The systematic evaluation of coupled climate models was only beginning to emerge at the time of the IPCC WGI Second Assess-ment Report (IPCC, 1996) (hereafter SAR). Suitable formalisms for evaluating fully coupled models were in very early stages of development whereas considerable progress had been made in the evaluation of the performance of individual components (atmosphere, ocean, land surface and sea ice and their interactions).
The need for flux adjustment and the widely varying spin-up methodologies in
coupled models were areas of concern and the need for a more systematic evaluation
of these on the simulated climate was expressed. It was noted that, although
flux adjustments were generally large in those models that used them, the absence
of flux adjustment generally affected the realism of the simulated climate and
could adversely affect the associated feedback processes. It was hoped that
the need for flux adjustment would diminish as components were improved.
A new feature of coupled model evaluation was an analysis of the variability of the coupled system over a range of time-scales. The new opportunities that provided for more comprehensive evaluation were an important component of the overall assessment of coupled model capabilities.
Evaluation of the performance of individual components of the coupled system (especially for atmosphere-only models) was much more advanced than previously. Results from the first phase of the Atmospheric Model Intercomparison Project (AMIP) demonstrated that "current atmospheric models generally provide a realistic portrayal of the phase and amplitude of the seasonal march of the large-scale distribution of temperature, pressure and circulation". The simulation of clouds and their seasonal variation was noted as the major source of uncertainty in atmospheric models. In general it was found that "atmospheric models respond realistically to large-scale sea surface temperature (SST) patterns" and hence can reproduce many facets of interannual variability. In the case of the land-surface component, however, it was noted that "the general agreement found among the results of relatively simple land-surface schemes in 1990 has been reduced by the introduction of more complex parametrizations". Ocean and sea-ice models were found to "portray the observed large-scale distribution of temperature, salinity and sea ice more accurately than in 1990", but some reservations were expressed regarding the possible adverse effects of the relatively coarse resolution of the ocean components of current coupled models.
The overall assessment of coupled models was that "current models are now able to simulate many aspects of the observed climate with a useful level of skill" and "model simulations are most accurate at large space scales (e.g., hemispheric or continental); at regional scales skill is lower".
Other reports in this collection