The U.S. Army has operational requirements for soil moisture profiling capabilities using advanced satellite data assimilations techniques. Colorado State University is using a four dimensional variational (4DVAR) data assimilation system that requires quantitative knowledge of the first guess background covariance fields. To evaluate the accuracy of the satellite-measured soil moisture, a statistical analysis of the USAF AGRMET land surface model output was performed. The AGRMET model soil moisture outputs are compared with in situ soil moisture data from two different sources for the month of September 2003. In situ data was provided by the U.S. Army Engineer Research and Development Center (ERDC) from their site near Mound, LA. In addition, sites from the USDA's Little Washita watershed and the Oklahoma Mesonet are included. Volumetric water content measurements at various levels are compared. Precipitation measured at the sites are also summarized. These results will be used to specify the background covariances of the 4DVAR data assimilation system, and also serve to estimate system biases.