合成臺灣河川之河性資料分析HEC-6模式不確定性

Abstract

本研究目的在於以臺灣地區無因次化河川縱剖面與坡型群集分析成果為基礎,進行臺灣河 川不定坡度資料之合成,首先對無因次化高程變化量作對數轉換,計算轉換結果之前四階動 差與相關矩陣,利用Johnson分佈與Nataf轉換建立常態空間下多變量Johnson分佈系,應用 蒙地卡羅模擬術以亂數繁衍隨機樣本並逆求各河段坡度,將合成坡度代入河道基本參數對坡 度因子之迴歸式以求得各河道基本參數,並轉為模式輸入資料.以此合成資料對美國陸軍 工兵團水文工程中心所發展的HEC-6輸砂模式進行模擬.模擬結果選取水面高程.底床變化量 .輸砂量.中值粒徑.粒徑分佈標準偏差等五項模式輸出,針對福祿數.雷諾 數.寬深比.中值粒徑 .砂石幾何標準偏差等五項基本河道參數,建立一敏感度與不確定性分析架構,分析基本參數 對模式輸出之敏感度,推導輸出對輸入之迴歸式,並評估模式不確定性與各 參數影響不確定性 的輕重比例. The purpose of this study is to generate synthetic data for rivers in Taiwan a nd to perform sensitivity and uncertainty analysis of HEC-6Model with the synt hesized data.Based on the non-dimensional channleprofile, constrained elevatio n changes are transformed into unconstrainedvariables with log-ratio method. S tatistical moments and correlation matrixin original space are then computed t o fit Johnson curves and to construct a Multivariate Johnson Distribution Syst em using Nataf Bivariate Model. Finally, an algorithm using Monte Carol Simula tion and inverse of orthogonaland log-ratio transformations is applied to gene rate stochastically randomslopes. As a result, key hydraulic channel paremeter s can be derived throughregression equations relating the slope to the paramet ers, hence enablingsimulation of HEC-6 Model with the synthesized river data. From the modeloutputs, regression models with respect to hydraulic parameters are established.A framework of sensitivity and uncertainty analysis is also es tablished toassess sensitivity and degree of uncertainty of each hydraulic par ameter.