Atmospheric Single Column Models (SCMs) provide an efficient modeling framework in which the vertical profiles of temperature and humidity evolve in response to diabatic interactions within the column and advection by the large scale circulation. The advective tendencies are either prescribed or neglected. This decoupling of the column physics from its large-scale environment can lead to rapid spurious error growth in SCM experiments, especially in the tropics. In particular, the temperature profiles from a SCM can become unrealistic enough within just a few hours to render any meaningful diagnosis of GCM physics difficult, if not impossible.

We have developed an SCM framework in which the vertical advective tendencies are coupled to the column physics. The vertical velocity at any instant is given by a formula that links the vertical temperature advection to the history of the SCM-generated diabatic heating rates upto that instant. The parameters in any such coupling formula should depend in principle depend upon the zonal, meridional, vertical and temporal scales of the heating. In practice, however, we find that the dependence is weak over a wide range of zonal and meridional scales; the vertical dependence is accounted for in the formula itself, as is also the temporal dependence by considering the time history of the diabatic forcing rather than just instantaneous values.

The effect of this dynamical coupling on the behavior of an SCM extracted from the NCAR CCM is investigated here. Because of the coupling, only the mean temperature and humidity profiles for the environment in which the column is embedded need to be explicitly specified; all other quantities are generated by the model. The coupled SCM is tested in tropical conditions during the TOGA COARE period. Control runs and 100-member ensembles, in which initial temperature and humidity profiles are perturbed, are run for environmental conditions taken from 85 sets of observed temperature and humidity profiles. The same data are also used to force the original, dynamically uncoupled, SCM.

Coupling substantially reduces the bias and variability of the SCM. Temperature and humidity profiles for each of the 85 sets of runs are maintained within realistic values in the coupled SCM, whereas in the uncoupled SCM the bias often exceeds 10 K. However, the sensitivity to initial perturbations in the two models is similar. Perhaps the most important benefit of the coupling is that it allows us to diagnose error growth in the SCM while maintaining a realistic atmospheric state.