Spatial smoothing can be applied to reduce a three-dimensional (3D) fuel cell model to a computationally cost-efficient two-dimensional (2D) counterpart. In this study, the averaged dependent variables in the streamwise direction are augmented with a statistical measure of their distribution in the spanwise direction in the form of standard deviations. A functional form of the latter is postulated and justified a posteriori by comparing the predictions from the 2D spatially-smoothed model with the original 3D counterpart: good agreement is found. This measure of the variability thus complements the average predictions of existing spatially smoothed models. Copyright (C) 2014, Hydrogen Energy Publications, LLC. Published by Elsevier Ltd. All rights reserved.