Turbulent length scales and LES grid size in ALM simulations

I have been searching through the research done on ALM and noticed that most of the researchers use 30 to 60 cells per rotor radius in LES simulations. This looks like an extremely coarse grid size for LES, not to mention that I am quite confused about the calculation of turbulent length scales in ALM simulations and justification of the grid size.
Any references on this issue would be great.