Abstract. The Intertropical Convergence Zone (ITCZ) is a key circulation and precipitation feature in the tropics. There has been a large spread in the representation of the ITCZ in global weather and climate models for a long time, the reasons for which remain unclear. This paper presents a novel approach with which we disentangle different physical processes responsible for the changeable behavior of the ITCZ in numerical models. The diagnostic tool is based on a conceptual framework developed by Emanuel (2019) and allows for physically consistent estimates of convective mass flux and precipitation efficiency for simulations with explicit and parameterized convection. We apply our diagnostic tool to a set of tropical aquachannel experiments using the ICOsahedral Nonhydrostatic (ICON) model with horizontal grid spacings of 13 and 5 km and with various representations of deep and shallow convection. The channel length corresponds to the Earth's circumference and has rigid walls at 30∘ N/S. Zonally symmetric sea surface temperatures are prescribed. All experiments simulate an ITCZ at the Equator coinciding with the ascending branch of the Hadley circulation and descending branches at 15∘ N/S with subtropical jets and easterly trade wind belts straddling the ITCZ. With explicit deep convection, however, rainfall in the ITCZ increases and the Hadley circulation becomes stronger. Increasing horizontal resolution substantially reduces the rainfall maximum in the ITCZ, while the strength of the Hadley circulation changes only marginally. Our diagnostic framework reveals that boundary-layer quasi-equilibrium (BLQE) is a key to physically understanding those differences. At 13 km, enhanced surface enthalpy fluxes with explicit deep convection are balanced by increased convective downdrafts. As precipitation efficiency is hardly affected, convective updrafts and rainfall increase. The surface enthalpy fluxes are mainly controlled by mean surface winds, closely linked to the Hadley circulation. These links also help understand rainfall differences between different resolutions. At 5 km, the wind–surface-fluxes–convection relation holds, but additionally explicit convection dries the mid-troposphere, which increases the import of air with lower moist static energy into the boundary layer, thereby enhancing surface fluxes. Overall, the different model configurations create little variations in precipitation efficiency and radiative cooling, the effects of which are compensated for by changes in dry stability. The results highlight the utility of our diagnostic tool to pinpoint processes important for rainfall differences between models, suggesting applicability for climate model intercomparison projects.