WIREs Climate Change | Farneti 
To estimate the anthropogenic contribution to climate signals in the recent past and future decades implies a certain degree of confidence in both understanding and simulating natural internal variability at interdecadal time scales. If we are to embark on the challenge of decadal prediction, we must be able to mechanistically attribute events to known processes and phenomena, and reproduce their features and statistics within our models. To date, models have succeeded in reproducing only partially spatial patterns, statistics and climatic impacts of interdecadal modes of variability. Reasons for the partial success and agreement among models are to be attributed to the short observational record, the different and complex flavours of coupling between the many subcomponents of the climate system, and the present inability to resolve all climate processes. At an even more fundamental level, this difficulty is aggravated by the limited understanding of the physical mechanisms involved. Here, we review the proposed mechanisms giving rise to interdecadal climate variability, we discuss the hypotheses explaining the main interdecadal modes of variability, and present an overview on the ability and level of agreement in their simulation by the latest generation of coupled climate models. To achieve any progress, the modeling community should focus on both improving the representation and parameterization of key ocean physical processes and obtaining a firmer grasp on the physical mechanisms generating the variability. Both goals can benefit from process studies, intercomparisons with perturbation experiments to study model’s sensitivities, and the use of a hierarchy of climate models.
Full text can be found here.