| 摘要: | 本研究採用多都卜勒雷達風場合成方法(WInd Synthesis System using DOppler Measurement, WISSDOM)及熱動力反演技術(Terrain-Permitting Thermodynamic Retrieval Scheme, TPTRS)，建立複雜地形上的三維風場、壓力場、溫度場與水氣場，並將其納入WRF模式同化中。以2022年TAHOPE (Taiwan-Area Heavy rain Observation and Prediction Experiment)梅雨鋒面降水個案(#IOP1)為例，探討雷達資料同化後對定量降水預報(Quantitative Precipitation Forecast, QPF)及熱動力場結構模擬的改善情形。
研究結果顯示，同化雷達反演資料能使降雨分布的預報更符合實際觀測，但在雨量強度方面仍有低估的情形，且改善效果主要集中在預報前 4 小時。雷達反演後的熱動力初始場及回波分布貼近觀測之外，模式對於近地層鋒面移速的模擬亦較未同化時準確；若進一步加入地面觀測風場，改善效益的持續時間則更為延長。然而，地面氣壓場的誤差改善有限，顯示模式在熱力結構及地表通量的處理上仍存在不足，而近地面氣溫及相對溼度則有些微改善。整體而言，雷達資料的同化有助於提升降雨空間分布及低層風場的模擬品質，同時改善模式對鋒面系統位置與移動速度的掌握。但由於效果仍偏短期，加上降雨量級的低估問題尚未完全解決，後續可從邊界層參數化、水氣場調整或同化策略等方向進一步改進，以提升預報的穩定度與實務應用價值。
;This study applies the Wind Synthesis System using Doppler Measurements (WISSDOM) and the Terrain-Permitting Thermodynamic Retrieval Scheme (TPTRS) to reconstruct three-dimensional wind, pressure, temperature, and moisture fields over complex terrain, and assimilates these fields into the WRF model. Using the 2022 TAHOPE Mei-yu frontal case (IOP1) as an example, we examine how assimilating radar-derived fields influences quantitative precipitation forecasts (QPF) and the simulation of thermodynamic and dynamic structures.
The results show that assimilating radar-retrieved fields improves the spatial distribution of rainfall forecasts, bringing them closer to observations, although the rainfall intensity remains generally underestimated. The improvements are most evident within the first four hours of the forecast. With initial thermodynamic fields and radar echo structures closer to observations, the model better captures the near-surface frontal propagation speed; furthermore, incorporating surface wind observations extends the duration of the positive assimilation impact. However, improvements in surface pressure remain limited and slight improvements are observed in near-surface air temperature and relative humidity, suggesting deficiencies in the model’s representation of thermodynamic structure and surface flux processes.Overall, radar data assimilation enhances the simulation of rainfall distribution and low-level wind fields, and improves the model’s ability to represent frontal position and movement. Nevertheless, the effect diminishes after a short lead time, and the underestimation of rainfall magnitude persists. Future work may benefit from refinements in boundary layer parameterization, moisture adjustment strategies, or assimilation techniques to further improve forecast stability and operational applicability. |
