Sampling Error in the Ensemble-Based Radar Data Assimilation System and Its Impact on Convective-Scale Precipitation Prediction - A Case Study of IOP#8 During SoWMEX
Sampling errors in the ensemble-based data assimilation (EDA) can result in spurious background error correlations, leading to false analysis corrections. To investigate the sampling error in the convective-scale EDA and its impact on precipitation prediction, the WRF-LETKF radar data assimilation system (WLRDAS) with ensembles of 256 and 40 members are performed in this study. The comparison of the background error correlation revealed that the sampling error is generally more serious over the intense reflectivity area. In addition, at the observation points where the radar cannot observe the radial velocity (Vr), the sampling error between Vr and model variables may be more serious. On the other hand, the water vapor of the model suffers from the sampling error severely even over the weak reflectivity area. We found that this can result in under-estimating of the rainfall significantly from the results of the precipitation prediction, especially when the horizontal localization radius is larger. The experiments with different vertical localization setting also revealed that a larger vertical localization distance is helpful for capturing more reasonable vertical structure and further shows the positive effect on precipitation prediction.
Ms. Pin-Ying Wu Kyoto University, Japan