The simulated soil moisture from the Florida State University (FSU) climate model coupled to the National Center for Atmospheric Research community land model (CLM2) is evaluated using two reanalyses (the European Centre for Medium-Range Weather Forecasts 40-year Reanalysis (ERA40) and the National Centers for Environmental Prediction/Department of Energy Reanalysis 2 (R2)) and in situ observations in Illinois and China from the Global Soil Moisture Data Bank. While the soil moisture was prescribed in the previous FSU climate model, the implementation of the CLM2 within the FSU model (hereafter FSU-CLM) provides a prognostic soil moisture (both soil liquid water and soil ice) in 10 vertical layers. For the first layer (~0--10 cm) the comparison of the soil moisture annual cycle between the FSU-CLM and the two reanalyses shows that in the Northern Hemisphere (NH) the FSU-CLM and the ERA40 are in phase but they are out of phase with the R2. While a parameterization of the soil ice process is included in both the FSU-CLM and ERA40, it is not included in the R2. In this study, we found that the soil ice plays a key role in determining the soil moisture variation over the NH. The two reanalyses and the FSU-CLM are also compared with in situ observations. The FSU-CLM turns out to follow the annual cycle of the in situ observations better than both reanalyses.