本研究探討如何使用都卜勒雷達資料改善複雜地形區域之降雨預報，透過觀測系統模擬實驗(Observing Systems Simulation Experiment, OSSE)測試得知水氣變數之重要性，因此本研究設計了一系列程序用來調整水氣。透過多都卜勒雷達風場合成方法(Wind Synthesis System using Doppler Measurement, WISSDOM)、熱動力反演與水氣溫度調整，反演複雜地形上的三維風、壓力、溫度與水氣場並同化進WRF，分別使用兩個真實個案探討反演變數是否可提升極短期降雨預報能力，第一個個案是2008年6月14日在台灣南部SoWMEX(South West Monsoon EXperiment)實驗期間的西南季風劇烈降水個案，而第二個是2014年8月19日在台灣北部發生的午後熱對流個案。 由第一個個案預報結果顯示，同化雷達觀測與反演變數可維持一段時間的雨帶結構並改善模式降雨強度與分布；第二個個案則顯示水氣調整可重現對流系統之細微結構，雖然預報結果有捕捉到台北地區的強降雨，但局部地區仍存在預報錯誤。 本研究的優點是所有方法皆可直接應用於地形上。此外，僅使用來自兩次體積掃描的雷達資料，即可改進模式極短期降雨預報。 ;This research discusses rainfall forecast improvement over mountainous regions using Doppler radar data. Through OSSE(Observing Systems Simulation Experiment) tests the importance of vapor is first identified. As a result, a procedure for adjusting vapor is designed. By assimilating three-dimensional meteorological fields obtained from multiple-Doppler radar wind synthesis and thermodynamic retrieval as well as vapor adjustment into WRF, the impact on rainfall forecast for two real cases are investigated. The first case is the heavy precipitation of a southwest monsoon event occurring on 14 June, 2008 in southern Taiwan during SoWMEX (South West Monsoon EXperiment), while the second case is an afternoon thunderstorm event occurring on 19 August, 2014 in northern Taiwan. The forecast results of the first case show that the rainbands’ structure could be maintained for a period of time and the rainfall intensity and distribution are also improved after assimilating the radar observed and retrieved parameters. The second case demonstrates that the vapor adjustment scheme is able to reproduce some of the fine structures of the convective systems. Extreme rainfall in Taipei is well captured, but false alarm does exist. The advantage of the methods developed in this research is that they can be applied directly over terrain. In addition, radar data from only two volume scans are sufficient and can be efficiently used to improve the model very short-term rainfall forecast.
