In-situ N2O and N2 data improved N budget simulation with APSIM and LandscapeDNDC in tropical sugarcane systems

Denitrification is a key process in the global nitrogen (N) cycle, causing nitrous oxide (N2O) and dinitrogen (N2) emissions. Biogeochemical models allow field-scale estimates of N2O and N2, extrapolating important yet often limited experimental results. However, such predictions rely mostly on N2O data, and the lack of N2 data hinders validating total denitrification, which remain a major uncertainty for N budgets. This study investigated denitrification losses and N budgets in two tropical sugarcane systems using the Agricultural Production Systems sIMulator (APSIM) and the LandscapeDNDC (LDNDC) simulation framework using a unique dataset of both N2O and N2 emissions measured in the field over a complete growing season. Key soil N parameters influencing N2O and N2 emissions in APSIM and LDNDC were identified via global sensitivity analysis, followed by generalised likelihood uncertainty estimation to determine their posterior distributions using (i) N2O data only and (ii) both N2O and N2 data. The simulation of N2O emissions in APSIM and LDNDC were improved in both calibration approaches, resulting in 0.7–1.3 kg N ha−1 of RMSE. However, simulated N2 emissions increased and agreed better with the observed values only when calibrated with both N2O and N2 (RMSE 30.1–45.0 kg N ha−1 before calibration and 19.3–19.9 kg N ha−1 after). The simulated N loss pathway shifted from leaching to N2 emissions after calibration including N2. The simulated N balance was larger when sugarcane residues were retained as compared to burning consistently across the different soil N parameter configurations. These findings indicate that biogeochemical models, when used with default soil N parameters or calibration limited to N2O data, are likely to underestimate denitrification losses (>50 %), leading to a bias in N budgets simulation. Accurate N loss estimates are essential for understanding the long-term management impacts on soil organic matter dynamics, as demonstrated by the improved N budgets from both simulation models denote N mining when sugarcane is burnt, and the potential to sequester N when cane residues are retained. These outcomes emphasise the importance of integrating in-situ measurements of N2O and N2 in simulation exercises, ensuring more accurate N budget estimates across scales.