Marine cold air outbreaks (CAOs) frequently occur over the North Atlantic during winter, producing characteristic transitions from cloud streets to closed-cell stratocumulus and, eventually, to more broken open cells. The transition to open cells reduces cloud fraction and thus plays an important role in the region's radiative energy balance. However, the drivers of this cloud break-up remain poorly understood.</p> <p>To investigate these drivers, we apply a newly developed machine learning cloud classifier to MODIS imagery, and combine it with MERRA-2 reanalysis and CloudSat/CALIPSO active remote sensing, enabling a detailed study of CAO cloud fields from the sea ice edge to break-up over the North Atlantic. By identifying break-up as the boundary between closed and open cellular convection, we relate its timing and location to precipitation, cloud phase, marine boundary layer (MBL) structure and instability.&nbsp;</p> <p>In contrast to earlier studies, we find no direct link between precipitation and break-up. Instead, MBL deepening emerges as the primary control on break-up timing and location. Specifically, break-up is observed only once the MBL deepens to approximately 2 km, a change we associate with reduced cloud-surface coupling, glaciation, and subsequent dissipation of the stratiform cloud layer. We further relate MBL deepening to inversion strength and instability, with weaker inversions and stronger instabilities promoting rapid deepening and break-up. This in turn controls the dominant cloud regime of CAOs, and thereby their radiative impact. &nbsp;