Fig. 2: The effect of ensemble size on NAO predictability.
From: NAO predictability from external forcing in the late 20th century
a We create an ensemble of increasing size (x-axis) by subsampling (with replacement) from all ensemble members, calculate the filtered ensemble mean, and finding the ACC with observations (black) and a random ensemble member (blue). For each ensemble size, we repeat this process 10,000 times and report the mean (thick dot) and 95% confidence interval (colored cloud). When considering a random ensemble member, that member is removed from the pool of potential ensemble members. b ACC of the individual ensemble members (open markers) and ensemble means (closed markers) for the individual models in the MMLEA compared to the best ACC for the NAO we produce (0.78 from Fig. 1b; blue dashed line).
