Numerical experiments were conducted to assimilate simulated and real reflectivity data for typhoon ‘In-fa’ (2021) and ‘Lekima’ (2019) cases every 12 minutes for 5 cycles by using a convective-scale (3km) EnKF-WRF system. The correlation analysis shows that radar reflectivity data have the greatest correlation with hydrometers, and also significant correlation with variables not directly related to the reflectivity operator as well, such as temperature and vertical velocity. Directly assimilating radar reflectivity with EnKF is able to develop reliable multivariate covariance among microphysical and other large-scale variables. Therefore EnKF updates well not only hydrometers but also other cross-variables (u,v,t,w) fields, resulting in obvious improvement in both structure analysis and typhoon forecast skills of intensity, and precipitation. If without updating the cross-variables, the positive impact of reflectivity assimilation vanishes quickly during the forecast process.
In addition, comparative experiments for using reflectivity operators based on SM (Lin scheme) microphysics were conducted. The observational operator based on SM is likely to introduce errors in the forward simulation, and the inconsistency between assimilation (SM operator) and forecast model (DM) will further aggravate the analysis and forecast performance.
Key Words: Radar Reflectivity; Direct Assimilation; Typhoon; Ensemble Kalman Filter

