Inter‐Model Variability of Tropical Total Ice Water Path Simulated by the Convection Permitting DYAMOND Models

Abstract Cirrus clouds have a large impact on radiation, yet coarse resolution models struggle to simulate their properties and structure realistically. Employing globally resolutions of below 5 km, the models from the Dynamics of the Atmospheric general circulation Modeled On Non‐hydrostatic Domains (DYAMOND) project do not require a parameterization for deep convection. This can be expected to lead to a better representation of cloudiness, particularly in the tropics where most of the total ice water path (TIWP) results from convection. We study the inter‐model variability of TIWP simulated by the DYAMOND models during boreal summer, compare to the TIWP from a lower‐resolution numerical weather prediction model and evaluate using satellite observations and ERA5 reanalysis data. We disentangle dynamical reasons for the inter‐model variability in monthly and daily mean TIWP from reasons connected with the cloud response to convective forcing. As expected, we find that differences in the models' cloud response are mainly responsible for the inter‐model variability. The ratio of TIWP to TIWP plus total liquid water path (TLWP), which is controlled by convective de‐ and entrainment, and the models' microphysical scheme lead to a large inter‐model variability in TIWP. Differences in convective dynamics also impact cloudiness with some models compensating a low TIWP in areas of low vertical ascent by an increased probability of large TIWP connected with large vertical velocities. Finally, the spatial distribution of convective events varies with models that parameterize convection simulating a larger fraction of the tropical TIWP in areas of high sea surface temperature.