Abstract The presence of Arctic clouds plays a crucial role in the evolution of the surface temperature of Arctic sea ice. However, large biases in cloud representation remain in state‐of‐the‐art weather and climate models. In this study, we use observational data from the one‐year Arctic ship campaign Multi‐disciplinary drifting Observatory for the Study of Arctic Climate (MOSAiC) to evaluate the Integrated Forecasting System (IFS), the global weather prediction model of the European Centre for Medium‐range Weather Forecasts (ECMWF). We find seasonal biases in the cloud liquid water path, which is underestimated in winter and overestimated in summer. We show that the occurrence of supercooled liquid clouds with strong supercooling (below ) is underestimated, while the occurrence of liquid‐containing clouds close to freezing temperature is overestimated. Focusing on winter, we investigate the sensitivity of supercooled liquid clouds to different model uncertainties in a single‐column model setup. We find the strongest sensitivity is to uncertainties in cloud microphysics and show that reducing the assumed ice particle number concentration (within the bounds of uncertainty) reduces the ice deposition rate and improves the underestimation of supercooled liquid clouds significantly. As a result, the long‐wave downward radiation at the surface improves. Positive effects on the boundary‐layer structure are shown in a case study. Furthermore, we document the sensitivity of clouds to ice‐particle fall speed and to small fractions of open ocean in the sea ice, which are often present in the model when satellite observations show full sea‐ice cover. Both sensitivities are minor compared with the sensitivity to uncertain assumptions that affect the ice deposition rate directly. Our results can guide future model development to reduce Arctic cloud biases.
Schulte et al. (Sun,) studied this question.