查詢結果分析
來源資料
頁籤選單縮合
題名 | 中央氣象局四維變分資料同化(4DVAR)之初步應用與評估=The Preliminary Assessment of Four-dimensional Variational Data Assimilation (4DVAR) Performance at Central Weather Bureau |
---|---|
作者 | 江晉孝; 張昕; 馮欽賜; 黃向宇; Chiang, Chin-hsiao; Zhang, Xin; Fong, Chin-tzu; Huang, Xiang-yu; |
期刊 | 氣象學報 |
出版日期 | 20140700 |
卷期 | 51:1 2014.07[民103.07] |
頁次 | 頁55-80 |
分類號 | 328.88 |
語文 | chi |
關鍵詞 | 變分資料同化; 與流場相關的背景場誤差統計特性; 颱風路徑預報; 多重增量; 格點校驗; 公正預兆得分; 偏離指數; Variational data assimilation; Flow-dependent; Typhoon-track forecast; Multi-incremental; Grid-point verification; |
中文摘要 | 中央氣象局 WRF模式目前使用之資料分析方法為三維變分資料同化 (Three-Dimensional Variational Data Assimilation, 3DVAR),此技術是假定觀測資料與模式控制變數皆屬同一時間,優點為所需之計算機資源較低,但其背景誤差協方差矩陣不隨時間變化與無法同化不同時間的觀測資料則為極大之劣勢。配合中央氣象局新一代超級電腦之建置,大幅提升的計算機資源,讓我們能嘗試運用四維變分資料同化 (Four-Dimensional Variational Data Assimilation, 4DVAR)技術評估與分析其對 WRF模式預報的效益。 研究結果發現, (1)由單觀測點實驗證明 4DVAR確實擁有與流場相關 (flow dependent)的背景場誤差統計特性,由實驗中亦了解觀測點位置與選取個案為影響與流場相關特性顯著程度之重要因素。 (2)在 2008年 (THORPEX Pacific Asian Regional Campaign¸ T-PARC)實驗期間,格點校驗結果顯示 4DVAR相較於 3DVAR得到略差的表現,但在颱風路徑預報上,則呈現優劣互現的結果。 (3)在評估同化如雷達等時空密集觀測資料能力方面, 4DVAR比3DVAR更能掌握較佳的降水分布型態與極值,而由公正預兆得分 (ETS)及偏離指數 (Bias)校驗結果, 4DVAR亦得到較高的得分值與較佳的偏離值,顯示4DVAR較有潛力與優勢在未來同化時空密集度較高的降水資料。 (4)在效能評估方面,由實驗顯示,多重增量 (Multi-incremental)4DVAR能在30分鐘內完成 45公里解析度 (東亞區域 )之同化過程,而在 5公里的高解析度網格 (臺灣區域)之設定下則耗時約 50分鐘即可完成同化流程,與 Full resolution 4DVAR相比較能符合作業上之需求與效益。 |
英文摘要 | The current data assimilation technique used by the Central Weather Bureau (CWB) in Taiwan is the three-dimensional variational data assimilation (3DVAR). Although 3DVAR is computationally economical in operational NWP, its background error covariance (BE) defined as static matrix and which can not evolve with the time is the major weakness. Therefore, all to-be-assimilated observations have to be presumed as the snapshot of the current atmospheric state. Recently, the CWB has upgraded its computational resources significantly; and the opportunities arise to test the advanced data assimilation methods, such as four-dimensional variational data assimilation (4DVAR), which has the capability to take into account the observation data within a certain timeframe. In this paper, the impact of 4DVAR on the Weather Research and Forecasting (WRF) model in the context of the CWB operational configuration has been evaluated systematically. The comparisons between the 3DVAR and 4DVAR experiments are summarized as follows: (1) The single-observation experiment confirms that the 4DVAR is able to evolve the BE implicitly and demonstrates the flow-dependent characteristics on the analysis; (2) With the data observed during 2008 THORPEX Pacific Asian Regional Campaign (T-PARC), the 4DVAR experiment preliminary results can not compete with which by the 3DVAR on grid-point verifications, but 4DVAR experiment demonstrates comparable typhoon-track forecasts with those of 3DVAR; (3) The 4DVAR experiment with high horizontal resolution usually produces better precipitation forecasts in terms of the pattern and peak value. This implies that the high temporal-and-spacial-resolution related observations, such as radar data, may be better assimilated in 4DVAR; (4) The multi-incremental 4DVAR is able to complete the analysis on the 45km horizontal-resolution CWB operational domain (East-Asia) in 30 minutes (2 outer loops), and complete the 5km Taiwan domain in 50 minutes. The computational costs of 4DVAR are affordable for operational purpose given the CWB computational resources. |
本系統之摘要資訊系依該期刊論文摘要之資訊為主。