| Abstract: | 水文頻率分析涉及決定資料所屬機率分布 之判定,由於水文資料缺乏易發生誤判,而導致 設計水文量不足或過高。 Houghton(1978)提出一具 五參數之威克拜分布(Wakeby Distribution),由於其 涵蓋度,應用於水文頻率分析可免除判定機率 分布之困擾,因此本研究將其與水文頻率分析 中常用之分布(TN、LN2、LN3、EV1、PT3及LPT3)作一 詳盡之比較。研究中使用合成資料分別以模擬 各迴歸週期之理論值及模擬隨機樣本值加以探討威克拜分布之特性,並應用於台灣年一日、 二日及三日最大暴雨量,探討威克拜分布之適 用性。合成資料之研究結果顯示威克拜分布無 論於推估各頻率理論值或擬合樣本值上,其精 確度均較六種常用分布為高。同時,實測資料 研究結果亦顯示威克拜分布較六種常用分布更 適合模擬台灣最大年一日、二日及三日暴雨資 料。
To determine the distribution obeyed by data plays an important role in hydrologic frequency analysis. Due to the small sample size of hydrologic data, the result is usually less accurate. This results in improper design. Wakeby distribution which has five parameters has been proposed by Houghton in 1978. Since Wakeby distribution is flexible, the intractable problem of determination the distribution obeyed by data could be avoided. It was investigated in depth in the present study. The results obtained by Wakeby distribution were compared with those of obtained by the distributions ( TN, LN2, LN3, EV1, PT3, and LPT3) frequently used in the hydrologic frequency analysis. In the present study, the synthetic data were used to investigate the properties of Wakeby distribution by the accuracy of fitted data and the theoretical values of some specified return periods. Meanwhile, the annual 1-day, 2-day and 3-day maximum rainfalls in Taiwan area were employed to study the aptness of Wakeby distribution. The results of synthetic data indicated that the accuracy of estimating the theoretical value of return periods or the accuracy of fitting the sample data was higher than those of six frequently used distributions. Furthermore, the results of observed data also indicated that Wakeby distribution was the most appropriate for analysis of maximum annual 1-day, 2-day and 3-day rainfall data in Taiwan. |
