Christopher O'Reilly

Long-range winter predictions over the Euro-Atlantic sector have demonstrated significant skill but suffer from systematic signal-to-noise errors. In this study we examine early winter seasonal predictability in 16 state-of-the-art seasonal forecasting systems. Models demonstrate skill in the hindcasts of the large-scale atmospheric circulation in early winter, which mostly projects onto the East Atlantic pattern. The predictability is strongly tied to the ENSO teleconnection to the North Atlantic, though the models' response to ENSO is systematically too weak. The model hindcasts of the East Atlantic index exhibit a substantial signal-to-noise errors, with the models predicted signal generally being smaller than would be expected for the observed level of skill. The signal-to-noise errors are found to be strongly dependent on the strength of the ENSO teleconnection in the models, with models with a weaker teleconnection displaying a larger signal-to-noise problem. It is demonstrated that the dependency on model ENSO teleconnection strength can be explained using a simple scaling relationship derived from a toy model. Further analysis reveals that the strength of the ENSO teleconnection in the model is linked to climatological biases in the behaviour of the North Atlantic jet. Models that better represent the dynamics of the jet over the northern part of the basin - with more frequent poleward jet excursions and less frequent Greenland blocking - are better at representing the ENSO teleconnection to the North Atlantic in early winter, with lower associated signal-to-noise errors.