Denitrification represents a major nitrogen loss pathway in agriculture, reducing plant N uptake, lowering crop N use efficiency, while emitting the potent greenhouse gas N2O. However, due to methodological challenges, field measurements of N2 emissions remain rare, leaving denitrification losses poorly characterised across most cropping systems and comprehensive N budgets lacking. This constrains parameterization and validation of biogeochemical models, impeding efforts to forecast denitrification losses under changing conditions and design effective mitigation strategies.
This study (i) reviewed the current literature on field-based N2 measurements, (ii) tested the ability of five different models to accurately simulate denitrification losses, and (iii) simulated denitrification losses from agricultural soils in Germany using gridded and point-based modelling frameworks.
Results show that field studies suitable for model calibration remain scarce and mostly limited to temperate systems. Simulated cumulative N2 emissions varied widely between models, ranging from 0.89 to 6.07 kg N ha⁻¹, far below the observed N2 emissions of 21±3 kg N ha⁻¹. The ratio of N2O to (N2O+N2) emitted (RN2O), a key determinant of the climate impact of denitrification, was not only strongly overestimated by all models but also exhibited considerable variations, ranging from 0.056 to 0.707, compared to the measured ratio of 0.006. Models also failed to capture expected responses of RN2O to environmental drivers. For Germany, simulated average N2 emissions from agricultural soils ranged from 1.7 to 14.5 kg N ha⁻¹yr-1, with corresponding RN2O ranging from 0.18 to 0.59. Spatial patterns of emissions also differed significantly across modelling frameworks, reflecting large structural uncertainties. 
These findings highlight that both N2 emissions and RN2O remain poorly represented in current models, severely limiting our ability to establish reliable N budgets and develop targeted N management strategies. Coordinated advancements in field-based measurements and modelling are urgently needed to improve the representation of denitrification in biogeochemical models.