在真實觀測實驗中，此雷達資料同化系統仍能有效改善定量降水即時預報。同化回波時須使用變數局地化法，只用來更新雨水混合比。使用觀測空間的統計方法，能診斷預報偏差和理想系集離散度。混合局地化法在真實觀測實驗的效益更加明顯，尤其能提升觀測資料稀疏或破碎處的風場準確度，進而改善降雨預報。;This study develops a Doppler radar data assimilation system, which couples the local ensemble transform Kalman filter with the Weather Research and Forecasting model. Its benefits to quantitative precipitation nowcasting (QPN) are evaluated with observing system simulation experiments (OSSEs) and real observation experiments on Typhoon Morakot (2009), which brought record-breaking rainfall and extensive damage to central and southern Taiwan. The purpose is to provide a useful plan of radar data assimilation for improving typhoon rainfall nowcasts in Taiwan, which are challenges due to complex terrain and the lack of in-situ observations over the surrounding sea.
In the OSSEs, the assimilation of radial velocity and reflectivity improves the three-dimensional winds and rain-mixing ratio most significantly because of the direct relations in the observation operator. For QPN, the positive impact of radar data lasts for 6 hours; the performance responds to reflectivity assimilation more quickly than radial velocity assimilation while assimilating both is most recommended. Increasing the observation coverage over upstream convection areas also largely enhances the QPN performance. For multi-scale interactions, we propose a mixed localization method, which yields further improvement.
Our system also improves QPN effectively with real observations. When real reflectivity data are assimilated, the variable localization method must be used to update only the rain mixing ratio. With observation-space statistics, the model bias and ideal ensemble spread can be diagnosed. The mixed localization method, which is more beneficial in the real case, enhances the accuracy of the wind field especially for the areas with sparse or discontinuous radar observations and also improves QPN.