The effect of grid resolution on a large eddy simulation (LES) of a wall-bounded turbulent flow is investigated. A channel flow simulation campaign involving a systematic variation of the streamwise (Îx) and spanwise (Îz) grid resolution is used for this purpose. The main friction-velocity-based Reynolds number investigated is 300. Near the walls, the grid cell size is determined by the frictional scaling, Îx+ and Îz+, and strongly anisotropic cells, with first Îy+ ⌠1, thus aiming for the wall-resolving LES. Results are compared to direct numerical simulations, and several quality measures are investigated, including the error in the predicted mean friction velocity and the error in cross-channel profiles of flow statistics. To reduce the total number of channel flow simulations, techniques from the framework of uncertainty quantification are employed. In particular, a generalized polynomial chaos expansion (gPCE) is used to create metamodels for the errors over the allowed parameter ranges. The differing behavior of the different quality measures is demonstrated and analyzed. It is shown that friction velocity and profiles of the velocity and Reynolds stress tensor are most sensitive to Îz+, while the error in the turbulent kinetic energy is mostly influenced by Îx+. Recommendations for grid resolution requirements are given, together with the quantification of the resulting predictive accuracy. The sensitivity of the results to the subgrid-scale (SGS) model and varying Reynolds number is also investigated. All simulations are carried out with second-order accurate finite-volume-based solver OpenFOAM. It is shown that the choice of numerical scheme for the convective term significantly influences the error portraits. It is emphasized that the proposed methodology, involving the gPCE, can be applied to other modeling approaches, i.e., other numerical methods and the choice of SGS model.