暴雨雨型模式有許多種，本研究採用具尺度不變性之高斯馬可夫雨型模式，該暴雨雨型模式視每一場降雨事件為一隨機歷程，考慮每個時刻之序列符合常態分佈，以一階高斯馬可夫歷程敘述臨前水文條件的遺傳效應，因此該雨型亦為具有最大概似度之設計暴雨雨型。但限於雨量測站為點的分佈，因此必須將雨型設計結果分類及區域化，並求得區域代表雨型，如此方能為未設站區域使用，亦能合理的應用於降雨逕流模式中。層級分析法(AHP)是最常用來協助決策者找出最佳策略方案的工具。透過可行方案及相關評估因素的擬定，AHP之決策分析模式可計算每個替代方案的權重來建議決策者最佳的策略選擇。本研究針對台灣北部地區提出一個新的方法來評估並建立區域代表雨型，首先選取各測站具尺度不變性之高斯馬可夫雨型，其次利用主成份分析法萃取出雨型中的五個重要成份，並以群集分析法將研究區域內之設計暴雨雨型分為三個均勻群集，但同一群集內之雨型並非唯一，故研究中最後利用層級分析法進行評估，選取並建立出區域代表雨型。
Several forms of design storm hyetograph have been developed in recent years. A Gauss-Markov hyetograph is based on non-stationary first-order Markov process. It is a dimensionless hyetograph and the most likely to occur is the average hyetograph. In order to reasonably use design storm hyetograph in uniform area, we need to select a regionalized representative hyetograph from alternatives in uniform area. The Analytical Hierarchy Process (AHP) is commonly used to assist decision makers to find out the best strategies. While the feasible alternatives and related criteria were established, the AHP module could compute the weights of alternatives for suggesting the optimal choices to the decision makers. In this study, we adopt a Gauss-Markov hyetograph. We also propose a new approach for selecting regionalized representative design storm hyetograph in northern Taiwan. By combining the principal component analysis and cluster analysis techniques, we select five principal components and group design storm hyetograph into three categories in the study area. Finally, we employ the method of analytic hierarchy process to establish the regionalized representative hyetograph for northern Taiwan area.
