Horizontal Grid Spacing Convergence of Aquaplanet Experiments Using a Global-Storm Resolving Model

Abstract Aquaplanet experiments are used to investigate the statistical convergence of the Global Storm-Resolving model (GSRM) ICOsahedral Nonhydrostatic (ICON) model, under successive, two-fold horizontal grid spacing refinements from 160 to 1.25 km. A methodology based on the Richardson extrapolation method is used with the aquaplanet hemispherical symmetry to quantify convergence. We use the symmetrical and anti-symmetrical solution components to estimate the asymptotic convergence pattern, the asymptotic estimate, and sampling uncertainty. Based on successive horizontal grid refinements, different climate statistics are explored to determine whether they enter a convergent regime and, if so, what their convergent value is. Our analysis focuses on global-mean statistics related to the general circulation and aspects that influence the climate: the meridional overturning circulation, the tropical structure (the Inter-Tropical Convergence Zone and the zonal mean thermodynamic state), and the energy and water budget. Our results show a kilometer and hectometer-scale horizontal grid spacing requirement for statistical convergence of the meridional overturning circulation structure and global mean statistics. Distinctively, the tropical structure is estimated to be very near its asymptotic values at km-scale grid spacing, but its intensity appears to converge more slowly, as do the storm-track and jet-stream. As we increase the horizontal grid spacing, cloud reduction and the zonal distribution of water vapor convergent pattern drive convergence in the energy and water budgets. We conclude that the ICON GSRM without convection parameterization exhibits statistical convergence at 10 km horizontal grid spacing in aquaplanet experiments across many of the metrics studied, specifically in the large-scale and tropical vertical structure.