Abstract. Accurate modeling of the downwelling longwave flux at the surface is critical to our understanding of a number of important issues: the earth's energy balance; processes at the atmosphere's lower boundary including ice melt and ocean forcing; and for evaluating our ability to model atmospheric fluxes for dynamical models including numerical weather prediction and climate models. Under the ARM program, there has been a concerted effort to evaluate the modeling of the downwelling zenith spectral radiances using AERI interferometric measurements and the LBLRTM line-by-line model. Up to this point, there has not been a comparable effort to extend this type of analysis to the downwelling fluxes at the surface, which are of interest to the broader community. Based on the data from SGP during 1997-1998, we found that the LBLRTM-computed fluxes are in almost all cases less than the AERI and pyrgeometer fluxes. Although this may be caused by a number of different biases in the measurements, models, and atmospheric state specification, one plausible reason for this behavior is that the model calculations lack a source of emitted radiation such as aerosols. This possibility has been investigated for a subset of the cases by performing MODTRAN calculations of longwave surface flux with and without aerosols. For each calculation, the longwave aerosol optical depth was determined from the visibility measured by the MFRSR and a rural aerosol was assumed. The result of these calculations suggests that the inclusion of aerosols in the model would increase longwave surface flux by less than 1 Wm-2 in most cases, not enough to explain all of the flux differences observed. The spectral regions in which the AERI-LBLRTM differences occur do correlate to some extent with the regions in which the increased flux due to aerosols occurs in the MODTRAN calculations.