Dominant microphysical processes for mixed-phase clouds across climate models

Global climate models poorly represent mixed-phase clouds in a realistic way, which leads to uncertainties in cloud radiative forcing and precipitation. In the FORCeS ice experiment (FOR-ICE) we compare three global climate models (ECHAM-HAM, NorESM, EC-Earth) and show which processes are crucial for a realistic representation of cloud ice and supercooled water in each global climate model framework using the factorial method as a statistical approach. A specific focus of the experiments is on secondary ice production (SIP) - which apart from one mechanism (rime splintering) is not represented in models, even if observations of ice crystal concentrations of ice crystal number in warm mixed-phase clouds often exceed available ice nuclei by orders of magnitude. We evaluate the importance of three SIP mechanisms combined (rime splintering, ice-ice collisions, and droplet shattering) compared to all other processes that can modulate ice mass and number in mixed-phase clouds: ice nucleation, sedimentation, and transport of ice crystals. Satellite observations are used to evaluate the representation of mixed-phase clouds. We found large discrepancies in dominant microphysical processes for mixed-phase clouds across the investigated climate models.