台灣雨量豐沛,各地區雨量相差甚大,其中又以山區與平地差異最爲明顯,目前台灣山區之雨量站分佈不均,且國內目前針對區域降水、暴雨頻率分析大多以徐昇氏多邊形法結合甘保氏極端值第一類分佈法或對數皮爾遜第三類分佈法計算爲多,並未對甘保氏極端值第一類分佈法或對數皮爾遜第三類分佈法進行正確性之檢驗,本研究針對上述兩種常用之頻率分析方法進行預估雨量之正確性檢驗且結合地理資訊系統分析與應用。所得結論有三:(一)甘保氏極端值第一類分佈法較爲正確但有高估之虞,對數皮爾遜第三類分佈法則稍低,結果約爲實際降雨量之90%,且兩法之5、10年頻率暴雨預估降雨量遠高於實際降雨量,但20年之暴雨頻率分析結果則較接近實測值。(二)以最佳趨勢曲線修正克利金法,可有效提高結果之正確性達6%。(三)以地理資訊系統可取代徐昇氏多邊形法等方式,地理統計則可節省人力,且可提高正確性。
Taiwan is arainfall abundant country, the rainfall for all regions has large difference, especially, the rainfall is always the highest in the mountain area. Currently, the station of rainfall is separated uneven, and lacks seriously in the mountain area. In Taiwan, the researchers are usually using Thiessen polygons method to used Extreme-value type Ⅰ distribution or log-Prearson type Ⅲ distribution for counting the precipitation of regions or analysis the frequency of cloudburst. However, they are always not check the accuracy of the result for Extreme-value type Ⅰ distribution or log-Prearson type Ⅲ distribution. This study tested the accuracy of the frequency analyses and it combined the geography information system analysis. We got three results as following. Firstly, although the estimate of Extreme-value type Ⅰ distribution is more accurate than log-Prearson type Ⅲ distribution, the estimate of Extreme-value type I distribution is higher than the origin estimate and the outcome of estimate of log-Prearson type Ⅲ distribution is about 90% that it tends conservative. Comparing the rainfall frequency in the fifth and tenth year with the twentieth, fiftieth, and one hundred year, the estimate of a year of rainfall frequency of fifth and tenth year is higher than the origin estimate, but it is more accurate in twentieth, fiftieth, and one hundred year. Finally, Thiessen polygons method can be replaced by using the geography information system, and the GIS can reduce the devoting of human resource and raise the accuracy of data.