Modeling N 2 O emissions of complex cropland management in Western Europe using DayCent: Performance and scope for improvement

Under the United Nations Framework Convention on Climate Change (UNFCCC), industrialized countries and countries with economies in transition (so called Annex 1 countries) are encouraged to move towards more sophisticated approaches for national greenhouse gas reporting. To develop a model-based approach for estimating nitrous oxide (N 2 O) emissions from agricultural soils, model calibration is one of the first important steps. Extensive multisite field observations are necessary for this purpose, as agricultural management in Western Europe is complex ( e.g., diverse crop rotations, different types of fertilizer and soil tillage). In the present study, we used ca. 24,000 daily N 2 O flux observations from six cropland sites, two in France and four in Switzerland, to conduct an automatic data-driven calibration of the biogeochemical model DayCent. This model is planned to be used for greenhouse gas reporting in the entire European Union as well as in Switzerland. After a site-specific calibration, a leave-one-out (LOO) cross-evaluation was conducted to assess the model’s ability to predict N 2 O emissions for sites it was not calibrated for. Mean observed N 2 O fluxes for 54 interactions of crop cycles, field studies and treatments were used to evaluate the model. The LOO cross-evaluation resulted in a R 2 of 0.63 for the prediction of mean N 2 O fluxes per crop cycle, compared to an R 2 of 0.51 obtained with default parameterization. Our results showed that the improvement in N 2 O predictions was associated with the adjustment of only seven parameters controlling the N cycle in soil ( e.g., the maximum daily nitrification amount and the inflection point for the effect of water-filled pore space on denitrification) out of several hundred parameters. These parameters showed a wide range of values between sites, revealing an important challenge for calibration-based improvement of N 2 O simulations. Despite the remaining uncertainty, our model-based estimates of N 2 O emission per crop cycle (2.64 kg N ha -1 ) were clearly closer to measurements (2.67 kg N ha -1 ) than commonly used emission factor approaches (1.60–1.71 kg N ha -1 ). Based on extensive field observations, our results suggest that, after data-driven calibration of only few N cycle parameters, DayCent simulations are useful for reporting N 2 O emissions of complex cropland management. These model based-estimates were more accurate, because they consider key drivers that are disregarded by simpler approaches. Moving towards more complex methods of N 2 O reporting, is therefore expected to improve the accuracy and additionally allows to assess mitigation options.