This study deals with the potential and the limitations of dynamic models for describing and predicting nitrous oxide (N₂O) emissions associated with biological nitrogen removal from wastewater. The results of a three-week monitoring campaign on a full-scale partial nitritation reactor were reproduced through a state-of-the-art model including different biological N₂O formation pathways. The partial nitritation reactor under study was a SHARON reactor treating the effluent from a municipal wastewater treatment plant sludge digester. A qualitative and quantitative comparison between experimental data and simulation results was performed to identify N₂O formation pathways as well as for model identification. Heterotrophic denitrifying bacteria and ammonium oxidizing bacteria (AOB) were responsible for N₂O formation under anoxic conditions, whereas under aerated conditions the AOB were the most important N₂O producers. Relative to previously proposed models, hydroxylamine (NH₂OH) had to be included as a state variable in the AOB conversions in order to describe potential N₂O formation by AOB under anoxic conditions. An oxygen inhibition term in the corresponding reaction kinetics was required to fairly represent the relative contribution of the different AOB pathways for N₂O production. Nevertheless, quantitative prediction of N₂O emissions with models remains a challenge, which is discussed.

This work demonstrates the usefulness of non-linear data reconciliation to evaluate available measurements and estimate unmeasured variables for a full-scale partial nitritation (SHARON) reactor for the treatment of wastewater with high ammonium concentrations. Despite a lack of measured data, the bilinear approach of formulating system constraints allowed to satisfy the requirements for data reconciliation and gross error detection, leading to a balanced data set and the estimation of unmeasured variables.