We have developed a physically-based parameterization for snow albedo, for the visible and near-infrared spectral regions used in general circulation models (GCMs). Snow albedo depends primarily on snow grain size, and also on solar zenith angle, snow thickness, impurity content, and atmospheric transmittance. This parameterization is now available as a Fortran subroutine. Simpler, but less accurate, parameterizations have also been developed which depend only on grain size or thickness. Since GCMs do not compute snow grain size, we also developed a method to estimate grain size based on the air temperature and the snow age.

Our parameterization for snow albedo is being incorporated in the NCAR Community Climate Model (CCM) in place of the existing empirical parameterization for snow albedo, to determine the effect of this improvement on the model's performance, and the results will be discussed. However, additional aspects of the treatment of the radiative properties of snow and ice were also capable of improvement and are being changed in the CCM. In particular, it is important to recognize that sea ice is often snow-covered and in that case has an albedo as high as that of snow, and that southern hemisphere sea ice is nearly always snow-covered, even through the melting season. The surface albedo for the Antarctic ice sheet should be about 0.83, but it had been set to 0.71 in the CCM, The CCM has been calculating temperatures too warm over Antarctica, and this low albedo contributed to that error.