Numerical Convergence of Shallow Convection Cloud Field Simulations: Comparison Between Double-Moment Eulerian and Particle-Based Lagrangian Microphysics Coupled to the Same Dynamical Core

The sensitivity of simulated nonprecipitating cumulus clouds to grid length was investigated using a large-eddy simulation model coupled to a particle-based Lagrangian cloud microphysical model (LCM) and an Eulerian cloud microphysical model (ECM). For the sensitivity experiment, the horizontal/vertical grid length was decreased from 100/80m to 6.25/5m. The results of the sensitivity experiment indicated a similar dependency of cloud cover (CC) on the grid length in the LCM and ECM, which is critical for the radiative properties of clouds. CC increased with a shorter grid length, and numerically converged with a horizontal/vertical grid length of 12.5/10m, although the three-dimensional cloud field and turbulence properties in the cloud layer did not numerically converge and the cloud fields simulated by the LCM and ECM differed. The dependency of CC on grid length originated from the dependency of the turbulence structure in the subcloud layer. Roll convection was clearly simulated in the subcloud layer using a short grid length, but it was gradually obscured with increasing grid length. With a long grid length, the shear production term of turbulent kinetic energy near the surface, which is critical for dominating roll convection, was not simulated because of insufficient vertical layers near the surface. On the other hand, with a short grid length, the number of layers close to the surface was sufficient to reproduce the shear production term, and roll convection was clearly reproduced.