In support of IPCC 95, PCMDI has produced two analyses of the AMIP for a chapter on model validation. These IPCC analyses have since been expanded to provide a fuller overview of model performance. This paper focuses on the "base" or "climate" state and systematic deviations of the 30 model solutions from "best" observations, i.e., "how good is the climate."
The base state is defined by seasonal means of the standard monthly mean output averaged from 1980-88. 1979 was not considered because validation data in the atmosphere was lacking, and to avoid contributions to the mean from potential transients forced by anomalous adjustments in the land surface hydrology. In addition to the climate average, the interannual variation of the seasonal mean is examined, as appropriate.
Model performance is evaluated along three broad categories: 1) "the climate," i.e., surface air temperature and precipitation climatology [as might be defined by physical geographers and the public]; 2) the "meteorology" or atmospheric general circulation features and synoptic activity as represented by monthly mean AMIP standard output variances; and 3) "energy" or the net energy fluxing into and out of the atmosphere.
The overarching conclusion is that variation among the 30 models is an order of magnitude greater than the variation among the observational estimates for many of the variables considered. Hence, while clear "losers" might be identified, clear "winners" cannot. Additional results regarding surface air temperature biases over the ocean and a severe net surface shortwave bias over the summer time midlatitude oceans will be covered at the conference.