雷達觀測具有高時空解析度的優點，常使用於劇烈天氣的監控與觀測。本研究主要目的為利用多部都卜勒雷達觀測資料，改善模式當時的初始場，增進模式降水定量預報(Quantitative Precipitation Forecast：QPF)之能力。此方法主要包含三大部分：(1)多都卜勒風場合成、(2)熱動力反演、(3)水汽調整。 吾人選取2008西南氣流實驗計畫(SoWMEX)中所觀測到的IOP8個案，作為本研究的實驗對象。使用中央氣象局七股雷達(RCCG)、墾丁雷達(RCKT)及美國國家大氣研究中心(NCAR)所屬的SPOL雷達，於2008年6月14日1200UTC當時的回波及徑向風觀測資料，反演出三維風場結構，接著透過動量方程計算大氣熱動力場，並且利用回波等條件對水汽進行調整，最後同化至模式中。本研究使用NCAR Weather Research and Forecasting (WRF) Model作為計算平台。 本研究設計了一系列實驗，主要的結果有：(1)Kain-Fritsch及WSM6 Scheme為最佳的積雲參數化與微物理組合、(2)水汽的調整有其必要性、(3)風場合成與熱力反演時需考慮雪的存在、(4)以多部雷達網連增加資料覆蓋量對同化結果有重要的影響。 經過本方法調整模式初始場，實驗顯示模式的預報能力可達三小時，雖然降水有高估之趨勢，但相較未同化前的降水分佈會更趨近於觀測。未來更可將本方法用於測試午後對流或甚至颱風降雨系統的預報上。 An important advantage of radar observations is their high temporal and spatial resolutions, which are suitable for heavy weather surveillance. The purpose of this study is to improve the initial field and hence the quantitative precipitation forecast (QPF) of the numerical model by using multiple-Doppler radar observation data. The assimilation technique includes three components: multiple-Doppler radar wind synthesis, thermodynamic retrieval and moisture adjustment. A case during IOP8, Southwest Monsoon Experiment (SoWMEX) 2008 is selected in this study. The radar data in use are the reflectivity and radial wind of the RCCG and RCKT radars from CWB and the SPOL radar from NCAR at June 14, 2008. The 3-D winds, retrieved from the radar and sounding data, are utilized to calculate thermodynamic fields by the momentum equations. The moisture field is updated if some conditions, including a minimum reflectivity of 30 dBZ, occur. The numerical model in use is the Weather Research and Forecasting (WRF) model from NCAR. Some conclusions are made after a series of experiments: (1) A combination of Kain-Fritsch cumulus parameterization and WSM6 microphysics schemes gives the best result; (2) The moisture adjustment is necessary; (3) Both wind and thermodynamic retrieval algorithms consider the effect of snow; (4) Using multiple-Doppler radar data is necessary because a larger data coverage leads to better results. The above assimilation technique in this case significantly improves the accuracy of the forecast for at least 3 hours compared with the one without data assimilation in spite of overestimated precipitation. We expect applications of this technique to the cases of afternoon convection and even typhoons in the future.
