台灣東部變質岩區溪流型土石流發生潛勢評估-基於坡面型土石流發生潛能

本研究建立台灣東部地區坡面型土石流判釋模式，進而探討坡面型土石流（HDF）發生潛能與溪流型土石流發生潛能兩者間之關係。本研究採用台灣東部地區變質岩區之坡面型土石流案例，總計發生坡面型土石流與未發生坡面型土石流（崩塌）案例各為31處，共62處樣本，以及30條土石流潛勢溪流作為研究案例。分別用於進行坡面型土石流判釋模式之研發及驗證、坡面型土石流發生潛勢分級模式，而後建立溪流型土石流發生潛勢分級模式。　　首先，本研究應用費雪區別分析，考量土石流案例集水區內地形特性因子，發展坡面型土石流判釋模式，並驗證判釋之成效及盲測。研究成果顯示坡面型土石流模式判釋結果，樣本區整體正判率達90.5%，驗證區正判率達90%，接續將此坡面型土石流判釋模式應用於台灣花東變質岩地區7處聚落及11條溪流進行模式盲測，共計35處坡面型土石流及50處崩塌盲測案例，取得整體捕捉率達88.2%，顯示模式判釋成效甚佳。　　坡面型土石流判釋模式為7項因子組合，分別為：集水區平均寬度、發生區形狀係數、發生區平均寬度、發生區起伏比、發生區深寬比、流動段高程差、坡面平均坡度比，並經相關性分析後確認無因子間相互共變之影響。以此模式之費雪區別函數作為指標，本文進一步建立坡面型土石流發生潛勢分級模式。於溪流型土石流發生潛勢分級模式研究方面，本文選定知本溪、卑南溪及花蓮溪等流域中，共計30條土石流潛勢溪流作為研究案例，建立溪流型土石流發生潛勢分級指標。結果顯示較佳之四組雙因子組合及其正判率為： (1)「溪流流域內HDF單元之平均隘口高程比」與「溪流流域內HDF單元之集水區總面積」，80％。 (2)「溪流流域內HDF單元之平均隘口高程比」與「溪流流域內HDF單元之發生區總面積」，80％。 (3)「溪流流域內HDF單元之平均隘口高程比」與「溪流流域內HDF單元之集水區總面積/總流域面積」，85％。 (4)「溪流流域內HDF單元之平均隘口高程比」與「溪流流域內HDF單元之發生區總面積/總流域面積」。，85％。　　研究結果顯示，坡面型土石流發生之潛能與溪流型土石流發生之間確實存在相關性。本文發展之三項要點：坡面型土石流判釋模式、坡面型土石流發生潛勢分級指標與溪流型土石流發生潛勢分級指標皆有良好的成效，可提供土石流事件之發生潛能參考依據，事前進行防災與疏散避難。

關鍵字

變質岩地質區；坡面型土石流；溪流型土石流；判釋模式；潛勢分級指標

並列摘要

This study mainly develops a new potential evaluation model of channelized debris flow (CDF), and also discusses the connection between hillslope type debris flow (HDF) and channelized debris flow. There are total 31 HDF cases ,31 landslide cases, and 30 CDF cases selected for developing the HDF recognition model, HDF occurrence potential model, and a new CDF occurrence potential model in metamorphic geologic region in Eastern Taiwan. All models were tested for its utility and verification. 　　The HDF recognition model was developed by considering Fisher’s discriminant analysis. Results show that rate of the HDF recognition model at the classification and verification are 90.5% and 90%. The HDF recognition model were composed by 7 factors which are: Average width of Watershed area, Form factor of the initiation region, Average width of initiation region, Ratio of heights and length of initiation segment, Depression ratio of the initiation segment, Elevation difference in stream catchment area, Gradient ratio of the initiation region. 7 factors are independent and also presented the physical meaning of HDF. Based on the HDF recognition model, Fisher’s function values were slected as the index and developed a HDF occurrence potential model 　　Furthermore, this study develops a new CDF occurrence potential model. A total of 30 potential CDF torrents are selected as research cases. The results show that the four best two-factor combinations of CDF potential assessment model are: 1. Average ratio of mountain pass elevation of total HDF units in stream basin and Total watershed area of HDF units in stream basin. 2. Average ratio of mountain pass elevation of total HDF units in stream basin and Total initiation region of HDF units in stream basin. 3. Average ratio of mountain pass elevation of total HDF units in stream basin and Ratio of total watershed area of HDF units in stream basin and basin area. 4. Average ratio of mountain pass elevation of total HDF units in stream basin and Ratio of total initiation region of HDF units in stream basin and basin area. 　　To sum up, the HDF recognition model, HDF occurrence potential model, and the CDF occurrence potential model tested for utility and verification are all positive. It can provide a reference basis for the potential of the debris flow event, preventing disasters and evacuating as soon as possible.