全球变化知识资源中心

Global mean surface temperature (GMST) fluctuates over decadal to multidecadal time scales. Patterns of internal variability are partly responsible, but the relationships can be conflated by anthropogenically forced signals. Here we adopt a physically based method of separating internal variability from forced responses to examine how trends in large-scale patterns, specifically the Interdecadal Pacific Oscillation (IPO) and Atlantic Multidecadal Variability (AMV), influence GMST. After removing the forced responses, observed variability of GMST is close to the central estimates of Coupled Model Intercomparison Project phase 5 simulations, but models tend to underestimate IPO variability at time scales > 10 years, and AMV at time scales > 20 years. Correlations between GMST trends and these patterns are also underrepresented, most strongly at 10- and 35-year time scales, for IPO and AMV, respectively. Strikingly, models that simulate stronger variability of IPO and AMV also exhibit stronger relationships between these patterns and GMST, predominately at the 10- and 35-year time scales, respectively.

Plain Language Summary Despite the smooth and steady increase of greenhouse gas concentrations, the rate of global warming has not been as stable over the past century. There are periods of stronger warming, or even slight cooling, in the global mean temperature record, which can persist for several years or longer. These changes have been linked to regional climate patterns, most notably within the Pacific and Atlantic Ocean climate systems. Climate models do not exhibit the same level of variations in these Pacific and Atlantic oscillations as compared to the observed record, and the connections between these oscillations and the global temperature are also diminished. However, there is a tendency for those models that show stronger Pacific and Atlantic oscillations to also exhibit stronger relationships between these patterns and global temperature changes.