will change as a function of energy consumption, land cover, and other drivers of change in Earth’s radiation balance. On such time-scales, the system loses its memory of the initial conditions, and the future trajectory of the coupled climate system is strongly determined by the external forcing.
Recently observations and models have suggested that it might be possible to make decadal climate predictions by using knowledge about regularities in the natural climate system on that time-scale, especially conditions involving the state of the ocean. However, climate projections based on emission scenarios indicate that decadal-scale variability is also influenced by the accumulated impacts of anthropogenic radiative forcing. Decadal-scale predictions therefore require information both about radiative forcing (e.g., levels of greenhouse gases and aerosols) and about the current observed state of the atmosphere, oceans, cryosphere, and land surface (Hurrell et al., 2010).
Understanding and making predictions of decadal variability is still very much in its infancy, and evaluating these predictions is a major challenge. It is straightforward to verify daily weather forecasts through statistical and historical data, but verification is more difficult for decadal forecasts. Observational records are simply not consistent enough or long enough to quantify prediction skill. Even the basic characteristics and mechanisms that describe climate on a decadal scale are poorly documented and not well understood. Testing models against observed climate variability provides some means of verification and thus can offer some confidence in using these models to simulate future climate, but even those efforts are hindered by a lack of subsurface ocean observations and satellite data.
As part of the work for the next Intergovernmental Panel on Climate Change (IPCC) report, modeling centers have coordinated decadal hindcast and prediction experiments covering the period from 1960 to 2035 in what is known as the Coupled Model Intercomparison Project Phase 5 (CMIP5). Part of the work involves an atmosphere–ocean general circulation model with a resolution of approximately 50 kilometers (31 miles) being run to make decadal predictions out to 2035. Lower-resolution versions of the same model will be run with coupled carbon cycles and biogeochemical processes to help quantify the magnitude of important feedbacks that will determine the degree of climate change in the second half of the 21st century (Hurrell et al., 2009).
A better knowledge of the initial conditions needed to initialize the models properly should increase the skill of predictions of the climate system on varying time-scales. An increased understanding of the physical mechanisms that govern climate variability is also critical. In sum, although physical climate science is able to project climate trends based on scenarios for increases in greenhouse gases (GHGs) and to estimate changes in the likelihoods of occurrence for some kinds of climate events in the coming