The Mark 1 version of the CSIRO9 GCM was used in the AMIP simulation. Three realizations of the simulation were made to give a preliminary assessment of the impact of chaos on climatic variability.
The current Mark 2 version of the CSIRO9 GCM exists in R21, R42, and
T63 horizontal resolutions. The major differences between the Mark 1 and
Mark 2 versions are:
i. A biospheric submodel, in place of the previously used bare earth
formulation, with 11 plant types and 3 soil types.
ii. A dynamic sea-ice submodel in place of the previously static sea-ice
representation.
iii. Semi-Lagrangian water vapor advection confined to the transform
grid instead of the previously spectral representation.
The Mark 2 model has been used in climatic change experiments involving equilibrium 1 x and 2 x CO2 runs with slab oceans, and also in a transient CO2 run with a fully coupled, flux-corrected dynamic ocean. Multi-seasonal predictions have been made with the R42 version of the model. The T63 version is being used currently in the GISST experiment involving running the model with observed SST for 1871-1993. The model has reached the 1920s at the time of writing.
A number of systematic errors exist in both the Mark 1 and Mark 2 versions of the model. A problem which exists across all model variants is the underestimation of the strength of the Northern Hemispheric jet in summer. This can be, at least partially, attributed to the surface temperatures being too high over land in summer thus reducing the latitudinal temperature gradient. This surface temperature problem may be associated with deficiencies in the cloud amount and evaporation rate. The model also fails to reduce the Gobi desert. Calculations of the model biome demonstrate the above deficiencies quite clearly.
Substantial problems exist at the model atmosphere-ocean interface. This is best illustrated by running the coupled atmospheric-oceanic model without flux correction. This reveals a systematic cooling over the Northern Hemispheric oceans and a warming over the Southern Hemispheric oceans. The magnitude and spatial patterns of the temperature errors vary somewhat with season but persist throughout the year. These errors can be attributed to deficiencies in the total cloud amount and also to the computation of the energy fluxes at the oceanic surface.
Consistent problems have also been noticed with the model monsoon systems.
It has proved difficult to simulate adequate Indian and Australian monsoonal
rainfall simultaneously. This appears to be related to the convective parameterization
used. A number of other less important but annoying problems persist in
the model. Development of a Mark 3 model has
commenced.