There is a need for numerical models capable of predicting local accumulation of hydrogen near stress concentrators and crack tips to prevent and mitigate hydrogen assisted fracture in steels. The experimental characterisation of trapping parameters in metals, which is required for an accurate simulation of hydrogen transport, is usually performed through the electropermeation test. In order to study grain size influence and grain boundary trapping during permeation, two modelling approaches are explored; a 1D Finite Element model including trap density and binding energy as input parameters and a polycrystalline model based on the assignment of a lower diffusivity and solubility to the grain boundaries. Samples of pure iron after two different heat treatments - 950C for 40 minutes and 1100C for 5 minutes - are tested applying three consecutive rising permeation steps and three decaying steps. Experimental results show that the finer grain microstructure promotes a diffusion delay due to grain boundary trapping. The usual methodology for the determination of trap densities and binding energies is revisited in which the limiting diluted and saturated cases are considered. To this purpose, apparent diffusivities are fitted including also the influence of boundary conditions and comparing results provided by the constant concentration with the constant flux assumption. Grain boundaries are characterised for pure iron with a binding energy between 37.8 and 39.9 kJ/mol and a low trap density but it is numerically demonstrated that saturated or diluted assumptions are not always verified, and a univocal determination of trapping parameters requires a broader range of charging conditions for permeation. The relationship between surface parameters, i.e. charging current, recombination current and surface concentrations, is also studied.