To overcome the difficulties in determining the optimal parameters needed for a radiative transfer model (RTM), which acts as the observational operator in a land data assimilation system, we have designed a dual-pass assimilation (DP-En4DVar) framework to optimize the model state (volumetric soil moisture content) and model parameters simultaneously using the gridded Advanced Microwave Scanning Radiometer–EOS (AMSR-E) satellite brightness temperature data. This algorithm embeds a dual-pass (the state assimilation pass and the parameter optimization pass) optimization technique based on an ensemble-based four-dimensional variational assimilation method and a shuffled complex evolution approach (SCE-UA). The SCE-UA method optimizes the parameters using observational information, thereby leading to improved simulations. The RTM is used to estimate brightness temperature from surface temperature and soil moisture. This algorithm is implemented differently in two phases: the parameter calibration phase and the pure assimilation phase. Both passes are applied in each assimilation time window during the parameter calibration phase. However, only the state assimilation pass is used in the pure assimilation phase after the parameters are determined during the parameter calibration phase. Several experiments conducted using this framework coupled partially with a land surface model (the NCAR CLM3) show that volumetric soil moisture content can be significantly improved to be comparable with in situ observations by assimilating only daily satellite brightness temperature. Furthermore, the improvement in surface soil moisture also propagates to lower layers where no observations are available.