頁籤選單縮合
題 名 | 具離群值資料之水文頻率分析=Hydrological Frequency Analysis with Presence of Outliers |
---|---|
作 者 | 鄭克聲; 連琮勛; | 書刊名 | 農業工程學報 |
卷 期 | 61:3 2015.09[民104.09] |
頁 次 | 頁29-41 |
分類號 | 443.046 |
關鍵詞 | 異常極端降雨; 水文頻率分析; 重現期雨量; Rainfalls of extraordinary amounts; Hydrological frequency analysis; Return period rainfalls; |
語 文 | 中文(Chinese) |
中文摘要 | 異常極端降雨會使水文頻率分析在判斷重現期雨量上有所失準,進而影響水文治理、水文建設等相關規範。為了能合理推估特定重現期之設計雨量,本文係就異常極端降雨(雨量資料離群值)對水文頻率分析的影響加以探討。 本文之內容,首先探討台灣南部五個縣市其28個雨量測站的區域頻率分析,判斷測站的分區方式與各分區極端降雨之機率分布型態。接著採用兩種區域調整變數進行頻率分析,形式一:測站雨量區域調整變數為各測站經過標準化的原始降雨量;形式二:測站雨量區域調整變數為各測站原始雨量值除以該站雨量平均值。估算出各測站在延時1、2、6、12、18、24、48與72小時下重現期5年、20年、50年、100年及200年的設計雨量。最後,針對兩種區域調整變數作1000次出象之區域頻率分析和單站頻率分析序率模擬,結果顯示以形式一之區域調整變數進行區域頻率分析模擬,結果有近8成的機率比單站頻率分析佳。 |
英文摘要 | Rainfalls of extraordinary amounts are outliers in the observed data series. Presence of outliers, particularly in short data series, often causes significant problems in goodness-of-fit tests and distribution parameter estimation which in turn can affect accuracies in estimation of return periods for hydrological observations in hydrological frequency analysis. A regional frequency analysis approach is proposed in this study to cope with problems induced by outliers. Annual maximum series (AMS) of rainfalls of 1, 2, 6, 12, 18, 24, 48 and 72-hour durations from 28 rainfall stations in southern Taiwan were collected. These rainfall stations were grouped into two homogeneous regions by a k-means cluster analysis which took into account the mean, standard deviation and coefficient of skewness of annual maximum rainfalls of individual durations. Two types of regional rescaled variables were considered for regional frequency: (1) regional variables rescaled with respect to the mean and standard deviation of AMS of individual stations and (2) regional variables rescaled with respect to mean AMS only. Parameters of the regional probability distributions of AMS were estimated independently for individual homogeneous and were then used for regional frequency analysis. A stochastic simulation process which involved 1000 simulation runs was also implemented to provide a quantitative and comparative evaluation of the results of regional frequency analysis and site-specific frequency analysis. It was found that there is an approxi-mately 80% chance that the regional frequency analysis using the first-type rescaled regional variable outperforms the site-specific frequency analysis. Such results provide a theoretical basis for adopting regional frequency analysis when outliers are present in observed data series. |
本系統中英文摘要資訊取自各篇刊載內容。