Low-frequency variability and CO 2 transient climate change. Part 2: EOF analysis of CO 2 and model-configuration sensitivity

Abstract

We used empirical orthogonal function (EOF) analysis to examine the monthly variance structure of several general circulation model (GCM) simulations to look for possible systematic changes of variability, not only due to increased carbon-dioxide (CO 2) concentration in the atmosphere but also due to model configuration. We evaluated four simulations which were present-day and doubled CO 2 experiments with the same atmospheric GCM coupled to (1) a simple nondynamic mixed-layer ocean (termed “mixed-layer model”) and (2) an ocean GCM (termed “coupled model”). Model-generated variability, as represented by EOFs of 700-mb height, is similar in all cases for global analyses and is mainly characterized by an opposition of sign between mid- and high latitudes in both hemispheres. This overall pattern does not appreciably change with a doubling of CO 2 in the models. However, there are regional changes between 1 × CO 2 and 2 × CO 2 runs which are similar for the mixed-layer and coupled models. These changes include shifts of centers of variability in the Pacific and Atlantic sectors of the Northern Hemisphere that are similar to changes in persistant height anomalies or “blocking” noted in a previous study. Changes in model configuration give rise to more extensive changes in the overall pattern of variation, with variability in Northern and Southern Hemispheres more tightly linked in the coupled model than in the mixed-layer model. We also computed EOFs using only model data for the tropics (between 30°N and 30°S). In these EOFs, differences between the two model configurations in terms of geographic centers of variability and time series power spectra are greater than between 1 × CO 2 and 2 × CO 2 cases. This is because the coupled model simulates some aspects of the El Niño-Southern Oscillation (ENSO) while the mixed-layer version does not. Consequently, different model configuration has a stronger effect on simulated interannual variability globally than does altered CO 2 forcing. Because ENSO is not represented in the mixed-layer model, CO 2-induced changes in variability are not credible in that model. For the coupled model, regional increases in variability, such as over the monsoon region of south Asia, are consistent with results from other analyses. We also evaluated CO 2 sensitivity of the coupled model's seasonal cycle of surface air temperatures using a harmonic analysis. Strongly different seasonal cycles appear in the high latitudes of the Northern Hemisphere in the coupled model under different CO 2 conditions. This phenomenon, noted in earlier studies with mixed-layer models, is apparently mostly due to snow and ice reductions with increased CO 2 that contribute to a smaller-amplitude annual harmonic and larger-amplitude half-year harmonic of surface air temperature. Similar reductions in amplitude of the annual harmonic at high northern latitudes are noted in observed surface air temperature data in a recent period compared to an earlier period in this century.