本文介紹一針對都卜勒雷逹觀測所設計的資料同化方案,並探討利用該方案將雷逹資料及其所反演的大氣狀態變數同化進入數值模式後,對改善劇烈天氣預報的可行性。此同化方案的流程共包含三部份:多都卜勒雷達風場反演、熱動力反演、及水氣調整。 為了能詳細地探討本方法的性質,研究中所做的實驗都是在Observation System Simulation Experiment (OSSE)的架構下進行,實驗內容包含:(1)研究方法完整流程的測試;(2)無探空資料輔助的同化實驗與同化次數測試;(3)有探空資料輔助時的同化策略。 OSSE實驗結果顯示,經過同化都卜勒雷逹觀測資料及其反演的大氣狀態變數後,的確可降低初始場中因擾動太過微弱所導致降水不足的預報誤差。無探空資料時,利用模式當時的預報場來當成額外的輔助資料來源,仍然可以改善對降水的預報。在同化次數的測試中,則發現同化次數較多的實驗結果較佳,但其效果要在同化結束後一段時間(約1小時)才會逐漸開始顯現。在同化過程中,給予探空資料可有效提升同化結果的準確度,而在此同化之後,如再進行一次沒有探空資料的同化,於預報結果會有更進一步的改善。
A data assimilating algorithm designed for Doppler radar observations is introduced in this manuscript. The feasibility of using this algorithm to improve the forecast of severe weather by assimilating radar observed and retrieved atmospheric state variables into a numerical model is investigated. This assimilation algorithm consists of three components. They are: multiple-Doppler radar wind synthesis, thermodynamic retrieval, and moisture adjustment. In order to study the performance of this method in details, all experiments are conducted under the Observation System Simulation Experiment (OSSE) framework. The experimental designs include: (1) Test of the entire assimilation algorithm; (2) Assimilation experiments without the auxiliary from an extra radiosonde, and tests of the assimilation numbers; (3) The assimilation strategy when radiosonde data are available. Experimental results show that the forecast errors introduced by incorrect initial condition can be reduced by assimilating Doppler radar observed and retrieved parameters into the model. Without sounding data, it is still possible to improve the quantitative precipitation forecast (QPF) using model-generated fields as an extra data source. More data assimilation produces better results, but the improvement won't appear until about one hour after the assimilation is completed. If the assimilation is combined with an extra sounding observation, the accuracy of the model forecast can be upgraded efficiently. After this, if one more assimilation is conducted, even without the information from an extra sounding, the model forecasts can be improved further.