This study compared forecasts of a squall line and a mesoscale convective system (MCS) over China produced by assimilating Doppler radar data using four-dimensional ensemble Kalman filter (4DEnKF) and EnKF methods. In the 4DEnKF approach, four-dimensional ensemble covariances were incorporated to consider the time differences of the radar data collected at different elevation angles within a volume scan. In comparison with EnKF, results showed that 4DEnKF exhibited substantially higher quantitative forecast skills of radar reflectivity for a 6-h forecast, and better representation of structure and intensity of the squall line and MCS; EnKF largely failed to capture the intense convection segments of both cases. 4DEnKF demonstrated greater skills than EnKF in probabilistic reflectivity forecasts for thresholds of moderate and heavy precipitation. It was also found that 4DEnKF strengthened the cold pool and wind speed in the convective regions, and outperformed EnKF in terms of all standard variables at most levels, especially middle and lower levels. Diagnosis indicated that 4DEnKF was capable of producing much stronger analysis increments of temperature, wind and moisture during radar data assimilation cycles, which was favorable for initialization and development of convection.