An efficient model development framework for an Eddy-Diffusivity/Mass-Flux cloud and precipitation parameterization with 20 free parameters

The physics represented by cloud and turbulence parameterizations are normally investigated according to either their performance against large eddy simulations (LES) or from GCM historical simulations. Among the benefits of these strategies lurk detriments, including but not limited to selection bias in field experiments and the considerable computational cost of GCM simulations. Here, we employ a complementary approach in which a Single Column Model (SCM) simulates atmospheres from thousands of diverse initial conditions in an effort to ensure that the model physics are applicable to a continuum of real-world conditions while also providing flexibility in testing many combinations of tunable parameters.

The SCM is composed of the JPL Eddy-Diffusivity/Mass-Flux (EDMF) and the RRTM-G radiation models and can be initialized by either observations of the atmospheric state or reanalyses, with model assessment performed by satellite observations from CALIPSO, CloudSat, MODIS, and AMSR-E. The efficiency of this framework permits a detailed investigation of free parameters within the parameterization through the Morrison One At a Time (MOAT) sensitivity tests.

MOAT testing of twenty parameters related to updraft speed and entrainment, turbulent mixing, aerosol size distribution, autoconversion, accretion, and raindrop size distribution reveals that simulated low cloud cover is most sensitive to parameters related to updraft speed and entrainment. Interestingly, simulated rain, LWP, and albedo are shown to be quite sensitive to the size distribution of sea salt, highlighting the roles of aerosol indirect effects in the evolution of the marine boundary layer. We also show that choosing between two autoconversion schemes does not significantly affect SCM clouds or rain.

More focused and rigorous testing is then performed on the parameters that the MOAT analysis identified as being the most relevant to our desired observables in order to gain a physical understanding of the processes that are important to the simulation of these complex environments. We place a focus on matching the observed rain rate, low cloud cover, and albedo as a function of lower tropospheric stability. The final state of the EDMF is also shown to produce simulations that closely resemble LES from the RICO and DYCOMS field experiments.