軟岩河床遷急點倒退率模式發展與參數率定

Abstract

1999年時由於車籠埔斷層活動引發集集大地震。在斷層上盤隆起造成台灣中西部之部分河道抬升，抬升之河道在地震後開始發生遷急點倒退情形。遷急點倒退往往加速河道的下切作用，造成跨河構造物毀損，然而目前仍缺乏有合適的模式詮釋國內軟岩河床上遷急點倒退之速率與歷程。本研究先使用歷年航照資料進行多時期之地形重建，基於歷史地形及影像資料分析、詮釋河道之變遷。大安溪自集集地震以來出現遷急點倒退，河道急速下切，直到2010年地形方才重新達到平衡狀態；大甲溪埤豐橋上游受到人為影響在2008年遷急點消失，但因為暴雨及設置中央排水道在河道順層處又出現新的遷急點；大里溪則是在震後便進行整治，因此未受到地震造成河道抬升影響；頭汴坑溪在近年影像可觀測到一江橋下游流況較不穩定，但是在震後無法確認河道抬升情況，無法肯定與九二一地震之相關性；乾溪則可觀測到遷急點倒退，但是並未確認倒退終止的時間。此外，本研究針對和弦搜尋演算法進行適度改良，在演算過程中強制新的和弦必須與和弦記憶不完全相同，並加入參數自由設定及粒子群法的概念，將改良後之和弦搜尋演算法在連續問題進行驗證分析後用於後續之模式參數率定。本研究接著基於觀測結果，擷取影響遷急點倒退相關之主要因子，包含河道坡度、抗沖蝕指數、河道落差、河道寬度及流量等，先透過定性分析發現，上述因子與遷急點倒退速率有強烈之相關性，再採用上述因子建立遷急點倒退模式，並藉由前述改良後之和弦搜尋演算法標定模式參數。經比較模式計算值與實際發生之遷急點倒退情況，本研究所提出之模式可大致描述國內軟岩河床上遷急點倒退之速率與歷程，應可作為後續進行遷急點相關研究之參考。

The rupture of the Chelungpu Fault in September 1999 resulted in some local riverbed uplift across a few rivers in the central western region of Taiwan; the local riverbed uplift created knickpoints on the river channels. The sudden presence of a knickpoints tends to retreat quickly so that accelerates the riverbed incision and endangers the structures crossing the river. Yet, there is no suitable model for describing the rapid knickpoint retreat rate (KPR rate) in soft rock riverbed in Taiwan. The first step of this thesis was terrain reconstruction and analysis: the tasks were to reconstruct digital surface model (DSM) from chronological aerial photographs and to analyze the geomorphology changes of river-channel from multi-year DSMs and images. The evolutions of river channels of several target rivers (namely, the Daan River, the Dajia River, the Toubiankeng river, the Dali River and the Gan River) were carefully examined; the relevant geomorphology data were then retrieved. Next, this study proposed a KPR-rate model by taking the major factors affecting KPR rate into account; these factors include slope, erodibility index, fall height, channel width and discharge. The study made use of an improved harmony search algorithm (HS) to calibrate the model parameters. The improvement of the original HS method was attempted with various schemes: including parameter-setting-free harmony search (PSF-HS), particle-swarm harmony search”(PS-HS) and limiting duplicate harmony memory. The improved HS method was tested and examined, then was used for the calibration of model parameters. Compared with the actual KPR data, it appears the proposed model is able to quantitatively describe the progressive knickpoints recession for soft rock riverbed in central western region of Taiwan.