From Marine Heatwaves Drivers towards Machine Learning Predictability of Air-Ocean Heat Extremes [AO1]Abstract Content· 2026 rules: 250 words

A lot of research has focused on exploring the sea-air interface and how Marine Heatwaves (MHW) result from the heat exchanges between both domains. Over the North Atlantic midlatitude region, these prolonged periods of anomalously warm ocean temperatures, usually detected through Sea Surface Temperature (SST), have already been shown to have a strong atmospheric signal, where the strength and position of high-pressure systems determine MHWs onset and persistence. In addition, several studies have been formulating hypotheses built upon the contribution from both climate modes and weather regimes in the prevalence and spatio-temporal characteristics of these events, attempting to typify them, in order to support our understanding and predictability of MHWs under a changing climate. Being based in the eastern midlatitude region of the North Atlantic basin, the Portuguese Exclusive Economic Zone (EEZ) is particularly subject to these influencing factors, knowing that the Azores anticyclone strongly determines the western Iberian climate and weather, affecting both the atmospheric and oceanic circulation. And as the national investment prioritises the country’s relatively big ocean domain, it becomes of utmost importance to have the capacity to recognise recent MHWs pattern changes and improve their predictability, especially at the seasonal scales. To address this, +ATLANTIC has built a portfolio of activities to establish the empirical relationship between MHWs and synoptic weather patterns, while emphasizing the multiple contributing factors that determine their position and intensity: particularly, the role of the North Atlantic Oscillation (NAO) modes has been studied to establish how they typify changes in the air-sea energy balance components, resulting in an excess net heat gain that shifts northwards or southwards, as a function of NAO’s signal. To attain this, a post-processing routine has also been developed to filter out smaller pixel-wise SST anomalies from the synoptic-scale signal, allowing for ranking and classifying MHWs according to their spatial similarity. Results have shown significant spatial dissimilarities between the positive and negative phases of NAO, and its relationships to synoptic weather regimes. These findings contribute to a better understanding of the mechanisms underlying MHWs, and support a larger purpose of ocean-atmosphere empirical coupling, in the sense of understanding how MHWs also provide a feedback mechanism to the atmosphere, which may result in teleconnections leading to extreme heat and drought over western Europe or fuelling cyclones and storms. Preliminary results already show the significance of such an empirical relationship, where causal inference algorithms support that there is a lagged relationship between the ocean surface thermal state and the prevalence of excessive warm and dry summers. The next steps will focus on training a machine learning algorithm that can predict near-surface temperature anomalies over Europe, as a function of the North Atlantic state, with preliminary results already showing promising outcomes.