降雨誘發崩塌潛勢區脆弱度評估模式之建置

Study on Vulnerability Assessment Model for Rainfall-induced Landslide Potential Areas

蔡惠雯(Hui-Wen Tsai)

指導教授 ： 陳怡睿 ； 蔡光榮

長榮大學/管理學院/土地管理與開發學系(所)/碩士(2017年)

摘要

台灣的地理環境特殊，地震活動相當頻繁，梅雨與颱風季節暴雨集中，使山坡地容易發生崩塌、地滑、沖蝕及淘刷等土砂災害。此外，人口的增長及社會環境的變遷，山坡地之開發利用不可避免，致使山崩敏感區之脆弱度不斷地提升，而調適能力卻相對下降；尤其當極端降雨事件的不確定性因素作用下，更造成坡地大量崩塌，致使山區聚落的自然環境條件產生劇烈變化。因此，當面對災害衝擊時，坡地受到外在擾動影響後，能夠承受脆弱的程度也值得加以分析探討。
本研究以南臺灣屏東縣三地門鄉及霧台鄉之部分地區為研究範圍，首先蒐集研究區之相關圖層及災害歷史資料後，運用地理資訊系統建置屬性與空間資料庫，並以基因演算法自動演化類神經網路技術，進行衛星影像之判釋，以擷取地表變遷與災害資訊。研究中，以多變量不安定指數法，建置崩塌潛勢評估模式，繪製研究區之坡地崩塌潛勢圖，再依本研究建立之脆弱度分析階層，篩選出包括暴露、敏感、調適、災害潛勢及聚落區域特性等五大面向之脆弱度指標。最後，運用關聯性矩陣法，結合地理資訊系統空間分析功能，進行脆弱度分析，並繪製脆弱度分級圖。
結果顯示，研究範圍歷經2009年及2013年共四期降雨事件侵襲前、後之衛星影像判釋分類皆達中高精確度以上。自然環境、坡地擾動及雨量觸發等崩塌影響因子中，以地質、坡地擾動程度、距水系距離及雨量等因子之影響較大。研究範圍坡地崩塌災害潛勢結果顯示，歷史崩塌災害點位約70%位於中高以上之崩塌潛勢。此外，本研究並合理地評估研究範圍崩塌潛勢區之脆弱度。研究之成果可供地區性防災計畫擬定時之參考。

關鍵字

脆弱度評估 ； 崩塌潛勢 ； 多變量不安定指數 ； 地理資訊系統 ； 衛星影像判釋 ； 關聯性矩陣法

並列摘要

The unstable geology in the slope and frequent seismic activities of Taiwan make these areas vulnerable to landslides and other sediment disasters during typhoons and heavy rains. Due to the growth of population and change of economic environment, the development and utilization of slope land is inevitable. Thus, in landslide sensitivity areas, the possibility of disaster increases, while the adjustment capacity relatively decreases. Especially, under the extreme rainfall, the large landslides of slope were developed and the natural environmental conditions of mountain areas were changed dramatically. Therefore, the assessment of vulnerability for rainfall-induced landslide potential areas is essential to disaster prevention.
Some areas of Sandimen and Wutai townships of Pingtung County in southern Taiwan were selected as study areas. Using GIS, the spatial and attribute database of disaster history records and relevant data layers were constructed. Based on the genetic adaptive neural networks with texture analysis, the satellite images were interpreted to obtain surface conditions and hazard log data. Meanwhile, using the multivariate hazards evaluation method, the landslide potential assessment model was constructed. Then, a landslide potential map was drawn combined with the GIS. In addition, according to the items of exposure, sensitivity, adjustment, landslide potential and mountainous village characteristics, a framework of vulnerability analysis was established. Finally, using the relational matrix analysis method combined the spatial analysis of GIS, the vulnerability analysis was performed and the vulnerability gradation map was drawn.
Results of image classification show that the values of coefficient of agreement for four different time periods in the years of 2009 and 2013 are at intermediate-high level. Considering the landslide potential influencing factors of natural environment, land disturbance and rainfall, the result indicates that weight of geology is the highest, followed by weights of slope disturbance, distance to the stream, effective accumulate rainfall, slope, elevation, and aspect. The accuracy of predicted potential of landslide is up to 70% compared with historical records. Furthermore, the results suggest that the proposed method and procedures can be an effective tool for vulnerability analysis of rainfall-induced landslide potential areas. The findings can be a reference for the regional planning of disaster prevention.

並列關鍵字

vulnerability assessment ； landslide potential ； multivariate hazards evaluation method ； GIS ； satellite image interpretation ； relational matrix analysis

參考文獻

林洧全(2012)。衛星影像判釋技術應用於山崩潛勢分析及風險評估模式建置之研究。未出版之碩士論文，長榮大學土地管理與開發研究所，臺南。

何宇麗(2014)。降雨誘發山崩潛勢與崩塌分佈之研究。未出版之碩士論文，長榮大學土地管理與開發研究所，臺南。

吳政芳(2011)。坡地災害潛勢區土地開發適宜性之評估。未出版之碩士論文，長榮大學土地管理與開發研究所，臺南。

黃孟璇(2015)。降雨誘發之坡地崩塌潛勢評估。未出版之碩士論文，長榮大學土地管理與開發研究所，臺南。

康焜堡(2009)。曾文水庫集水區之崩塌潛能評估。未出版之碩士論文，長榮大學土地管理與開發研究所，臺南。