• 作者： 易任
• 作者服務機構： 國立臺灣大學農業工程學系
• 中文摘要： 區域性之水文頻率分析，在利用流域之水文及地文特性，而對某一重現期距之年極值水量予以推估，並求出其可信限界。使大於一個標準差時對應上限之極值水量，其發生之可能率為16%（約1：6）。此種方法不僅可以正確推估某極值水量，並同時顯示推估之可靠程度。本研究雖僅作關於最大極值年水量之推估，但亦可應用於最小極值水量之推估，故對於區域性之系統規劃設計，當有其參考之價值。 關於極值年流量理論分佈檢定，由於流域水文站水文記錄年數過短或殘缺不全，致使作卡方檢定時，由於資料不足及不完整，而使木研究結果顯示：瞬時之尖峰流量分佈中僅以對數常態分佈（二參數）及對數伽瑪分佈（二參數及三參數）之適應性差異較小。如對尖峰流量、一日、二日、四日及六日等各期距作通盤之考慮，則以對數常態分佈（二參數）及對數伽瑪分佈（二參數）為佳，參閱表7，故在實際應用上，木區城似宜採用對數常態分佈（二參數）及對數伽瑪分佈（二參數）。 本研究為考慮區域中各集水區特性之計量參數，與關於所假定頻率分佈之多種動差，求出一簡捷之方法，使區域中全部或大部份可用之極值流量資料，可以作合理之分析與整理，而加以應用。又對某一指定重現期距（由設計之需要作決定）洪水流量之推估而附帶有關資料可靠性之研究者，本研究亦作有週詳之考慮。此外並對本省新店溪上游北勢溪之翡翠谷水庫之設計洪峰流量、大漢溪石門水庫壩址霞雲站之洪峰流量及其上游之稜角水文站洪峰流量，以及南非官方水文測站編號為 W5MO5 增加若干項之記錄資料而將短暫記錄作適度之延長所求得之洪峰流量等，均列舉責例作比較之研究，以增強本文之參考與實用價值。
• 英文摘要： A regioqalized study of maximum annualflows of different short durations including peakshas been carried out for a long time in theU.S.A. and other European countries. Toreview the limited length of record available atmost of the gauging stations in Taipei-Keelungregion, the author attempts to develop atechnique (or Modeling) to strengthen the dataavailable at any particular point of interest, byusing all available pertinent flow data in thisregion. Having chosen the extremal distributionbest suited to this region, the moments of thesample after adjustment are correlated withvarious catchment characteristics. This allowsestimation of flood magnitude frequency curvesat any site of interest within this region, withassociated confidence limits. Such frequencycurves are determined for various suitable timeintervals which then allows the synthesis ofcharacteristic flow hydrographs, with a specificprobability of occurrence attached to each,along with associated envelopes correspondingto specific confidence limits. Comparison withhydrographs derived from rainfall input depthswith specified probabilities, subtracting losses,and then using unitgraph methods, leads to theconclusion that a better relation betweenprobability of occurrence of a specific hydro-graph, and its magnitude, can usually beobtained by direct statistical methods, than bymore indirect deterministic techniques. An estimate of a flood with a specifiedrecurrence interval (determined by specifieddesign considerations) should be accompaniedby information concerning the reliability ofsuch estimate. This paper outlines a methodologyby mean of which all or most of the availableflood flow information in a region can berationally analysed and assembled, by takinginto account quantifiable parameters of charac-teristics of the various catchments in the regionand relating these to the moments of thefrequency distribution assumed. The applications of the normal, log normal(2-parameter), log normal (3-parameter), gamma(2-parameter), gamma (3-parameter), log gamma(2-parameter), log gamma (3-parameter) andGumbel distributions，χ -distribution，F-test,consistency and homogeneity test are alldescribed respectively. Many illustration ex-amples are also given in this article.
• 中文關鍵字： --
• 英文關鍵字： --