Abstract:Large eddy simulation (LES) of turbulence is expected to to be a more accurate and robust modeling approach than Reynolds averagedvNavier-Stokes (RANS) modeling, and indeed, provided sufficient resolution is used, LES can be very reliable. However, for LES to be a practical tool for engineering applications, it is important that it provide reliable model predictions with the coarsest possible resolution. Such coarse resolution in LES introduces a number of challenges in both subgrid modeling and numerical discretization that will be discussed in this talk. These include: 1) modeling the subgrid contribution to the mean Reynolds stress, 2) subgrid modeling in the presence of anisotropic LES resolution, 3) the effects of numerical dispersion on the energy cascade, and 4) the effects of inhomogeneous LES resolution. Also discussed are several modeling strategies to overcome some of these challenges and improve the reliability of coarsely resolved LES.