本篇論文第一部份是利用VDRAS的一系列OSSE實驗和真實個案模擬，來探討同化策略和微物理過程對都卜勒雷達資料同化於定量降雨預報的影響。 OSSE實驗指出直接使用模式輸出且帶有錯誤天氣系統為背景場，VDRAS預報會受小尺度錯誤天氣資訊所影響，而產生不必要零星降雨。若使用平滑後的背景場則可去除小尺度錯誤訊息，以改善降雨預報。當低層無雷達觀測資料情況，在缺乏低層資料的區域，VDRAS反演出過強的下降運動和冷池。越多同化循環未必得到較佳的分析和預報。故當低層資料缺乏時，同化循環的數目須審慎選擇。OSSE實驗結果顯示兩個同化循環設計分析和預報上有相對較佳結果。微物理非線性特性隨預報時間的發展，對同化系統表現有明顯負面影響，會導引極小化過程往錯誤方向。其中回波有較徑向風顯的非線性特性。使用多個短循環的同化策略，可幫助同化系統的極小化演算法得到較佳的初始場和預報結果。真實個案方面，則是選擇2008 SoWMEX IOP#8期間的一個強降雨個案，以驗證OSSE的實驗結果。 第二部分則主要是改進VDRAS的微物理過程的研究。吾人在維持VDRAS原本的暖雨微物理架構下加入含雪和冰的冰相微物理過程。OSSE實驗結果顯示，若用原本只有暖雨過程的VDRAS同化回波，會造成分析場的雪水量嚴重被低估。有冰相過程的VDRAS同化雷達資料時，若利用探空的結冰高度用來界定被同化回波中雨和雪，則用兩個4Dvar同化循環設計較單一循環可得到較合理分析場結果。OSSE實驗和2008 SoWMEX IOP#8真實個案的降雨預報分布上，增加冰相微物理可改善VDRAS只有暖雨過程所造成降雨過強且集中現象，使得降雨範圍較為廣。目前定量降水預報比較，冰相過程則對小雨的降雨有較明顯幫助。 A series of observation system simulation experiments (OSSEs) and real case study are conducted to investigate the application of the Doppler radar data assimilation technique for numerical model quantitative precipitation forecasts (QPF). A four-dimensional Variational Doppler Radar Analysis System (VDRAS) is adopted for all experiments. The first set of OSSEs demonstrates that when the background field contains the imperfect information predicted from a mesoscale model, the incorrect convective-scale perturbations in the background can result in spurious scattered precipitation. However, a smoothing procedure can be utilized to remove the fine structures from the primitive model output to avoid this over-prediction. Results from a second set of OSSEs indicate that the lack of low-elevation data due to beam blockage could significantly alter the retrieved low-level thermal and dynamical structures when different number of data assimilation cycles is applied. These impacts could lower the rainfall forecast capability of the model. The third set of OSSEs shows that, when the rainwater is assimilated over a long assimilation window, the nonlinearity embedded in the microphysical process could lead the minimization algorithm to a wrong direction, causing a further degradation of the rainfall prediction. However, using multiple short assimilation cycles produces better minimization and forecast results than those obtained with a single long cycle. A real case experiment based on data collected during Intensive Operation Period (IOP) #8 of the 2008 Southwest Monsoon Experiment (SoWMEX) is conducted to provide a verification of the conclusions obtained from OSSEs under a realistic framework. The microphysics scheme of VDRAS is extended from warm rain process to cold rain process. It is found that the retrieved water content would be underestimated if all radar reflectivities are assumed to be in the form of warm rain. This underestimation can be improved when the cold rain process is implemented into VDRAS. The VDRAS with ice physics can provide better rainfall forecast.
