本研究應用地理統計中之克利金及資訊熵(entropy)理論，提供一套評估集水區既有雨量站觀測網適用性之方法。克利金法用於推估空間中可能設站的降雨資料，而資訊熵則用以量測不確定性，經由熵值之計算可了解流域中各雨量站可透露降雨資訊之多寡，因此可藉由熵之方法以計算雨量站間之不確定性及資訊之轉移，進而排列流域中各雨量站之重要順序。此外，亦可藉由評估雨量站所含資訊之多寡，以推估集水區達飽和雨量站之數目及設站之位置。研究區域選定國立臺灣大學實驗林管理處轄區，將其區分為346個1平方公里之待選網格，並搜集區內自1992年起至2009年由中央氣象局直轄雨量站、農業氣象觀測網一級站、國立臺灣大學實驗林管理處所轄雨量站及作者架設之微氣象站和自記式雨量站之月累積雨量紀錄來進行分析。研究結果顯示，以地理統計法半變異元套配結果分析各個月份全實驗林轄區之降雨空間分布，可大致符合試驗半變異元之變化趨勢。對於降雨量豐沛之7月至9月，其平均距離參數較短，顯示降雨量較大之月份在空間上具有較小之影響半徑。而較乾旱之月份如11、12月及1月，其降雨量變化不大，擁有較大之影響半徑，亦即在空間上需較長之距離方能達到一定之變異值。站網分析結果顯示，僅需2個待選站即能掌握整個轄區62.60%之降雨不確定性，5個待選站即能掌握85.34%之降雨不確定性。研究結果亦發現優先選入之待選網格多位於該處或其對岸地形畸嶇與高差較大處，管理單位可視其實際需要酌予增站，排序較後之既有測站可考慮遷移位置或予以廢站。透過雨量站觀測網分析未來可有效增加集水區雨量觀測之精確性與代表性，為水土保持設計及坡地與土石流防災預警提供更精確之降雨資訊。
Accurate rainfall information both in spatial and temporal scale is crucial for hydrologic design and a well allocated raingauge network can provide the needed information. In this study, kriging and information entropy theory are used to estimate spatial rainfall distribution and uncertainty to decide the priority of the reallocated raingague network. National Taiwan University Experimental Forest (NTUEF) was selected as the study site. The area is delineated into 346 1 km^2 candidate grids and rainfall records between 1992 and 2009 in 50 existed raingauges are analyzed to demonstrate the proposed scheme. The results show that average influence parameter is shorter in rainy months while the value is comparatively larger in dry months. Only 2 and 5 reallocated raingauge can cover the 62.6% and 85.35% of rainfall uncertainty, respectively. In addition, it reveals that the candidate raingague with higher priority is mostly located in the area with steep terrain relief. The authorities for watershed management can adjust the existed raingauges to get more accurate rainfall information for future hydrologic design and early warning for mitigation of debris flow.
