The ensemble Kalman smoother (EnKS) is employed to estimate surface and subsurface soil moisture and surface energy fluxes during the Southern Great Plains 1997 (SGP97) experiment through the assimilation of observed L band radiobrightness temperatures. Previous work using the ensemble Kalman filter (EnKF) and a simple smoother demonstrated that soil moisture estimation is a reanalysis‐type problem. The EnKF uses observations as they become available to update the current state. The EnKS takes the EnKF estimate as its first guess. However, in addition to updating the current state it also updates the best estimate at previous times. The performance of the EnKS is compared to the EnKF and the ensemble open loop in which no measurements are assimilated. Estimated surface soil moisture is compared to gravimetric observations at three locations. Root zone (5–100 cm) soil moisture is evaluated by comparing the resultant latent heat flux to flux tower observations. In a fixed lag smoother, observations are used to update past estimates within a fixed time window. The EnKS can be implemented in a fixed lag formulation in problems with limited memory such as soil moisture estimation. It is shown that there is a trade‐off to be made between the improved accuracy with longer lag and the increased computational cost incurred. It is demonstrated that the EnKS is a relatively inexpensive state estimation algorithm suited to operational data assimilation.