Our goal is to develop an efficient coupled climate model that can be integrated
over many thousands of years for the purpose of long-term climate
studies.
Recognizing that the atmosphere and ocean are governed by equations that are
isomorphic to one another, we have built a coupled climate model
based on a single hydro dynamical core - the MITgcm. Separate atmospheric and oceanic models are rendered by use of
appropriate physics overlaid on the dynamics.

Novel features of our approach are:
- The dynamical core of the MITgcm exploits the isomorphism between the
equations of motion and boundary conditions that govern the atmosphere
and ocean, in pressure (p) and height coordinates (z) respectively.
- The vertical coordinate (generic name "r") is scaled by the total fluid
column thickness yielding a modified r* vertical coordinate analogous
to the atmospheric Eta coordinate. Combined with the finite volume method and a partial
cell representation of topography, the method remains accurate near
steep topography.
- An expanded cubic grid provides a global coverage of the sphere with a relatively uniform
resolution (2.8 degree at the equator) and is used in both
models, thus avoiding the problem of convergence or
meridians as the pole is approached. Careful discretization allows
one to accurately treat the whole domain, including the eight singular points at the corner of the cube.
- The use of the same horizontal grid simplifies the coupling procedure,
and conservation properties of the coupled model. The 2 components exchange fluxes and surface fields through a single processor coupling interface, using
MPI.
- An atmospheric physics package of intermediate complexity (called
SPEEDY, Molteni, 2003) enable extended (100's to 1000's of
years) integrations to be carried out. Although highly idealized the
simplified atmospheric model captures many realistic features of
atmospheric circulation. Other components include a simple land model (2
levels), an instantaneous runoff scheme and thermodynamic sea-ice (in progress).
- The model runs efficiently on a local PC cluster
enabling thousands of years of
integration to be carried out in days.
We present some preliminary results from our coupled model here.

|