Summary of The Dynamics of Idealized Convection Schemes and Their Effect on the Zonally Averaged Tropical Circulation
by Frierson, which appears
in Journal of the Atmospheric Sciences.
Primary arguments:
- The zonally averaged tropical circulation is quite sensitive to
convection scheme parameters: the Hadley circulation and ITCZ precipitation
can be easily increased by 50% by varying convection parameters.
- The presence of some kind of shallow convection parameterization is key
to creating a useful convective parameterization within the model.
- Aspects of the convective parameterization that change the ability to
build up and release CAPE strongly affect the zonally averaged circulation.
Discussion:
In our original studies of midlatitude dynamics with this idealized
moist GCM (FHZ06 and
FHZ07), we
presented the model without convective parameterization, i.e., with
grid scale condensation only. It's often surprising to people that a
model that's strongly heated from below can be run without blowing up
without a convection scheme. The model in this configuration can
eliminate the convective instability on its own however, and the
results are quite similar to running with the Manabe moist convective
adjustment. It's a perfectly well-posed model that gives reasonable
results.
However, in order to get a better understanding of the tropical
circulation in a model with a more realistic convection scheme. we have
developed a simplified parameterization in the style of Betts-Miller.
The basic principle of this scheme is relaxation to post-convective
equilibrium profiles, to a moist adiabat for temperature, for instance.
One has to specify a post-convective humidity profile for the scheme as
well, a clear disadvantage since there is likely no such universal
profile. We use a fixed relative humidity with respect to the moist
adiabat to keep this simple.
There are several other complicating factors with the scheme as well. One
must assure that energy is conserved by making the convective
heating by the scheme equal to the convective drying. We tried
several different ways of doing this, but settled on the original Betts-Miller
method of shifting the reference profile for temperature. Then, one has
to perform some kind of shallow convection when the predicted precipitation
is negative. This ends up being quite important: if you don't use some
kind of shallow convection, it's almost like using no convection scheme at
all.
We perform a sweep of the parameters in the model in the paper to get an
idea for how convection affects the Hadley circulation and precipitation
distribution within the model. Unfortunately (from the perspective of
coming up with simple theories), the circulation is quite sensitive to
convection, with the ITCZ precipitation, Hadley cell strength, etc changing
by 50% over different parameter sets. On the other hand, other parameter
changes alter the Hadley cell remarkably little (for instance, changing
convective relaxation times by nearly a factor of 10).
The general result is that simulations in which convective instability can be
built up and then is rapidly released have a stronger Hadley cell, while
simulations in which convection can occur to higher levels more easily have a
weaker circulation. In the latter set of simulations, we argue, convection
is more effective at stabilizing the tropical troposphere, meaning there is
a larger "gross moist stability," and a smaller mass flux is needed to
flatten temperatures in the tropics. In simulations with more bursty
convection, convection is less effective at stabilizing the atmosphere, and
a large mass flux is needed to create flat temperatures without the
tropics. Therefore, we argue that the gross moist stability is not just a
diagnostic in these simulations, but is actually set to some extent by the
convection scheme.
Another key diagnostic here is the amount of precipitation that is handled by
grid-scale condensation versus the amount handled by the convection scheme,
because the grid-scale condensation tends to be much more bursty. In the
code and in the paper, the grid-scale condensation is referred to as
"large-scale condensation," but I've found that this name creates confusion.
First of all, large-scale precipitation is often associated with stratiform
precipitation, but the physics for stratiform clouds is completely different
than for the grid-scale entities within this model. Further, the
grid-scale condensation is just as likely to be associated with deep
convection and convective instability as the convective precipitation. In
fact, quite often in midlatitudes there is CAPE but not enough moisture
for the convection scheme to be called, but there is still some grid-scale
condensation at certain levels. Finally, the notion of large-scale
condensation seems to imply that convection occurs at larger scales, but
convection actually tends to occur on smaller scales with this scheme
(at the grid scale!). Therefore, I'm trying to use the term grid-scale
condensation now, as a more precise term.
For further work on the Hadley cell width, see the paper of
Frierson, Lu and Chen
(2007). For work on how the convection scheme parameters and the gross
moist stability affect convectively coupled equatorial waves, see
Frierson (2007b) and
Lin et al (2007).
The figure above shows an instantaneous time slice of precipitation from
two simplified Betts-Miller simulations, with (bottom) and without (top) a
shallow convection parameterization. Without shallow convection, the
parameterization is clearly not effective in preventing precipitation at the
gridscale.
Full citation:
Frierson, D. M. W.
The Dynamics of Idealized Convection Schemes and Their Effect on the Zonally Averaged Tropical Circulation.
Journal of the Atmospheric Sciences, 64, 1959-1976,
2007.
The official journal link can be found
here.
A PDF download of the full paper can be found here.
This download is courtesy of the American Meteorological Society, who owns sole
rights to it.
The download is subject to copyright laws and statutes. For more
information, please visit the Allen Press/AMS Journals Online website.
|
|