Cold-season simulations of the Noah land-surface model have shown excessively large sublimation which results in lower snow water equivalent amounts and early springtime snowmelt when compared with observations from 110 Snotel stations and other land models (e.g., VIC and MOSAIC). Tests also show large negative sensible heat fluxes in stable conditions when compared to observations from four surface flux sites in the U.S. and those simulated by VIC and MOSAIC models which use different surface-layer schemes. Cold-season model biases may be due in part to the treatment of the surface-layer, and in part to the snow albedo. We use the North-American Land Data Assimilation System (NLDAS) 2005 annual forcing data for the NLDAS domain (continental U.S.) as a testbed, and select six different surface-layer schemes to test in cold-season simulations by the Noah model. Examining snow water equivalent, sensible heat, latent heat, skin temperature, and aerodynamic conductance show that different surface-layer schemes, as well as the snow albedo formulation, have a significant impact on cold-season simulations by the Noah model. After modifying the surface-layer scheme and adjusting the snow albedo formulation, Noah model tests are extended to coupled land-atmosphere simulations in the North American Mesoscale (NAM/WRF-NMM) model.