Differentiating molecular fingerprints of dissolved organic nitrogen between natural soil and iron-based sorption media for stormwater treatment.

Dissolved organic nitrogen (DON) remains a significantly under-investigated component in the disruptive nitrogen cycle especially in different soils or specialty adsorbents that can efficiently remove inorganic nitrogen after pollution when simultaneously retaining, removing and remaking DON. The complexity of DON fate, transport, and transformation processes in low impact development infrastructure design is still a pending question for stormwater treatment. To compare different DON degradation pathways and clarify inherent complexity, we integrated paired influent-effluent column experiments with Fourier transform ion cyclotron resonance mass spectrometry, molecular descriptor and zonation analyses, fate-resolved reaction inference, Kendrick mass defect-based network analysis, and depth-resolved nitrogen-cycling gene profiling to examine DON transformation in natural soil, biosorption activated media (BAM), and two iron-filings-based green environmental media (IFGEM1 and IFGEM3) under three nitrate-phosphate loading conditions. Across the three nutrient-loading conditions, IFGEM1 and IFGEM3 achieved NOx-N (NO₃⁻-N + NO₂⁻-N) removal efficiencies of 98.6-99.2% and 98.9-99.9%, respectively, while their effluents showed systematic shifts in the assigned CHON formula pool. Relative to natural soil, IFGEM effluents showed lower relative contributions of condensed aromatic-like formulas and higher contributions of protein/amino sugar-like formulas, together with greater formula richness and stronger separation in molecular descriptor space, most notably for IFGEM3. Natural soil increasingly retained a lignin-rich, oxidizing-saturated DON pool as loading increased, whereas IFGEM3 promoted greater removed-to-produced turnover and concentrated newly produced DON in reducing-saturated chemical space under higher loading. Increasing N and P loading also redistributed inferred transformation portfolios: natural soil and BAM retained broader, more mixed reaction spectra, whereas IFGEM1 and IFGEM3 shifted towards more selective pathway combinations. KMD-derived formula-association networks showed broader cross-class connectivity in IFGEM3 than in natural soil, with the network edges interpreted as putative molecular linkages rather than confirmed transformations. Pre-spiking qPCR profiles showed depth-dependent differences in baseline nitrogen-cycling gene abundances. Overall, iron-based engineered media were associated with systematic restructuring of the detectable DON molecular composition across the tested nutrient-loading conditions.