by A. Bellucci (CMCC), R.J. Haarsma (KNMI), N. Bellouin (), B. Booth (), C. Cagnazzo ()B. van den Hurk (KNMI)N. Keenlyside ()T. Koenigk ()F. Massonnet ()S. Materia ()M. Weiss (KNMI)
Progresses in understanding the role of sea-ice, land surface, stratosphere and aerosols in decadal scale predictability are reviewed, and the perspectives for improving the predictive capabilities of current Earth system models (ESM) are discussed. Numerous extra-oceanic processes active over the decadal range are arguably established, but state-of-the-art climate models only reproduce a few of these with some degree of accuracy, while the vast majority is still either under-represented or totally missing from the spectrum of model-resolved processes.
Potential predictability associated with the afore-mentioned, poorly represented/scarcely observed constituents of the climate system, has been primarily inspected through numerical simulations performed under idealized experimental settings. Their impact, however, on practical decadal predictions, conducted with realistically initialized full-fledged climate models, is still largely unexploited.
Enhancing initial-value predictability through an improved model initialization appears to be a viable option for land surface, sea-ice and, marginally, the stratosphere. Similarly, capturing future aerosol emission storylines might lead to an improved representation of both global and regional short-term climatic changes.
In addition to the obvious benefits stemming from the refinement of the initial and boundary-value predictability, a key role on the overall predictive ability of ESMs is expected to be played by an accurate representation/inclusion of processes associated with specific components of the climate system. These act as 'signal carriers', transferring across the climatic phase space the information associated with the initial state and/or boundary forcings, and dynamically bridging different (otherwise unconnected) sub-systems. Through this mechanism, Earth-system components trigger low-frequency variability modes, thus extending the predictability beyond the seasonal scale.