Modeling nitrous oxide emissions from agricultural soil incubation experiments using CoupModel

Abstract

Abstract. Efforts to develop effective climate mitigation strategies for agriculture require methods to estimate nitrous oxide (N2O) emissions from soil. Process-based biogeochemical models have been often used for field- and large-scale estimates, while the sensitivity and uncertainty of model applications to incubation experiments are less investigated. In this study, a process-oriented model (CoupModel) was used to simulate N2O and CO2 fluxes and soil mineral nitrogen (N) contents in a short-term (43 d) factorial incubation experiment (16 treatments). A global sensitivity analysis (GSA) approach, “Morris screening”, was applied to quantify parameter sensitivity. The GSA suggested that a higher number of sensitive parameters was associated with N2O flux estimates and that inter-treatment variations in parameter sensitivities were distinguished by soil moisture levels or NO3- content and residue types. Important parameters regarding N2O flux estimates were linked to the decomposability of soil organic matter (e.g., organic C pool sizes) and the denitrification process (e.g., Michaelis constant and denitrifier respiratory rates). After calibration, the model better captured temporal variations and magnitude of gas fluxes and mineral N in unamended soils than in residue-amended soils. Low-magnitude daily and cumulative N2O fluxes were well simulated with mean errors (MEs) close to zero, but the model tended to underestimate N2O fluxes, as observed daily values increased by over 0.1 g N m−2 d−1, in which the major mismatch was due to limited success of the model to describe the high emissions during the first few days after crop residue addition. A larger uncertainty was also seen in the magnitude of pulse emissions by the posterior simulations. We also evaluated ancillary variables regarding N cycling, which indicated that more frequent measurements and additional types of observed data such as soil oxygen content and the microbial sources of emitted N2O are required to further evaluate model performance and biases. The major challenges for calibration were associated with high sensitivities of denitrification parameters to initial soil abiotic conditions and the instantaneous residue amendment. Model structure uncertainties and improved modeling practices in the context of incubation experiments were discussed.