摘要本研究為國內首次利用WRF 4DVAR (Weather Research and Forecasting four-dimensional variational data assimilation system) 探討同化虛擬渦旋及氣象雷達資料 (五分山、七股、墾丁、花蓮等四個雷達) 對於颱風凡那比 (2010) 模擬的影響。實驗設計以cold start及hot start分別使用不同的背景誤差 (cv3及cv5) 進行同化，討論四維變分資料同化對於颱風凡那比 (2010) 初始場修正及模擬的影響。此外，也於同化虛擬渦旋實驗中測試水平影響尺度對於同化分析場及模擬的影響。 經由四維變分資料同化虛擬渦旋，可以修正颱風凡那比的中心位置，改善颱風的強度及結構，使環流結構更為完整。同化過程中使用背景誤差cv3對於模式初始場會造成大範圍的修正，若使用背景誤差cv5，修正範圍明顯集中於植入虛擬渦旋處，兩者在路徑模擬上cv5明顯優於cv3，但仍不及未同化虛擬渦旋之實驗。不同的水平影響尺度測試實驗顯示，降低水平影響尺度能縮小同化修正的範圍，在本研究中降低水平影響尺度對於背景誤差cv3之同化模擬可降低路徑誤差。 同化雷達資料的部份在cold start實驗路徑模擬中，使用背景誤差cv5之路徑北偏，而cv3於未登陸台灣前於外海打轉，兩者移動速度緩慢，路徑誤差高。hot start實驗中，使用背景誤差cv3或cv5經雷達資料同化後，路徑模擬都有不錯的表現，cv3更在模擬24小時後路徑誤差低於未同化的模擬，改善路徑模擬誤差約23%，而同化對於路徑影響的原因有待未來作更進一步的探討。AbstractThis study uses the Weather Research and Forecasting (WRF) model and its four-dimensional variational data assimilation system (4DVAR) to improve the model initial conditions for numerical simulations. A series of experiments were conducted to investigate the impact of bogus vortex and radar reflectivity and radial wind from Wufenshan, Cigu, Kenting and Hualien Weather Radars on assimilation of typhoon Fanapi (2010) using WRF 4DVAR. Both cold start and hot start experiments are compared, with different background errors (cv3 and cv5). In addition, we investigate the effect of reducing the horizontal scale of influence in 4DVAR. To better understand the performance of WRF 4DVAR, the model increments at the initial time and 48-h typhoon track forecasts were analyzed. The assimilation results demonstrate that bogus data assimilation tends to intensify the initial typhoon and better relocate the typhoon center of Fanapi, but does not improve the track effectively. The initial analysis using cv3 was modified more widely compared to that using cv5. Consequently, cv5 is better than cv3 in the track simulation. The experiment with a reduced horizontal scale of influence in 4DVAR with cv3 shows that the lateral range of initial increment is reduced and leads to smaller track prediction errors. Radar data assimilation results are associated with higher tracking errors in the cold start experiment. However, the track prediction is improved by 23 % in the hot start experiment, as compared to the simulation without radar data assimilation.
