In the stable atmospheric boundary layer (SBL), the interaction of processe
s in numerical weather prediction (NWP) models seems to test their represen
tation more stringently than their separate 'validation'.
Some SBL schemes derived from micrometorological research seem to allow a '
decoupling' behaviour when implemented in NWP. That is, turbulence dies out
from the ground upwards. Such 'decoupling' of the surface from atmospheric
fluxes can permit dramatic and possibly unrealistic falls in surface tempe
rature. This study traces the mechanism of model decoupling, asks whether t
his behaviour is in any sense correct and considers the implications.
It is shown that decoupling can occur in idealized single-column models, or
iginating from an unstable boundary-mode. This behaviour can depend critica
lly on parameters such as surface roughness and soil thermal diffusivity as
well as turbulence. But the turbulence dependence arises through the respo
nse of the boundary layer as a whole, and not just the surface-layer scheme
. Such decoupling arises from the 'physics', rather than the finite-differe
nce schemes, and appears to occur sometimes in the real atmosphere.