the existing assessment models for landslides, including debris flow, debris slide (shallow landslide) and rock slide, and the susceptibility proposed by Tsai (2012) and the Central Geological Survey (2009), the susceptibility of landslide of the Kao-Ping River catchment could be calculated. Also, the susceptibility of landslide dam formation after slope failure was determined from the proposed model for the Kao-Ping River catchment. Then, the susceptibility of landslide dam development was determined by combining two sets of susceptibility mentioned above for the Kao-Ping River catchment. The potential landslide dam forming map along the river was also presented.

並列關鍵字

landslide dam ； landslide dam inventory ； susceptibility analysis for landslide dam development ； logistic regression

參考文獻

95. 陳昆廷、臧運忠、郭玉樹、呂明鴻、謝正倫(2010)，莫拉克颱風引致堰塞湖之案例分析，中華防災學刊，第2期，第1卷，第43-50頁。

97. 陳樹群(1999)，堰塞湖潰決機制與減災工法研究，中國水土保持學報，第30卷，第4期，第299-311頁。

98. 陳毅青(2012)，降雨誘發崩塌侵蝕之規模頻率及其控制因子，國立台灣大學土木工程學系博士論文。

109. 史天元(2000)，橫麥卡托(TM)座標與經緯度轉換，地籍測量，第19卷，第2期，第1-9頁。

110. 臧運忠、陳冠廷、陳昆廷、王昭堡、謝正倫、郭玉樹(2012)，堰塞湖應變處理機制建立與應用，中華防災學刊，第4卷，第2期，第153-160頁。