## Abstract

The 200-hPa divergence and streamfunction from the 30 models of the Atmospheric Model Intercomparison Project (AMIP) are compared. The data used are in the form of monthly averages and are filtered to a spatial resolution of T10, although the actual spatial resolution of the models varies from R15 to T42. The tools of the analysis are principal components analysis (PCA) and common principal components (CPC). These analyses are carried out on the 120 months of data with the climatological annual cycle removed and in the case of the streamfunction with the zonal average also removed. The AMIP period (1979–88) encompasses two El Niño–Southern Oscillation (ENSO) events (1982–83 and 1986–87), and as could be expected the ENSO characteristic response has a prominent impact in the model simulations.

The results indicate the following.

The PCA of the divergence has a dominant mode that is similar for all the models and has the signature of an ENSO response. It has an east–west dipole of divergence anomaly centered on the equator in the western Pacific. The streamfunction PC analysis also exhibits an ENSO-type response as the dominant mode, but this accounts for only 8%–21% of the variance.

The CPC analysis allows a direct comparison of the data from all the models on a common set of vectors. These results indicate that the models share a basic common pattern but there is a strong variation in the amplitude of the corresponding modes. There is less commonality in the higher components for the CPC streamfunction than seen in the divergence. This appears to be related to the stronger streamfunction response in the midlatitudes, which is presumably more affected by nonlinearity and intrinsic variability of the model integrations.

Based on results using an ensemble of five decadal runs using the European Centre for Medium-Range Forecasts (ECMWF) GCM an estimate is made of the variation of explained variance due to intrinsic variability for a single model. It is found that in general the intermodel variation is somewhat greater than the intramodel ensemble variation using the ECMWF model.

A probability density function (PDF) analysis in the space spanned by the first two CPCs for the velocity potential (which explain over 70% of the variance for all but one model) yields distinctive dynamical signatures. Some models populate a somewhat larger PDF space than others.

There is an implication that the models differ beyond the variations due to intrinsic variability in the dynamical system. Some of the models have distinctly different responses to a common SST forcing. The disparate results indicate that consensus on the representation of the physics of the atmosphere has not been reached, and the present uncertainty in the parameterizations is greater than the intrinsic uncertainty of the model system as shown by ensemble simulations.

*Corresponding author address:* Dr. James S. Boyle, Prog. for Clim. Model Diagnosis and Intercomparison, Lawrence Livermore National Laboratory, Mail Stop L-264, P.O. Box 808, Livermore, CA 94550.

Email: boyle@cobra.llnl.gov