A statistical comparison of observations and forecasts using the Regional Atmospheric Modeling System (RAMS) for operations at the Savannah River Site (SRS) is discussed. Simulated and observed values of wind direction, wind speed, and temperature, collected twice daily for a two-year period from April 1998 through March 2000, are compared in a variety of ways for 5 different locations in the southeast United States. Turbulence quantities are also compared for a one-year period beginning in February 1999 for the SRS. Both surface and upper-level model predictions are compared with observations taken from both the National Weather Service and local SRS tower locations (surface measures only). Variability based on the time of year, the forecast hour, the location of the observations, and the height above ground for each of these variables is discussed. Statistics of accuracy used for comparison include absolute mean bias, relative bias, root-mean-square error, and standard deviation. The largest errors occur during the transition periods of early evening (~19 to 20 LST) and late morning (~07 to 08 LST), especially for temperature.