使用四維變分與都卜勒雷達資料改進短期定量降雨預報－2008年SoWMEX實驗期間一個鋒面系統的個案研究

Using Four Dimensional Variational Method and Doppler Radar Data to Improve Short Term Quantitative Precipitation Forecast-A Case Study of a Frontal System Observed during 2008 SoWMEX Field Experiment

戴聖倫(Sheng-Lun Tai) ； 廖宇慶(Yu-Chieng Liou) ； 張少凡(Shao-Fan Chang)

《大氣科學》 38卷4期 (2010/12) Pp. 269-287

摘要

前人之研究中曾經利用美國國家大氣研究中心(National Center for Atmospheric Research; NCAR)所發展之都卜勒雷達變分分析系統(Variational Doppler Radar Analysis System; VDRAS)，即時分析低層風場及輻合／散場以進行雷暴(thunderstorm)之預報，也曾同化多部都卜勒雷達觀測資料以分析並預報超級胞(supercell)與颮線的演化，但大多使用於廣大開闊的平原地區，本研究則首次將VDRAS應用在臺灣及其鄰近地區。由於臺灣複雜的地形，且儀器的觀測受到四面環海的限制，這些條件預期會對VDRAS的適用性帶來相當大的挑戰，所以本研究也嘗試為其尋求一適當的同化策略。
為了檢視此同化系統應用於本地的表現，吾人選取2008年SoWMEX (Southwest Monsoon Experiment；西南氣流實驗）IOP8（Intensive Observation Period 8；第八次密集觀測期）中6月14日的鋒面系統作為研究個案，匯集探空、地面測站、與來自歐美作業中心的再分析場等資料建構背景場，並且以四維變分法(4DVAR)同化中央氣象局七股及墾丁兩座S-band都卜勒雷達之回波以及徑向風觀測資料，藉此完成雲模式初始化得到一最佳的分析場，再由此分析場進行預報。同時，為了較為妥善地處理地形對於降水預報的影響，吾人將VDRAS之最佳分析場與具有地形解析能力的WRF模式進行結合，並比較結合前後對預報降雨的差異。
檢視同化觀測資料後的分析場，顯示出VDRAS能夠反演出對流系統中合理的動力及熱力結構。吾人亦發現低層幅合場的分佈與當地山脈走向呈現出一致性，表示VDRAS可能具有反映地形效應的潛力。在預報方面，VDRAS模式對於造成降雨的主要線狀對流之移動方向有不錯的掌握。針對兩小時累計降水的預報，其Equitable Threshold Score (ETS)得分在0.1~0.2之間。如將VDRAS與WRF結合後再進行預報，則此得分較單獨使用VDRAS或WRF的預報都有很顯著的改善。
本研究中的同化策略，可做為在其它具有類似地理環境與觀測限制的區域，進行資料同化與預報時的參考。

關鍵字

四維變分同化 ； 都卜勒雷達變分分析系統

並列摘要

The Variational Doppler Radar Analysis System (VDRAS), developed by National Center for Atmospheric Research (NCAR), had been used to analyze low-level wind and convergence field in order to forecast thunderstorms at a real-time base. It was also applied to predict the evolution of super cells or squall line systems by assimilating multiple Doppler radar data. However, those studies were mostly performed over a wide open plain. In this research, it is for the first time that VDRAS is applied in the Taiwan and vicinity area. Since the complex terrain and limited observations due to the surrounding oceans pose great challenges, it is attempted in this research to find an appropriate strategy for using VDRAS under such conditions.
A real case observed during IOP8 (Intensive Observation Period 8) of SoWMEX (Southwest Monsoon Experiment) on 14, June, 2008 is selected. VDRAS uses the data collected by radiosondes, surface stations, and re-analysis data to construct a background field, followed by assimilating the radial winds and reflectivity detected by two Central Weather Bureau S-band Doppler radars (RCCG and RCKT). Through Four-dimensional variational (4DVAR) adjustment, one obtains an optimal initial field, from which the VDRAS starts to make forecast. In addition, for a proper treatment of the influences from the complex terrain, it is also attempted to combine the VDRAS analysis field with WRF, and let the latter continue the forecast.
It is found that VDRAS is able to retrieve reasonable kinematic the thermodynamic fields. The low layer convergence is consistent with the orientation of the local mountains, which implies that it is possible for VDRAS to reflect the topographic effects of the terrain. In terms of model forecast, VDRAS can correctly capture the movement of the major precipitation system. The Equitable Threshold Scores (ETS) of predicted 2-hour accumulated rainfall are between 0.1 to 0.2. However, if VDRAS is merged with WRF, the resulting rainfall forecast skill can be significantly improved than that from using WRF or VDRAS alone.
This research provides a possible alternative if VDRAS is to be applied in another region with similar geographic environment and observational limitations.

並列關鍵字

Four-dimensional Variational Method ； Doppler radar data assimilation system