A pore network model is presented to predict permeability and diffusivity in porous bodies with a relatively high porosity. The model application is exemplified on a macroporous sample of alpha-alumina with the porosity of approximate to 0.4. The network model is constructed on the basis of proximity of the computed data on its total porosity, pore-size function, and simulated mercury intrusion curve to the respective experimental data. The experiment related pore-size function is extracted from a 3-D stochastic replica of the alpha-alumina sample obtained by stochastic reconstruction. The reconstruction technique employs morphological information based on a set of 2-D cuts through the porous medium. The only free parameter in the network construction is connectivity. The impact of the connectivity adjustment on the transport properties is studied. If the calculated mercury intrusion curve is forced to fit the experimental one, changes in connectivity are counterbalanced by the presence of slightly wider or narrower throats in the network. This compensation effect decreases the span of calculated values of permeability and diffusivity, which agree well with experiment. (c) 2007 Elsevier Ltd. All rights reserved.