An analysis of the surface net radiation (Rn) distribution and factors which determine its representativeness and accuracy has been carried out as part of the Boreal Ecosystem-Atmosphere Study (BOREAS). A set of three calibrated net pyrradiometers has been used to intercalibrate 22 Rn measuring systems at 21 sites. GOES satellite-based fields of Rn are also compared with the associated measured fields to quantify the effect of sparse sampling on the spatial and temporal error structures of Rn. The results show that a standard make of net pyrradiometer, used at 15 of the 21 measuring sites, underestimates net radiation by about 5% in the daytime and about 45% at night. The distribution of measuring sites and intercalibration errors are shown to produce distinctive patterns of bias error in the space-time fields of Rn over the large-scale study area. However, we find that sparse sampling of measured Rn has a far more significant effect in generating errors in the large-scale space-time framework. At the small scale, considering the distribution of net pyrradiometers within the southern study area, we find that intercalibration is the more serious factor when considering time-averaged gradient structure. Although the changing ecotone pattern over the BOREAS region introduces gradients in the Rn fields, we find that in characterizing the space-time sampling error structures, cloud effects dominate ecotone effects. The study concludes with a demonstration of how to take advantage of the GOES satellite's capability of representing between-site Rn gradients and the net pyrradiometer network's capability of making accurate point-source measurements, by producing optimally averaged maps of Rn. ¿ 1997 American Geophysical Union