運用統計或頻率分析之方法於乾旱分析方面由來已久,然而由於受限於乾旱現象是屬於持續性水文量偏低之極端事件,無法以一般之機率分布分析,因此在評估乾旱程度及研判結果之應用上也受到限制。 本研究目的係衍導三參數之通用極端分布之LL-moments,並利用合成資料及台灣南部及北部實測資料,以探討LL-moments於乾旱頻率分析之可行性。 由研究結果顯示,LL-moments適用於左偏之機率分布與有較小之極端事件的樣本,不論在資料長度較短或是高重現期距之推估時,因給予低極端值部分更高之權重,再加上台灣地區水文資料長度之不足及近年氣候異常現象頻仍等因素,因此應用LL-moments於乾旱頻率分析之推估方法,相較於L-moments,具有更高之穩定性,值得未來於台灣地區之乾旱頻率分析所採用。 Drought events have been analyzed by using statistics or frequency analysis for a long time. However, drought happened due to the extreme hydrology phenomena that have a continuity of lower-tail events. It can not be analyzed by using ordinary probability distributions. Therefore, the estimation on drought condition and its applications in performance analysis are restricted. This study aims to derive the LL-moments of the generalized extreme value (GEV) distribution, all with three parameters, by using synthetic data and real data which comes from the northern and southern Taiwan area. Furthermore, the feasibility of using LL-moments for drought frequency analysis is also discussed in this paper. The result shows that LL-moments are applied to characterize the data with left-skewed distribution and small sample size of extreme events. For the real hydrological data; however, there are short records in Taiwan and abnormal phenomenon caused by climate change. By giving a high weight to the lower-tail of distribution in estimation, no matter shorter length or high return period of the hydrological data, LL-moment shows high stability in estimation on drought frequency analysis when compared to L-moment. It should be useful for getting evaluation of drought frequency analysis in Taiwan.