The DOE Atmospheric Radiation Measurement (ARM) Program operates a network of surface radiation measurement sites across north central Oklahoma and south central Kansas. This Southern Great Plains (SGP) network consists of 21 sites unevenly spaced from 95.5 to 99.5 degrees west longitude, and from 34.5 to 38.5 north latitude. We use the technique outlined by Long and Ackerman (2000) and Long et al. (1999) to infer continuous estimates of clear sky downwelling shortwave (SW) irradiance, SW cloud effect, and daylight fractional sky cover for each site, then interpolate these quantities to a 0.25 degree grid for the SGP area. We will present an analysis, using the gridded product, of spatial variability across the SGP area on various time scales. We will show that as the spatial domain increases, the variability increases, the frequency distributions tend toward median values, and the data collected at the SGP Central Facility is less representative (correlated with) the larger area. The SGP Central Facility is also less representative of the larger area as the temporal averaging time decreases. To date, most comparisons between measurements and models have used only data from the SGP Central Facility. These results show that this practice can and often does result in large uncertainties inherent in the comparison which temporal averaging of Central Facility data cannot address.