In order to estimate the upper limit of decadal predictability or the so called "potential predictability", perfect ensemble simulations are performed with EC-EARTH.
Six ensembles with 5 members each are started from different years of a 200 year long present day climate simulation. The ratio between the variance of the ensemble members and the variance in the long climate simulation is a measure for the potential predictability (P(t)=var(ens) (t)/var(climsim)). A value near 0 means that the spread between the ensemble members is small compared to the variability in the long climate simulation and thus the potential predictability is high. A value near 1 means that the spread is of the same size than that in the long simulation and thus there is no predictability.
The results show that the atmospheric circulation (SLP) is only in very few regions significant predictable in the first decade after start of the prediction. Temperature shows significant predictability particularly over the North Atlantic and over the southern Pacific and Atlantic. Even some areas on the continents like parts of Africa, China and the Middle East show a potential for decadal predictability.