透過您的圖書館登入
IP:3.17.184.90
  • 學位論文

應用斜坡單元及分離元素法探討大型崩塌之演育

Investigation of the Large Landslide Evolution Using Slope Unit and Discrete Element Method

指導教授 : 林銘郎

摘要


民國98年8月7日至9日間中度颱風莫拉克襲台,所累積的鉅量降雨造成台灣中南部流域多處崩塌面積超過10,000平方公尺之大型崩塌,本研究地點高雄市桃源區之布唐布那斯溪(Butangbunasi river)為其中之一,其上游源頭處之崩塌面積約350萬平方公尺,經災害前後DEM估計其崩塌量體超過8,000萬立方公尺,此崩塌規模接近小林村山崩事件,然而兩者並未造成相當的影響範圍。本研究認為於布唐布那斯溪於莫拉克期間之大型崩塌可能是由多次的小規模崩塌事件階段性所造成,相較於小林村災害屬於一次性的劇烈式山崩,分次崩塌中較早發生的事件其堆積材料會對後續的崩塌量體產生消能作用,即各次崩塌事件之間具有複合性與連動性,會關係到大型崩塌最終的災害影響範圍。 為了解布唐布納斯溪之大型崩塌演育過程,本研究應用物件式影像分析自動產製斜坡單元先建立本區域之災害模式與破壞機制。理想的斜坡單元尺度應小到能區分不同運動方向之崩塌事件,因此以坡向作為分割的主要因子,引入計算局部方差(local variance, LV)選擇適合的分割尺度取代過去的試誤法,並透過由莫拉克後之山崩目錄統計的平均山崩面積,以及野外調查的觀察結果來選擇適合的斜坡單元尺度,即同一斜坡單元內的崩塌材料可視為一次性運動,且若斜坡單元發生崩塌,其材料體積具有相當的影響性。 以此斜坡單元為框架,配合三維分離元素法軟體PFC3D探討崩塌材料之運動過程與最終堆積結果。由斜坡單元之坡向將此大型崩塌區由東向西分為A、B、C、C、E共五處崩塌事件,根據研究結果顯示,於莫拉克期間,左岸A、B、C與右岸之源頭D處為最先發生的事件,其材料往下游運動的同時刮蝕兩側河道,使右岸主要崩塌E區的量體與左岸近下游A、B兩處之坡腳受到淘刷,進而形成大型崩塌;另外,在數值模型中材料因不同次序的崩塌時序產生的消能作用,使山崩材料在中游河道轉折處便形成天然壩,且天然壩的堆積層序又與崩塌事件的先後有關,因天然壩的形成,造成後續的崩塌材料運動至與荖濃溪匯流處之量體減少,因而堆積影響範圍也因此減小;而在崩塌塊體的運移過程中,最高的滑移速度可以達到每秒60公尺以上,在中游河道運移之平均速度則約每秒40公尺,而根據速度之監測資料的波動也顯示崩塌塊體彼此的碰撞作用與河床上下、左右蜿蜒之起伏地形,尤其先滑落的材料在運移過程中若形成堆積,則會造成後面的崩滑體岩塊相當大的消能效果。未來於大型崩塌處之山區工程應討論不同情況之山崩次序下崩塌材料於河道中下游的堆積形貌,整體評估其災害可能之影響範圍。

並列摘要


During the Typhoon Morakot, which happened on August 7 to 9 in 2009 in Taiwan, the heavy rainfall caused many more than 10,000 m^2 area of large landslides. One of them is the Butangbunasi river large landslide in Kaohsiung City. According to the DEM analysis, the Butangbunasi river large landslide source area is about 3.5 million m^2, and the total landslide material is more than 80 million m^3 through DEM analysis. The source materials of the Butangbunasi river landslide is four times of the Hsiaolin landslide, but the severity of damage did not cause the same. The previous research proposed the Hsiaolin landslide is catastrophic landslide, but this study consider that the Butangbunasi river large landslide may be caused by several small-scale landslide in sequence. The earlier deposit materials would produce the energy dissipation to the subsequent landslide block, and reduce the final disaster region of the large landslide. In this study, we generated the slope unit by the object-based image analysis to recognize the mechanic of the different small-scale landslide events of the large landslide evolution. The ideal size of the slope unit should separate the direction of movement of different landslide events and the material volume has an influence. This study used the aspect and basin layers in the segmentation process, and carried the concept of local variance (LV) for defining meaningful segmentation scale parameters. Segmentation results had been evaluated visually, based on field work, expert knowledge and landslide inventory. Compared to previous researches have been heavily dependent upon trial-and-error exploration, the LV method can speed up the process and have quantified standard in the selection of appropriate scale parameters. According to the slope unit aspect, we can divide the Butangbunasi river large landslide into five landslide events, and number the events as A, B, C, D and E from east to west. With the slope unit as a framework, the software PFC3D, which is based on the discrete element method, was simulated the complex behavior in the kinematic process and the geometry of deposition. Base on numerical model, the friction coefficient for each slip face should decrease into 0.06 that the landslide block will start sliding and the friction coefficient of each particle as 0.09 can obtain the best deposition regions which were closely matched with satellite photos after the Typhoon Morakot. The bonding strength as 60MPa will cause the disintegrative rockslide to develop into fragmentation and debris avalanches, which demonstrated by the field work. Those best fitting microcosmic parameters will be used to simulate the kinematic process of the Butangbunasi river large landslide. Based on simulation results, the large landslide started in 11 to 30 sec and the rock block become the debris avalanche. After failure initiation, the average velocity was about 40 m/s which permits the debris to move in Laonung river. At 452 s after the event, the debris avalanche came to rest, forming a landslide dam in the middle of the Butangbunasi river. In the 3-D model most of the energy will dissipate because of the collision interaction of landslide blocks and deposition during the movement. The result helps with reconstructing historical large landslide events, which is due to the sequence of small-scale landslide events, and also predicting the affect region of the disaster.

參考文獻


蒲淵明 (2012) 「地滑/岩崩形成堰塞湖之運動模擬及參數探討」,國立交通大學土木工程研究所,碩士論文,共149頁。
唐昭榮 (2010) 「臺灣遽變式山崩傳送與堆積之顆粒流離散元素模擬」,國立台灣大學地質科學研究所,博士論文,共240頁。
簡瑋廷 (2011) 「應用物件導向分類方法自動產製斜坡單元」,國立中央大學應用地質研究所,碩士論文,共74頁。
翁晨桓 (2011) 「分離元素法山崩模型建置及震滑機制行為初探」,國立臺北科技大學土木與防災研究所,碩士論文,共109頁。
莊心凱 (2012) 「結合地貌主題圖層及物件導向式影像分析方法應用於山區氾濫原及周邊區域特徵判釋」,國立臺灣大學土木工程研究所,碩士論文,共120頁。

被引用紀錄


翁正學(2017)。向上裂隙水對順向坡穩定性及破壞行為之影響〔碩士論文,國立臺灣大學〕。華藝線上圖書館。https://doi.org/10.6342/NTU201703292
彭厚仁(2016)。不同尺度山崩潛感圖製作方法之研究〔碩士論文,國立臺灣大學〕。華藝線上圖書館。https://doi.org/10.6342/NTU201601421

延伸閱讀