Abrupt shifts in Subpolar Gyre deep convection under stable climate conditions

The potential collapse of North Atlantic subpolar gyre (SPG) deep convection under global warming has emerged as an increasingly important research topic and a significant source of public concern within the context of climate risk. While both conceptual and coupled climate models have indicated the possibility of abrupt changes in SPG circulation, a comprehensive understanding of the mechanisms behind convection shutdown remains incomplete, despite existing dynamical interpretations. Pre-industrial control simulations from coupled climate models, designed to simulate a stable pre-industrial climate state over time periods of the order of 10^3 years, have been shown to provide meaningful insights about the behavior of SPG in absence of anthropogenic global warming. In this study, we investigate the potential collapse of SPG deep convection in the pre-industrial control simulation of the Community Earth System Model 2 (CESM2) developed by the National Center for Atmospheric Research (NCAR). By analyzing the time series of mixed layer depth, we identify 15 events of winter SPG shallow convection. The temporal evolution of the SPG states leading to convection shutdown exhibits common features across different events. Notably, a positive sea ice cover anomaly east of Greenland emerges four years before the event, coupled with a negative anomaly west of Greenland and strong negative phase of the North Atlantic Oscillation the year of the event, with an abrupt sea surface cooling in the Labrador sea. Defining a causal chain, as presented in this work, could be valuable for spoiling the major feedback mechanisms involved in the process as well as for detecting dynamical early warning signals, with a possible improvement in the predictability of such convection collapse events. This working hypothesis will be tested in other models for cross-validation and compared with similar events in forced simulations to explore parallels between the autonomous (pre-industrial) and non-autonomous case.