A key challenge in the use of simulations to determine transport properties of PEMFC catalyst layers is the computational reconstruction of the catalyst layer microstructure. In this work, a number of different algorithms incorporating different assumptions are used to computationally reconstruct a large number of catalyst layer microstructures. In particular, the different algorithms use a variety of methods to account for agglomeration and distribution of carbon black spheres and ionomer. A pore scale model is then used to compute effective transport properties for each microstructure. It is found that the choice of the considered reconstruction algorithms does not have a significant effect on effective transport properties in most cases. Finally, the model assumptions which account for Knudsen diffusion are analyzed and modified to account for non-cylindrical pore structures. When cases are run using the Derjaguin correction for Knudsen diffusion, the obtained computational results are much closer to experimental data. Published by Elsevier B.V.