Assessing Potential Seasonal Predictability with an Ensemble of Multidecadal GCM Simulations

Abstract

A global search for areas where seasonal prediction may be feasible has attracted scientific interest for many years. This contribution is based primarily on data from a six-member ensemble of 45-yr climate runs, each of which is forced by observed sea surface temperatures (SSTs) and sea-ice extents and are unique only in their initial atmospheric conditions. The potentially predictable component of atmospheric interannual variability is assumed to be that due to oceanic forcing, and using ?analysis of variance,? this is separated from the unpredictable internal component; potential predictability is measured as the ratio of ocean-forced variance to total variance. Significance levels and confidence intervals are calculated, both of which are essential for a meaningful interpretation of predictability estimates; the latter show there is low susceptibility to sampling problems here because of the large data source available. Global maps of potential predictability, for simulated seasonal mean precipitation and mean sea level pressure (MSLP), are shown for both the solstitial and equinoctial seasons. In most regions, this model-based predictability estimate has large variations through the annual cycle. Not surprisingly, the highest predictability occurs over the tropical oceans, particularly the Atlantic and Pacific, for which a better knowledge of the influence of SSTs on diabatic heating is important for understanding the variability of teleconnected regions. Land areas displaying high predictability tend to support existing empirical studies, although over Australia and parts of Africa the model?s response to SSTs seems erroneously weak. In the midlatitude Northern Hemisphere, a winter?spring peak of predictability is confirmed, but a notable autumnal minima of predictability is also proposed. At polar latitudes, there is a small but significant influence of SSTs on spring MSLP, and in some localities a moderate influence on precipitation. Further work is required with observational data to properly assess these findings.