In this paper, verification and validation of a turbulence closure model is performed for an experimental compound channel flow, where the velocity and turbulent fields were measured by a Laser Doppler Velocimeter (LDV). Detailed Explicit Algebraic Reynolds Stress Model (EARSM) simulations are reported. There are numerous methods and techniques available to evaluate the numerical uncertainty associated with grid resolution. The authors have adopted the Grid Convergence Index (GCI) approach. The velocity components, the turbulence kinetic energy (TKE), the dissipation rate and the Reynolds stresses were used as variables of interest. The GCI results present low values for the u velocity component, but higher values in what concerns the v velocity component and w velocity component (representing secondary flows) and for Reynolds stresses RSxy and RSyz. This indicates that the mean flow has converged but the turbulent field and secondary flows still depend on grid resolution. Based on GCI values distribution, the medium and fine meshes were further refined. In addition to GCI analysis, the authors have performed linear regression analysis for estimating the mesh quality in what concerns small value variables. Comparison of numerical and experimental results shows good agreement.