Large-Eddy Simulations of Dust Devils and Convective Vortices
Space Sciences Series of ISSI
In this review, we address the use of numerical computations called Large-Eddy Simulations (LES) to study dust devils, and the more general class of atmospheric phenomena they belong to (convective vortices). We describe the main elements of the LES methodology. We review the properties, statistics, and variability of dust devils and convective vortices resolved by LES in both terrestrial and Martian environments. The current challenges faced by modelers using LES for dust devils are also
... vils are also discussed in 1 2 LES methodology 2.1 Dynamical integrations LES consist of four-dimensional (horizontal+vertical+temporal) numerical integrations of atmospheric fluid dynamics equations, usually the fully compressible, non-hydrostatic Navier-Stokes equations. While the vast majority of GCM simulations are performed under the approximation of hydrostatic balance between gravity and pressure forces in the vertical, such an approximation is not appropriate in LES because vertical wind accelerations in CBL turbulent structures might become comparable to the acceleration due to gravity. For hydrodynamical integrations, Martian LES are based on similar solvers as terrestrial LES: e.g., for their Martian studies, Rafkin et al. (see Section 4.1) used the RAMS terrestrial model (Pielke et al., 1992) , and Spiga and Forget (2009) used the WRF model (Moeng et al., 2007; Skamarock and Klemp, 2008) , also used by Klose and Shao (2013) to simulate terrestrial CBL vortices (see Section 3 and Tables 1 and 2). A remaining cornerstone for the proper simulation of dust devils, even in high-resolution LES, is the application of suitable numerical schemes. The numerical approximations used by the schemes cause characteristic numerical errors due to their specific construction (e. g., Durran, 2010, Chapter 3). Numerical diffusion is usually caused by schemes of odd order of accuracy, whereas numerical dispersion is caused by schemes of even order. Numerical dispersion is triggered by strong gradients of the advected quantities, which typically exist in the center of dust devils. The resulting wiggles are cumbersome for the development and evolution of dust devils. By comparing a standard advection scheme to a scheme which is especially designed to avoid these errors, Weißmüller et al. (2016) showed that the average central pressure drop intensity increases by 8 % if the latter scheme is used. Generally speaking, every LES model is capable of carrying out dynamical integrations at very high horizontal resolution. However, the increasing number of simulated grid points is accompanied by increasing demands on the model's high-performance computing capabilities, such as its efficient parallelization, which is necessary to distribute the simulation on up to thousands of computing nodes, and the optimized handling of data, simply because a higher number of grid points produces more data, which has to be processed and analyzed.