Water management is still a key challenge for optimal performance and durability of polymer electrolyte membrane (PEM) fuel cells. Water levels along the channel in a PEM fuel cell present important spatial variations that should be taken into account to avoid both local flooding and local drying. In this work, a decentralised model predictive control scheme is designed to maintain the water activity on both anode and cathode sides of the PEM at appropriate levels. The proposed strategy tackles the accumulation of liquid water on the surface of the catalyst layers, and the possibility of local drying, by controlling observed water activity spatial profiles. Classic PEM fuel cell issues like reactant starvation are also considered. High control performance is achieved. The strategy is applied to a validated distributed parameter PEM fuel cell model. Results show increased cell power density in comparison to non-spatial control strategies. (C) 2017 Hydrogen Energy Publications LLC. Published by Elsevier Ltd. All rights reserved.