A new ground heat flux parameterization for land surface schemes, such as those used in climate and numerical weather prediction models, is described. Compared with other approaches that lump the canopy layer and ground surface, or empiricially based approaches that consider the effect of radiation attenuation through the canopy layer, the new parameterization has several advantages. First, the reduction of radiation available for conducting soil surface exchange under vegetated areas is represented in a manner that assures that heat is conserved in the long term. Second, problems in representing properly the phase of the ground heat flux are alleviated. Finally, the approach is relatively simple and is computationally efficient, requiring only two soil thermal layers. Comparison of the method with analytical solutions for special cases shows that the new method approximates the analytical solution very well for different conditions, and that the new method is superior to the force-restore and the Crank-Nicholson method. Model-derived ground heat heat fluxes for the French HAPEX-MOBILHY (Hydrology-Atmosphere Pilot Experiment-Modelisation de Bilan Hydrique) site and the Brazilian ABRACOS (Anglo-Brazilian Amazonian Climate Observation Study) cleared ranch land site are shown to be in close agreement with observations. Sensitivity analyses show that if the attenuation of radiation under vegetation and soil heat storage are ignored, the daytime peak and nighttime minima of ground heat flux, latent and sensible heat fluxes, and surface temperature can be significantly in error. In particular, neglecting the radiation attenuation through the canopy layer can result in significant overestimation (underestimation) of daytime (nighttime) ground heat flux, while neglecting soil heat storage can result in significant phase errors. ¿ 1999 American Geophysical Union