An ongoing challenge for hydrologists and remote sensing scientists is the design of experiments to demonstrate the value - if any - of remote sensing observations for efforts to monitor and/or predict surface hydrologic processes. The need is especially pressing for remote observations of surface geophysical state variables like soil moisture and skin temperature. One efficient utilization of remote surface state observations is within the context of a data assimilation system designed to merge surface state predictions from numerical models with remote observations of the land surface. Such systems contain at least three components: a numerical land surface model, an emission model to convert land surface model predictions into observable quantities (e.g. brightness temperature), and an assimilation algorithm. The "value" of remote sensing observations depends therefore on a myriad of factors including the quality of non-updated open-loop model predictions, the optimality of the data assimilation approach, and the accuracy of the observational model. One basic benchmark for data assimilation approaches should be the accuracy of model predictions (e.g. evapotranspiration or streamflow) obtainable from non-updated open-loop model simulations. This presentation will address some of the basic issues surrounding such evaluations and examine ways in which remote sensing observations can add skill or value to land surface model predictions. Three key weaknesses of land surface hydrology models will be highlighted: their reliance on uncertain measurements of meteorological forcing data (e.g. rainfall), parameter selection ambiguities presented by their complex representation of surface processes, and difficulties associated with the modeling of deep hydrologic states (i.e. water table depths) required for streamflow forecasting. The prospects for addressing each weakness using data assimilation techniques and remote sensing data will be examined. Preliminary results for prototype data assimilation systems based on the Ensemble Kalman filter and adjoint-based variational assimilation will be presented.