The NAO decadal predictability determined by initial ocean heat content anomalies in the subpolar North Atlantic — SST gradients playing a key role.

In recent studies using large ensembles, the North Atlantic Oscillation (NAO) has been shown to exhibit significant decadal predictability stemming from skillfully predicted sea surface temperature (SST) anomalies in the subpolar North Atlantic (SPNA).  In turn, various studies have demonstrated that the decadal SST predictability in this area is dominantly due to ocean initialization. It remains unclear, however, which component of the oceanic initial conditions determines the evolution of the SPNA SSTs and the NAO in the following years, and through which physical processes this is accomplished.

Here we assess the role of initial upper-ocean heat content (OHC) anomalies in the SPNA in four decadal prediction systems (DPSs) exhibiting significant skill for the wintertime NAO. First, using observations, it is found that the NAO averaged in several successive winters is significantly correlated with the SPNA OHC in the November preceding the first winter.  Second, it is shown that this relationship holds also in the DPSs, and it is stronger in the systems that exhibit higher skill for the NAO itself.  Finally, we discuss the causal chain that leads from skillfully predicted SSTs to the NAO predictability via changes in low-level baroclinicity and a key positive feedback internal to the atmosphere.

Even though multi-decadal variations in the Atlantic Meridional Overturning Circulation (AMOC) may play a key role in determining respective historical variations in the SPNA OHC, no AMOC anomalies were found in the initial conditions of the hindcasts that could explain the subsequent evolution of the NAO.  Of course, this result does not preclude an important role for the AMOC in real-world NAO predictability.  Our findings advance the understanding of the mechanisms underlying decadal predictability and raise new questions regarding the role of model fidelity and ocean–NAO feedbacks in relation to the signal-to-noise problem.