Simulating the winter North Atlantic Oscillation: the roles of internal variability and greenhouse gas forcing
- First Online:
- Cite this article as:
- Osborn, T.J. Climate Dynamics (2004) 22: 605. doi:10.1007/s00382-004-0405-1
- 498 Downloads
Analysis of simulations with seven coupled climate models demonstrates that the observed variations in the winter North Atlantic Oscillation (NAO), particularly the increase from the 1960s to the 1990s, are not compatible with either the internally generated variability nor the response to increasing greenhouse gas forcing simulated by these models. The observed NAO record can be explained by a combination of internal variability and greenhouse gas forcing, though only by the models that simulate the strongest variability and the strongest response. These models simulate inter-annual variability of the NAO index that is significantly greater than that observed, and can no longer explain the observed record if the simulated NAO indices are scaled so that they have the same high-frequency variance as that observed. It is likely, therefore, that other external forcings also contributed to the observed NAO index increase, unless the climate models are deficient in their simulation of inter-decadal NAO variability or their simulation of the response to greenhouse gas forcing. These conclusions are based on a comprehensive analysis of the control runs and transient greenhouse-gas-forced simulations of the seven climate models. The simulations of mean winter circulation and its pattern of inter-annual variability are very similar to the observations in the Atlantic half of the Northern Hemisphere. The winter atmospheric circulation response to increasing greenhouse gas forcing shows little inter-model similarity at the regional scale, and the NAO response is model-dependent and sensitive to the index used to measure it. At the largest scales, however, sea level pressure decreases over the Arctic Ocean in all models and increases over the Mediterranean Sea in six of the seven models, so that there is an increase of the NAO in all models when measured using a pattern-based index.