Nitrate pollution of water bodies is a critical issue in many parts of the world because of its negative effects on aquatic ecosystem and human health. Effective management of pollution, such as the continuous or instantaneous release from point-sources, requires an understanding – with high spatial and temporal resolution – of how nitrate is dispersed and cycled within rivers. Nitrate sensing data show promise for this purpose, but their integration into numerical models is scarce; thus, questions remain regarding the necessary spatial grid size and temporal resolution required to resolve sensor readings. In this study, we developed an unsteady two-dimensional model to simulate nitrate transport, dispersal, and cycling along a 33-km stretch of the Kansas River (USA), following a strategic release of nitrogen from a decommissioned fertilizer plant. To validate modeled estimates of dispersion and uptake, we integrated 15-minute nitrate and temperature data from two aquatic sensors, one located proximal to the fertilizer release point and a second further downstream after complete lateral mixing. Model results at the site near to the contamination (0.4 km) were highly sensitive to river grid size and turbulent mixing, but insensitive to uptake. Results at the site far downstream of the contamination (31 km) were unaffected by grid size or mixing parameterization but were very sensitive to selection of uptake rate. High-frequency sensors allowed us to resolve diel variability in nitrate signals, which we incorporated into the model to improve performance and model realism. The 33-km study reach assimilated 14% of the total nitrate load in the river, or approximately half of what was contributed by the fertilizer release, during the two-month study period. Regarding nitrate cycling, modeled C/C ranged from 0.04 to 0.11 whereas sensor observations showed much higher C/C values of 0.11 to 0.25. Disagreements between data observations and model simulations in cycling are hypothesized to exist due to potential breakdown of the first-order rate kinetics. Together, our study shows the potential of combining numerical models and high-frequency data for a better understanding of the physical and biogeochemical processes that control nitrate dynamics in aquatic environments.

中文翻译：

---

使用高频传感器和二维建模揭示大型河流系统中硝酸盐的吸收和扩散动态

水体硝酸盐污染是世界许多地区的一个严重问题，因为它对水生生态系统和人类健康产生负面影响。有效的污染管理，例如点源的连续或瞬时释放，需要以高空间和时间分辨率了解硝酸盐如何在河流内分散和循环。硝酸盐传感数据显示出实现此目的的希望，但将其集成到数值模型中的情况很少；因此，关于解析传感器读数所需的空间网格大小和时间分辨率仍然存在问题。在这项研究中，我们开发了一个非稳态二维模型来模拟在废弃化肥厂战略性释放氮后，硝酸盐沿堪萨斯河（美国）33 公里河段的输送、扩散和循环。为了验证扩散和吸收的模型估计，我们集成了来自两个水生传感器的 15 分钟硝酸盐和温度数据，一个位于肥料释放点附近，第二个位于完全横向混合后的下游。靠近污染地点（0.4 公里）的模型结果对河流网格尺寸和湍流混合高度敏感，但对吸收不敏感。污染下游较远的地点（31 公里）的结果不受网格大小或混合参数化的影响，但对吸收率的选择非常敏感。高频传感器使我们能够解决硝酸盐信号的昼夜变化，我们将其纳入模型中以提高性能和模型真实性。在两个月的研究期间，这个长达 33 公里的研究河段吸收了河流中硝酸盐总量的 14%，或者说大约是化肥释放量的一半。关于硝酸盐循环，模拟的 C/C 范围为 0.04 至 0.11，而传感器观察结果显示 C/C 值要高得多，为 0.11 至 0.25。假设由于一级速率动力学的潜在破坏，自行车运动中的数据观察和模型模拟之间存在分歧。总之，我们的研究表明了将数值模型和高频数据相结合的潜力，可以更好地理解控制水生环境中硝酸盐动态的物理和生物地球化学过程。