| 摘要: | 本博士論文旨在利用系集卡曼濾波器，同化福爾摩沙衛星三號掩星觀測資料所反演之電離層電子濃度剖面，以結合美國國家大氣研究中心高層大氣實驗室所發展的熱氣層與電離層耦合環流電動模式，求得熱氣層與電離層的最佳化狀態。福衛三號提供高空間、時間解析度之電子濃度剖面，接近均勻全球分布的觀測資料，有助於重建全球電離層電漿三維結構。熱氣層與電離層耦合環流模式是透過數值方法解析非線性物理方程式，模擬熱氣層與電離層中的中性大氣與電漿的分佈與特性。美國國家大氣研究中心所發展的資料同化開發模組是一套以系集卡曼濾波器為架構的開放軟體，能夠簡化資料同化的開發過程。
於本耦合模式中，由於熱氣層與電離層的狀態受到外在環境參數控制，因此無法僅以擾動模式初始場的方式產生系集樣本，必須進一步以擾動環境參數的方式產生具有代表性的系集樣本。為了瞭解模式中主要的環境參數與模擬後之電離層狀態相關性，首先選定太陽輻射強度、高緯度電位強度、大氣潮汐變化等模式中的主要環境參數進行相關性研究。結果發現電離層電子濃度變化受到大氣潮汐中的全日潮與半日潮的影響相對較小，主要仍然是受到太陽輻射強度與高緯度電位強度的作用。有鑑於此，本論文所使用的系集樣本，將針對太陽輻射強度與高緯度電位強度以高斯擾動的方式的產生。此外，為了讓系集樣本的電子濃度剖面除了在電離層層峰濃度有顯著變化外，也能在電離層層峰高度上有著明顯的差異，更引進了一個倍增參數調整全球的電磁交互作用漂移，使得系集樣本中的電離層層峰高度亦能夠具有顯著的差異性。
在同化福衛三號真實觀測資料之前，先建立觀測系統模擬實驗來設定環境參數與檢驗系集卡曼濾波器同化熱氣層與電離層耦合模式之表現。實驗結果顯示，本同化系統需要採用90個以上的系集樣本才能夠顯著降低均方根誤差，並表現出系統特徵。為了降低採樣誤差，除了在水平方向設定限地化參數，也須要在垂直方向上設定對應之限地化參數，並規範各個模式狀態的極值範圍以避免同化結果出現不具物理性的數值。模擬實驗結果更指出即使系統中只有同化電離層觀測資料，但透過系集卡曼濾波器亦可同時調整模式內其他不具觀測資料之中性大氣參數。
在完成觀測系統模擬實驗之相關測試後，真實福衛三號電離層觀測資料即可同化至系統中進行相關研究。福衛三號自2006年4月發射升空後，任務主要涵蓋太陽活動極小時期。為了評估同化系統在不同地磁活動強度下的表現與差異，於衛星進入任務軌道完成佈局後，分別選擇一個地磁寧靜時期與一個地磁擾亂時期進行分析。除此之外，為了檢視同化觀測資料後所出現的現象是否真實存在於電離層中，同化結果將與地面電離層雷達觀測與全球二維全電子含量相互比較。結果發現模式在觀測資料的輔助之下，能夠重現電離層中時空尺度較小的電漿結構，並在接近春分時期呈現出南北半球不對稱的現象。然而，在同化結果中仍然有部分違背物理規範的現象發生，且有多數發生於夜間的福衛三號觀測資料在同化過程中遭到此同化系統排除在外。這些仍然需要透過相關的研究分析並且逐步改善模式內部的物理方程式或參數設定，才能在未來避免這些情形發生。
由於福衛三號任務在國際間獲得廣大迴響與正面評價，台灣國家太空中心規畫於2016年開始分兩階段發射福爾摩沙七號衛星，以接替已逐漸接近衛星設計年限的福衛三號，以延續中性大氣與電離層之掩星觀測任務。在福衛七號的任務規畫中，共計將有12顆衛星分布於兩組軌道面，接收三組不同國家的全球導航系統衛星之訊號，估計可提供的資料量約為現今福衛三號所能提供的四倍。在大量觀測資料的優勢下，可在短時間累積觀測資料重建全球電離層三維電漿結構，縮短電離層資料同化的時間間隔。從福衛七號的觀測系統模擬實驗結果中發現，福衛七號發射後將有助於研究電離層中時空尺度較小的結構與變化，有效監測太空天氣變化與強化預報結果。
本論文內容涵蓋此同化系統自發展初期測試結果與同化實際觀測資料後的現象研究，並進一步驗證同化結果的真實性與評估福衛七號未來在科學研究上的重要性。未來更待相關的研究分析與進一步改善，讓此熱氣層與電離層資料同化系統有機會能夠成為真正用來監控與預報太空天氣的作業系統。
This dissertation presents efforts to assimilate FORMOSAT-3/COSMIC (F3/C) GPS Occultation Experiment (GOX) observations into the National Center for Atmospheric Research (NCAR) Thermosphere Ionosphere Electrodynamics General Circulation Model (TIE-GCM) by means of ensemble Kalman filtering (EnKF). The F3/C electron density profiles (EDPs) globally distributed which provide an excellent opportunity to study the three-dimensional (3D) ionospheric electron density structure. NCAR TIE-GCM simulates the Earth's thermosphere and ionosphere by self-consistently solving for the coupled nonlinear equations of hydrodynamics, neutral and ion chemistry, and electrodynamics. F3/C EDPs are combined with the TIE-GCM simulations by an EnKF algorithms implemented in the NCAR Data Assimilation Research Testbed (DART) to obtain 'the best' estimate of the current ionospheric state.
Since the thermosphere and ionosphere are dominated by external forces, it is difficult to generate ensemble members simply by perturbing the model initial conditions. First of all, model sensitivity experiments are designed to examine the model responses associated with forcing parameters such as the solar 10.7cm flux, the hemispheric power, the cross-tail potential, and the phase and amplitude of migrating tides. Furthermore, multiplying factors of the vertical E × B drift are introduced to diversify peak height altitude of electron density distribution over height. According to the model sensitivity experiments, the ensemble generation strategy is decided for following comprehensive EnKF study.
Observing system simulation experiments are conducted in the primary stage to investigate the optimal ensemble size, suitable horizontal and vertical localization lengths, and performance of assimilation system for F3/C EDPs as well as approaches to prevent unphysical values of state vectors. The results suggest that this EnKF assimilation system for coupled thermosphere-ionosphere model can be effectively applied to assimilate the F3/C EDPs by using 90 ensemble members, setting the vertical localization length equal to 1,000 km with additional simple bounds for particular state vectors.
Assimilation analyses obtained with real F3/C EDPs during geomagnetic quiet time and disturbed conditions are further compared with independent ground-based observations and global ionospheric maps as well as the F3/C profiles themselves. The comparison shows the improvement of the primary ionospheric parameters such as NmF2 and hmF2 and furthermore suggests that some small scale physically meaningful structures that are absent in TIE-GCM simulations are resolved. Nevertheless, some unrealistic signatures appear in the results, and high rejection rates of observations due to the outlier threshold and quality control are also found in the assimilation experiments. These issues are analyzed to identify the model biases, for example, the nighttime O+ flux at the upper boundary in TIE-GCM is decreased to reduce the nighttime rejection rates.
In the near future, the FORMOSAT-7/COSMIC-2 (F7/C2) constellation will be deployed in 2016 and provide more than 12,000 occultation soundings per day. An additionally OSSE is conducted in this dissertation to determine the impact of F7/C2 on ionospheric weather monitoring and forecasting. Results show that the denser F7/C2 observations can reconstruct 3D ionospheric structure with a dramatically shorter data accumulation period, and significantly reduce the assimilation analysis error. The results also represent a major advance in ionospheric weather monitoring of the F7/C2 mission.
Finally, this dissertation presents comprehensive investigation of the DART/TIE-GCM assimilation system for the F3/C ionospheric observations. There are a number of promising results as reported herein. However, it still needs more investigations and analysis to improve the system to be used as an operational system for monitoring and forecasting the ionospheric space weather in the future. |
