Understanding of the coupled transport processes that occur in thin gas diffusion layers (GDLs) is necessary to develop improved designs. The traditional technique used to model GDLs is the volume-averaged approximation. However, the applicability of this approach has been long questioned, and the error in the results is unclear. In this work, the limitations of GDL volume-averaged models are examined under single-phase conditions. The lattice Boltzmann method is combined with tomography images of carbon-paper GDLs to assess the existence of a representative elementary volume (REV) in terms of various effective properties. Then, the predictions of GDL volume-averaged and pore-scale formulations are compared by using a CFD model of a differential cell. The results show that a REV cannot be clearly defined. This leads to inhomogeneities in the pore-scale model that the volume-averaged model is not able to capture despite the overall flux through the GDL is similar in both cases.