Establishing denitrification (N2O+N2) budgets – a data driven scaling approach in grains systems

Accounting for denitrification losses from agroecosystems remains challenging, particularly due to methodological challenges regarding measurements and upscaling of highly episodic nitrous oxide (N₂O) and dinitrogen (N₂) emissions to seasonal and system levels. Denitrification losses, and thus nitrogen (N) budgets remain therefore poorly constrained across a wide range of agro-ecosystems, hindering targeted development of N loss mitigation strategies. Here we present an innovative data driven upscaling approach using high-frequency N₂O datasets, and the XGBoost model, establishing seasonal N₂O and N₂O emissions. Emissions of N₂O and N₂ were measured in-situ using the ¹⁵NGF method. The XGBoost model was trained using in-situ N₂O and N₂ data across eight site-seasons in different grains systems in southeastern Australia, expressing the ratio of N₂ to N₂O emitted as a function of water-filled pore space, nitrate content and soil temperature. The trained model was then applied to additional 38 site-season-treatment high-frequency N₂O datasets to estimate daily N₂ emissions, followed by aggregation to seasonal scales. Emissions N₂O over the cropping season accounted for only 3.2% (on median, 2.1–4.1% at quartiles) of denitrification. Seasonal denitrification losses ranged from 2.6 to 66.1 kg N ha⁻¹ and were dominated by N₂, exceeding N₂O emissions by a factor of ~30 (on median, 23–46 at quartiles). Our approach delivers for the first time denitrification budgets for a range of different grains systems, providing a blueprint to investigate the effects of environmental drivers and management on denitrification. Extending this approach to other soil types and production systems offers the opportunity to derive more generic relationships between drivers and the N₂O:N₂ ratio as way forward to improve N budgeting for agronomic and environmental benefits.