Why did deep convection persist over four consecutive winters (2015–2018) southeast of Cape Farewell?

Abstract. After more than a decade of shallow convection, deep convection returned to the Irminger Sea in 2008 and occurred several times since then to reach exceptional convection depths (> 1500 m) in 2015 and 2016. Additionally, deep mixed layers deeper than 1600 m were also reported southeast of Cape Farewell in 2015. In this context, we used Argo data to show that deep convection occurred southeast of Cape Farewell (SECF) in 2016 and persisted during two additional years in 2017 and 2018 with a maximum convection depth deeper than 1300 m. In this article, we investigate the respective roles of air–sea buoyancy flux and preconditioning of the water column (ocean interior buoyancy content) to explain this 4-year persistence of deep convection SECF. We analyzed the respective contributions of the heat and freshwater components. Contrary to the very negative air–sea buoyancy flux that was observed during winter 2015, the buoyancy fluxes over the SECF region during the winters of 2016, 2017 and 2018 were close to the climatological average. We estimated the preconditioning of the water column as the buoyancy that needs to be removed (B) from the end-of-summer water column to homogenize it down to a given depth. B was lower for the winters of 2016–2018 than for the 2008–2015 winter mean, especially due to a vanishing stratification from 600 down to ∼1300 m. This means that less air–sea buoyancy loss was necessary to reach a given convection depth than in the mean, and once convection reached 600 m little additional buoyancy loss was needed to homogenize the water column down to 1300 m. We show that the decrease in B was due to the combined effects of the local cooling of the intermediate water (200–800 m) and the advection of a negative S anomaly in the 1200–1400 m layer. This favorable preconditioning permitted the very deep convection observed in 2016–2018 despite the atmospheric forcing being close to the climatological average.