In this research, a reanalysis approach is taken to land surface data assimilation. Soil moisture and surface energy fluxes are estimated using an ensemble-based batch estimation algorithm. In filters such as the extended Kalman filter and ensemble Kalman filter, observations are processed sequentially as they become available. As such, filtering is ideally suited to forecasting or control problems in which observations are available in real-time. In batch estimation techniques, all available observations are used to estimate the state. This includes observations from times later than the estimation time. By including information on how the state evolves beyond the estimation time, batch estimation techniques yield improved estimates over sequential filtering. Batch estimation techniques are thus more appropriate for problems where all relevant observations are available at the time of estimation. Consequently, they are particularly suited to the soil moisture estimation problem. Previously, soil moisture observations have been gathered during field experiments using instruments on aircraft. These observations have subsequently been used to determine soil moisture. In the future, data will be available from pathfinder missions such as HYDROS and SMOS. As these are exploratory missions, data will not be available in real-time but rather as historic data. Thus, the problem lends itself to a reanalysis or batch estimation approach. The batch estimation algorithm used here has previously been used in a one-dimensional problem. Results showed that smoothing (reanalysis) rather than filtering resulted in an improved estimate of volumetric soil moisture at the surface and at depth. The improvement was apparent in the reduced errors as well as the reduction in the uncertainty associated with the estimate. Here, the batch estimation algorithm is applied to a 3D problem. Instead of estimating the soil moisture profile in a single independent problem, the current research expands the problem to estimating the spatial distribution of soil moisture. In an Observing System Simulation Experiment (OSSE), active and passive L-band observations are simulated over the Red River Basin (Arkansas), and then used in the data assimilation framework to estimate the known true soil moisture conditions. The HYDROS mission will provide both passive and active L-band observations, which are at different spatial resolutions and have different errors associated with them. In a so-called Observing System Simulation Experiment (OSSE), these observations are simulated over the Red River Basin (Arkansas), and then used in the data assimilation framework to estimate the known true soil moisture conditions. Smoothing rather than filtering yielded improved results over filtering in the 1D case, and allowing the state vector to include spatially correlated states is expected to further improve the estimate. Combining higher resolution active L-band observations with lower resolution but more accurate passive observations will provide information on the spatial distribution of soil moisture. Smoothing is more computationally expensive than filtering as there are more states and observations, and consequently larger covariance matrices to be calculated and a larger number of ensemble members. As the state vector consists of soil moisture profiles at a number of pixels, there will be a considerable increase in the computational burden in increasing the dimension of the problem, in terms of both processing and memory requirements.