Is the NAO signal-to-noise paradox exacerbated by severe winter windstorms?

The signal-to-noise paradox is known to be a limitation in multiple seasonal and decadal forecast models where the model ensemble mean predicts observations better than individual ensemble members. This ‘paradox’ occurs for different parameters, like the NAO, temperature, wind speed or storm counts in multiple seasonal and decadal forecasts. However, investigations have not yet found the origin of the paradox. First hypotheses are that weak ocean – atmosphere coupling or a misrepresentation of eddy feedback in these models is responsible.

Our previous study found a stronger signal-to-noise error in windstorm frequency than for the NAO despite highly significant forecast skill. In combination with the underestimation of eddy feedback in multiple models, this led to the question: Might the signal-to-noise paradox over the North-Atlantic be driven by severe winter windstorms?

To assess this hypothesis, the signal-to-noise paradox is investigated in multiple seasonal forecast suites from the UK Met Office, ECMWF, DWD and CMCC. The NAO is used to investigate the changes in the paradox depending on the storminess of the season. The results show a significant increase of the NAO-signal-to-noise error in stormy seasons in GloSea5. Other individual models like the seasonal model of the DWD or CMCC do not show such a strong difference. A multi-model approach, on the other hand, shows the same tendency as GloSea5. Nevertheless, these model differences mean that more hindcasts are needed to conclusively demonstrate that the signal-to-noise error arises from Atlantic windstorms.