The authors propose a simple theory to explain a climatic drift previously
found in a series of coupled general circulation model (GCM) experiments. T
heir theory places the GCM results on a simple stability diagram for the th
ermohaline circulation in the Atlantic. Ir is argued that the drift is due
to Stommel's salt advection feedback and arises because the particular spin
up procedure of the coupled model has in some cases resulted in an initial
state on an unstable solution branch. Such drift may result from model init
ialization with unsuitable relaxation boundary conditions, rather than bein
g a climatic transition that could occur in the real world.