A computational study is performed to analyze oxygen transport in dry and wet stochastically reconstructed catalyst layers (CLs). CL stochastic reconstructions are generated using random penetrating spheres of a given particle size that agree with the statistical correlation functions of a reference 3D FIB-SEM reconstruction. A nucleation-based full morphology approach is used to partially flood the CLs from within the structure in an attempt to reproduce the conditions in an operating fuel cell. In order to validate the 3D numerical model and reconstruction method, the dry effective diffusivity of CLs and its variation with porosity are obtained and shown to be in agreement with reported literature data. Simulations are then performed for CLs with varying porosity and saturation. Statistical analysis is used to estimate an expression for effective Knudsen radius as a function of porosity and particle size, and then the computed dry effective diffusivities are used to develop a generalized percolation-based correlation function to estimate dry and wet effective diffusivities. The effective diffusivity of CLs with different pore size distribution are also obtained from 3D simulations and using the proposed correlation function and shown to differ by less than 10% for porosities in the range of 0.4-0.7. (c) The Author(s) 2019. Published by ECS.