The trouble with these emissions is that they are notoriously difficult to pin down. Nitrous oxide from treatment plants emerges from the biological machinery that removes nitrogen from sewage, and that machinery involves at least three distinct microbial production pathways operated by two different bacterial groups. Ammonium-oxidizing bacteria can release the gas as a byproduct of hydroxylamine oxidation, the first step in converting ammonium to nitrite, or through nitrifier denitrification, in which they reduce nitrite to nitrous oxide using ammonium as the electron donor. Meanwhile, ordinary heterotrophic bacteria perform classical denitrification, coupling the oxidation of organic matter to the stepwise reduction of nitrate and nitrite, with nitrous oxide appearing as an obligatory intermediate. Crucially, the same heterotrophs can also act as a sink, reducing nitrous oxide all the way to harmless dinitrogen gas. Which of these processes dominates at any given moment depends on oxygen levels, substrate availability and reactor operation, and until now the tools for watching them in real time simply did not exist.

A team of researchers from Empa, Eawag, Vito and the University of Münster, led by Hannes Keck of Empa’s Laboratory for Air Pollution and Environmental Technology, has now demonstrated a way to do exactly that. Writing in the journal Atmospheric Measurement Techniques, they describe the first long-term application of an off-axis integrated cavity output spectrometer, or OA-ICOS, for continuous isotopic analysis of nitrous oxide at a pilot-scale wastewater treatment plant. The instrument, equipped with a quantum cascade laser probing absorption lines near 2192.3 inverse centimeters, recorded the abundance of the rare nitrous oxide isotopologues carrying nitrogen-15 in the central alpha position, nitrogen-15 in the terminal beta position, and oxygen-18, at a temporal resolution of up to one spectrum per second. Over twelve months and 998 hours of measurement time at two eight-cubic-meter sequencing batch reactors fed with real municipal wastewater in Dübendorf, Switzerland, the setup delivered isotopic data fast enough to follow the daily rhythms of the microbial community itself.

The power of the approach lies in what the isotopes reveal. Because the nitrous oxide molecule is asymmetric, the position of a nitrogen-15 atom within it carries information about how the molecule was made. Each production pathway leaves a characteristic imprint on the difference between the two nitrogen positions, a quantity known as site preference, and on the bulk nitrogen and oxygen isotope ratios. Hydroxylamine oxidation, for instance, typically produces high site preference values, whereas denitrification pathways yield lower ones. Reduction of nitrous oxide to dinitrogen adds a further twist: enzymes preferentially cleave bonds between lighter isotopes, so as reduction proceeds the remaining nitrous oxide pool becomes progressively enriched in heavy isotopes. By plotting site preference against the oxygen isotope ratio, researchers can read off both the source of the gas and how much of it has already been destroyed by the microbes before it escaped to the atmosphere.

Getting such measurements to work outside the laboratory required considerable engineering. Laser-based isotope analyzers are sensitive to more than just the target gas: pressure broadening by the surrounding gas matrix, spectral interference from carbon dioxide and methane, and non-linear detector responses all distort the apparent isotope ratios. The team built a custom dynamic dilution system with five mass flow controllers that diluted the reactor off-gas with nitrous-oxide-free synthetic air to a stable target of twelve parts per million, a compromise between analyzer sensitivity and the need to cover real emission events. Off-gas was dehumidified by permeation drying, stripped of carbon dioxide by chemical absorption, and filtered before entering the spectrometer. Calibration gases bracketed every sample measurement to correct for instrumental drift, and dedicated experiments quantified the analyzer’s dependence on nitrous oxide mole fraction and on oxygen content, which varies as aeration bubbles through the reactors. The resulting combined uncertainties, between 0.5 and 1.1 per mil across all measured quantities, met the team’s target of one per mil for source attribution work.

The first major result came from four months of monitoring under standard operating conditions, with dissolved oxygen held at two milligrams per liter. The isotopic signatures pointed clearly to denitrification, either nitrifier denitrification or heterotrophic denitrification, as the dominant source of nitrous oxide, consistent with earlier full-scale studies. Just as telling was what was absent: no elevated site preference values above thirty per mil appeared, indicating that hydroxylamine oxidation played no significant role during the measurement periods. The data also traced a clear reduction line, a linear relationship between site preference and oxygen isotopes with a slope of 0.37, falling within the range expected for progressive microbial reduction of nitrous oxide to dinitrogen. In other words, the plant was simultaneously producing and destroying the greenhouse gas, and the balance between the two determined what ultimately reached the sky.

To probe that balance directly, the researchers ran a paired experiment in which one reactor operated at high dissolved oxygen and the other at a low setpoint of 0.5 milligrams per liter, with nitrite added to stimulate nitrous oxide production. The low-oxygen reactor progressed much further along the reduction line, and calculations based on the isotopic enrichment showed that the fraction of nitrous oxide converted to dinitrogen increased by forty percent compared with the high-oxygen reactor. Over the course of a single aeration phase, the low-oxygen reactor’s reduction fraction climbed from thirty-five to fifty percent before settling near forty, while the high-oxygen reactor’s fell steadily from ten percent to zero. The explanation is enzymatic: the copper-sulfur cluster of nitrous oxide reductase, the enzyme that performs the final step of denitrification, is notoriously oxygen-sensitive, so suppressing aeration gives the microbes a chance to finish the job they started.

This finding carries immediate practical weight. Dissolved oxygen setpoints are among the most controllable parameters in plant operation, and the new data suggest that lowering them during critical periods could tip the microbial balance toward destruction of nitrous oxide rather than its release. The isotopic approach makes this actionable in near real time, because operators can see not just how much gas is escaping but which process is responsible and whether mitigation is working. The team also demonstrated a third capability: low-level nitrogen-15 labeling, in which small amounts of tagged ammonium are added to raise isotope values just above natural abundance. When they spiked a reactor with ammonium enriched to about one hundred per mil, the emitted nitrous oxide showed a corresponding jump in bulk nitrogen isotope values, from around minus forty per mil under standard operation to 6.5 per mil, tracing the labeled nitrogen through nitrite and nitrate into the gas phase and confirming the setup’s suitability for tracing specific microbial conversions.

The researchers are candid about the limitations. More than half of the raw data had to be discarded because nitrous oxide mole fractions drifted outside the narrow eleven-to-thirteen-ppm window where corrections are reliable, particularly during rapid emission spikes. Carbon dioxide traps occasionally broke through before reaching capacity, and suitable isotopic calibration standards at process-relevant concentrations are not yet commercially available. Yet the authors argue that the alternative, manual bag sampling followed by laboratory analysis, carries comparable costs while sacrificing the temporal resolution needed to capture the strong daily and seasonal dynamics of plant emissions. The spectroscopic platform is also scalable: measurement frequency can be increased, multiple reactors can be analyzed sequentially, and with minor adaptations the system could move from pilot scale to full-scale plants, where isotopic data could feed directly into operational control strategies.

What emerges from this year-long experiment is a template for turning wastewater plants from opaque emission sources into observable, optimizable systems. The isotopic fingerprints showed that denitrification, not hydroxylamine oxidation, drives the bulk of nitrous oxide production, that oxygen availability strongly controls how much of the gas the microbes manage to destroy themselves, and that operational levers like dissolved oxygen setpoints can be tuned with evidence rather than guesswork. As countries revise their emission inventories upward and plants face pressure to cut their climate footprint, tools that reveal the invisible microbial accounting of nitrogen in real time may prove essential. The laser in a Swiss pilot reactor has, for the first time, made that accounting visible, one five-minute measurement at a time.

Subject of Research: Real-time isotopic analysis of nitrous oxide production and reduction during biological nitrogen removal in wastewater treatment

Article Title: On-line analysis of N2O isotopic composition during biological nitrogen removal in wastewater treatment to disentangle production and reduction processes

Article References: Keck, H., Strubbe, L., Magyar, P. M., Joss, A., Froemelt, A., Kupferschmid, A., Knorr, K.-H., & Mohn, J. (2026). On-line analysis of N 2 O isotopic composition during biological nitrogen removal in wastewater treatment to disentangle production and reduction processes. Atmospheric Measurement Techniques, 19(19), 6311-6326. https://doi.org/10.5194/amt-19-6311-2026

Image Credits: AI Generated

DOI: 10.5194/amt-19-6311-2026

Keywords: nitrous oxide, wastewater treatment, stable isotopes, laser spectroscopy, denitrification, nitrification, greenhouse gas emissions, site preference, dissolved oxygen, OA-ICOS, nitrogen removal, isotopocule analysis