颱風是臺灣重要水資源的來源之一，然而卻常常造成農業、經濟上的損失。雷達資料具有高時空解析度的優點，適合研究中小尺度的天氣系統。目前臺灣擁有十座都卜勒雷達，本研究使用臺灣環島雷達網，配合上Liou and Chang(2009)提出的多雷達風場合成方法來反演凡那比颱風(2010)風場結構，此方法具有許多的優點，最重要的是解決雷達基線上風場無法反演的問題，傳統風場合成需要給定的上下邊界垂直風場，此方法也可以視情況選擇，因此可以減少許多誤差。另外本研究可以彈性加入任意數目雷達同時做合成，在反演風場過程中可以得到渦度收支，且反演三維風場後可以直接做熱力場反演。到了Liou et al.(2012)加入Immersed Boundary Method (IBM)方法，使該風場合成程式又具備在複雜地形上合成能力以及不須轉換成地勢追隨座標的優點，減少座標轉換上的誤差。 Liou and Chang(2009)所發展出來的風場合成方法除了用到雷達資料，還加入了背景場來補足沒有雷達覆蓋地區的風場結構。本研究比較不同背景場對風場合成結果的影響，背景場的製作分為兩種，第一種是直接使用NCEP資料作為背景場，第二種為放入WRF(Weather Research and Forecast)模式執行十二小時後，以模式不同預報時間輸出風場投影到各雷達上，再與雷達觀測徑向風比較差異最小的模式預報時間風場做為背景場。此外針對有無雷達資料交界處的異常運動場作探討，分別討論改變damping層數、改變合成層數以及增加水平風場平滑係數，改善異常運動場的產生。 Liou and Chang (2009)提出的風場合成方法並不希望對背景場太敏感，因此只要大環流結構正確，不管使用上述任一種背景場，有雷達資料覆蓋地區結果都不會有太大差異，儘管原始背景場渦旋中心各不相同，加入雷達資料後，都會將渦旋中心重置搬移到相似的位置。因此雷達資料越密集，減少背景場主導會對風場合成有較好的結果表現。 Typhoon is one of the most important water resources of Taiwan. Doppler radar is a good observation instrument for mesoscale weather system with its high spatial and temporal resolution. Presently Taiwan radar network have ten Doppler radars. In this research, the radar data sets from seven radars of Taiwan radar network during the Typhoon Fanapi (2010) land fall period were selected. The multiple-radar synthesis method which demonstrated by Liou and Chang (2009) was applied to retrieve the three dimension wind field of Taiwan and its vicinity area. The first advantage of this multiple-radar synthesis method is that the wind field can still be recovered well on the radar base line, second is that there is no need to prescribe the top or bottom boundary condition of vertical velocity. This variation method can also acquire the vorticity budget with no residual term, finally it has the flexibility to add any numbers of radar to synthesize the wind field. Liou et al. (2012) add IBM(Immersed Boundary Method) into the multiple-radar synthesis method, thus it has the ability to deal with complex terrain, another feature is that it has no need to change Cartesian grid into terrain-following coordinate system and reduce the error from interpolation. There are two methods for making background field. First is using the NCEP FNL data directly. Second way is running the WRF model and selecting the forecast time period wind data which has the least difference between the projected radial wind field from model output and the observed radial wind. Both types background provided wrong locations of typhoon vortex, the Multiple-Radar Wind Synthesis method proved that with proper coverage of radar data, the more realistic vortex center can be relocated to more correct position.
