We tested three atmospheric surface-layer parameterization schemes (Mellor-Yamada level 2, Paulson, and modified Louis), both in a 1-D mode in the new NCEP land-surface scheme against long-term FIFE and HAPEX observations, and in a coupled 3-D mode with the NCEP mesoscale Eta model. The differences in these three schemes and the resulting surface exchange coefficients do not, in general, lead to significant differences in model simulated surface fluxes, skin temperature, and precipitation, provided the same treatment of roughness length for heat is employed. Rather, the model is more sensitive to the choice of the roughness length for heat. To assess the latter, we also tested two approaches to specify the roughness length for heat: 1) assuming the roughness length for heat is a fixed ratio of the roughness length for momentum, and 2) relating this ratio to the roughness Reynolds number as proposed by Zilitinkevich. Our 1-D column model sensitivity tests suggested that the Zilitinkevich approach can improve the surface heat flux and skin temperature simulations. A long-term test with the NCEP mesoscale Eta model indicated that this approach can also reduce forecast precipitation bias. Based on these simulations, in January 1996 we operationally implemented the Paulson scheme with the new land-surface scheme of the NCEP Era model, along with the Zilitinkevich formulation to specify the roughness length for heat.
