本研究針對Cressman客觀分析法，分別就影響半徑、修正次數以及測站與格點關係進行數值分析，說明逐次修訂法如何改善Cressman 客觀分析的效能，以及系統性地探討如何建構一個有效且快速的設計，並提出其之所以可行的理由。同時將研究結果，評估其在網格化臺灣地區自動雨量觀測資料之效能。研究結果指出，Cressman逐步修正法將使測站的觀測增值逐步收斂為0，影響半徑會決定觀測增值收斂的速度，影響半徑越小，觀測增值和分析場收斂的速度越快，可更快在分析場中得到觀測所呈現的極值。Cressman 逐步修正法無可避免地會導致過度內插（overshooting）的問題發生，此一過度內插的問題和觀測點與格點位置的分佈有關。影響半徑越小，越少的疊代次數即可能出現過度內插的現象。當觀測點越接近分析格點時，分析值越接近觀測值，因此過度內插的問題會受到抑制。最佳的Cressman 逐步修正法應考慮在修訂過程中逐次減少影響半徑，如此可取得觀測所掌握的大尺度結構，並在分析場中得到觀測所呈現的極值，同時也能有效抑制過度內插的問題。

This study is to investigate the impact factors of the Cressman objective analysis method to grid the rain gauge observations from Automatic Rainfall and Meteorological Telemetry System managed by Central
Weather Bureau. The impact factors include the influence radius, number of successive correction, and the distance between the observation and grid point. The results show that the successive correction will result in
the observation innovation to converge to 0, and the overshooting problem. The influence radius will speed up the convergence, while the longer distance between the observation and grid point and the larger observation gradient will enhance the overshooting problem. The optimal solution to apply the Cressman objective analysis method is to reduce the influence radius during the successive adjustment to capture the larger scale structure, the observation extreme value, and well control the overshooting problem.