Agricultural production's sustainability and productivity nexus requires comprehensive nitrogen budget (NB) assessments. The global agricultural sector is a major source of anthropogenic greenhouse gases (GHGs), primarily due to increased synthetic nitrogen (N) fertilizer use. The Intergovernmental Panel on Climate Change (IPCC) suggests alternative cropland nutrient management strategies to mitigate GHG emissions. Croplands are a substantial N₂O emission source, contributing significantly to global emissions. Denmark, with its intensive agriculture and ambitious emission reduction targets (70% by 2030 from 1990 levels and net-zero by 2050), provides a suitable case study. Process-based models simulating biogeochemical cycles offer cost-effective frameworks for assessing management practices' impact. However, data limitations often necessitate aggregating field-scale information, potentially affecting simulation accuracy. Previous studies have examined the effects of aggregating soil and climate data, revealing variable impacts depending on factors such as region and agro-climatic conditions. However, the influence of spatial aggregation of field-activity data (crop rotations, fertilization) on simulated NBs at the regional level remains largely unexplored. This study uses the Danish LOOP program (2013-2019) data to evaluate the impacts of different data aggregation approaches on NB simulations using the LandscapeDNDC (LDNDC) model. The study compares a detailed, field-level management data approach (A) with two common aggregation approaches: (B) sequential mono-cropping, and (C) simulation of dominant crop rotations. The analysis uses reported crop yields and N balances from six well-monitored catchments in Denmark's National Program for Monitoring Aquatic Environment and Nature (NOVANA).

Existing research has investigated the effects of aggregating soil and climate data on the accuracy of process-based models. Studies have shown that aggregation can lead to significant deviations in simulated biomass production and nitrate leaching, particularly in water-scarce regions. However, there is limited understanding of the impact of aggregating field-activity data, such as crop rotations and fertilization practices, on regional-scale nitrogen budget estimations. Recent studies have highlighted the importance of detailed management inputs for accurate yield and GHG simulations. For instance, Constantin et al. (2019) demonstrated the value of detailed management data for simulating winter wheat and maize in Germany, while Butterbach-Bahl et al. (2022) emphasized the significance of such data in identifying hotspots of CH4 and N2O emissions from rice systems in Vietnam. These studies underscore the need for investigating the effects of data aggregation on the accuracy and reliability of regional-scale nitrogen budget assessments. The Danish LOOP program, with its comprehensive data on agricultural practices, offers a valuable dataset for this investigation.

This study utilized data from six Danish agricultural catchments within the National Monitoring Program for Water Environment and Nature (NOVANA) – the LOOP program (2013-2019). These catchments represent a range of soil types, climates, and agricultural practices in Denmark. The LOOP program provides comprehensive field-scale data including crop types, cropping calendars, fertilizer applications (synthetic and organic), and catchment-scale NB estimates using a semi-empirical approach. The LandscapeDNDC (LDNDC) model, a process-based model simulating C, N, and water processes in various ecosystems, was employed. Three approaches were compared: 1. **Approach A (Detailed):** Utilized high-resolution field-level management and soil data (30.4 m resolution) without aggregation. This approach simulates realistic crop rotations for the 20 crops within the six catchments. 2. **Approach B (Mono-cropping):** Simplified the catchments to six mono-cropping systems (one for each major crop) on a representative soil type for each catchment. Simulation results were averaged according to the proportional area of each crop. 3. **Approach C (Dominant Rotations):** Simulated the 20 most dominant crop rotations in Denmark, deployed on a representative soil type for each catchment. Annual simulation results for main crops were averaged across all rotations and then post-processed using long-term average fractional cover. The model was run from 2001 to 2019 (with data from 2011-2020 for the final 10 years), incorporating soil properties, climate variables (from the Danish Meteorological Institute), and management data. Yield and NB components (gaseous N fluxes, nitrate leaching, N harvest) were compared across the three approaches and against the reported NOVANA values. Model calibration was performed, and sensitivity analysis was conducted to evaluate uncertainties and potential biases within the model.

Approach A (detailed field-level data) showed the highest correlation with observed yields (r² = 0.93), outperforming approaches B (r² = 0.92) and C (r² = 0.77), although all were statistically significant (p < 0.05). Significant differences emerged in simulated NB components. Compared to Approach A, Approach B overestimated total gaseous N fluxes by 31.4% (+7.6 kg-N ha⁻¹ year⁻¹) and showed a similar average nitrate leaching with a distinct pattern. Approach C overestimated total gaseous fluxes by 17.6% (+4.3 kg-N ha⁻¹ year⁻¹) and nitrate leaching by a substantial 204.9% (+30.2 kg-N ha⁻¹ year⁻¹). Approach A's average gaseous N losses were 24.3 kg-N ha⁻¹ year⁻¹ (N₂O: 1.5 kg-N ha⁻¹ year⁻¹; NO: 0.6 kg-N ha⁻¹ year⁻¹; N₂: 16.6 kg-N ha⁻¹ year⁻¹; NH₃: 5.6 kg-N ha⁻¹ year⁻¹), and nitrate leaching was 14.7 kg-N ha⁻¹ year⁻¹. LOOP6 exhibited the highest N loss, while LOOP2 displayed the highest N-use efficiency (2013-2019). Around 30% of the total area across all catchments showed signs of N depletion (negative net N budget), and ~43% had N-use efficiency between 60% and 80%. Spatial variations in NB components were evident at the sub-field level, mainly due to soil differences. Figure 5a visually compares N fluxes across the three approaches, showing consistent overestimations of gaseous N losses and nitrate leaching by approaches B and C. Figure 5b illustrates the temporal variation of NB components, revealing that approach B overestimates gaseous N fluxes while underestimating other components.

This study demonstrates the significant impact of data aggregation on the accuracy of regional-scale nitrogen budget simulations. The use of detailed field-level data (Approach A) provides more accurate estimates of both yield and nitrogen budget components. Aggregation methods (Approaches B and C), while simpler and computationally less demanding, introduce substantial uncertainties, particularly in the estimation of nitrogen losses to the environment. The overestimation of gaseous nitrogen fluxes and nitrate leaching observed in Approaches B and C highlights the importance of considering the spatial variability of management practices when assessing regional nitrogen budgets. These findings have implications for national and international nutrient budget estimations and comparisons among different sources using varied management representation approaches. The discrepancies observed in this study emphasize the need for careful consideration of the trade-offs between data resolution, computational resources, and the accuracy of nitrogen budget estimations. The study underscores the importance of incorporating detailed, high-resolution data whenever possible for accurate regional assessments. For policymakers and researchers interested in understanding the environmental impact of agricultural practices, choosing an appropriate aggregation method is crucial to ensure the accuracy and reliability of the findings.

This study demonstrates that data aggregation methods for representing crop management significantly impact the accuracy of regional nitrogen budget estimations. Using detailed field-level data yields the most accurate yield and NB simulations, while aggregation approaches introduce substantial uncertainties, particularly overestimating N losses. These findings emphasize the importance of considering data resolution and its trade-offs in regional modeling and highlight the need for detailed data when assessing the environmental impact of agricultural practices. Future research could explore more sophisticated aggregation methods or investigate the sensitivity of the results to different model parameters and configurations.

The study's findings are specific to the Danish context and might not be generalizable to other regions with different agro-ecological conditions or farming practices. The accuracy of the simulations is also dependent on the accuracy of the input data, including soil properties, climate variables, and management practices, which can be subject to uncertainties. The model parameters used in the LDNDC model could influence the simulations, and it is possible that other model parameterizations could lead to different results. Additionally, the number of catchments included in the study could limit the generalizability of the findings. Future studies with a wider range of catchments and regions would enhance the robustness of the conclusions.