透過您的圖書館登入
IP:44.200.74.73
  • 學位論文

應用Logistic回歸法建立崩塌風險模式 -以高屏溪為例

The Application of Logistic Regression for Landslide Risk Model Construction in Koaping River Basin

指導教授 : 李鴻源
若您是本文的作者,可授權文章由華藝線上圖書館中協助推廣。

摘要


崩塌地發生常造成嚴重的土砂災害,危及產業與居民生命財產安全,政府日前已投入相當的經費及人力進行崩塌地治理,但每當颱風挾帶大量豪雨行經台灣時,仍無法避免各地災害發生。有鑑於此,建立崩塌地潛勢評估模型,在災害發生前,研判高崩塌風險區位,準確評估崩塌地發生位置及機率,實為一重要研究課題。 崩塌發生潛勢的計算大致可分為物理模型及統計模型兩種方法。由於衛星影像與電腦計算技術的發達,資料取得容易,以統計模型中多變量分析為最普遍使用的工具,本研究以其中的羅吉斯回歸法,於高屏溪流域建置崩塌地潛勢評估模型。使用的資料包含2001年至2010年11場降雨事件造成的崩塌地與水文、地文及人文因子。參考行政院公共工程委員會(2010)調查結果,將11場降雨事件以平均累積雨量800公釐為基準分為大、小颱風兩類別建立模型,並結合保全對象,模擬四種降雨情境下崩塌地發生潛勢與風險的改變。 兩模型率定準確率平均為79%及71.7%,驗證結果準確率分別為78%及71%;另外以崩壞比概念探討兩模式中的顯著因子與崩塌地發生之關聯,發現大颱風(以莫拉克率定參數)及小颱風(以碧利斯率定參數)造成的崩塌地多分布在東北至東南坡向,且莫拉克颱風崩塌地發生位置靠近山稜線,碧利斯颱風崩塌地則多靠近河道,可能原因與莫拉克雨量過大有關,大量雨水入滲至土壤中直接在離山稜線較近的區域造成崩塌,導致較少雨水透過入滲或地表逕流匯集至河道邊,使得靠近河道邊坡的崩壞比例較少;長延時降雨易造成大面積規模崩塌,短延時且高強度降雨則容易產生小面積規模崩塌。在情境四模擬(平均雨量1,535公釐)時,隘寮溪、荖濃溪、旗山溪子集水區分別產生24.5%、49.3%、47.4%的崩塌地坡單元,且在荖濃溪子集水區有較高的崩塌風險。本研究提供的崩塌潛勢評估模型,期能事先防範災害,減少人命及財產損失,達到減災之目標。

並列摘要


Landslides usually cause serious damages to industry, lifes, and properties. The government has invested considerable funds and manpowers for landslide managements, but people can`t avoid hazards during typhoon events. The construction of landslides-risk-assessment model, which can identify occurrence probabilities and locations of landslides, is an important topic. The calculations of landslide occurrence probability can be divided into physical and statistical models. As the satellite images and computer technology developed, data can be obtained more easily. Logistic regression is one of well-know multivariate analyses methods and applied to construct landslide modeling in the Kaoping River Basin. We interpreted and delineated new landslides triggered by eleven rainfall events from 2001 to 2010 and considered hydrological, humanity and landscape factors as the variables we used. We divided eleven rainfall events into two clusters according to the investigation of Public Construction Commission, Executive Yuan. Then, we simulated four precipation scenarios to represent landslide probability and risk analyses. The averages of two models accuracies are 76.5% and 71.1% in calibration, and 76% and 71% in validation, respectively. According to the results, the landslides triggered by Morakot and Bilis typhoons tend to northeast to southeast aspect. In addition the landslide triggered by typhoon Morakot located near ridgeline,but near river by typhoon Bilis. The reason is that typhoon Morakot has most precipation so that lots of water infiltrates near the ridgeline. Then, we found more landslides near the ridgeline then the river. In the scenario four, which average rainfall is 1,535mm , the percentages of landslides in Ailiao、Laonong and Cishan River subwatershed are 24.5%,49.3%,and 47.4%, respectively. The Laonong River subwatershed has most landslides. This study provides landslide-risk-assessment model to prevent disasters and loss of lives and properties.

參考文獻


36. 蕭煥章(2008) 「鄉鎮市區水災脆弱性評估模式之研究」,華岡地理學報第二十一期。
22. Lee,S.,Min,K.(2001)“Statistical analysis of landslide susceptibility at Yongin,Korea“Environmental Geology(40),p1095~1113
38. 張石角(1987) 「山坡地潛在危險之預測及其在環境影響評估之應用」,中華水土保持學報,18(2):41-62。
42. 吳亭燁(2006) 「土石流防災社區承受度評估之研究」,國立中興水土保持學研究所碩士論文。
1. Ayalew, L., Yamagishi,H. (2005) “The application of GIS-based logistic regression for landslide susceptibility mapping in the Kakuda-Yahiko Mountains, Central Japan” Geomorphology (65),p15~31

被引用紀錄


沈哲緯(2017)。河道彎道水力侵蝕崩塌預測暨連結度之研究〔博士論文,國立臺灣大學〕。華藝線上圖書館。https://doi.org/10.6342%2fNTU201702718

延伸閱讀