The convergence behavior of a cloud resolving simulation across kilometer-scale grid spacings is illustrated. Panel (a) shows snapshots of the cloud liquid water content of clouds within a 350 km x 350 km large subdomain of the full modeling domain (located over the southwestern parts of the European Alps). The refinement of the grid spacing decreases the size of individual clouds and increases the number of convective clouds. Panel (b) shows the nine-day mean diurnal cycle of surface rain rate averaged over a larger domain covering the whole Alpine mountain range. The magnitude and timing of surface precipitation are largely insensitive to the horizontal grid spacing.
The explicit treatment of convection using models with grid spacings less than 4 km has led to considerable improvements of quantitative precipitation forecasts and climate simulations and provided remedy to issues that plagued convection-parameterizing models (CPMs) for too long. However, there is a continued quest for even finer grid spacings to resolve the energy spectrum of atmospheric motion across the kilometer scales. This study explores convergence characteristics of bulk heating and moistening rates related to an ensemble of deep convective clouds in real-case simulations in order to address resolution requirements in regional climate/weather simulations.
Why do we care?
Ideally, a cloud resolving model operates at a grid spacing fine enough such that the bulk feedback of convective clouds to the larger scale is unaffected by any further refinement of the numerical mesh. If - at a given resolution - a cloud resolving simulation has not yet converged, we can put little faith into its projections of, e.g., the future water cycle. A demonstration of convergence is thus key to underline the credibility of cloud resolving simulations.
How do we approach this?
A nine-day long summer period is simulated with widespread deep convection occuring over the European Alps each day. During this high-pressure period the flow remains unaffected by synoptic-scale disturbances. The domain is large enough to capture the whole Alpine mountain range. Convergence is tested across the kilometer scales using grid spacings of 4.4, 2.2, 1.1, and 0.55 km. The metric under investigation is the mean diurnal cycle of bulk heating and moistening rates that relate to the ensemble of several convective cells (and days).
What do we find?
Although individual convective clouds become smaller and smaller with smaller grid spacings (Fig. a), little sensitivity is found in terms of average surface precipitation (Fig. b). The simulated feedback to the large-scale flow (as measured by the bulk tendencies) converges across the investigated range of scales. A 4.4 km run is largely indifferent to a 0.55 km run if such bulk metrics are considered. Although these results might hold only for orographically triggered convection, these findings are particularly encouraging as it seems unnecessary to resolve small-scale turbulent eddies in order to simulate the area-averaged precipitation. Equivalent studies are required over flat terrain to confirm this insensitivity for climate regimes less influenced by the persistent forcing from thermally-driven flows.