The main objective of this work concerns the evaluation of the biological aerated filtration model (BAF) found in GPS-X®, which had never been evaluated with adequate data. This model is interesting since it integrates the physical and biological phenomena involved during filtration with a low complexity of use. The validation of the model parameters combines experimental and theoretical approaches. Experimental data were recorded at a semi-industrial pilot scale submerged biofilter operated at a tertiary nitrification stage, receiving the effluent of a medium loaded activated sludge process for municipal wastewater. Also, several protocols were regularly applied to characterize the biofilm and the nitrogen removal performances: dry density and thickness of biofilm, nitrification rates and corresponding quantity of autotrophic biomass accumulated inside the filtering media, quantity of extracted autotrophic bacteria in the backwash water, nitrification capacity along the biofilter, as well as nitrogen compounds in the effluent. For short-term dynamic conditions, a set of reliable parameter values has been used to predict nitrogen removal for different data sets. For long-term dynamic periods, the need to adapt some of the parameters from one set of data to another is demonstrated. It is shown that the hydraulic loading rate and the backwashing frequency are the main parameters responsible for these modifications.