Abstract
<title>Abstract</title> <p>North Atlantic sea surface temperature (SST) variability is known to influence weather patterns over Europe and the Mediterranean. Yet the causal links between ocean conditions and atmospheric circulation remain debated. Using Convergent Cross Mapping, which tests whether changes in one variable reliably predict changes in another, we find that the ocean actively shapes atmospheric states at sub-seasonal timescales, a directional signature established by the asymmetric way SST and atmospheric height predict each other and sustained across daily-to-subseasonal lags. We identify four recurring winter SST patterns in the North Atlantic, including an "unprecedented" state that has appeared only since 2021. These patterns are extracted using Self-Organising Maps and k-means clustering, machine-learning techniques that group similar SST configurations into a small number of representative regimes. Each pattern is associated with mechanistically distinct shifts in the large-scale circulation: one promotes zonal, west-to-east flow through enhanced baroclinic activity; another suppresses Atlantic wave activity and favours atmospheric blocking. Dynamical-systems metrics, which quantify the local predictability of atmospheric states, show that the blocking-favouring regime exhibits enhanced atmospheric persistence, suggesting "windows of opportunity" for forecasts at 3 to 4-week lead times, beyond the typical 14-day deterministic limit. Overall, our results highlight an actionable pathway by which ocean conditions can extend the predictability of weather and climate extremes.</p>