Published work has shown that GRACE water storage estimates are consistent with water storage observations for many river basins. GRACE data can therefore serve as a proxy for water storage data. In this analysis, we compare estimates of total water storage (TWS) anomalies from the GRACE mission to soil moisture data from IPCC AR4 simulations for the Mississippi River Basin (MSRB). IPCC models do not carry a TWS variable for direct comparison; therefore we use the IPCC models' soil moisture content parameter to compare to the GRACE data, since total water storage variability in the mid-latitudes is mostly due to soil moisture variability. GRACE data is thus used to validate hydrological output from IPCC models, and trends are then examined to project future water storage for this river basin for the 21st century. Results of our analysis show that the composited seasonally-varying TWS anomaly fields for the MSRB projected by IPCC models correlate well in time with the composite GRACE anomaly field, with most correlations greater than 0.9. This indicates that the models do fairly well at projecting the phase of the seasonal change in water storage in the study area. However, of the eight model simulations examined (assuming scenario A2), some models over- or under-estimate the amplitude of the annual variability significantly (e.g., by a factor of two). Most fail to capture the annual cycle closely. In addition, the spatially distributed water storage values from the IPCC simulations vary considerably among the models. These differences are in contrast to the broadly similar precipitation and evaporation distributions among the models examined. Our validation efforts (using GRACE data) suggest that the soil models in many codes need improvement before we can accurately examine soil moisture projections through the next 100 years.