| Abstract: | 台灣的坡地山崩災害常伴隨著豪雨發生，故山崩和降雨有著密切的關係。降雨在時間軸上可視為一非線性動力系統，降雨序列是一個非線性動態行為，因此描述降雨的動態行為可更完整瞭解降雨特性。本研究係利用降雨的靜態、動態特性來探討降雨行為與山崩之關係：(1)以碎形維度描述池谷浩所提三類降雨型態的差異，並利用山崩前降雨量與降雨天數等條件界定山崩之可能度；(2)以敏督利颱風降雨動態行為之盒維度表示降雨在時間序列上的碎形特性，並以此結果說明降雨的複雜程度，同時利用山崩前降雨的混沌動力行為，說明山崩可預報時間的精度觀念。最後分別考慮南投縣之動態或純量降雨誘因與山崩潛能聯合效應，重新評估南投縣山崩之危險程度。
研究結果得致下列主要結論：(1)池谷浩定義之第I類降雨類似土壤粒徑分佈曲線中之躍級配，而第II類和第III類型則類似優良級配，證實碎形維度Db可用以描述降雨型態。(2) 台灣各縣市91年至93年間山崩可能度Lf平均介於26.67%~50﹪，表示台灣大部分縣市平常處於易觸發山崩之觸發狀態；其中，南投縣Lf介於40%~60%間，觸發山崩之可能度又高於其它縣市。(3) 在南投地區，描述七二水災期間之降雨動態特性需用4或5個相空間，而長期的降雨行為只需2或3個相空間就足以描述，顯見七二水災之降雨動態行為確實較平時複雜。（4）累積降雨量與山崩發生並無明顯相關性(5)由能資所評估溪頭地區之山崩潛感中，加入考慮短期降雨動態因子之盒維度Db(T)誘因，經以敏督利颱風降雨期間已發生崩塌點檢視之，發現各已崩塌點之山崩可能度確實皆高於原山崩潛感的判定，說明增加降雨誘因之山崩評估更符合實際狀況，而且納入考慮短期降雨動態誘因Db(T)較考慮長期降雨純量誘因Lf更符合實際崩塌現況。
This research is doing on the relationship between raining behaviors and landslides by using the scalar and dynamic properties: (1)depicting the differences of the three types of raining which brought up by Hiroshi Ikeya by fractal dimension, and using some conditions as the amount of of rainfall before landslide and days of raining to attribute the likelihood of occurrence of failure；(2)indicating the fractal properties of rainfall on the time series by box dimension of dynamic raining behavior of Typhoon Mindulle，and show the complicacy of raining by this result, on the same time, using the raining chaos dynamical behavior before landslide to explain the precise respect of forcasting time by the landslide. Finally, we consider by parts of static or dynamic reasons and the landslide susceptibility union effect in Nantou，and reconsider the dangerous of landslide in Nantou.
The following are the main results:(1) Hiroshi Ikeya defined the first kind of raining condition similar to the gap graded of the soil grain-size distribution curve, but the second and the third condition were similar to the well graded. That proved “Db” could be used to describe the raining condition.(2) During 91 years to 93 years,The average of landslide impossibility Lf are situated between 26.67%~50%,in various counties of Taiwan indicated that Taiwan partial counties are usually easy to trigger the landslide condition; Among them,The Lf in Nantou County is situated between 40%~60%, The impossibility of triggers thelandslide is higher than other counties.(3) In Nantou area, it needs four or five phase space to discribe the rainfall dynamic characteristic during Typhoon Mindulle,but for long-term rainfall just needs two or three, therefore the behavior of rainfall dynamic characteristic during Typhoon Mindulle is more complicated then usual obviously.(4) Accumulation rainfall and landslides are irrelevant.(5)Energy & Resources Laboratories considered box dimension Db(T), the short-term rainfall dynamic factor, to assess landslide susceptibility of Hsitou. Survey the avalanche data during Typhoon Mindulle, it appear that Likelihood of failure occurrence of each avalanche point is higher than original landslide susceptibility, indicating that assessment of landslide subsumed rainfall incentive tally with real condition. In addition, it’s more tallying with real condition considering Db(T) than that consider Lf, long-term rainfall incentive. |
