Handout (2.5 MB)
A new coupling parameterization was proposed by Machulskaya and Mironov (2018), referred to here as the MM scheme. This scheme used aggregated (co)variances coupled to the sub-grid shallow convection parameterization, Clouds Layers Unified by Binormals (CLUBB); CLUBB prognoses variables in subsequent timesteps using multivariate, sub-grid probability density functions (PDF) of temperature (T), moisture (q), and vertical velocity (w). Using the MM scheme, the atmospheric response to SGS heterogeneity was increased, but only for the lowest few levels. Furthermore, Witte et al. (2022) proposed an augmented CLUBB parameterization, CLUBB+MF, which uses discrete “plumes” from the Mass-Flux (MF) approach for additional vertical transport of T and q. Plumes may be tile dependent; initiated over some tiles, while lying dormant (or weaker) over others. Preliminary results combining CLUBB+MF and the MM scheme indicate an increased vertical propagation of the signal from surface heterogeneity. This is more representative of observations and cloud resolving models (CRMs) such as large-eddy simulations (LES). However, these promising new parameterizations require a thorough assessment before they can be widely incorporated into models.
For this work, we generate statistics to measure the variation between sub-grid probability density functions (PDFs) generated explicitly through WRF-LES, a CRM, and those indirectly derived within the Community Earth System Model v2 (CESM2) using several parameterization approaches: the default, the MM scheme only, CLUBB+MF only, and both novel schemes in combination. This hierarchy of parameterizations is compared to two LES simulations for each case, one with a homogeneous surface and one with a heterogeneous one. Large-scale forcing data for both models is sourced from the Atmospheric Radiation Measurement (ARM) Southern Great Plains (SGP) site in Northern Oklahoma. A total of 92 days are analyzed; a much larger dataset than previous case-specific studies. We detail the differences and agreements between PDFs, thereby providing deeper insights into the accuracy of CESM2 for these circumstances and potentially suggesting new avenues for improvement.

