Membrane distillation is an emerging technology to separate non-volatile components from an aqueous feed stream. Mathematical models have proven useful to pursue breakthrough in the economics of the technology and for further improvement through module design and operational optimization. However, before this can be done, all of the resistances in the system must be identified correctly and the model must be carefully calibrated to ensure its predictive power. In this work the typical structure of a direct contact membrane distillation (DCMD) model is studied, where the mass transfer inside the membrane is simulated using the Dusty Gas Model and Nusselt type equations are used to simulate the heat transfer inside the channels. We demonstrate that an off-the-shelf Nusselt equation cannot directly be applied to simulate the heat transfer in the spacer filled channels. Instead, the equations should be calibrated to match the behaviour of the particular spacer. A Monte Carlo filtering method was applied to calibrate and study the structure of the DCMD model for the membrane region. The method proved useful to identify which parameters need to be included in the calibration as it highlighted parameter correlations. Additionally, a submodel selection was performed for the heat and mass transfer inside the membrane. A simple, yet physical method for the simulation of supported membranes was tested and validated on 3 supported membranes, resulting in an excellent fit. (C) 2016 Elsevier B.V. All rights reserved.