Solar properties of snow can be computed by the SNow ICe and Aerosol Radiative (SNICAR) model widely used in land models, and by Icepack, the column physics used in CICE and MPAS-seaice. These models adopt 2-stream approximations (TSAs) with different radiative transfer techniques; as a result, the same snow has different solar radiative properties depending whether it is on land or on sea ice in Earth System Models such as the Energy Exascale Earth System Model (E3SM) and the Community Earth System Model (CESM). A unified cryospheric surface radiative model in Earth System Models is thus crucial for constraining modeling uncertainties and studying climate/hydrology. We evaluate the current TSA models in E3SM (SNICAR and MPAS-seaice) and a 2-stream discrete ordinate model (2SD) against two benchmark radiative transfer models, for their simulations of snow and sea-ice solar properties.

Compared with a 16-stream benchmark model, the errors in snow visible albedo for a direct-incident beam from all three TSA models are small (<±0.005) and increase as snow becomes shallower, especially for aged snow when the Sun is low. The errors in near-IR albedo are small (<±0.005) for solar zenith angles smaller than 75°, and increase (up to 0.1 or larger for melting snow) as solar zenith angle increases. For diffuse incidence under cloudy skies, MPAS-seaice produces the most accurate snow albedo for both visible and near-IR (<±0.0002) with the lowest underestimate (-0.01) for melting thin snow. SNICAR performs similarly to MPAS-seaice for visible albedos, with a slightly larger underestimate (-0.02), while it overestimates the near-IR albedo by an order of magnitude more (up to 0.04). 2SD overestimates both visible and near-IR albedo by up to 0.03. Unlike SNICAR, MPAS-seaice and 2SD can also simulate bare, ponded, and snow-covered sea ice. Compared to MPAS-seaice, 2SD produces higher sea ice albedos, except for bare sea ice thinner than 1 meter. More tests are in progress to evaluate their performance against a benchmark sea-ice Monte Carlo model.

Based on these offline tests, we correct the snow-covered sea ice albedo for solar zenith angles larger than 75°, merge the SNICAR snow properties into MPAS-seaice, and keep its radiative algorithm. Preliminary 1-year tests of the modified snow treatments in the MPAS-seaice driven by offline analyses show reduced summer sea-ice areas of approximate 0.5 million km2 in both hemispheres.