Leveraging the signal-to-noise paradox to improve seasonal forecasts of the AMOC and its impacts

It has been shown that predictability of the North Atlantic Oscillation (NAO) in seasonal forecasts is better than models suggest, a consequence of the signal-to-noise paradox, whereby individual ensemble members contain a smaller proportion of the predictable variance than seen in observations. We intend to use two seasonal forecast models, GloSea6 and CESM-SMYLE, to study whether ‘NAO-matching’, where we select only the ensemble members that most closely resemble the ensemble mean NAO, can produce more accurate seasonal forecasts of the Atlantic Meridional Overturning Circulation (AMOC) than the full seasonal forecast ensemble. This method has been shown to improve predictability of other aspects of the North Atlantic climate, such as the Atlantic Multidecadal Variability pattern and Northern European Precipitation. The skill of AMOC predictability in seasonal hindcasts will be assessed against the RAPID array observations as well as historical reconstructions of the overturning circulation to determine whether it too is subject to signal-to-noise errors, and consequently if ‘AMOC-matching’ is a potentially useful calibration tool for improving predictability of its related climate impacts.