| 摘要: | 傳統觀測資料大部分集中於陸地，故可藉助雷達、衛星和飛機觀測來彌補資料稀少的海洋觀測之不足。本研究利用WRF之三維變分資料同化系統(3DVAR)和系集調整卡爾曼濾波器(EAKF)同化FORMOSAT-3 GPSRO觀測和GTS資料，針對2009年侵台的莫拉克颱風探討GPSRO資料在不同資料同化系統下對颱風預報路徑、降水、颱風結構及環境場等之影響。 本研究四個實驗組名稱為GTS、GPS、GTSe及GPSe，前兩個使用3DVAR同化GTS資料和GPSRO+GTS資料，後兩個實驗和前兩個實驗之設計相同，但使用EAKF同化系統。由11個資料同化預報週期的平均路徑誤差顯示GPS實驗在預報前24小時的平均路徑誤差比GTS實驗小。EAKF無論有無同化GPSRO觀測在預報48小時內的平均路徑預報皆比3DVAR好。將重力位高度、溫度、濕度和風場的預報與全球分析場和探空觀測比較發現使用EAKF同化系統的RMSE比3DVAR大，推測原因可能為GPSRO觀測資料太少及模式模擬範圍太小，造成分析場與預報場和真實大氣差異較大。降雨得分方面，對3DVAR之同化系統而言，GTS同化對中雨之00小時至24小時之預報較好；同化GPS之預報，大雨的ETS得分比GTS高。若使用EAKF同化GTS之GTSe實驗在大豪雨與超大豪雨得分比較高，24-48小時之累積降水預報，以EAKF同化GPSRO之GPSe實驗的表現明顯比其他組實驗好。 將預報與觀測24小時累積降水比較，發現利用3DVAR同化之實驗GPS和GTS之實驗能提早12小時預報到台灣西南部強降水，而使用EAKF實驗則可提早30小時。使用EAKF同化系統時，GPSRO增量大部分都是貢獻在颱風環流，對颱風結構有正的作用，因此雖處於較不利的環境仍能預報得到較小的路徑誤差、較好的降水預報及颱風結構。由深層大氣的駛流分析得知3DVAR預報的颱風路徑通常在最佳路徑的右側、EAKF在最佳路徑的左側，原因為3DVAR的東風駛流較快減弱加上南風增強，造成颱風較快往北移，預報路徑就在最佳路徑的右側，EAKF則是駛引的東風維持較久，把颱風往西帶後再北轉，因此得到的預報路徑都在最佳路徑左側。比較3小時累積降水時間序列分佈可看到使用3DVAR同化系統累積降水都只發生在受到地形抬升造成的山區降水，而EAKF除了也能預報到地形抬升所造成之降水外，也能預報到在外海和沿岸明顯的輻合降水外，EAKF模擬結果和觀測的降水分佈較為一致。 Most conventional observations are located on land, so observation over data-sparse ocean can be compensated by radar, statellite, and aircraft. In this study, the case of typhoon MORAKOT in 2009 is selected and FORMOSAT-3 GPSRO and GTS data assimilation with WRF 3DVAR and EAKF are utilized to investigate the impact of GPSRO data on forecasting of typhoon track, precipitation, typhoon structure and environment in different DA system. Four experiments in the study are named GTS, GPS, GTSe and GPSe, the former 2 experiments assimilate GTS data only and both GPSRO and GTS data with 3DVAR respectively, the latter two experiments are same as the former 2 experiments but with EAKF. The mean track error in 11 DA forecast periods indicates the GPS experiment has smaller track error in first 24-hr forecast and the EAKF assmilation has the smallest track error at 36 hr forecast. Mean predicted track in 48 hours of the experiments of with EAKF DA system are better than 3DVAR no matter what data was assimilated. Compared forecasts of geopotential height, temperature, moisture and wind fields with sounding and global analysis field from NCEP GFS and ECMWF, using EAKF has larger RMSE than 3DVAR. The presumed reasons possibly result from lack of GPSRO data or small model domain which could cause significant difference between model analysis field (or prediction field) and true state. In the forecast rainfall skill of 3DVAR, GTS has higher skill over 1~50 mm.in the first 24 hours prediction. After assimilating GPSRO data, the skill of the GPS has larger ETS than the GTS at the the threshold of heavy rain. The GTSe experiment get larger skill when the accumulated rainfall is over 200 mm. The GPSe has better performance than the others in 24-48 hours forecast. Compared predicted and observed 24-hr accumulated rainfall, the GPS and GTS experiment can predict heavy rain of southwestern Taiwan advanced by 12 hours, and EAKF can be advanced by 30 hours. Using EAKF DA system, the increment of GPSRO assimilation most contributed to typhoon circulation implying positive effect on typhoon structure. Although the environment is not favorable, it still has smaller track error, well rainfall forecast and typhoon structure simulation. Typhoon tracks of 3DVAR prediction always deviate to right-side of best track; however the EAKF predicting tracks deviate to the left. The steering flow of deep layer atmosphere in 3DVAR DA system reveals much weaker easterly and strong southerly resulted in typhoon moving northward rapidly, so the prediction tracks are on the right side. Strong easterly steering flow in EAKF system makes typhoon move westward, so the prediction tracks deflect to the left of best track. Time serie of 3-h accumulated rainfall indicate the 3DVAR predicted rainfall always happen in the mountain area. On the other hand, GPSe has captured convergence well occurred in mountains and coastal area, so rainfall occures in the offshore zone besides mountain. It is much better consistent with the observations. |
