https://scholars.lib.ntu.edu.tw/handle/123456789/161968
標題: | 台灣地區降雨等級分類之研究─子計畫:台灣北部地區降雨等級分類之研究(III) | 作者: | 鄭克聲 | 關鍵字: | 設計暴雨雨型;高斯馬可夫歷程;群集分析;區域化雨型;hyetograph;Gaussian-Markov process;cluster analysis;regionalization hyetograph | 公開日期: | 2001 | 出版社: | 臺北市:國立臺灣大學水工試驗所 | 摘要: | 為符合頻率分析是針對年最大值序列計算的特性,且希望反應出真實暴雨事件特性,在研究中選取年最大值事件進行雨型設計。本研究擬以具自我相似性之高斯馬可夫模式模擬暴雨事件之降雨歷程,該模式為非定常性之一階馬可夫歷程。經由此模式我們可推演出具有最大概似度之設計暴雨雨型,並進而利用自我相似性,建立不同延時暴雨雨型間之轉換模式。為使未設站地區在做雨型設計時,能有一合理的依據,本研究以群集分析將北部所有雨量站分成三類,並利用類別指標變數與考慮地理距離之相關 性,進行空間網格推估,建立區域化雨型。 In this study we propose a simple-scaling, Gaussian-Markov model for rainfall process of storm events. We explain the simple-scaling characteristics in terms of the IDF curves. Rainfall depths of storm events were initially normalized with respect to storm duration and total depth. Our Gaussian-Markov model is a nonstationary first-order Markov process. We proved that, under simple-scaling assumption, the normalized rain rate (expressed in percentages ) process is an IID random process and thus normalized rainfall data of different storm durations can be combined together for parameter estimations. We showed that the maximum likelihood estimator of the dimensionless hyetograph is the average hyetograph. We also propose a method for regionalization of design hyetographs. By combining the Ward’s cluster analysis technique and a indicator-variable-based probabilistic algorithm, we demonstrate that a map of regionalized hyetographs could be developed and used to determine the design hyetograph of any ungaged site in the study area. |
URI: | http://ntur.lib.ntu.edu.tw//handle/246246/9896 | 其他識別: | 892625Z002065 | Rights: | 國立臺灣大學水工試驗所 |
顯示於: | 生農、工學院附設水工試驗所 |
檔案 | 描述 | 大小 | 格式 | |
---|---|---|---|---|
892625Z002065.pdf | 78.62 kB | Adobe PDF | 檢視/開啟 |
在 IR 系統中的文件,除了特別指名其著作權條款之外,均受到著作權保護,並且保留所有的權利。