透過您的圖書館登入
IP:3.17.150.89
  • 學位論文

崩塌潛勢分析方法之研究-以高屏溪流域為例

Landslide susceptibility mapping methodologies for the Kaoping River basin, Taiwan

指導教授 : 林國峰

摘要


台灣平均每年遭受3至4個颱風侵襲,颱風除了帶來豐富的水資源也造成了許多坡地災害,其中崩塌為坡地災害中最具破壞力的災害型態。崩塌的發生不僅造成經濟上的損失,也使人民性命受到威脅。本研究藉由崩塌潛勢分析,評估崩塌風險區域,提供防災參考依據以達到減災之目的。 本研究選擇台灣南部高屏溪流域為研究區域。蒐集2008年到2011年的崩塌事件,將前三年2008年至2010年作為訓練資料,2011年作為測試資料。十四個可能造成崩塌的影響因子被挑選為崩塌潛勢評估模式的候選因子,包含了坡度、坡向、曲率、平面曲率、剖面曲率、高程、坡長、與道路距離、與河川距離、與斷層距離、地形濕度指數、岩性、24小時累積雨量和48小時累積雨量,利用K-S檢定 (Kolmogorov-Smirnov test) 進行因子篩選,選出研究區域適合之崩塌影響因子。研究中選用了三種不同的方法來建立崩塌潛勢評估模式,第一種為過去崩塌研究中最常被使用的方法邏輯斯回歸,另外選擇了兩種類神經方法,分別為支援向量機和改良式自組織線性輸出映射圖。 以2011年崩塌事件驗證各模式崩塌判別成效,並比較邏輯斯回歸、支援向量機以及改良式自組織線性輸出映射圖這三種方法建立的崩塌潛勢式評估模式其總準確率。另外,利用接收者操作特徵曲線與其曲線下面積,評鑑模式之識別力。各個模式在崩塌潛勢判別上都有不錯的表現,其中又以改良式自組織線性輸出映射圖所建立之崩塌潛勢評估模式表現最佳,能準確模擬研究區域崩塌分布之特性。最後利用所建立之崩塌潛勢評估模式映射不同重現期雨量的崩塌潛勢圖,結果顯示,隨重現期增加崩塌面積有逐漸增加之趨勢,代表高屏溪流域之崩塌地受雨量影響大。未來可以根據本研究所發展之崩塌潛勢評估模式映射崩塌潛勢圖,協助相關管理機關擬訂適當的防災策略。

並列摘要


On average, three to four typhoons attack Taiwan each year. Although typhoon rainfall is an important source of water resources, the heavy rainfall brought by typhoons frequently result in serious disasters. Landslide is one of the most destructive slope disasters. Therefore, to establish a landslide susceptibility model, which can efficiently mitigate the disaster, is always an important task of slope disaster management. In this study, three methods are employed to construct landslide susceptibility models for the Kaoping River basin in southern Taiwan, and then the model performances of these three models are compared. The three methods include the conventional logistic regression (LR) and two novel machine learning methods, namely, Support Vector Machine (SVM) and Improved Self-organizing Linear Output Map (ISOLO). Landslide events from 2008 to 2011 are collected. The first three-year data from 2008 to 2010 are used in the training phase of the models, and the remaining data are for testing. Moreover, fourteen landslide-related factors are used in the landslide susceptibility analysis, such as slope, slope aspect, elevation, curvature, profile curvature, plan curvature, slope length, topographic wetness index, distance to river, distance to road, distance to fault, 24-hour rainfall and 48-hour rainfall. The performances of three models are checked by the accuracy and the area under the receiver operating characteristic curve (AUC). The results show that the ISOLO model outperforms over the LR and SVM models in the study area. Landslide susceptibility maps obtained from the proposed model are expected to be helpful to local administrations and decision makers in disaster planning.

參考文獻


44. 楊智翔,范正成,2013,「氣候變遷對坡地災害發生潛勢之影響評估」,國立臺灣大學生物資源暨農學院生物環境系統工程學研究所博士論文。
43. 張家銓,林國峰,2009,「改良式自組織映射線性輸出模式於水庫入流量預報之研究」,國立台灣大學土木工程學研究所碩士論文。
45. 廖珮妤,詹勳全,2012,「阿里山森林鐵路事件型山崩潛感分析」,國立中興大學水土保持學系碩士學位論文。
1. Agresti, A. (2002). Categorical Data Analysis. (2nd ed.). John Wiley & Sons, New York, NY, USA.
2. Alkhasawneh, M. S., U. K. Ngah, L. T. Tay and N. A. M. Isa (2014). "Determination of importance for comprehensive topographic factors on landslide hazard mapping using artificial neural network." Environmental Earth Sciences 72(3): 787-799.

延伸閱讀