Title

應用聯合機率分析方法於淡水河系之重現期洪水位預測

Translated Titles

The Application of Joint Probability Method to Predict Water Levels for Different Return Periods in the Danshuei River System

DOI

10.29974/JTAE.201806_64(2).0006

Authors

柳鴻明(Hong-Ming Liu);柳文成(Wen-Cheng Liu)

Key Words

淡水河系 ; 一維變量流 ; 聯合機率 ; 蒙地卡羅法 ; 拉丁超維取樣法 ; 頻率分析 ; Danshuei River system ; One-dimensional unsteady flow model ; Joint probability ; Monte Carlo simulation ; Latin hypercube sampling ; Frequency analysis

PublicationName

農業工程學報

Volume or Term/Year and Month of Publication

64卷2期(2018 / 06 / 01)

Page #

75 - 91

Content Language

繁體中文

Chinese Abstract

台灣因為河川坡陡流急,於颱洪期間洪流溢堤時有所聞,加上外海暴潮位與內陸河川洪水宣洩的交互作用有密切相關性,暴潮與河川流量間具有相當的不確定性。本研究建置河系一維變量流模式,分析天文潮、暴潮偏差量及河川流量的機率密度函數,應用聯合機率分析之蒙地卡羅法與拉丁超維取樣法,對於各參數進行取樣,再以取樣結果進行河系之洪水位模擬預測。本研究分別以2011年南瑪督颱風、2012年蘇拉颱風與2013年蘇力颱風,三場颱風事件進行一維變量流模式之水理模式驗證,結果顯示模擬水位與實測水位大致上吻合,經驗證後之模式進行不同聯合機率分析法於淡水河系之洪水位預測。聯合機率分析法則選擇常用之蒙地卡羅法與拉丁超維取樣法進行比較,蒙地卡羅法取樣2,000組、拉丁超維取樣法取樣200組與400組,顯示應用拉丁超維取樣法取樣400組之洪水位預測結果與蒙地卡羅法取樣2,000組之預測結果相當接近,意即拉丁超維取樣法較蒙地卡羅法可減少許多模擬組數及模式模擬計算時間。另應用拉丁超維取樣法取樣與頻率分析法進行淡水河系不同地點之重現期水位預測,結果顯示拉丁超維取樣法取樣預測之不同重現期洪水位,大致皆低於頻率分析法所預測之洪水位,表示頻率分析法所預測的洪水位可能會有高估之現象。

English Abstract

Due to the steep topography and flash flood in Taiwan, massive flood events and recurring overtopping flow occurred frequently. The interaction between storm surge in coastal ocean and inland river flood displays a high correlation. However, storm surge and river discharge exist uncertainties. In the present study, one-dimensional unsteady flow model is established for the Danshuei River system of northern Taiwan. The astronomical tide, surge height, and river discharge were analyzed with historical data to build the probability density functions of each parameter. The joint probability method including Monte Carlo simulation (MCS) and Latin hypercube sampling (LHS) was adopted to yield samples of each parameter. Then the sampling scenarios were served as boundary conditions to drive the model simulations with one-dimensional unsteady flow model. The model was validated with observational water level data of three typhoon events. The overall model simulation results are in quantitative agreement with the available measured data. The validated model was then applied to predict water levels with different sampling scenarios obtained from joint probability method. The predicted water levels using MCS 2000 samplings, LHS 200 samplings, and LHS 400 samplings were analyzed and compared. The results revealed that the predicted water levels for different return periods using MSC 2000 samplings and LHS 400 samplings were quite similar. It means that the predicted water levels using LHS 400 samplings can save the computational time comparing with using MCS 2000 samplings. Moreover, the predicted water levels at different locations for different return periods using LHS and traditional frequency analysis. We found that the predicted water levels for different return periods using LHS were less than using traditional frequency analysis. The traditional frequency analysis used to predict water levels for different return period would be overestimated.

Topic Category 基礎與應用科學 > 永續發展研究
生物農學 > 生物科學
生物農學 > 農業
生物農學 > 森林
生物農學 > 畜牧
生物農學 > 漁業
生物農學 > 生物環境與多樣性
工程學 > 水利工程