降雨誘發之坡地崩塌潛勢評估

Evaluation of the Potential of Rainfall-induced Landslides

黃孟璇(Meng-Hsuan Huang)

指導教授 ： 陳怡睿

長榮大學/管理學院/土地管理與開發學系(所)/碩士(2015年)

https://doi.org/10.6833/CJCU.2015.00052

摘要

台灣山地與丘陵共約佔總面積的三分之二，且地形陡峭；近年來，由於全球氣候變遷，降雨大多時間短而集中，極易造成坡地崩塌，山區居民常常遭受財產損失及生命安全之威脅。因此，探究降雨誘發山崩之潛勢實為防災工作重要之一環。
本研究以南台灣曾文溪流域上中游之達邦村及大埔鄉大埔村、和平村為範圍，採用2013年3期颱風侵襲前後之福衛二號衛星影像，透過基因演算自動演化類神經網路技術及紋理分析之應用，進行影像判釋分類，獲得山崩發生前後的地表變遷及災害資訊。本研究並透過雨量資料的蒐集及地理資訊系統平台的運用，推估研究區內之有效累積雨量及最大3、6、12及24小時滾動降雨強度。研究中藉由不安定指數，建置災害潛勢評估模式，且繪製研究區山崩潛勢圖，並探討降雨特性、坡地擾動與坡地崩塌之關聯性。
結果顯示，本研究採用基因演算自動化類神經網路技術結合紋理分析所判釋之衛星影像分類之一致性達高精確程度。由不安定指數分析評估後得知，坡度、坡地擾動程度及雨量因子之影響最大，且崩塌潛勢圖與歷史災害點為之吻合度達83%。
此外，由研究結果亦發現，不論降雨之多寡，皆有不安定指數越高，崩塌發生點位也越多之趨勢。當有效累積雨量或滾動降雨強度值越大，且坡地擾動程度越大時，崩塌再發生的情況也越多。
崩塌地空間分布情形可發現，降雨後，新增加之崩塌地分布區位距離山脊及距河道所佔之比例較平均，但對崩塌再發生之情況，則有較多崩塌再發生之區位置是偏向河道，且有規模較大的崩塌地，當降雨量較大時，會有再次發生崩塌的情形。

關鍵字

山崩 ； 衛星影像判釋 ； 降雨強度 ； 不安定指數 ； 類神經網路

並列摘要

About two-thirds of Taiwan’s total area is covered by mountains and hills. Coupled with the global climate change, most of the time the rainfall is of short duration but with high intensity. Due to Taiwan’s steep terrain, rainfall-induced landslides often occur and lead to human causalities and properties loss. Therefore, the evaluation of potential disaster due to rainfall-induced landslides is indeed an important task.
The study areas in this research are Tapang village, Tapu village, and Hoping village, which are located in the upstream and middle stream of Tsengwen River Basin in the south Taiwan. The FORMOSAT satellite images before and after three typhoons in 2013 were acquired and used. The Genetic Adaptive Neural Network (GANN) and texture analysis were implemented in the analysis techniques for the interpretation of satellite images and to obtain surface information and hazard log data. The effective accumulative rainfall and maximum rolling 3-hour, 6-hour, 12-hour, and 24-hour rainfall intensity of rainfall in the research region was estimated by using data from the Bureau of Meteorology and geographic information system. The relationship among rainfall characteristics, slope land perturbation and landslide was explored. The instability index method was employed to establish evaluation model of landslide potential. The geographic information system was used to draw potential map of landslide in the study area.
The results of image classification show that the values of coefficient of agreement are at high level. By instability index analysis, slope, slope disturbance, and rainfall are the most important factors. The accuracy of predicted potential of landslide is up to 83% compared with historical records. Furthermore, under the different rainfall, the greater the instability index, the more the landslide occurrences. The greater the effective accumulative rainfall, maximum rolling rainfall, or slope disturbance, the more the landslide recurrence. The results also show that the ratios of new landslides near ridges and rivers are more even in the distribution of landslide spots. However, most of the locations of landslide recurrence and larger scale landslides are near rivers. The landslide will happen again when rainfall is heavy.

並列關鍵字

landslide ； satellite image classification ； rainfall intensity ； instability index ； neural networks

參考文獻

11. 林正杰(2012)。運用多期福衛二號高時空分辨率影像建立崩塌災害預警模式。國立成功大學環境工程學系碩士論文，台南。

45. 游佳靜(2015)。最佳數值搜尋原理應用於降雨誘發之山崩潛勢評估。長榮大學土地管理與開發學系未完成之碩士論文(初稿)，台南。

13. 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。長榮大學土地管理與開發研究所碩士論文，台南。

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

62. 謝正倫、黃敏郎、蔡在宗、張維恕(2010)。運用福衛二號影像進行莫拉克颱風崩塌地判釋。中華防災學刊。2(1)，35-42。