This study compares different simple mixing schemes for one-dimensiona
l models and then focuses on the two-scale mixing approach. Two-scale
mixing consists of local diffusion between adjacent grid levels and no
nlocal mixing over the bulk of the boundary layer (nonlocal mixing). T
he latter represents nonlocal mixing by the boundary-layer scale eddie
s. A common example of two-scale mixing is the formulation of the turb
ulent heat transport in terms of an eddy diffusivity to represent smal
l-scale diffusion and a ''countergradient correction'' to represent bo
undary-layer scale transport. Most existing two-scale approaches are a
pplied to heat and moisture transport while momentum transport is simu
ltaneously parameterized only in terms of a local diffusivity without
nonlocal mixing. This study attempts to correct this inconsistency. Th
e resulting model is compared with Lidar observations of spatially ave
raged winds which are found to be superior to radiosonde and aircraft
data for determining the mean structure. The two-scale mixing correctl
y predicts the observed well mixed conditions for momentum while the o
riginal model based on a local diffusivity for momentum fails to produ
ce a well mixed state. Unfortunately, the ''best'' value for the adjus
table coefficient in the nonlocal mixing part of the two-scale approac
h appears to depend on baroclinity in a way which can not be completel
y resolved from existing data.