On the Influence of Grid Resolution and Domain Size on the Structure and Evolution of the Stratocumulus-Topped Boundary Layer: A Large-Eddy Simulation Study

Abstract:

Improved understanding of processes related to the evolution of stratocumulus clouds is needed, e.g. for more accurate prediction of weather and climate. As a supplement to measurements, numerical simulation is a widely used and valuable tool in stratocumulus studies. However, due to limited computational resources, simulations are often run at resolutions too coarse to account for the smallest eddies involved in e.g. the entrainment process, and possibly in computational domains too small to contain the largest relevant flow structures in the boundary layer. Here we investigate how changes in domain size and spatial resolution affect key parameters such as cloud cover and liquid water path in large-eddy simulations of the stratocumulus-topped boundary layer (STBL). Details of the entrainment process and subsequent mixing within the STBL is studied by adding a passive scalar to the flow. We use a modified version of the 3D nonhydrostatic anelastic Eulerian-semi-Lagrangian (EULAG) model, and perform both simulations including an explicit sub-grid scale turbulence model and simulations in which the effect of unresolved turbulence is accounted for implicitly by the applied numerical scheme. The simulations are based on measurements from the second Dynamics and Chemistry of Marine Stratocumulus (DYCOMS-II) and Physics of Stratocumulus Top (POST) field campaigns. We show how refining the horizontal resolution facilitate development of small-scale turbulence in the cloud-top region, which enhance entrainment and tends to dissolve the cloud. Refining the vertical grid spacing, on the other hand, allows for stronger vertical temperature gradients which tend to strengthen the capping inversion and inhibit entrainment. The statistics of the flow and the evolution of the cloud is found to be more sensitive to changes in resolution than to changes in domain size. We do however observe still larger flow structures as the horizontal extent of the computational domain is increased.