AQUA-PLANET EXPERIMENT (APE)

Report of Workshop at the University of Reading UK, 20-22 April 2005

                  Submitted to WGNE, January 2006


WGNE continues to recognize the value of applying atmospheric models
to very simplified surface conditions for examining the behaviors of
physical parameterizations and the interactions of parameterizations
with the dynamical cores. In particular, "aqua-planet" experiments
with a basic sea surface temperature distribution offer a useful
vehicle in this regard.  Thus WGNE endorsed an intercomparison, the
Aqua-Planet Experiment (APE), being led by staff from the University
of Reading, NCAR and PCMDI.  The details of the experiment and
schedule are available at http://www.met.reading.ac.uk/~mike/APE

The experiment is designed to provide a benchmark of current model
behavior and to stimulate research to understand differences arising
from: (1) different models, (2) different subgrid-scale
parameterization suites, (3) different dynamical cores, and (4)
different methods of coupling model dynamics and parameterizations.  A
Workshop was held 20-22 April 2005 at the University of Reading, UK to
discuss the results, summarize current model behavior and produce a
summary of research questions arising from the experiment.

Fourteen groups have now submitted their simulations to the APE
database.  Many groups had representatives at the workshop. However,
at that time the data for many of the experiments were not yet
available. Therefore comparative analyses are only now beginning on
the complete database. More complete analyses are needed before we can
say what we have learned and what are the follow-on questions.
Understanding will undoubtedly require numerous exploratory
experiments.  Some more constrained experiments to aid understanding
were discussed at the workshop and are planned.  Analysis of the APE
experiments is continuing for another year. A second workshop is
planned to discuss the more complete analyses in the Fall of 2006 or
Spring of 2007 at the University of Tokyo.

Some points brought out at the workshop are that APE is useful to test
model changes.  It can help us understand how models work, establish
sensitivities, and provide information about the coupling of
parameterizations and dynamics.  It was noted that one cannot conclude
whether a model change is beneficial solely on the basis of idealized
configurations such as APE: real-world experiments are also needed to
establish whether they produce the same signal.  The APE system is
still very complex and difficult to understand, because the
experimental strategy simplifies the surface boundary conditions but
retains all the complexities of the moist processes and feedbacks:
i.e. the models themselves are not simplified.  Caution is required to
avoid over simplistic deductions, especially when changing single
parameterizations as opposed to entire suites of
parameterizations. Nevertheless, the system can be useful to examine
mechanisms and how components interact.

The models show a wide range of behaviors with resolution and
parameter changes, both within a single modelling environment, and
across different models.  For example, there are notable differences
in tropical convective behaviour which manifest themselves in
differences in the latitudinal structure of the ITCZ, the spectra of
precipitation intensity, and the wave modes arising from convective
organization.  In mid-latitudes many models exhibit very low-frequency
zonal wave number 5 phenomena but the details differ among the models.
The basic experiments are deliberately done at "climate model"
resolutions. It was speculated that the extreme variation in behaviors
might arise because the models are not near a convergent regime.  A
few groups are examining convergence with resolution and more
resolution work is needed.

The discussions at the workshop led to plans for further diagnosis and
journal papers to be developed before the next workshop. Leaders have
been identified for each category of analysis. Action items for all
categories are to "document" where we are and "catalogue" model
simulation characteristics. The categories include tropical
variability broken down into diurnal cycle, tropical wave activity,
and asymmetric behavior arising from the SST anomaly experiments; the
mean state and meridional transports; and mid-latitude variability,
both low frequency modes and storm tracks.  There was some discussion
of future directions, such as models coupled to swamp oceans, with
aqua planet mirror runs using SST averaged from the swamp runs, to
study the role of transients and intra-seasonal variability. But these
will be left to a new project in the future.