We propose a mathematical model to describe the microstructure of the gas diffusion layer (GDL) in proton exchange membrane fuel cells (PEMFCs) based on tools from stochastic geometry. The GDL is considered as a stack of thin sections. This assumption is motivated by the production process and the visual appearance of relevant microscopic images. The thin sections are modeled as planar [two-dimensional (2D)] random line tessellations which are dilated with respect to three dimensions. Our 3D model for the GDL consists of several layers of these dilated line tessellations. We also describe a method to fit the proposed model to given GDL data provided by scanning electron microscopy images which can be seen as 2D projections of the 3D morphology. In connection with this, we develop an algorithm for the segmentation of such images which is necessary to obtain the required structural information from the given grayscale images. (c) 2008 The Electrochemical Society.