本研究應用多變量統計分析方法探討台灣北部地區設計暴雨雨型之特性。研究中採用具尺度不變性高斯馬可夫雨型（Scale-Invariant Gauss-Markov，簡稱SIGM），並將此資料進行多變量統計分析，以主成份分析法萃取出設計雨型之五個主要成份，此五個主成份具有解釋設計雨型總變異數之程度約爲90%。其次，再以此五個主成份進行群集分析，將台灣北部地區雨量站分成三類，並建立均一雨型區域圖。
The method of multivariate statistical analysis was applied to study on the characteristics of design hyetograph in Northen Taiwan. In this study, we adopt a Scale-Invariant Gauss-Markov hyetograph that show the maximum likelihood estimator of the dimensionless hyeto graph. There are five significant principal components were extracted which explained total variance about 90%. By using cluster analysis techniques, we demonstrate that a map of three homogeneous regional hyetographs could be developed.
