A hybrid 3DVAR-ensemble (denoted as 3DVAR hybrid) data assimilation based on the Gridpoint Statistical Interpolation (GSI, a 3DVAR algorithm) and the Ensemble Kalman filter (EnKF) has been developed and successfully tested for the NCEP Global Data Assimilation System (GDAS) using the Global Forecast System model (GFS). Various results have shown that the 3DVAR hybrid provided better analyses and subsequent forecasts than the current operational 3DVAR system (GSI). An ensemble 4DVAR (ENS4DVAR) system is further developed based on this 3DVAR hybrid system to further improve upon the 3DVAR hybrid. Like the traditional 4DVAR, four dimensional analyses are obtained with ENS4DVAR by fitting observations within the assimilation window. Unlike the traditional 4DVAR, ENS4DVAR does not require the tangent linear and adjoint of the forecast model. The forecasts initialized by the analyses generated by the ENS4DVAR were compared with the 3DVAR hybrid for both a winter month and a summer month assimilating all operational conventional and satellite observations. Verification against in-situ observations has shown that the ENS4DVAR improved upon the 3DVAR hybrid especially at the early forecast lead time. Verification of hurricane track forecasts over the 2010 hurricane season has also shown that the track forecasts by the ENS4DVAR were better than that by the 3DVAR hybrid. In addition, experiment results comparing the ENS4DVAR and 3DVAR hybrid for the 2011 hurricane season with the operational resolution and configuration will be presented in the conference also.