This study evaluates the ability of global climate models that comprise the Coupled Model Intercomparison Project phase 3 (CMIP3) to simulate intrinsic decadal variations detected in the observed North Atlantic sea-surface temperature (SST) record via multi-channel Singular Spectrum Analysis (M-SSA). M-SSA identifies statistically significant signals in the observed SSTs, with time scales of 5–10, 10–15, and 15–30 yr; all of these signals have distinctive spatiotemporal characteristics and are consistent with previous studies.

Many of the CMIP3 20th-century simulations are characterized by quasi-oscillatory behavior within one or more of the three observationally motivated frequency bands specified above; however, only a fraction of these models also capture the spatial patterns of the observed signals. The models best reproduce the observed quasi-regular SST variations in the high-frequency, 5–10-yr band, while the observed signals in the intermediate, 10–15-yr band have turned out to be most difficult to capture. A handful of models capture the patterns and, sometimes, spectral character of the observed variability in the two or three bands simultaneously.

These results imply that the decadal prediction skill of the models considered — to be estimated within the CMIP5 framework — would be stratified according to the models’ performance in capturing the time scales and patterns of the observed decadal SST variations. They also warrant further research into dynamical causes of the observed and simulated decadal variability, as well as into apparent differences in the representation of these variations by individual CMIP3 models.

Last Updated: 2009-12-17