Surface soil moisture from satellite retrieval and land surface modeling usually exhibit a significantly different dynamic range and variability in time and space. Compared to in situ measurements and land surface simulations, the satellite-based data are subject to a distinct systematic bias due to errors arisen from the retrieval algorithm, calibration method, instrument failures, measurement degradation, and so on. Recently, their negative impact on data assimilation has gained some attention. In order to develop a bias-reduced climatology of satellite products and/or mitigate the negative impact of systematic observation bias in data assimilation applications, we extend the EnKF for state estimation rooted in a land data assimilation system with different methods of modeling the observation bias. This assimilation system assumes that the land surface model is unbiased (it is only prone to random error), while the biased satellite soil moisture retrievals are corrected using a Kalman Filtering approach where the observation biases are estimated by making use of correction information from the model. Essentially, the observation bias is estimated and corrected parallel to the state estimation computation through the data assimilation process. Two bias correction schemes are proposed and tested in a real-world environment. Assimilation experiments are performed by assimilating real satellite soil moisture data from AMSR-E into the Noah land surface model over the North American domain for a warm season. We will analyze the impact of the dynamic observation bias estimation and compare it with a static scaling approach (e.g. CDF matching).