| 摘要: | 模式的初始場對於颱風的預報扮演著關鍵的角色。本研究利用模式動力上調整颱風的渦旋透過全球模式(Model for Prediction Across Scales-Atmosphere , MPAS-A)的1小時來回積分而產生虛擬渦旋，同時將氣象變數重新移位到初始位置上，當渦旋的強度達到與最佳路徑的強度一致，或是達到給定循環的次數就停止來回積分的動作，而後再進行預報。
本篇研究使用的60-15公里的可變解析度，15公里是涵蓋整個西北太平洋及大部分的亞洲，三組實驗是建構在不同的模式初始場，分別是NCEP FNL 1度(CTRL)、0.25度(CTRH)以及0.25度搭配的渦旋初始化方案(NT)，探討2015年蘇迪勒(Soudelor)、杜鵑颱風(Dujuan)及2016年尼伯特(Nepartak)、梅姬(Megi)颱風，不同初始場對颱風模擬的影響。
在spinup的過程我們發現蘇迪勒颱風的渦旋結構從NCEP全球分析場進來後，大約進行40至80個循環，颱風眼牆結構有逐漸改善並且更有組織性。此外，蘇迪勒個案中，在預報前期，颱風的垂直結構NT實驗發展的最好，從垂直結構的相對濕度、位溫及水平風速來看，有移植動力結構的渦旋，相對濕度明顯的偏高、暖心結構有加深，且增溫的狀態以及整層水平風速較強且呈現更為對稱的情形，使得颱風的強度與觀測最為接近。而路徑方面，蘇迪勒、尼伯特及梅姬颱風的NT實驗在預報前期的路徑有改善，而杜鵑颱風則是在預報中期以前有明顯改善。此外，蘇迪勒颱風在NT模擬的雨量，出現與觀測相似的兩個極端降雨分布，但是東北側的極端值是略微偏南，而定量降水校驗的部分，NT實驗在較弱的降水門檻(50mm/24hr)以及極端降水的門檻(800mm/24hr)，公正預兆得分、偏倚得分及偵測率是比其他實驗來的好一些，但其餘的降水門檻則是CTRH分數來的好。再看到降雨的空間相關係數，在不同時間的累積雨量，CTRH是比其他實驗來的好一些，但是整體而言，三組實驗的空間相關性都很接近且是不錯的，都有0.5以上。
目前4個個案研究中，我們發現在72小時預報期間，NT的平均路徑誤差大部份是優於其他實驗組，而平均最低氣壓及最大風速的誤差，NT實驗幾乎是最小的，明顯優於其他實驗組，這也顯示出模式初始動力上的虛擬渦旋在MPAS高解析度的預報是有正面的影響，特別是颱風強度與結構。
;The initial condition of a model plays an important role in typhoon forecasts. In this study, the initial typhoon vortex is replaced by a dynamically adjusted typhoon vortex, which is derived from cycling runs of one-hour MPAS model integration. In this study, the MPAS variable-resolution of 60-15 km is used, and the region with 15-km resolution covers the East Asia and North Western Pacific. Three experiments are conducted for each typhoon case, i.e., Typhoon Soudelor (2015), Dujuan (2015), Nepartak (2016) and Megi (2016). All of the initial conditions come from National Centers for Environmental Prediction (NCEP) Final (FNL) analysis but in different resolutions, i.e., 1 degree (CTRL), 0.25 degree (CTRH), and cycling runs with resolution of 0.25 degree (NT).
After about 40-80 cycles for typhoon Soudelor, the typhoon structure is greatly improved than the NCEP global analysis by providing a more organized eyewall. The results show that NT has improvements in early track predictions for typhoons Soudelor, Nepatak and Megi, but in the later simulation for typhoon Dujuan. The NT run for typhoon Soudelor obtains a consistent rainfall pattern with a double-peak precipitation, which in a better agreement with the observation compared to the no-bogussing run, even though the extreme at the northeast side is slightly southward. For the rainfall verification, NT shows better score with extreme (smaller and larger) thresholds in equitable threat score, bias score and probability of detection, but the CTRH has higher score in the middle thresholds. As for spatial correlation coefficient, the accumulated rainfall in different periods is better for CTRH. In general, all the spatial correlation coefficients are very similar and more than 0.5. Generally speaking, NT has the best simulation in the 72-h mean track error, minimum pressure, and maximum wind speed. Therefore, the dynamically adjusted vortex has a positive impact on the typhoon simulations, especially in the intensity and structure. |
