| 摘要: | 本篇論文主旨在於研究有地形冷池以及無地形風暴兩種不同的初始條件下，如何透過一熱動力反演方法求得所需反演資料，再將反演出來的熱動力場重新置入模式中，進行資料同化的步驟，並藉此探討此一流程是否可行，且對資料同化後的結果有何助益。本研究使用RASTA（Radar Analysis System for Taiwan Area）的三維變分分析，其反演所用的所有動量及熱量方程均以追隨地勢座標表示；模式部分則是使用ARPS（Advanced Regional Prediction System）模式，是由美國CAPS（Center for Analysis and Prediction of Storms）以及奧克拉荷馬大學（University of Oklahoma）所研發的區域氣象模式。 由於真實大氣中缺乏高解析度的熱動力場觀測資料，本篇研究使用高解析氣象模式在理想地形下所模擬之冷池與無地形的風暴系統來驗證RASTA的熱動力反演方法，並估計其誤差。結果顯示，本方法反演的壓力與溫度場，其水平及垂直結構相較於模式結果都有不錯的一致性。 在資料同化結果的部分，將RASTA反演結果帶回ARPS模式中進行資料同化的方法是可行的，且同化後的結果也能模擬出不錯的p′和θ′，不過同化時需要有較準確的p′和θ′。要提高資料同化後結果的準確度，需要先有準確性佳的熱動力場，故在利用三維風場反演熱動力場的部分需要有更精準的計算公式。位溫擾動場θ′對於同化資料的誤差接受度較高，壓力擾動場p′則不是很理想，特別是在p′的SCC值低於0.9以下的時候。本研究的資料同化方法針對兩種不同的天氣系統雖然有顯著的改善結果，但還是有所限制；若初始場的差異過大，儘管有準確的熱動力場資訊，還是無法將兩者之間的差異修正回來。對於總降水量的同化模擬結果並沒有顯著的改變，還是跟每個個案的原始結果相似，推測其可能原因為加入同化的資訊沒有更新水氣場的訊息，故要改善降雨的部分，應該還要同化雷達回波等的資訊，才能達到此目的。 整體來說，個案一(冷池)與個案二(Del City storm)在本研究的資料同化方法中，對於改善p′和θ′的模擬都有不錯的表現。 Under two conditions of a cold pool with terrain and a storm without terrain, this study attempts to establish a thermodynamic retrieval method to get all datasets we needed. After getting all retrieval data, we put back all the thermal retrieval fields (p′ and θ′) into model again and continue doing the simulation. Finally, we discuss whether the procedure is feasible and how much advantages we get from the results in the data assimilation. This research uses RASTA (Radar Analysis System for Taiwan Area) 3-D variational method, and all momentum and thermodynamic equations used in the retrieval method are expressed in a terrain-following coordinate system. The model we use here is called ARPS（Advanced Regional Prediction System）, it is created by CAPS (Center for Analysis and Prediction of Storms) and University of Oklahoma. Due to lack of high-resolution thermodynamic observational data in real atmosphere, this research uses ARPS model to simulate a cold pool with ideal terrain and a storm without terrain, and then tests the performance of this method and estimates the error. The results show that, the thermodynamic fields retrieved from this method have good consistencies in horizontal and vertical structures. For the data assimilation results, putting the RASTA retrieval data back to the ARPS is feasible, and this method can simulate reasonable p′ and θ′ if we have good initial p′ and θ′. In order to enhance the accuracy of the data assimilation results, first we need to have better p′ and θ′. Therefore, a finer computational formula is required in retrieving thermodynamic structure by the use of three-dimensional wind field. The θ′ can accept higher errors in the assimilation data, however, the p′ is not very ideal, especially when the SCC value of p′ is lower than 0.9. Although the data assimilation method in this research has improvement in the results for two different conditions, but there still has limit. If the difference in the initial fields is large, it is difficult to amend the difference even we have correct thermodynamic fields. In the data assimilation of precipitation, the simulation result changes only a little. We guess the possible reason is that the assimilation does not update the information of moisture. Therefore, if we want to improve the rain field, radar reflectivity should be taken into account in the assimilation procedure. Over all, the data assimilation method in this research improves both the p′ and θ′ in both case one (cold pool) and case two (Del City storm). |
