Although it is tempting to surmise that a boundary condition that includes the effects of surface drag would be the most realistic (the third one listed above), the assumption of a logarithmic surface-layer wind profile is unjustified within and near convective storms. M-O similarity theory requires horizontal homogeneity and stationarity, neither of which are satisfied within the outflow of a convective storm. These conditions may not even be met within the near-storm environment, especially in baroclinic boundary layers (i.e., boundary layers with a large variation of the geostrophic wind with height), within boundary layers in which winds are rapidly accelerating toward a storm updraft, or in boundary layers within complex terrain. Moreover, the M-O predictions of near-surface vertical shear, upon which drag boundary conditions are based, apply to mean quantities, not the vertical profiles at every instant at every grid point, which is most often how the boundary conditions are specified in convective storm simulations.
The presentation also will discuss which aspects of storm simulations are likely to be most influenced by the choice of lower boundary condition, as well as some techniques that can be adopted from the engineering community to improve the handling of near-surface turbulence in convective storm simulations.