Estimating Spatial Subgrid-Scale Vertical Velocity Variability from Profiles at a Single Point

Abstract:

Turbulent updrafts and aerosols go hand in hand in producing cloud droplets: updrafts generate a supersaturation of water vapor, and aerosols provide nuclei on which the water vapor can condense; together, they control the cloud droplet number concentration. The droplet number concentration is a key variable in all aspects of aerosol–cloud interactions, both the Twomey forcing and rapid adjustments of liquid water path and cloud fraction. Although updraft speed is often the limiting factor in droplet activation, general circulation and convection resolving models usually do not explicitly represent the cloud-scale updraft speed in their activation parameterizations. Instead, they rely on the boundary layer turbulence scheme to provide an updraft speed estimate with the correct statistical properties such as variance and skewness.

How to evaluate the representativeness of the parameterized updraft speed is not a straightforward problem. Although Doppler lidars can retrieve the cloud-base updraft speed and its temporal variability at a single location, this is not necessarily a good proxy for the instantaneous spatial variability over a model grid box. We use a combination of large eddy simulations and distributed Doppler lidars at the ARM Southern Great Plains site to investigate the conditions under which (1) sampling the updraft distribution at a small number of distributed locations is representative of the gridbox mean and (2) the single-point temporal variability is representative of the instantaneous spatial variability.

Our results demonstrate that a small number of distributed point measurements is generally sufficient to represent the domain mean updraft statistics for a large range of domain sizes across a variety of cloud and dynamical regimes. One exception occurs when the domain contains multiple cloud regimes at once. The correspondence between spatial and temporal variability is regime-dependent, indicating that care must be taken in observational evaluation of subgrid parameterizations in models.