研究中使用四種微物理參數化方案: Goddard (GCE)、WRF SM 6-category (WSM6)、WRF DM 6-category (WDM6)與Morrison (MOR)，四種方案皆進行系集預報，欲瞭解不同微物理參數化方案於系集預報中不確定性之表現。利用2008年6月16日台灣北部熱對流個案，探討在對流成熟期時，強對流區之系集離散程度與協方差，討論不同微物理參數化方案間之不確定性差異。不同參數化方案模擬結果顯示，GCE有量值最顯著的冰相混合比，因此回波發展最高；在低層暖雨過程中，雖然WDM6有最大的雨水混合比，但回波卻是最弱的，而MOR之雨水混合比並非特別顯著，但回波強度卻是最大的，其原因為WDM6的預報產出大量的粒子數量，而MOR的粒子數量則是最少，因此導致上述的回波特徵。此結果顯示使用雙矩量微物理方案時，不可忽視粒子數量所帶來的影響，而雨滴粒子數量不只影響回波，降雨與蒸發效率也可能因為不同雨滴大小、數量而有不同。以系集法(Ensemble-based method，Houtekamer et al. 1996)估計背景場之不確定性，根據不同微物理參數化設定，方差分布也有不同的特徵，GCE之系集預報在高層有較大的系集離散度，WDM6的最大系集離散度則在低層，而MOR則是出現在融化層附近。研究中發現GCE在冰相有較多不確定性，WDM6則在暖雨過程離散程度較大，因此在進行系集資料同化時，考慮計算資源受限的情況，推測使用GCE與WDM6可在有限的系集個數下，有效地增加系集間的離散度。研究中亦討論背景場之誤差相關性，在MOR的回波自相關中也可以看到粒子數量所帶來的影響，使得其背景誤差相關性在融化層附近較弱。此外，對流區之垂直風速與潛熱釋放作用具有高度相關性。

To understand the characteristics of different microphysics schemes and investigate the forecast uncertainty structure in very short-term forecast, four microphysics schemes are used in the study. They include two single-moment schemes: Goddard (GCE)、WRF SM 6-category (WSM6), and two double-moment schemes of WRF DM 6-category (WDM6) and Morrison (MOR). A thunderstorm case in northern Taiwan on June 16, 2008 is selected. The results show that GCE has the most ice-related mixing ratio, so the reflectivity development is the highest. In the low-level warm rain process, WDM6 (MOR) has the most (fewest) rain mixing ratio and the weakest (strongest) reflectivity due to large (small) number of rain total number concentration. It is found that when using the double-moment microphysics scheme, the influence of the total number concentration cannot be ignored. According to different microphysics scheme settings, the variance also has different characteristics. With the same ensemble members (36), it is found that GCE (WDM6) has more uncertainty in ice-related processes (warm rain processes). Therefore, using combination of these two schemes can effectively increase ensemble spread and improve the benefits of data assimilation. The error correlation between different variables is also discussed in the study. In the convective zone, the vertical wind and the latent heat release are highly correlated. In addition, the reflectivity auto-correlation in MOR is greatly affected by the number of particles around melting layer.