Discontinuous Galerkin spatial discretizations of compressible flows allow to perform local degree adaptation (shortly, p-adaptation) in a very straightforward way and almost without computational overhead, as shown e.g. in [6]. Dynamical adaptation was also applied successfully to inviscid geophysical flows in [11, 12]. All the previous works relied however on a refinement criterion which essentially estimates the L 2 norm approximation error. In [10], we have argued that such a criterion may not be optimal for LES and we have proposed a different, physically based criterion that was shown to be more effective in a number of numerical experiments. The goal of this work, which summarizes some of the results presented in [9], is to extend the above approach to dynamical adaptation and to test the new criterion also in a dynamically adaptive framework. 2 The DG-LES Approach and Its Numerical Implementation The DG-LES model for compressible flows employed in this work, based on a Local Discontinuous Galerkin (LDG) discretization of the viscous terms [3], is fully described in [1], to which we refer for all the details on the model