Figure 2 shows the evolution with time of the maximum azimuthal velocity in a control run, and compares it with the result of running the original model of E89. Both have been initialized using the warm core vortex and boundary conditions described in E89, with parameter values listed in Table 2. The evolutions, as well as the radial distributions of variables (not shown), are very similar in the two cases. Given that the only substantive difference here is the use of a cumulus scheme based on subcloud-layer equilibrium, the similarity of the results of running both models supports the conclusion of E89 that the hurricane subcloud layer is very nearly in equilibrium, even during rapid development.
In other respects, as well, the reformulated model behaves very
similarly to the original, displaying the same sensitivity to the
control parameters and initial conditions. The sensitivity to the
convective relaxation time scale, 
, which appears in (32) is weak. On
the other hand, the model proves very sensitive to the parameters 
and 
that are present in the lower-tropospheric entropy equation, (30), as
illustrated in Figure 3. This is hardly surprising in view of the arguments
set forth in section 2.
As suggested by E89, the near saturation of a mesoscale column
of the troposphere at the cyclone core is a necessary condition for
intensification. This conjecture is also supported by the results of
a recent field experiment (Emanuel, 1994b). Only when the
troposphere is nearly saturated are the downdrafts that normally
accompany deep convection suppressed; this allows surface fluxes to
actually increase the entropy of the subcloud layer and, through moist
adiabatic adjustment,
the temperature of the troposphere. This conclusion is also
supported by an experiment in which the initial vortex is very weak,
but a mesoscale column is saturated initially. As shown in Figure 4,
the small initial disturbance grows rapidly, whereas the same
disturbance in the normal tropical atmosphere, with low 
aloft everywhere, dies.
One interesting aspect of the model's behavior is the appearance of
multiple eyewalls when the initial relative humidity of the
troposphere is high. Figure 5 shows the updraft mass flux as a
function of radius at a particular time for an experiment in which
(compared to 
in the control case). Outer eyewalls form and
gradually move inward, while the inner eyewalls dissipate, resembling
the observed behavior of concentric eyewall cycles (Willoughby, et
al., 1982). The mean position of the eyewall as well as the entire
storm circulation expand very gradually in the radial direction
during these model cycles.