Comparative modeling of nitrogen losses in a tile-drained watershed using SWAT model: uncertainty and calibration considerations

Authors: QI, JUNYU - University Of Maryland; Malone, Robert; LIANG, KANG - US Department Of Agriculture (USDA); Cole, Hailey; Emmett, Bryan; Moriasi, Daniel; SHAHID, MUHAMMAD - Iowa State University; CASTELLANO, MICHAEL - Iowa State University

Submitted to: Frontiers in Environmental Science
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 10/8/2025
Publication Date: 10/27/2025
Citation: Qi, J., Malone, R.W., Liang, K., Cole, H.D., Emmett, B.D., Moriasi, D.N., Shahid, M.R., Castellano, M. 2025. Comparative modeling of nitrogen losses in a tile-drained watershed using SWAT model: uncertainty and calibration considerations. Frontiers in Environmental Science. https://doi.org/10.3389/fenvs.2025.1651136.
DOI: https://doi.org/10.3389/fenvs.2025.1651136

Interpretive Summary: Computer models are necessary to fully understand the complex dynamics in tile-drained watersheds and to assess different management strategies. We assessed the performance of the Soil and Water Assessment Tool (SWAT) model in predicting nitrogen (N) losses within a typical Midwestern tile-drained watershed, focusing mainly on comparing original and new tile drainage and N modules. Daily streamflow and nitrate load data from three monitoring stations were used for model assessment. We found that the new tile and N modules generally improved daily streamflow and N load predictions compared to the original. However, the new N module did not accurately predict nitrous oxide without field measurements to restrain the model. These results suggest that more comprehensive/detailed field observations and estimates will help improve the revised SWAT predicted N budgets. Our findings are important to scientists and policy makers using models to evaluate agricultural management practices and N fluxes in the Midwestern U.S.

Technical Abstract: Ecohydrological models are crucial for understanding the complex hydrological and biogeochemical dynamics in tile-drained watersheds and assessing various management strategies. Using a comparative modeling approach to assess the performance of SWAT in predicting N losses within a typical Midwestern tile-drained watershed, we evaluated eight model configurations by combining original and new tile drainage modules, calibrated and non-calibrated tile parameters, and original and new N modules. Daily streamflow and NO3' load data from three monitoring stations were used for model calibration, validation, sensitivity analysis, and uncertainty assessment. The results indicated that all model configurations adequately simulated daily streamflow and NO3' loads for the three stations, although peak flows and NO3' loads were underestimated. We found that the new tile and N modules generally improved daily streamflow and NO3' load predictions compared to the original, such as average increases in percent bias of 1–15% across the three stations. Calibrating tile parameters did not notably improve daily streamflow prediction, while increasing parameter sensitivity and uncertainty, particularly for NO3' load predictions. Overall, we recommend using the new tile and N modules for simulating streamflow and NO3' fluxes in tile-drained watersheds. However, the new N module introduces additional parameters, increasing uncertainty in simulating N gas fluxes due to limited observational data for parameter calibration. This highlights the challenge of modeling overall N fluxes using only outlet NO3' load data without N2O measurements, suggesting that estimated or observed field data will help constrain soil N cycling and improve overall N budget estimates.