Several models commonly employed to represent the mass transfer in osmotic distillation (OD) systems are applied to the results of pure water OD experiments carried out with two commercial asymmetric porous membranes. Molecular and Knudsen diffusion mechanisms are tested to model the vapour transport across the membrane. When using the global structural characteristics specified by the membrane manufacturer, both diffusion models underestimate the membrane permeability to water vapour. The exceptionally high experimental permeability can be predicted by a Knudsen mechanism when considering the Teflon top layer alone. The membrane support is envisaged as an additional resistance to water transfer in the liquid form, leading to splitting of the asymmetric membrane into a series of two resistances: one resistance to gas transfer in the top layer and another to liquid transfer in the support layer. In this model, the gas membrane contribution is estimated to cover 40-70% of the total mass transfer resistance; the film of diluted brine entrapped in the membrane support can cover up to 30% of the total mass transfer resistance and the diluted brine boundary layer up to 60%, indicating the sensitivity of the OD system to concentration polarisation. Classical empirical correlations of dimensionless numbers are fitted to the experimental flux results to try and predict the mass transfer coefficients of the brine boundary layer in the OD system. The poor quality of the model is attributed to the special hydrodynamics of the membrane module whose geometry does not fit in the reference of the correlations, i.e. straight circular ducts. The heat transfer associated with water transport is integrated into the mass transfer equations. The thermal effect due to evaporation and condensation at both liquid-membrane interfaces appears to be significant: a high vapour flux of 12 kg m-² h-' generates a transmembrane temperature difference of approximately 2°C inducing a 30% driving forc