Dynamic sub-filter closures for artificially thickened flame (ATF) combustion models for large eddy simulation (LES) are investigated with consistent a-priori and a-posteriori analyses. The analyses are based on a flame resolved simulation (quasi DNS) and large eddy simulations of the bluff body burner experiment by Hochgreb and Barlow with premixed flamelet generated manifolds (PFGM). Both flame resolved simulation and LES are performed under the conditions of a single real flame experiment, using the same domain size, filter sizes, boundary conditions and numerics, all with an additional validation by comparison to experimental data. For the evaluation of the sub-filter wrinkling factor, the well-established model by Charlette et al. (2002) in the modified version by Wang et al. (2011) is used with a static and with a dynamic model parameter, a new dynamic power-law model is discussed additionally. In the a priori analysis, the probability density functions (PDFs) of the sub-grid scale (SGS) wrinkling factor are compared against the modeled ones based on the flame resolved simulation data. These a-priori modeled wrinkling factor PDFs are then compared against the a-posteriori ones from the LES results, where a similar shape is observed for all models. The static model tends to over-predict the wrinkling factor, a better agreement is found for the dynamic models for the medium and small filter width, where the new formulation improves the results for the latter. For the largest filter width, the wrinkling factor is under predicted by the dynamic models. This under-prediction is, however, compensated by larger gradients of the progress variable field so that the mean flame surface density conditioned on the progress variable is in closer agreement with the flame resolved simulation than the wrinkling factor PDFs are. Finally, radial profiles of the time-averaged temperature from the LES, flame resolved simulation and experiment are compared against each other. With the dynamic SGS wrinkling models, the LES results converge with grid refinement against the flame resolved simulation results. (C) 2017 The Combustion Institute. Published by Elsevier Inc. All rights reserved.