Characteristics of precipitating convection and moisture-convection relationships in global km-scale simulations

In this study, we compare convection characteristics in three models that are at the forefront of global km-scale modelling, the ICON model developed by the Max Planck Institute for Meteorology (MPI-M) and German Weather Service (DWD), the IFS developed by the European Centre for Medium-Range Weather Forecasts (ECMWF), and the NICAM model developed by the University of Tokyo, the Japan Agency for Marine-Earth Science and Technology (JAMSTEC) and the National Institute for Environmental Studies (NIES). For IFS and ICON, we analyse 1-year coupled simulations at 4.4 and 5 km resolution, respectively, which stem from Cycle 3 of the H2020 Next Generation Earth Modelling Systems (nextGEMS) project. For NICAM, we analyse a 1-year AMIP-type simulation at 3.5 km resolution. Convection schemes have been switched off in ICON and NICAM, while in the IFS the deep convection schemes cloud base mass flux is strongly reduced. Modelling convection at km-scale resolutions is both exciting and challenging because some important processes are already resolved at these scales (e.g., deep convection) but other important processes remain under-resolved (e.g., mixing of grid-scale updrafts with their environment). Thus, we analyse in this study what common issues exist in ICON, IFS and NICAM with respect to the convection characteristics in the tropics, in what respects all models do well and where there are substantial inter-model differences. Specifically, we analyse local convection characteristics and show that compared to satellite observations, the models tend to overestimate precipitation intensity (NICAM and ICON), while they underestimate precipitation cell size and precipitation duration. We study mesoscale organisation by using different organisation metrics and show that the models tend to underestimate organisation, even though they all consistently show that when organisation is enhanced, heavy precipitation is enhanced as well. We also investigate moisture-convection relationships and show that the models generally do not moisten enough during a convective event compared to ERA5 reanalysis data. Consistently, the sensitivity of lower-tropospheric moisture variations to the life cycle of deep convection over ocean looks too weak in ICON and IFS. Finally, we look at land-ocean differences of the convection characteristics and show that while all models capture the diurnal cycle of precipitation over ocean well, there are some substantial differences over land, even though biases are not consistent between the models. Over coastal regions of the Maritime Continent, ICON has too strong mean precipitation and a too strong diurnal cycle, whereas IFS overall underestimates both, connected to a too weak propagation of convection onto the ocean during nighttime, potentially connected to too weak cold pools. Meanwhile, NICAM has more realistic convection characteristics in these coastal regions.