Unraveling Nitrate Source Dynamics in Megacity Rivers Using an Integrated Machine Learning–Bayesian Isotope Framework

Rapid urbanization has intensified nitrate pollution in megacity rivers, posing severe challenges to urban water governance and sustainable nitrate management. This study presents nitrate dual-isotope signatures (δ¹⁵N-NO₃⁻ and δ¹⁸O-NO₃⁻) from surface water samples collected during the wet season from the Yongding River (YDR) and Chaobai River (CBR) in the Beijing–Tianjin–Hebei megacity region of North China. Average concentrations of nitrate (as NO₃⁻) were 8.5 mg/L in YDR and 12.7 mg/L in CBR. The δ¹⁵N-NO₃⁻ and δ¹⁸O-NO₃⁻ values varied from 6.1‰ to 19.1‰ and −1.1‰ to 10.6‰, respectively. The spatial distribution of NO₃⁻/Cl⁻ ratios and isotopic data indicated mixed sources, primarily sewage and manure in downstream sections and agricultural inputs in upstream areas. Isotopic evidence revealed widespread nitrification processes and could have potentially localized denitrification under low-oxygen conditions in the lower YDR. Bayesian mixing model (MixSIAR) results indicated that sewage and manure constituted the main nitrate sources (49.4%), followed by soil nitrogen (23.7%), chemical fertilizers (19.2%), and atmospheric deposition from rainfall (7.7%). The self-organizing map (SOM) further revealed three nitrate regimes, including natural and agricultural, mixed, and sewage dominated conditions, indicating a clear downstream gradient of increasing anthropogenic influence. The results suggest that efficient nitrogen management in megacity rivers requires improving biological nutrient removal in wastewater treatment, regulating fertilizer application in upstream areas, and maintaining ecological base flow for natural denitrification. This integrated framework provides a quantitative basis for nitrate control and supports sustainable water governance in highly urbanized watersheds.