The usefulness of a Land Data Assimilation System (LDAS) to provide land surface states to NWP and climate models to improve weather and seasonal climate predictions presupposes that the surface states estimated by land surface models (LSMs), forced with observations, are sufficiently accurate. The North American LDAS (N-LDAS) project has been running in real-time at 1/8th-degree spatial resolution, hourly since April 1999, and using the same data sources, retrospectively back to October 1996 over the CONUS region of North America. It is forced with Eta/EDAS-based meteorology, GOES-based 0.5 degree resolution incoming solar radiation, and gage-based daily liquid precipitation analysis at 1/8th-degree resolution, which is disaggregated to hourly amounts using the 4-km Stage IV merged WSR-88D/gage hourly precipitation estimates. In this paper the validation of the N-LDAS land surface models surface energy budget terms is presented using data from the NOAA SURFRAD sites, Oklahoma state mesonet sites, ARM/CART sites, and GOES satellite-based surface temperature estimates. The analysis is carried out across a range of temporal scales, from averaged diurnal cycles to monthly averages, and includes both the forcing data as well as model outputs.