Abstract We compare three global kilometer‐scale models (ICON, IFS and NICAM) to clarify the advantages and challenges of high‐resolution global weather and climate modeling, using different approaches to represent convection, from fully parameterized to fully explicit. Our analysis focuses on tropical precipitation characteristics spanning a wide range of spatio‐temporal scales—including the diurnal cycle, extreme precipitation, convective organization, and the Madden‐Julian Oscillation (MJO)—along with interactions between convection and the thermodynamic environment. All three models commonly show weaker convective organization with smaller precipitation cells than observed, though the strength of the bias varies by model. This diversity is introduced by differences in the representation of (a) convective initiation affected by the convective sensitivity to moisture and (b) tropospheric moistening associated with deep convection. Models with stronger thermodynamic‐convection coupling increase environmental moisture near convection, thereby enhancing convective organization. This has important upscale effects on the MJO; while IFS and NICAM capture its eastward propagation well, ICON has difficulty reproducing it. The amplitudes and phases of precipitation diurnal cycles over land show much greater disagreement among the models than over ocean, influenced by how convection is initiated. Biases in rain evaporation and cold pool formation hinder the propagation of mesoscale convection, leading to errors such as the misrepresentation of nocturnal convection moving off the coast of Sumatra in IFS and ICON. These results highlight the importance of thermodynamic‐convection coupling in realistically simulating tropical convection across scales. To improve this coupling, kilometer‐scale models require better representation of the interaction between resolved convection and three‐dimensional turbulent mixing.
Takasuka et al. (Sun,) studied this question.