Comparison of large-eddy simulations with a single-column model: Implications for mid-level cloud parameterization

General circulation models (GCMs) tend to underpredict thin mid-level clouds. To improve prediction, we are developing a single-column model (SCM) that has been previously tested for boundary layer clouds. Using this same SCM without any case-specific adjustments, we have simulated an observed altocumulus cloud. Our ultimate goal is to use these and other results to eventually implement this SCM as a generalized layer-cloud parameterization in a GCM.

To provide a benchmark, we have simulated the observed altocumulus cloud using a high-resolution, three-dimensional large-eddy simulation (LES) model. This model outputs mean fields such cloud fraction and liquid water mixing ratio, along with various moments, such as turbulent fluxes of heat and moisture. We also simulate the same cloud using our one-dimensional SCM. The SCM predicts the same mean fields and moments as the LES by use of the assumed probability density function (PDF) method. That is, the SCM prognoses various low-order moments and uses an assumption about the shape of the PDF to close the higher-order moments.

Aircraft observations show that the observed altocumulus cloud dissipated with time. Both LES and SCM models exhibited similar time evolution, including nearly equal cloud lifetime. Additionally, both models had similar profiles of cloud fraction, liquid water, and turbulent fluxes of heat and moisture. These results indicate that our SCM simulates the cloud accurately and similarly to the LES model.

In addition to the control altocumulus case, we have also performed a sensitivity study containing multiple simulations. Each sensitivity simulation involves changing one parameter, such as the solar zenith angle, ice number concentration, or large-scale subsidence. Both models show remarkable similarity between corresponding sensitivity runs, regardless of the physical parameter being varied. This shows that our SCM behaves reasonably over a range of conditions.

All SCM altocumulus simulations were performed with the same parameter values that we use for boundary layer cloud simulations not reported here. Because the model allows us to simulate both types of clouds, we can use a single set of parameter values to simulate altocumulus and boundary layer clouds. In future studies, we hope to simulate other mid-level cloud cases using this set of parameters.