It is found that the EnKF performs consistently better than the 3DVar method by assimilating either individual or multiple data sources (i.e., sounding, surface and wind profiler) for the MCV event and in the month-long experiment which is performed by assimilating 12-hourly in-situ sounding data. Proper covariance inflation and using different combinations of physical parameterization schemes in different ensemble members (the so-called multi-scheme ensemble) can significantly improve the EnKF performance. Result also shows that the EnKF seems to benefit more from the ensemble-based prior estimate than from using a flow-dependent background error covariance. The 3DVar system can benefit substantially from using short-term ensembles to improve the prior estimate (with the ensemble mean). Noticeable improvement is also achieved by including some flow dependence in the background error covariance of 3DVar. Besides, similar error statistics and weather system structures are observed for a particular period of time between experiments starting at different times. This result suggests that long-term experiment may not be very sensitive to the initial condition and 12-h pre run may be long enough to provide a reasonable background error covariance structure.