CEDAR email: special issue - Towards_a_better_representation_of_turbulence_in_GCMs

[Wayne Hocking]

For many atmospheric researchers, the holy Grail of research is the development of a computer model which can reproduce the entire behavior of the atmosphere. Every now and then, programmers feel that they have achieved (or approached) that goal. Yet, is it achievable? Some authors of computer models developed in the 1970's felt that they had "nailed" the problem of a realistic mesopheric flow. In their models, substantial meridional airflow was found not to exist, and they considered the basic state of the atmosphere in the mesosphere to be a perfect balance between Coriolis forces and pole-to-pole heat flow. For a time, observational data which produced meridional velocities over long term were rejected for publication in certain journals as they disagreed with these models. Then in 1981 Lindzen produced his paper demonstrating that meridional did not just exist, but was crucial to understanding the existence of the cold summer mesopause. In that case the unique characteristics of gravity waves, with their ability to have critical -level interactions  and directed forcing, were found to be key components of a good model. The earlier confidence had been misplaced.
Turbulence is particularly poorly represented in many large models. It is highly parameterized, and often the parameterization is simplistic, such as Rayleigh drag, or perhaps simple relations like K ¡Ö e/N2 are employed. Problems of scale are particularly acute, since a serious treatment requires studies on scales from millimetres to thousands of kilometres. While nested models, in which smaller scale models are used to build large-scale models, are useful, some degree of parameterization is still necessary. The use of artificial intelligence seems to have helped here, and instead of solving the Navier¨CStokes equation, Newtons second law, the first law of thermodynamics, the continuity equation, and Fick¡¯s law for heat transport using differential-equation solvers, AI is used as a error-reduction process on the basic equations. But if the fundamental equations are oversimplified or limited, then the solution must likewise be flawed. And even here, problems of covering multiple scales exist.
Key parameters of turbulence in large scale models include energy dissipation rates and diffusion rates. But  their application is simplistic, and does not recognize that there are different forms of turbulence; for example, as Dewan showed very clearly, turbulent diffusion in the stratosphere is very different in nature to diffusion in the troposphere, with necessarily different ways to calculate diffusion coefficients. Incorporation of Stokes diffusion is rare in most GCM's, yet the joint incorporation of Stokes diffusion on scale of hundreds and thousands of kilometres, coupled with small-scale diffusion at scales of centimetres, is crucial to understanding atmospheric diffusion processes.
Here, we invite contributions to  a special journal issue which seeks contributions on what is missing in regard to modelling turbulence in atmospheric models, as described at this link.
https://www.mdpi.com/si/233021
Papers are sought which present meaningful  new ways to improve representation of turbulence in GCM and computer models generally. There is no need to produce a full solution; rather, innovative ideas are sought that might help contribute to better representations of turbulence in large-scale models.
Examples of relevant topics are given in the link.
Thanks

Wayne Hocking, Professor Emeritus.


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mailman.ucar.edu/pipermail/cedar_email/attachments/20250611/d73c54d3/attachment.htm>