中央氣象局全球模式(CWB GFS)引進美國NCEP全球模式使用之SAS(Simple Arakawa Schubert)積雲參數化方案，希望藉由該積雲參數法對雲水逸出的考慮，增加模式雲水收支計算的完整度。同時並更新模式邊界層參數法原使用的局部(local)方案，以非局部(nonlocal)方案取代，嘗試解決在使用新版積雲參數法後，所發現之模式大氣底層偏濕及因而產生的模式雲底偏低現象。本研究針對此二項模式更新，進行冬夏各二個月資料同化數值預報實驗，以評估其對全球模式預報的影響，論文評估的重點放在預報之大範圍系統性偏差及統計預報技術得分的變化。結果顯示，無論在高度場的距平相關得分，或是溫度場及風場的均方根偏差，新版模式在南北半球均有顯著的進步；其中第五天預報之距平相關得分進步達2%-4%，模式原本預報過高之全球平均降水量，亦得到改善。

The SAS (Simple Arakawa Schubert) cumulus parameterization scheme from NCEP GFS is introduced to Global Forecast System at Central Weather Bureau (CWB GFS). The detrainment of cloud water on cloud top in SAS scheme is expected to have improvement in the prognostic calculation of cloud water. The local pbl (planetary boundary layer) scheme applied in CWB GFS is also upgraded by a nonlocal version from NCEP GFS in order to decrease moist bias in the lower levels by transporting moisture away from surface more efficiently. Data assimilation experiments are conducted to assess the impact of new pbl and cumulus parametrization schemes. Assessment is focusing on forecast skill score and root mean square error statistics. Results show new packages have significant improvement on forecast skill including anomaly correlation (AC) and root mean square errors for wind field and temperature field. Quantitatively, the AC improvement on the forecast day 5 can reach 2%-4%. The overestimates of forecast precipitation are also reduced.