A Hypothesis for the Sensitivity of AGCMs to Horizontal Resolution at Hydrostatic Scales

Variable and adaptive grids are a promising method for improving weather prediction and regional climate projections using global atmospheric general circulation models (AGCMs). These non-uniform grids require closure models that can faithfully represent unresolved scales for a range of truncation scales represented on the grid. The widely reported sensitivity of AGCM solutions to horizontal resolution indicates that present closure formulations are inadequate for non-uniform grids. An explanation for the sensitivity of the Community Atmosphere Model (CAM) to horizontal resolution is developed from a convergence study in an idealized aqua-planet configuration. We hypothesize that an increase in horizontal resolution results in a reduction in horizontal scale of the diabatic forcing from the physics package, facilitating fine scale flow and faster resolved vertical motion by the dynamical core, consistent with linear Boussinesq theory. A set of dry dynamical core tests have been developed to test our hypothesis, using a simple warm bubble as a surrogate physics forcing. The results of our dry tests are consistent with Boussinesq theory and our hypothesis. We discuss the implications of our hypothesis, including the development of `scale-aware' closures for hydrostatic and non-hydrostatic models.