Nitrogen fertilization is a key agronomic lever for high crop productivity, but also an important source of N 2 O emission, a potent greenhouse gas. Process-based agroecosystem simulation models are popular tools for managing the timing and amount of fertilization, and help reduce N 2 O emissions. However, accurate simulation of N 2 O emissions at field scale is still a challenge due to the spatial and temporal variability of the soil conditions. In this study, we investigated the sources of structural uncertainty in predicting N 2 O emissions under a wide range of pedo-climatic conditions using a representative field data set. We implemented the same nitrification/denitrification/N 2 O emission formalism in three different agroecosystem models and analyzed how the inter-model variability of variables involved in nitrification and denitrification processes, affected the simulated N 2 O emissions. We characterized the dispersion of the key variables (water-filled pore space, NO − 3 and NH + 4 concentration, and soil temperature) between models and we evaluated the effect of variable uncertainty on N 2 O emissions uncertainty using a sensitivity analysis. We also analyzed model errors over a wide range of soil-climate conditions to identify the most challenging conditions for simulation, which require further model improvement. Our results highlighted that the simulation of the timing and amplitude of the NO − 3 and NH + 4 peaks was highly variable between agroecosystem models, with an important impact on N 2 O emission. These peaks occurred mainly after fertilization or incorporation of crop residues, and the different representations of fertilization and mineralization between the models had a major effect on the simulation of N 2 O emissions. Our analysis also emphasized that wet acidic soils with high denitrification potential are more challenging for models to simulate.