Impact of Climate Variations on Nitrous Oxide Emissions during Spring Wheat Growing Seasons in Eastern Canada—Micrometeorological Measurements, STICS Model Verification, and Long-Term Simulations

Capturing the variability of nitrous oxide (N₂O) fluxes during the growing season in response to synthetic fertilizer application and climate variations is quite challenging for process-based models. Indeed, nitrous oxide fluxes are very sporadic and heterogeneous. This variability is not well captured yet in the inventories based on emission coefficients. Verifying process-based model prediction of N₂O emissions is a top priority if we want to reduce the uncertainty in our regional and global estimates and if we want to make sound assessments of beneficial management practices over space and time.

The STICS crop model can simulate the soil–crop system with a daily time step by individual year (i.e. with reinitialization) or linked over multiple years to account for crop rotation (i.e. no reinitialization). The daily N budget takes into account mineralization, denitrification, nitrification, NH₃ volatilization, and N absorption. Recently, new nitrification and denitrification formalisms were added to STICS crop model to estimate N₂O emissions, based on experimental results collected mostly from western Europe. Denitrification and nitrification are assumed to occur in the biologically active layer (i.e., 30 cm in the present study). The N₂O predictions of STICS were evaluated against field-scale fluxes measured using micrometeorological towers equipped with a tunable diode laser to measure fast-response N₂O gradients. The N₂O fluxes were measured in spring wheat (Triticum aestivum L.) fields (Ottawa, ON, Canada) during 5 growing seasons between 2001 and 2014. The experimental fields were tiled drained and had homogeneous soil properties (silty clay loam and clay loam soil textures). Different mineral N fertilization rates (40-80 kg N ha^-1) and forms (urea, ammonium nitrate) were applied. The study focused on growing season N₂O emissions following mineral fertilization, which were divided between the vegetative and reproductive stages. In humid climate regions such as eastern Canada, nitrous oxide emissions are mostly driven by denitrification and to a lesser extent by nitrification. After completing the model performance verification, long-term simulations (1953-2012) were performed at Ottawa and Quebec City for three N fertilization rates (100%, 80% and 60% of the recommended N rate) and on two contrasted soil textures (sandy loam and clay loam in Ottawa; sandy loam and silty clay in Quebec City). Simulation results were analyzed to evaluate the impact of climate variability on N₂O fluxes.

Overall the STICS model predictions were in the same range than the observations for each growing season, except for 2014 resulting in a normalized root mean square error of 25.4% for all years and 11.5% when 2014 was excluded. Compared to observations, STICS predictions were usually smaller for the vegetative stage when denitrification was dominant (mean error of -0.26 kg N ha^-1). During the reproductive stage, the predictions were closer to the observations (mean error of 0.06 kg N ha^-1). The best results were obtained in 2005, when a dry spell occurred during the vegetative stage. Although the temporal dynamic of N₂O fluxes was not always well captured by the model, the satisfactory results obtained for cumulative emissions over the entire growing season allowed to perform long term simulations over 60 years using the STICS model.

As expected the long-term simulation results showed that N₂O fluxes were greater on more clayed soil and for the higher N fertilization rates. The N₂O fluxes of the recommended N fertilization treatments were 15 to 32% greater than those of the treatments with 60% of the recommended N rate. The N₂O fluxes were also greater in Quebec City (47^oN) than in Ottawa (45^oN), as a result of the more humid climate favorable to denitrification processes. In Ottawa, the fluxes during the vegetative stage were mainly controlled by the N fertilization rate. On the other hand, the fluxes during the reproductive stage were not affected by fertilization rate, but a strong linear relationship was found with cumulative precipitation (R² ranging from 0.48 to 0.65). These results could be explained by the fact that in the spring, during the vegetative stage, soil moisture was usually high and soil nitrate was then the main factor controlling soil N processes and N₂O fluxes. In summer, during the reproductive stage, soil moisture was much more variable and became the main factor controlling soil N processes and N₂O fluxes. Weaker similar results were found in Quebec City for the sandy loam soil (R² ranging from 0.23 to 0.28). However, on the silty clay soil texture no clear relationship between precipitation and N₂O fluxes was found, most likely because soil water retention was greater for this texture in response to the elevated precipitation. Further analyzes are planned to evaluate the effect of growing degree-days and crop growth on N₂O fluxes.

This study showed that the recent improvement of the STICS crop model allowed to simulate quite accurately the cumulative N₂O fluxes during the growing season under variable climate conditions of eastern Canada. Accurate simulation of soil moisture during the reproductive stage and soil mineral N content during the vegetative stage were found to be critical for obtaining accurate predictions. The next phase of the project will be to evaluate the model performance over the entire year, from spring crop seeding until the next spring crop seeding, thus including winter with snow cover and the high N₂O emission period following snow melt and spring thaw.