本研究主要是針對WRF模式12種物理參數法組合，就2004年梅雨季(2004年5月15日~6月15日)，以美國NCEP(National Center Environment Prediction)GFS全球模式分析場為初始場，進行連續32天、每天2次，每次48小時的數值預報；並利用東亞地區探空店觀測之氣象場(包括高度場、溫度場和水氣場)進行校驗，以評估WRF(Weather Research and Forecast model)模式不同物理參數法組合在東亞地區的預報能力。 分析結果顯示，模式預報誤差特性不因不同解析度(45公和15公里)和預報時間長度而有明顯差異；選擇不同邊界層參數法對模式預報誤差影響最顯著、積雲參數法之次之，微物理參數法更次之。低對流層水氣和溫度場之預報誤差的垂直分佈，因選用不同邊界層參數法而顯著的差異，並連帶反映在高度場預報誤差的垂直結構。 本研究的實驗範疇中，就探空站上的氣象場校驗而言，當應用WRF模式於東亞地區進行數值天氣預報時，總結建議以YSU+WSM5+KF為最佳的物理參數法組合。

This paper presents the verification over radiosound observations of the Weather Research and Forecasting (WRF) model during the 2004 Mei-yu season. Twelve members of WRF simulations, with different physics combinations, were run twice daily using the initial condition form NCEP (National Center for Environmental Prediction) GFS model from 15 May to 15 June 2004. The verification results show that the model forecast error tended to be insensitivity to the model resolution and forecast periods. The choice of PBL scheme is the most significant to the model forecast, cumulus scheme the second, and the microphysics the third. Different PBL scheme results in different vertical profile of temperature and moisture fiels within boundary layer, and then project the vertical structure of the forecast of geopotential height. It overall, the best choice of the combination of the physical parameterization package in WRF model applying over southeast Asia area is suggested to be YSU+WSM5+KF.