A systematic analysis on the use of the lattice Boltzmann method (LBM) for predicting the permeability of packed beds is presented. A filtration rig is used to obtain permeability measurements of beds of glass beads, sand and crushed minerals. Subsequently, X-ray micro-tomography is employed to image bed samples for use as input to LBM calculations. Uncertainties in accuracy and reliability of predictions arising from pixel resolution, sample size and digitisation errors in the simulations are evaluated through sensitivity studies and assessments of the representativeness of the bed samples. For beds of spherical and near-spherical particles, any bed sample is capable of providing reproducible and reliable simulation results provided that the particles are adequately resolved within an LBM simulation. For more complex beds of polymorphous, polydisperse particles estimation of the permeability of representative samples of the entire bed is required before average results comparable with data are obtained. (C) 2014 Elsevier Ltd. All rights reserved.