The remarkable oxygen transfer efficiencies attainable in membrane aerated biofilm reactors (MABRs) are expected to favor their prompt industrial implementation. However, tests in clean water, currently used for the estimation of their oxygen transfer potential, lead to wrong estimates once biofilm is present, significantly complicating reactor modeling and control. This study shows for the first time the factors affecting oxygen mass transfer across membranes during clean water tests and reactor operation via undisturbed microelectrode inspection and bulk measurements. The mass transfer resistance of the liquid boundary layer developed at the membrane-liquid interface during clean water tests accounted for two thirds of the total mass transfer resistance, suggesting a strong underestimation of the oxygen transfer rates when it is absent (e.g. after biofilm growth). Reactor operation to attain partial nitritation showed that predicted oxygen transfer rates are enhanced up to six Limes with biofilm activity. The higher availability of ammonia at the biofilm base drives this process. Such behavior can be captured with the addition of two terms (depending on system characteristics and reactor loading) to existing model structures. Overall, we provide tools to better estimate, model, and optimize oxygen transfer supporting a more energy-efficient approach to MABR operation. (C) 2013 Elsevier B.V. All rights reserved.