Background
The authors indicate that the planetary-scale mode of sea surface temperature swings in the North Atlantic basin, known as the Atlantic Multi-decadal Oscillation or AMO, "has attracted considerable attention in recent years due to its extensive impact on regional as well as global weather and climate," citing Ting et al. (2011). And in light of this fact ...

What was done
... Kavvada et al. compared and evaluated "the representation of AMO-related features both in observations and in historical simulations of the twentieth century climate from [climate] models participating in the IPCC's CMIP5 project," namely, CCSM4, GFDL-CM3, UKMO-HadCM3 and ECHAM6/MPI-ESM-LR.

What was learned
The three researchers report that most of the models exhibit a region of positive anomalies in the mid-latitudes, but that (1) they place "the maximum of the anomalies further to the east (southeast-ward of Greenland) than observations show," that (2) "the models also exhibit weaker positive anomalies over the Davis Strait and the Labrador Sea," that (3) they show "a weaker secondary maximum off of the northwestern African coast, in comparison to observations," that (4) "anomalies are also shown over the equatorial Pacific, a feature not present in observations," that (5) the GFDL-CM3 and CCSM4 models depict "anomalies over the equatorial Pacific that are not present in observations," that (6) both the ECHAM6/MPI-ESM-LR and the UKMO-HadCM3 models "appear unable to capture the magnitude of the observed anomalies in the subtropical/tropical Atlantic," that (7) "the latter two models also "appear unable to capture the magnitude of the observed anomalies in the subtropical/tropical Atlantic," that (8) in terms of the temporal features of the AMO indices, "the majority of the models have poor correlation with observations and under-estimate the observed variability," that (9) "higher frequency variability remains present in the model indices, in contrast to the observed AMO index," that (10) "even though the models do capture the northward focus of the observed SST anomaly maxima, they lack the ability to effectively reproduce their structure and evolution," that (11) "none of the models is able to simulate the positive salinity anomalies over the Straits around Greenland," that (12) "the models appear challenged in portraying the position and magnitude of AMO-related salinity anomalies," that (13) "most of them are unable to capture the atmospheric seasonality that is characterized by a summer minimum in the anomalies," that (14) none of the models captures the fall wave pattern over North America and parts of the northeastern Atlantic," that (15) most of the models "tend to place the maximum SST anomalies in the North Atlantic too far to the east of the Labrador Sea, in comparison to observations," that (16) the broad region of enhanced rainfall over the Guinean zone in Africa is also problematic for all four models," that (17) "AMO-related fall surface air temperature anomalies are not being fully captured, in magnitude or position, by any of the four model simulations," that (18) "the broad extension of the warming over northwestern Africa is also weakly and sparsely simulated by all four models," that (19) the models "underestimate the life span of the phenomenon by increasing variability in the 10-20 year range, to the extent that it becomes more dominant than variability in the 70-80 year range," that (20) the four CMIP5 models are "unable to portray the extension of same-sign anomalies into the tropics, during the pre- and post-mature phases of the AMO," that (21) "none of the four models employed in this study is able to capture the anomalous circulation pattern that is seen in observations during the fall season," and that (22) "the models remain unable to efficiently depict a holistic perspective of the AMO-related oceanic and atmospheric features."

What it means
In their closing remarks about the far-less-than-perfect abilities of the four AR5 models to adequately replicate real-world observations of the AMO, Kavvada et al. almost too graciously state that "without a proper incorporation of low-frequency natural variability in climate simulations, decadal predictability and the accuracy of climate projections under different climate change scenarios remain compromised."

Reference
Ting, M. Kushnir, Y., Seager, R. and Li, C. 2011. Robust features of the Atlantic multi-decadal variability and its climate impacts. Geophysical Research Letters 38: 10.1029/2011GL048712.