Two contrasting representations of land surface variability used in ge
neral circulation models (GCMs) are compared through an analysis of th
eir corresponding surface energy balance equations. In one representat
ion (the ''mixture'' approach), different vegetation types are assumed
to be homogeneously mixed over a grid square, so that the GCM atmosph
ere sees near-surface conditions pertaining to the mixture only. In th
e second representation (the ''mosaic'' approach), different vegetatio
n types are viewed as separate ''tiles'' of a grid-square ''mosaic,''
and each tile interacts with the atmosphere independently. The mosaic
approach is computationally simpler and in many ways more flexible tha
n the mixture approach. Analytical solutions to the linearized energy
balance equations and numerical solutions to the nonlinear equations b
oth demonstrate that the mixture strategy, when applied to two coexist
ing vegetation types that differ only in canopy transpiration resistan
ce, promotes both total turbulent flux and latent heat flux relative t
o the mosaic strategy. The effective differences between the strategie
s, however, are small over a wide range of conditions. In particular,
the strategies are effectively equivalent when the transpiration resis
tances of the different vegetation types are of the same order of magn
itude.