A soil moisture assimilation scheme based on the extended Kalman filter（EKF）and an unsaturated soil water flow model is developed and numerical experiments using synthetic data and numerical simulation with in-site observations are presented. The numerical experiments show that the assimilation scheme improve estimation of soil moisture, the frequency of observation, the depth and the layers at which observation is introduced have the influence on assimilation. The assimilation experiments with in-site observations and infiltration derived from meteorologic forcing condition and the land surface model VIC show the monthly and seasonal variation of soil moisture, which show that the assimilation scheme is reasonable.