由於3DVAR的背景誤差協方差是不具有flow-dependent的特性，但一般認為，如果背景誤差協方差具有流場相依的特性，可能會比較符合大氣變數的實際狀況，於是本篇論文研究挑選了兩個具有flow-dependent特性的同化系統，EAKF是屬於系集卡曼濾波器的一種，因為其背景誤差協方差是利用完全非線性模式預報出來的背景場計算得到的，所以具備了明顯的flow-dependent的特性，4DVAR則是四維變分方法的一種，由於其cost function加入了時間的維度(所以稱為四維)，所以相較於3DVAR來說，具備了隱性flow-dependent的特性。本篇論文於是想要比較這兩種同化系統的表現。
主要研究實驗計劃是在2010101900開始進行同化，由於NCEP全球分析場資料已經包含了GPSRO和GTS觀測資料在其中，為了避免分析場本身已有的觀測資訊干擾實驗，也為了讓初始場跟模式進行一定的平衡，所以在EAKF方面，先在2010101818進行預報六小時，把此預報場當作初始猜測場進行同化，再把每同化完的 Ensemble Mean進行72小時的模擬預報，另外把每個member進行六小時預報當作下個階段的背景場以及初始猜測場，再進行同化得到分析場。依序在10/19/00、10/19/06、10/19/12/、10/19/18、10/20/00，進行五次的同化，4DVAR則是先在2010101818進行預報三小時，把此時的預報場當作是初始猜測場，進行六小時窗區的資料同化，得到分析場，把分析場預報3小時得到10/19/00的初始場，利用此初始場進行72小時的模擬預報，再把同化後的分析場進行六小時的預報得到下個階段的初始猜測場，在進行資料同化，依序在10/18/21、10/19/03、10/19/06/、10/19/15、10/19/21，進行五次的同化。
研究發現，EAKF、4DVAR經過cycle run的同化策略，颱風路徑的表現上EAKF在經過多個cycle窗區之後，有慢慢把資料同化的效果表現出來，4DVAR則較不一定，在颱風強度的部分，4DVAR在五個cycle窗區裡面，颱風強度整體來說是偏低的，在觀測資料部分發現EAKF因為是系集初始場的關係，較可以表現出GPSRO掩星觀測的效益，4DVAR則是比較跟觀測資料的遠近有關，在環境場的表現，EAKF整體來說都要比4DVAR還要來的優異。
針對EAKF進行了同化策略差異的實驗，以及系集數的敏感度測試，發現經過spin up 72小時的EAKF在颱風路徑會比spin up 6小時的EAKF有較好的模擬，但颱風強度有低估的情形。在系集數的敏感度測試中，發現當系集數增加不夠明顯時，對背景誤差斜方差的估計改善程度並不會有多大的幫助。

The background error covariance (COV) in 3DVAR does not have the flow-dependent quality. It will have a higher possibility to fit better to the realistic atmosphere data if the COV in other systems is flow-dependent. This study selects two data assimilation systems with the flow-dependent quality. EAKF is a kind of ensemble Kalman filter with the background error based on a calculation of various member spreads predicted by the perfect nonlinear model, therefore contributing to the flow-dependent quality. On the other hand, 4DVAR is a kind of four-dimensional variational method based on the model adjoint. Compared to 3DVAR, 4DVAR contains an implicit flow-dependent quality. In this study, we attempt to compare the performances of these two assimilation systems with the WRF model.
The first assimilation starts at 0000 UTC 19 Oct. 2010 as the Typhoon Megi is west of Philippines. Due to that fact that the NCEP global analysis has already contained the observations of GPSRO and GTS, EAKF has been conducted for six hours in order to avoid the influences of GPSRO and GTS, which also allows the initial field to have a better balance with the model, and then this forecast field is used as the initial field to proceed the next assimilation. After each cycle, the model simulation is integrated for 72 hours using the ensemble mean for the EAKF experiments.
For the 4DVAR experiments, the WRF model with the NCEP analysis at 1800 UTC 18 Oct. 2010 is integrated for three hours to obtain the initial field as the background with assimilation of observation data in next six hours to obtain the analysis field which is then used as the initial field for a simulation of three hours. The simulated field is then integrated for 72 hours for comparison with EAKF experiments. In total, there are five cycles for both EAKF and 4DVAR, separated by six hours.
On verification of the analyses, EAKF is found to be better than 4DVAR. The model results after cycle run show that the simulated typhoon track from EAKF is improving as cycling increases, however the 4DVAR performances are less improved at later cycles. EAKF also gives a weaker intensity of Megi than 4DVAR. For impact of observation data, EAKF shows bigger benefits of GPSRO in typhoon predictions than 4DVAR.
For the sensitivity of spin-up period in EAKF experiments, it was found that the simulated typhoon track is improved as the spin-up of 6 h for the EAKF experiments is increased to 72 h. However, an increase in the ensemble members doesn’t lead to significant improvement on both model analysis and typhoon forecasts.

巫佳玲，2011年:利用WRF 3DVAR 與EAKF 探討GPSRO 資料同化對莫拉克颱
風模擬之影響。國立中央大學，大氣物理研究所，碩士論文，134頁。
張志謙，2013年:利用系集重新定位法改善颱風路徑預報-2011年南瑪都颱風
個案研究。國立中央大學，大氣物理研究所，碩士論文，93頁。
吳俊澤，2007年:利用MM5 4DVAR 模式同化掩星折射率資料及虛擬渦旋探討
颱風數值之影響。國立中央大學，大氣物理研究所，碩士論文，70頁
吳家苓，2008年:FORMOSAT-3 GPS RO 觀測資料之校驗及其對東亜地區梅雨
季天氣分析與預報之系統性影響。國立中央大學，大氣物理研究所，碩
士論文，106頁。
林妗庭，2012年:WRF 4DVAR 虛擬渦旋及雷達資料同化對於颱風凡那比(2010) 數值模擬之影響。國立中央大學，大氣物理研究所，碩士論文，92頁。
黃振星，2011年:同化FORMOSAT-3/COSMIC 及Follow-on 掩星觀測資料對颱
風預報之影響。國立中央大學，大氣物理研究所，碩士論文，108頁。
陳舒雅，2008年:GPS 掩星觀測資料同化及對區域天氣預報模擬之影響。國
立中央大學，大氣物理研究所，博士論文，154頁。
陳怡孜，2013:GNSS RO 觀測資料對颱風莫拉克預報之影響:觀測系統模擬實
驗。國立中央大學，大氣物理研究所，碩士論文89頁。
交通部中央氣象局委託研究計畫期末成果報告地形影響颱風異常變化之探
討:郭鴻基:計畫編號：MOTC-CWB-100-M-02
江琇瑛 技士101 年9月:臺美氣象先進資料同化與預報模式系統發展技術合
作協議-系集調整卡爾曼濾波( Ensemble Adjustment Kalman Filter)
同化系統
江琇瑛 技士100 年12 月臺美氣象先進資料同化與預報模式系統發展技術
合作協議-系集調整卡爾曼濾波( Ensemble Adjustment Kalman
Filter)同化系統
林昀瑱 國防訓儲人員97年11月12日:臺美氣象先進資料同化與預報模
式系統發展技術合作協議－系集調整卡爾曼濾波( Ensemble
Adjustment Kalman Filter)
Andrew, C. Lorenc* and F. Rawlins Met Office, UK. Why does 4D-Var beat
3D-Var? (Received 16 May 2005; revised 13 October 2005)
Caya, A., J. Sun, and C. Snyder, 2005: A Comparison between the 4DVAR
and the Ensemble Kalman Filter Techniques for Radar Data Assimilation. Mon. Wea. Rev., 133, 3081–3094.
Evaluation of the operational 4D-Var at the Meteorological Service of Canada S.
Laroche*, P. Gauthier, M. Tanguay, S. Pellerin, J. Morneau Meteorological Service of Canada.
Houtekamer, P. L. , and Herschel L. Mitchell, 1998: Data Assimilation
Using an Ensemble Kalman Filter Technique. Mon. Wea. Rev., 126,
796–811.
Houtekamer, P. L., Herschel L. Mitchell, G. Pellerin, M. Buehner, M. Charron, L.
Spacek, and B. Hansen, 2005: Atmospheric Data Assimilation with an Ensemble Kalman Filter: Results with Real Observations. Mon. Wea. Rev., 133, 604–620.
Huang, X.-Y., Q. Xiao, D. M. Barker, X. Zhang, J. Michalakes, W. Huang, T.
Henderson, J. Bray, Y. Chen, Z. Ma, J. Dudhia, Y. Guo, X. Zhang, D.-J. Won, H.-C. Lin, and Y.-H. Kuo, 2009: Four-Dimensional Variational Data Assimilation for WRF: Formulation and Preliminary Results. Mon. Wea. Rev., 137, 299–314.
Jeffrey, L. Anderson, 2001: An Ensemble Adjustment Kalman Filter for Data
Assimilation. Mon. Wea. Rev., 129, 2884–2903.

Meng, Z. , and F. Zhang, 2007: Tests of an Ensemble Kalman Filter for Mesoscale and
Regional-Scale Data Assimilation. Part II: Imperfect Model Experiments. Mon. Wea. Rev., 135, 1403–1423.
Meng Z. , and F. Zhang, 2008: Tests of an Ensemble Kalman Filter for Mesoscale and
Regional-Scale Data Assimilation. Part III: Comparison with 3DVAR in a Real-Data Case Study. Mon. Wea. Rev., 136, 522–540.
Meng Z. , and F. Zhang, 2008:Tests of an ensemble Kalman filter for mesoscale and
regional-scale data assimilation. Part IV: Comparison with 3DVAR in a month-long experiment.Mon. Wea. Rev. ,136 (10), 3671-3682
Xiang-Yu Huang*? Qingnong Xiao, Wei Huang, Dale Barker, John
Michalakes, John Bray, Zaizhong Ma, Yongrun Guo, Hui-Chuan Lin,
Ying-Hwa Kuo National Center for Atmospheric Research, Boulder,
Colorado Preliminary results of WRF 4D-Var
Zhang, F. , Z. Meng , and Altug. Aksoy, 2006: Tests of an Ensemble Kalman Filter for
Mesoscale and Regional-Scale Data Assimilation. Part I: Perfect Model
Experiments. Mon. Wea. Rev.,134, 722–736.
Zhang, F. , Z. Meng , and J. Poterjoy, 2013: E3DVar: Coupling an Ensemble Kalman
Filter with Three-Dimensional Variational Data Assimilation in a Limited-Area Weather Prediction Model and Comparison to E4DVar. Mon. Wea. Rev., 141, 900–917.
Zhang, M. , and F. Zhang, 2012: E4DVar: Coupling an Ensemble Kalman
Filter with Four-Dimensional Variational Data Assimilation in
a Limited-Area Weather Prediction Model. Mon. Wea. Rev., 140,
587–600.
Zhang, M. , F. Zhang, X.-Y. Huang , and X. Zhang, 2011: Intercomparison of an
Ensemble Kalman Filter with Three- and Four-Dimensional Variational Data
Assimilation Methods in a Limited-Area Model over the Month of June 2003. Mon. Wea. Rev., 139, 566–572.