Abstract
On the basis of the evidence available to date it is concluded that the two most important faults of early baroclinic models, 1) overdevelopment in terms of a general increase in kinetic energy and 2) failure to amplify cyclone-scale, baroclinic wave disturbances which amplify in the atmosphere, were due, respectively, to the absence of a dissipation term to balance the kinetic energy generated in the model and to space truncation which imposed too large a minimum scale for amplification.
The solution to these problems is to reproduce in a model the scale dependence in the atmosphere of net development (development minus dissipation). The three paths available to achieve this goal are: 1) decrease the minimum resolvable scale (grid size), 2) reduce the space truncation of finite difference operators, and 3) increase the scale dependence of the dissipation term so that it removes energy only from the smallest permitted scales.
The current best estimate of the residence time for the total kinetic energy of the atmosphere is 2–4 days. Since about 70% of the total kinetic energy dissipation in the atmosphere occurs above the Ekman layer, it is unlikely that this can be adequately simulated by a surface friction term alone.
A possible source for a characteristically distinct behavior of 2-level models is proposed.
