Computational Fluid Dynamics coupled with Discrete Element Method (CFD-DEM) is a commonly used numerical method to model gas-solid flow in fluidised beds and other multiphase systems. A significant limitation of CFD-DEM is the feasibility of the realistic simulation of large numbers of particles. Coarse-graining (CG) approaches, through which groups of multiple individual particles are represented by single, larger particles, can substantially reduce the total number of particles while maintaining similar system dynamics. As these three CG models have not previously been compared, there remains some debate, however, about the best practice in the application of CG in CFD-DEM simulations. In this paper, we evaluate the performance of three typical CG methods based on simulations of a bubbling fluidised bed. This is achieved through the use of a numerical validation framework, which makes full use of the high-resolution 3D positron emission particle tracking (PEPT) measurements to rigorously validate the outputs of CFD-DEM simulations conducted using various different coarse-graining models, and various different degrees of coarse-graining. The particle flow behaviours in terms of the particle occupancy field, velocity field, circulation time, and bubble size and velocity, are comprehensively analysed. It is shown that the CG simulation starts to fail when the size ratio between the bed chamber and the particles decreases to approximately 20. It is also observed, somewhat surprisingly, that the specific CG approach applied to interparticle contact parameters does not have a substantial effect on the simulation results for the bubbling bed simulations across a wide range of CG factors.