Predictability and Mechanisms of North Atlantic Oscillation Variability—A Signal-to-Noise Paradox

Recent advances in climate modelling have resulted in the achievement of skilful seasonal prediction, particular that associated with the winter circulation over the North Atlantic. The North Atlantic Oscillation (NAO) is now a significant source of predictability for seasonal forecasts in the UK, Europe and North America for winters a few months to a year ahead (e.g. Scaife et al 2014, Stockdale et al 2015, Dunstone et al 2016), with teleconnections that can influence much of the Northern Hemisphere. However, while highly significant and potentially useful skill exists, the signal-to-noise ratio of the ensemble mean to total variability in these ensemble predictions is anomalously small (Scaife et al 2014) and the correlation between the ensemble mean and historical observations exceeds the proportion of predictable variance in the ensemble (Eade et al 2014). This means the real world is more predictable than our climate models. Here we investigate how well global climate models can simulate the NAO variability, and discuss possible causes of the anomalously weak predicted signals in our seasonal hindcasts.