台灣地處亞熱帶，位於西太平洋北迴歸線上之區域，造成颱風災害頻繁，且降雨在時間與空間分佈上呈現不均勻現象，因此本研究將針對颱洪期間區域降雨量進行一系列的研究，以瞭解颱洪期間區域降雨量分佈之特性，希望研究結果能夠提供石門水庫集水區防災工程之用。 本研究將以交通部中央氣象局從1986年至2001年間，所發佈的侵台颱風作為基礎，以石門水庫集水區內15個雨量站進行研究。先以具時變性及具定常性之半變異圖模式分別描述不同降雨延時之降雨空間變異特性，並以1996年賀伯颱風、2001年桃芝颱風與2001年納莉颱風進行驗證，接著比較算數平均法（Arithemtical averaging method）、徐昇式多邊形法（Thiessen polygons method）與克利金法（Kriging method）估算不同降雨延時平均雨量之差異性，並討論區域內雨量站數多寡對算數平均法、徐昇式多邊形法推估平均雨量準確度之影響。 由分析結果顯示，當降雨延時很短(如：1hr降雨)時，並不建議以克利金法描述區域降雨特性，因驗證結果發現在短時間內，推估點之降雨量誤差相當明顯，但隨降雨時間愈長，推估點之誤差範圍逐漸變小且變化較穩定。而在一般傳統平均雨量方法推估區域平均雨量部份，其方法適用性應考慮降雨延時長短，因在短時間(如：1hr降雨)以算數平均法、徐昇式多邊形法與克利金法三者方法差異性比較明顯，隨降雨時間愈長則以三種方法任何一種方法估算，其差異性並不大。本研究發現在石門水庫集水區內，以誤差範圍在10%以內為標準，以算數平均法估算區域平均雨量，若擁有區域內三分之二以上之雨量站，其推估之雨量可代表區域平均雨量特性，而徐昇式多邊形法推估，若擁有二分之一以上雨量站，估算平均雨量之準確度亦可代表區域平均雨量。 The weather in Taiwan, being located at the subtropical zone in western Pacific Ocean, is the typical island climate. The damage of typhoon events happened to Taiwan can be related to its characteristics such as the temporal-spatial disproportionate distribution in precipitation. This study not only focuses to identify the precipitation in typhoon events but also provides in sights to Shih-Men Watershed for engineering design。 The precipitation record of typhoon events form 1986 to 2001, based on the Central Weather Bureau and 15 rain-gages in Shih-Men Watershed is used for analysis on this study. Firstly, the time-variant model and non-variant model are employed to discuss the spatial variability of different rainfall duration, and the typhoon of HEB(1996)、TORAJI(2001) and NARI(2001) is taken for case study. Secondly, the difference of regional precipitation have been compared by applying Arithmetical averaging method、Thiessen polygons method and Kriging method. Finally, this study estimates the rainfall of the Shih-Men Watershed by using different sets of rain-gage record and to detect the effective of statistical accuracy. From the results, it is not suggesting to describe precipitation properties by applying Kriging method in the case of short duration, because there exists obviously error of counting value. But the errors are becoming less with increasing of rainfall duration. Therefore, applying traditions precipitation methods to estimate original area mean rainfall should be considered carefully. For short rainfall duration cases, the results show obviously difference among Arithmetical averaging、Thiessen polygons and Kriging method. However, as the increasing of rainfall duration, the difference decreases. To obtain a acceptable accuracy, the method of Arithmetical averaging can be used with more than one-third rain-gages on this watershed, while Thiessen polygons method should be with more than half of total gauges to estimating area mean rainfall.
