本研究係應用基於集對分析之相似及秩次預測模式於台灣 東部流域之年逕流量預測。流域可視為水文系統，其具有隨機性、 混沌性、模糊性及灰色性等多種不確定性特質，因此如何應用過去 之水文觀測資料以及預測理論，針對水文系統未來之可能變化進行 預測，以降低對水文系統未來瞭解之不確定性，實為重要課題。集 對分析係處理不確定性問題之新方法，其利用聯繫度來表示研究對 象間之結構關係。集對分析方法其原理簡單且易於應用，因此本研 究基於集對分析之原理，從同、異、反三個觀點來建立相似及秩次 預測模式，並應用於年逕流量之預測。為驗證模式之合適性，共選 取台灣東部4 個測站之年逕流量進行分析，並與傳統之自迴歸模式 進行比較。研究結果顯示，以基於均方根誤差之平均值而言

This study describes a new framework for the application of set pair analysis (SPA)-based similarity and rank forecast model to annual runoff forecasting of basins in Eastern Taiwan. A basin can be regarded as a hydrological system, which has the characteristics of random, chaos, fuzzy and gray uncertainty. To reduce the uncertainty of the upcoming state of the hydrological system, it is important to apply the observed hydrological data and prediction theories to forecast hydrological system change in a certain period of time into the future. SPA, a new method for solving uncertainty problems, can express the detailed structure among studied variables using the connection degree. Therefore, this SPA theory and predictive model-based study will be applied from the iso-, hetro- and contra-viewpoints to build the similarity and rank forecast model, which can be applied to predict the annual runoff time series. To verify the accuracy of the proposed method, the annual runoff data of 4 stations in Eastern Taiwan were analyzed. The results obtained using SPA-based models are compared with those obtained using Autoregressive (AR) model. Based on a root mean square error (RMSE) criterion, the analytical results demonstrated that the average results obtained using SPA-based models outperformed those obtained using AR model. Moreover, for 3 stations, the similarity forecast model outperformed the rank forecast model. Hence, the annual runoff time series can be accurately predicted using the SPA-based similarity forecast model, thereby used to provide useful references and applications for the water resources planning.