A hybrid ensemble square root filter and three-dimensional ensemble-variational(EnSRF-En3DVar) data assimilation system is developed based on WRF model.The EnSRF is used to update ensemble perturbations in this system.Aiming at Typhoon Saomai,two sets of sensitivity experiments are conducted to test this system.One is choosing different weight coefficients of ensemble covariance,the other is assimilating radar radial velocity(V_r) and/or reflectivity(Z).The first set of experiments is to answer how the flow-dependent covariance estimated from the forecast ensemble works best in the hybrid EnSRF-En3DVar system.The results show that the assimilation effects from experiments with 30 ensemble members and ensemble covariance weight coefficients of 0.25,0.5 and 0.75 are better than those from 3DVar experiment.The best(worst) one is from the experiment when weight coefficient of ensemble covariance is 0.75(1.0).The second set of experiments is to examine the impact of assimilating radar V_r and/or Z on the analysis using EnSRF-En3DVar system.The results show that the combined assimilation of V_r and Z can improve the initial analysis of humidity and wind fields,but the only assimilation of V_r can improve the analysis of wind field most.Finally,the methods to generate the ensemble perturbations are compared.The results show that the “perturbed observation” method gives smaller ensemble spread and needs more expensive computational cost than the EnSRF method.