| 摘要: | 海洋是個非常巨大的能源儲存庫，對熱帶氣旋而言海洋提供能量的多寡將決定其強度的變化趨勢。為了能夠預測熱帶氣旋的強度，除了海表溫度之外還需要上層海洋的溫度結構來判斷海洋熱含量的多寡。目前的衛星觀測技術能夠快速地測量到海表溫度，但海洋的溫度結構卻無法如海表溫度一般快速地獲取，增加獲取海洋溫度結構的方法是當今重要的課題。
本研究使用Argo Floats與衛星數據在西北太平洋和南海建立回歸模型，利用海表面高度距平(Sea Surface High Anomaly，SSHA)來估計海面下的垂直溫度結構。本研究建立了三組不同的回歸模型，分別為：(1) 使用2000~2008年颱風季節(5月~10月)的Argo Floats建立回歸模型(簡稱Pun2014)；(2) 使用2000~2017年颱風季節的Argo Floats建立回歸模型(簡稱Y18_FW)；(3) 改善Y18_FW搜尋範圍解析度的回歸模型(簡稱Y18_DY)。探討海洋資料量及搜尋範圍對回歸模型的影響，並評估三組回歸模型估計海洋溫度結構的表現。
西北太平洋的驗證裡中，三個回歸模型表現都非常良好而且估計的溫度結構也都非常相似。而在不同緯度的驗證中我們發現，模型的表現會因為Argo Floats在各緯度的數量及分佈產生差異。原本Argo Floats較少的低緯度地區(0~10°N)會因為數據量的提高而有著明顯的改善，緯度較高的地區(25~40°N)則是受到模型的搜尋範圍解析度影響。緯度越高Argo Floats的數量就越多，解析度較低的Y18_FW增加了5~10%的均方根差，Y18_DY則能有效地降低此誤差。
本研究進一步在南海建立Y18_DY回歸模型，探討模型在南海的適用性。研究結果顯示，回歸模型估計的海洋熱含量及T100 (上層100米的平均海水溫度)與Argo Floats都有著很高的相關性，其相關係數也都高於0.8。與特定Argo Floats進行長時間的溫度剖面比較，衛星反演的溫度結構與Argo Floats的溫度剖面相當一致，其趨勢變化都能清楚地顯示。說明有了回歸模型之後，我們能夠利用SSHA有效地估計出海洋的溫度結構。;Ocean is a very important source of energy for tropical cyclones, the amount of energy provided by the ocean will determine the trend of its intensity. In order to be able to predict the intensity of tropical cyclones, the temperature structure of the upper ocean is needed to determine the amount of ocean heat content. With the current technology, we cannot quickly obtain the ocean thermal structure, so increasing the method of obtaining the ocean thermal structure is an important topic today.
In this study, we have established three regression models: (1) Use Argo Floats from 2000 to 2008 (19,915 Argo Floats) to build the regression model, called Pun2014, (2) Use Argo Floats from 2000 to 2017 (87,580 Argo Floats) to build the regression model, called Y18_FW, (3) Use the dynamic window method to improve the spatial resolution of Y18_FW, called Y18_DY. Then use 6502 independent Argo Floats in 2018 to verify the three regression models.
In the verification of the Northwest Pacific, the three regression models performed very well in estimating the temperature profile, and their results were also very similar. In the verification of different latitudes, we found that the performance of the model will be different due to the number and distribution of Argo Floats in each latitude. The low latitude areas that originally had fewer Argo Floats will be significantly improved due to the increase in data volume, and the higher latitude areas will be affected by the spatial resolution. When the latitude is higher, the number of Argo Floats increases. Y18_FW with lower spatial resolution increases the root mean square error by 5-10%, and Y18_DY can effectively reduce this error.
We also established Y18_DY in the South China Sea and discussed the applicability of the model in the South China Sea. The research results show that the estimated thermal structure has a high correlation with Argo Floats, and the correlation coefficients of ocean heat content and T100 (average sea temperature of the upper 100 meters) are also higher than 0.8. The long-term temperature profile comparison with specific Argo Floats shows that Y18_DY can estimate the underwater temperature well, and the temperature change trend observed by Argo Floats can be effectively estimated. It shows that with the regression model, we can use SSHA to effectively estimate the ocean thermal structure. |
