石門水庫集水區崩塌地產砂與後續沖刷之量化研究

Estimation of Sediment Yields from Landslide and its Post Failures Soil Loss in Shihmen Reservoir Watershed

蔡宗賢(Zong-Xian Tsai)

指導教授 ： 李鴻源

國立臺灣大學/工學院/土木工程學研究所/博士(2012年)

https://doi.org/10.6342/NTU.2012.01139

摘要

台灣的水庫集水區地質脆弱、雨量豐沛、降雨集中，故容易造成崩塌地並帶來大量的土砂，要評估崩塌土砂對下游庫區的影響是一困難的目標，本研究就崩塌地本身量體估算、到河道之崩塌運移率推估、河道輸砂計算、以及崩塌地後續產砂量化，結合各種方法，從一個整合的觀點作出發，在時間尺度與空間尺度來探討整個集水區內崩塌地整體的產砂貢獻。本研究先利用歷史航拍圖所判釋之新增崩塌地 (1968-1972、1972-1976、1976-1986、1986-1998、1998-2003以及2003-2004)資料，分析其崩塌頻率-面積關係，以三參數反伽瑪函數(three-parameter inverse-gamma function) 結合最大概似法進行擬合，配合崩塌體積-面積關係，估算石門水庫集水區崩塌體積量。另外在崩塌土砂進到河道之運移率推估上，運用七種簡化驗證法則公式(simple rule-based)之平均進行崩塌泥砂運移到河道距離的估算，配合崩塌地離河道距離，計算崩塌地運移率。河道輸砂方面，本研究以HSPF模式模擬石門水庫各子集水區之河道年輸砂量，在這些基礎上，從集水區崩塌地的產生、移動到對下游接近庫區的影響作一連串的整合。結果顯示石門水庫集水區在1986-2003年之崩塌率(landslide erosion rate)為每年0.4-2.2mm，但2004年受艾利颱風影響則迅速增加為7.9mm；崩塌泥砂遞移率則由1986-1998年的78%降低至2004年的55%，雖然遞移率有減少的趨勢，但並不代表崩塌土砂進入河道的量也減少。比較上游崩塌產砂與HSPF模式模擬河道輸砂結果顯示大約僅有一小部分崩塌泥砂可運移到下游，在1986-1998年間大約13%；2004年艾利颱風則約有23%。在石門水庫集水區內的泥砂問題主要受到河道輸送能力所限制(transport capacity limited)而非上游的泥砂供應太少(sediment supply limited)，造成此原因與集水區內大量興建攔砂壩有關，而兩者間的不平衡也造成日後集水區內仍有相當可觀的泥砂停留在上游河道內，在下次豪雨發生後可能會對下游造成嚴重的衝擊。
在後續崩塌地上的沖刷量估算上，本研究以全站儀(total station)於2006-2007年間針對集水區內十三處崩塌地於雨季前後進行量測與分析，研究顯示裸露崩塌地的產砂率主要受到降雨條件、崩塌地的幾何形狀以及地形特徵所影響。而崩塌地的變遷則主要為降雨所主導，但崩塌面上的植生復育面積卻對時間長度較為敏感。為了量化後續崩塌土砂沖刷與雨量變化的關係，本研究提出兩條迴歸公式，結果顯示崩塌土壤的沖刷深度約為0.78~1.06 m yr-1，相當於20,670至28,090 t ha-1 yr-1。本研究針對石門水庫集水區整體評估泥砂來源比率發現，崩塌沖蝕量、後續崩塌沖刷量以及土壤沖蝕量比約為66:31:3。顯示集水區產砂問題中，崩塌沖蝕是一最主要的來源，但後續崩塌地上的沖刷量亦為一不可忽略的角色，也需要相當程度的重視。

關鍵字

崩塌地 ； 崩塌頻率-面積統計 ； 泥砂運移率 ； HSPF ； 崩塌坡面後續產砂 ； 石門水庫 ； 集水區產砂

並列摘要

Landslides generate enormous volumes of sediment in mountainous watersheds; however, quantifying the downstream transport of landslide-derived sediment remains a challenge. This study estimated landslide volumes for the Shihmen Reservoir watershed in Taiwan, for six different time periods, using empirical landslide frequency-area and volume-area relationships. The percentage of total landslide volume reaching streams was estimated using empirical landslide runout models. Fluvial transport of landslide sediment to the reservoir was simulated using the Hydrological Simulation Program- FORTRAN (HSPF). Results indicate that landslide erosion rates ranged 0.4 mm yr-1 to 2.2 mm yr-1 during the periods of 1986-1998 and 1998-2003 , but rapidly increased to 7.9 mm yr-1 following Typhoon Aere in 2004. The percentage of landslide sediment delivered to streams decreased from 78% during the period of 1986-1998 to 55% in 2004. Although the delivery ratio was lower, this decrease is not necessarily associated with a decrease in the volume of landslide sediment reaching streams. Model simulations indicate that only a small proportion of the landslide material was delivered downstream. An average of 13% of the landslide material in the river was moved downstream during the period of 1986-1998. In 2004, the period including Typhoon Aere, the annual channel sedimentation accounted for approximately 23% of the landslide material moved to streams. Channel sediment yield in the study area was determined by transport capacity rather than sediment supply, due to a reduction in sediment transport capacity resulting from existing check dams. The imbalance between sediment supply and transport capacity has resulted in a considerable quantity of landslide material remaining in the upper-stream regions of the watershed.
Landslides have both immediate and extended-duration effects on sediment discharge. Examining the role of post-failure soil loss/sediment yield of landslides is essential for basin management purposes. The comprehensive measurement of sediment yield and the production of existing landslides could provide a better understanding of post-failure behavior. This study investigated the post-failure sediment yield of 13 landslide sites in the Shihmen reservoir watershed during the rainy seasons of 2006 and 2007 using a total station. Our study showed that soil loss is dominated by rainfall conditions and that the geometric and geographic characteristics of each landslide can have significant influences on the efficiency of sediment delivery from fields. We observed that rainfall is important in determining landslide retrogression/enlargement, while the recovery of vegetation in the landslide area increases with time. Intuitive knowledge of these phenomena was confirmed by our data. We obtained valuable landslide measurement data, which could be useful for future landslide modeling. In addition, our results provide more general information about the behavior of landslides.
Finally, we suggest two regression models for the estimation of gross soil loss from cumulative rainfall. We estimated gross soil loss of 0.78 to 1.06 m yr-1, which is equivalent to 20,670 to 28,090 t ha-1 yr-1. Using these estimates, we assessed the sources of sediment in the watershed. We found that the ratio of the landslide erosion, soil loss after landslide, and watershed soil erosion was approximately 66:31:3. The primary implication of this study is that remedying existing landslides might be more effective than any other solution in reducing the sources of sediment in the Shihmen reservoir watershed.

並列關鍵字

landslides ； landslide frequency-area statistical estimation ； sediment delivery ratio ； HSPF ； post-failure sediment yields ； Shihmen reservoir watershed ； Watershed sedimentation

參考文獻

(115) 李昶谷，「蝕溝沖蝕產砂量推估之研究」，國立台灣大學土木工程學系碩士論文，2007。

(117) 趙倬群，「石門水庫集水區崩塌土砂產量推估之研究」，國立台灣大學土木工程學系碩士論文，2004。

(119) 陳樹群、翁愷翎、吳俊鋐，「玉峰溪集水區崩塌特性與崩塌體積之探討」，中華水土保持學報，第四十一卷第三期，2010。

(116) 謝正倫，「桃芝颱風災區土石流災害潛勢分析成果報告書」，行政院農委會水土保持局，2000。

(1) Acharya, G., Cochrane, T., Davies, T., Bowman, E., 2009. The influence of shallow landslides on sediment supply: A flume-based investigation using sandy soil. Engineering Geology 109, 161-169.