Marine cold air outbreaks (MCAOs) create conditions for hazardous maritime cyclones known as polar lows, which pose risks to marine infrastructure. For marine management, MCAO predictions would be highly beneficial. Previous studies explain the genesis of MCAOs, while predictability and large-scale causal drivers of MCAOs remain largely unstudied. We investigate (i) the ability of the Earth System Model from the Max-Planck Institute for Meteorology (MPIESM) to predict MCAOs and (ii) options to improve predictability of MCAOs through their large-scale causal drivers. To identify MCAO preconditions, we utilize the atmospheric reanalysis ERA-Interim in the lagged crosscorrelation analysis, composite analysis, and causal effect network (CEN). The results show that the MPI-ESM has high prediction skill for MCAOs over the Barents Sea (BS), Greenland-Iceland-Norwegian Seas and the Labrador Sea for about 2.5 weeks ahead starting from the November initial conditions. Whereas the lagged cross-correlation analysis indicates relationship between the early-fall atmospheric and sea-ice conditions and the late-fall BSMCAO index, CEN identifies the causal link only from the Arctic sea ice concentration (SIC).