The dynamic chamber system uses two identical commercially available units (eosAC-LT Eosense, Dartmouth, NS). These are modified and coupled to programmable valves that control sample gases delivered to a suite of instrumentation (Fig. 1A).

The dynamic chambers are constructed with transparent polyacrylate walls and lids, with an internal volume of 0.072 m³ (72 L) and bottom surface area of 0.21 m². When used on soils the chambers are secured with collars and custom-made polytetrafluorethylene (PTFE) rings (Fig. S1 in the Supplement). A built-in fan ensures uniform distribution of gases inside the chamber. Air temperature is measured inside from the fan arm, pressure is from the control box outside the chamber, and two auxiliary ports, one internal and external, can collect environmental properties such as relative humidity (RH), photosynthetically active radiation (PAR), soil temperature, and/or soil volumetric water content (VWC).

For an Nr sampling approach, one chamber is used as the measurement (MC) from an experimental surface while the second is a reference (RC) sealed at the bottom with a 51 µm (0.002^′′) film of perfluoroalkoxy alkane (PFA; McMasterr-Carr^®, PN: 84955K24). The inert PFA film is held in place between the chamber collar and our custom-made PTFE rings (Fig. S1). The RC acts as a negative control for physical interactions and/or associated chemistry of reactive gases on chamber and gas transfer line surfaces. The use of the RC, therefore, is to facilitate the correction of surface-mediated effects, reactions, and reduction of bias when determining flux values.

The RC component of this system is designed to continuously baseline the physical interactions and chemistry happening on its surfaces both before and after quantifying reactive gas fluxes with the MC. Here, the flux measurements are made every 30 min, where one chamber is closed for gas analysis while the other is open to the ambient atmosphere. The sampling time interval was determined based on (i) obtaining enough measurements at 1 min time resolution to perform a reliable accumulation or loss linear regression, and (ii) an ability to detect the lower limit of field HONO flux values previously reported in the literature for our pilot field study (see Sect. 2.7). For the first criterion, this includes an exclusion of the first and last few measurements (3 to 5) to allow complete gas replacement in the chamber lines and analyzers, as well as the disruption of the sealed environment when the chamber cycle alternates. The resulting accumulated mixing ratios of HONO at the lower literature limit in the chamber, closed for 30 min, are well above the 1.4 parts per billion (ppbv) mixing ratio detection limit (LOD) of even a modified NOx analyzer (Crilley et al., 2023; Lao et al., 2020; Zhou et al., 2018; Nodeh-Farahani et al., 2021).

Headspace recirculation to facilitate analyte mixing ratio accumulation or depletion for non-destructive spectroscopic GHG analysis is a common measurement approach to decrease flux observation times. Reactive nitrogen measurements, in contrast, are typically destructive techniques that change the identity of the target analyte in the act of quantifying its abundance. To interface with such instruments, the sampled air needs to be replaced (Linde Canada Plc, PN: NI LC250-230) to balance the flow demand in a closed chamber. This balance is delicate even when using mass flow controllers (MFCs) on both incoming and outgoing flows, and the best solution we identified is to provide a slight overflow that takes advantage of the chamber design to vent excess pressure through a short length (∼ 15 cm) of 1/8^′′ ID (3.175 mm), 1/4^′′ OD (6.35 mm) tubing that keeps the internal pressure equivalent to ambient. The flow differential between make-up gas and sampling is roughly 400 cm³ min⁻¹. Such a supply of make-up gas was explored across a range of potential flow rates when using destructive gas analyzers (e.g., three instruments each sampling at 1–4 standard litres per minute, L min-1) to find that 6 L min⁻¹ is the upper limit of flow-through where the chamber pressure is not substantially perturbed from ambient and the chamber lid retains its seal.

In the field, a flux measurement cycle begins with closing the RC while the MC is open. At defined intervals, they alternate their open-closed states. Flows of make-up gas to each chamber are modulated with a pair of two-way solenoid valves. When sampling from the RC, one valve (V1; Fig. 1) is open to permit make-up gas flow while the other valve (V2) is closed to prevent the flow from being directed to the MC. On the sampling lines, a three-way solenoid valve (V3) alternates to guide flow from whichever is closed to the suite of gas analyzers. All instrument and operational details are provided in Sect. 2.1.3.

The chamber eosLink-AC software (Eosense Inc., Dartmouth, NS) is used to define the duration of chamber opening and closing cycles, and logs chamber temperature, pressure, and auxiliary sensor data associated with a given eosAC-LT chamber. Each chamber requires a 12 V DC power supply connected by USB to a laptop through a weatherproof communication cable, controlling the chamber lid and data transfer.

When a chamber cycle begins, a text file is generated and includes measurement time elapsed, chamber lid status, and sensor data. This data file is updated at least once every 10 s, varying between 2 and 8 s intervals, which we average onto a 1 min time base to match measurements from the slowest gas analyzers. The solenoid valves are modulated by the electrical circuit shown in Fig. 1B. Automation can be facilitated by a microcontroller (NI-6509i, National Instruments) programmed with a custom-scripted LabVIEW VI (LabVIEW version 2020). Further design information and full details of this sampling strategy and script can be found in Sect. S1 of the Supplement and is available on the GitHub repository alongside our VI (https://github.com/fjs-vdblab/fluxchamber.git, last access: 2 April 2026).

The mixing ratios of NO and NO2 were measured using a commercially available chemiluminescent NOx analyzer (EC 9841, American Ecotech, Warren, RI). The calculated LOD determined from sampling dry zero air was 0.8, 0.7 and 1.1 ppbv for NO, NOx, and NO2 (or HONO when using the denuder as described below), respectively. The instrument has an operating range of 0–20 parts-per-million by volume (ppmv), a sample flow rate of 0.5 L min⁻¹, and reports measurements at a time resolution of 1 min. To quantify NO2, it is reduced to NO on a heated molybdenum catalyst (325 °C). To prevent interferences reported by others from atmospherically-relevant acidic species in this system (e.g. HONO, HNO3, and N2O5) the sampled air from the chambers during field experiments was passed through a sodium carbonate (Na2CO3) coated annular denuder to reduce bias in the NO2 measurement, as these species and other components of NOy (e.g. peroxyacetyl nitrate; PAN) may also be reduced to NO (Villena et al., 2012). The Na2CO3 denuder was prepared according to the EPA Compendium Method IO-4.2 (Winberry et al., 1988; U.S. EPA, 1999) to remove atmospheric acids by reactive uptake to the basic coating. As part of our controlled laboratory and pilot field study experiments, this denuder was also used to selectively measure HONO by scrubbing this target gas for a specified period, but it would include the other known interferences. If the NOy term is depositing, it could include other NOy detected by the same conversion mechanism, but if emitting we expect it to be dominated by HONO. Ideally, a platform like time-of-flight chemical ionization mass spectrometry (ToF-CIMS) would be used for disambiguation, but was not available at the time of this work.

A commercial O3 analyzer (Serinus 10, American Ecotech, Warren, RI) was used to measure mixing ratios, quantify O3 loss to surfaces, and constrain the reaction of O3 with NO to form NO2 in the pilot study sampling. This analyzer employs a non-dispersive UV absorption cell to quantify O3 in the sampled air. The calculated LOD from sampling zero air is 0.95 ppbv at 1 min time resolution, with an operating range of 0 to 20 ppmv and a sampling flow rate of 0.5 L min⁻¹. Quality control procedures for the NOx and O3 instruments can be found in Sect. S2.

The mixing ratios of the GHGs N2O, CH4, CO2, H2O, and NH3 sampled from the automated chamber system were measured using a Picarro G2509 which uses cavity ring down spectroscopy. The analyzer has a time response of ∼ 8 s for N2O, CH4, CO2, H2O and < 2 min for NH3. It was used for the lab experiments and the pilot field study. The customized version of the instrument sampled at ∼ 0.23 L min⁻¹ and was equipped with an inlet filter. To minimize adsorption and chemical interactions of NH3 on instrument surfaces, stainless steel gas handling components, including the inlet bulkhead, were replaced with PFA counterparts. The instrument cavity material was treated with a SilcoNert^® coating by the manufacturer. We did not observe changes in its performance for the measured gases when operated according to the manufacturer guidelines. The Picarro G2509 analyzer spectroscopic mixing ratio determination means a full span calibration is not a regular necessity. Despite this, we validated its calibration and performed quality control checks in the lab to ensure the accuracy and stability of the analyzer for all aspects of this work (Sect. S2).

Control experiments for the transmission of GHG and Nr gases were conducted by filling and emptying the chambers with known quantities at mixing ratios relevant to the atmosphere, as well as quantities expected to accumulate during real observations of modest emission fluxes (e.g. from a fertilized agricultural field). All assessments herein matched: the standard configuration of the chambers with all fittings, 15 m of 1/4^′′ O.D. (6.35 mm) PFA gas transfer tubing, flow rates, valves and gas transfer lines to instrumentation, with line/fitting/valve/instrumentation surfaces included. Details to control gas concentrations can be found in Sect. S3.

The positive control experiments filled the chambers in both modified and unmodified configurations and time constants were calculated from the measurements. In each filling experiment, the chamber was flushed with pure N2 from a dewar (Linde Canada, PN: NI LC250-20) until a stable baseline level of each gas mixing ratio was reached; typically, these were values at the analyzer detection limits. Then, a blend of GHGs or one of the Nr analytes was delivered into the chamber with N2 dilution at 2 L min⁻¹, which was sampled at 1.8 L min⁻¹ by the analyzers (Fig. S4). The gases were added to the chamber until the observed concentration (C) reached the known value being delivered (C0), within error. Since different mixing ratios of the gases were added for these control experiments, normalized concentrations facilitate data analysis and visualization. Where surface interactions could be identified (e.g. for NH3), the role of surfaces versus air exchange was explored using double exponential fits (see Eqs. S1 and S2 in Sect. S3) (Crilley et al., 2023; Ellis et al., 2010; Moravek et al., 2019). The gases were emptied back to the initial baseline level before starting the next replicate or a new experiment with a different target gas. Time constants for filling and emptying were determined by fitting the observations in Igor Pro 8 (Wavemetrics, Portland, OR, US).

Similarly, the Eosense eosMX multiplexer is designed to coordinate chamber flux measurements using eosAC chambers with non-destructive analyzers, such that headspace can be recirculated while the chambers are closed. One of these devices was characterized following similar modifications. This system is appealing as it has the eosLink-MX software (Eosense, V1.9.07), for communication, scheduling actions, and logging peripheral data from all connected eosAC chambers. It features dedicated chamber tubing inlets and outlets, along with a COMM port supporting up to 12 eosAC chambers. Each chamber channel includes two Swagelok gas fittings for transport to analyzers and for recirculating, or in the case of Nr measurements, supplying a dilution gas to the chamber headspace.

To optimize the performance for Nr species, the original stainless-steel (SS) Swagelok fittings and solenoid valves were compared against replacement PFA bulkhead unions (Swagelok, PFA-420-61), and a PTFE 3-way valve (Clippard, NR1-2-12-G2). A 2 L min⁻¹ flow of dry zero air containing the target compounds was passed through the SS valve with SS fittings or the PTFE valve with PFA fittings for 30 min each. The flow was measured before and after the valves to ensure the setup was free of leaks. The ratio of the transferred gas amount to the nominal identified surface interactions impacting downstream gas analyzers. Details are presented in Sects. 2.7 and S4.

Randomized soil samples weighing 4–5 kg were collected into Ziploc^® bags from an eight-plot grid established at an agricultural field site in Lambton County, ON, Canada (43°09^′36.0^′′ N 81°55^′48.0^′′ W). The samples were used to investigate emissions in the lab with and without the addition of fertilizers. Individual and pooled samples from the plots were used. Bulk soil samples were prepared for lab-based chamber measurements by removing debris, roots, and seeds, followed by drying at 35 °C for 24–72 h on a stainless-steel mesh tray covered with aluminum foil, to prevent alteration of the microbial community from exposure to unrealistic temperature regimes. After drying, samples were stored in Ziploc^® bags at room temperature until use.

Ultrapure water (18 MΩcm; Milli-Q^®, Sigma-Aldrich, St, Louis, US) was added to approximately 350 g of a dry soil sample to achieve ∼ 28 % volumetric water content (VWC). The soil sample was loaded into the chamber on a foil-lined tray, and the water content was measured using a soil moisture probe inserted fully into the sample (TEROS 11, VWC range for mineral soils: 0.00–0.70 m³ m⁻³; accuracy: ±0.03 m³ m⁻³; resolution: 0.001 m³ m⁻³, METER Group Inc., WA, USA). Zero air modified to 65 % RH was delivered at 3.6 L min⁻¹ to the chamber headspace where the soil was contained. The soil started the drying process from a VWC of approximately 25 % with the flows held constant for around 4 d or until the VWC reached 15 %. The chamber was sealed to conduct the drying cycle while our modified NOx analyzer and the Picarro G2509 measured fluxes. Unamended soil samples fertilized with urea (CO(NH2)2), ammonium carbonate (AC, (NH4)2CO3), and ammonium bicarbonate (ABC, NH4HCO3) at mass surface densities of 100 kg N ha⁻¹ were assessed. A similar experiment was conducted with ammonium nitrate (NH4NO3) at the same fertilizer mass surface density using only the modified NOx analyzer. These fertilizer application values are at the upper end of those used in North American and European agriculture at present. Soil VWC and headspace RH were recorded using auxiliary sensors within the chamber.

The RC and MC setup was deployed to make automated Nr flux measurements from the same agricultural field as in the prior section. The observations took place in early September 2022 at the end of a soybean (Glycine max) cropping season. A detailed description of the campaign and its results is the subject of a separate work, so we only provide a brief overview here. The total measurement period was approximately two weeks in duration to test system performance. Generally, conditions were hot and dry, without precipitation, and the soybeans surrounding the observed soils were undergoing senescence during the measurement period. The chambers were deployed only on the soil, between crop rows, and operated to quantify fluxes as outlined in Sect. S7. After an initial 7 d period of observing baseline fluxes from the field, an experimental perturbation was conducted to stimulate Nr emissions through the addition of an aqueous urea solution equivalent to 22 kg N ha⁻¹ of fertilizer added by broadcast application, followed by washing into the soil by an equivalency of 2.5 cm (1^′′) of rain depth. The modified NOx analyzer and the Picarro G2509 were used to measure the Nr and GHG fluxes continuously.

Determination of the GHG time constants benchmarks the chamber performance before and after modifications. In both configurations CO2, CH4, and N2O were anticipated to behave as non-reactive trace gases with little to no physical interactions on chamber surfaces.

The theoretical fill and empty rates for the chambers with a flow rate of 2 L min⁻¹ are 36 min. The average measured time constants of filling for CH4, CO2, and N2O in the unmodified chamber were 37 ± 1, 37 ± 2, and 37 ± 1 min, respectively (Table 1). During emptying, they were 37 ± 1, 37 ± 1, and 38 ± 2 min, respectively. These measurements are not different from theory within the limits of experimental accuracy (Fig. 2). Since the GHGs are effectively transferred through both the modified and unmodified configurations of the chamber, the baseline performance of the chambers was not affected by the hardware modifications. Therefore, comparison with the time constants of Nr gases provides a description of their interaction or transformation processes on the modified chamber surfaces.

In the modified chamber, the time constant of filling or emptying with NO was 38 ± 1 min. The obtained value is similar to GHGs, as NO is not expected to have strong surface interactions. The slower time response of the NO2, HONO, and NH3 measurements results from two processes: (1) the exchange of the sample air volume in the gas transfer lines and the chamber, and (2) the adsorption and desorption of the gas onto and from their surfaces (Whitehead et al., 2008). Increases in the extent of surface interactions followed the increasing polarity, reactivity, and/or ionizability of gases in the order of NO2, then HONO, and most for NH3 (Fig. 2).

For example, NO2 is lost more readily than NO, possibly through its known reaction on surfaces to make HONO and HNO3, being lost itself in the process (Finlayson-Pitts et al., 2003). It also has higher water solubility than NO, but lower than for HONO or NH3. Similarly, a decrease in transmission efficiency for HONO could be explained by its weakly acidic nature (pKa of 3.16) (Finlayson-Pitts et al., 2003) and solubility in water (Henry's law constant of 0.48 mol m⁻³ Pa⁻¹; Schwartz and White, 1981) that facilitate partitioning and dissociation in surface water films, which could generate non-volatile nitrite (NO2-) on chamber surfaces. This chemistry will slow the transfer of gas-phase HONO through the chambers, as the NO2- would need to protonate before being lost as neutral HONO when repartitioning to the gas phase (Reactions R4, R5). Finally, NH3 has the most delayed transmission, likely because it undergoes strong inter-molecule interactions and ionization on the chamber surfaces and/or with any interfacial water (Henry's law constant of 5.9 × 10⁻¹ mol m⁻³ Pa⁻¹, pKa of 9.25, Lide, 2009). The same interactions on tubing surfaces and potential partitioning into the tubing material may also occur before reaching the analyzer (Pagonis et al., 2017). The best improvement between the modified and unmodified configurations was 2.6 ± 2.5 min for HONO, with smaller improvements observed for NO2 and NH3 during the filling process. Improved time constants when emptying Nr from the chambers had similar trends (Table 1). As a result, increasing delays from NO through NH3 exist in our Nr gas suite due to increasingly stronger interactions with chamber surfaces and gas handling lines.

The determined surface interaction values (D; Table 1; Eq. S2, Sect. S3) demonstrate the expected greater impact of surfaces when no reactive gas is present in the headspace prior to filling, and a lesser effect during emptying as the exposed surface has equilibrated with the analyte, which is commonly referred to as passivation. For NH3 specifically, the fill has a D value of 89 %, while during emptying it is only 23 %, similar to our findings with NH3 transfer for other Nr instruments (Crilley et al., 2023). The surface interactions for these gases are minimized in the modified chambers to facilitate more time-efficient measurements of surface exchange. However, they necessitate the use of the λ term when deployed in the MC-RC configuration for those Nr species which experience partial transmission, such as NH3. The λ term is required to obtain accurate values, as the enclosed flux measurement surface should be perturbed for the least amount of time possible when making field measurements, and the chambers cannot be closed for several hours to allow surface-active gases to passivate the lines. One potential option to improve the system performance further for NH3 could be to heat the gas transfer lines between the chambers and gas analyzers. In addition, minimizing the potential for transformations reduces the frequency required for in-field characterization of these processes through positive and negative gas delivery controls. For NO2, specifically, we sought to quantify this as a function of modifying components of our chambers, as NO2 is the most reactive gas in our suite (Sect. S5).

Substantial reduction in NO2 loss fraction (fNO2) and transformation was observed from the implemented PFA and PTFE modifications under conditions of 83 % RH and 5 ppb of NO2. In the original unmodified configuration of the chamber, fNO2 was 0.36 ± 0.02 (Fig. S7) which was reduced to 0.22 ± 0.03 with the PFA film, a relative decrease of 18 %. This is consistent with the acrylic chamber surfaces and fasteners to the chamber frame facilitating physical and/or chemical loss of NO2.

The film of PFA, as with other fluoropolymers, is known to have excellent chemical resistance and low reactivity towards a range of chemicals, including NO2 (Scheirs, 2005). In addition, the superhydrophobic nature of these materials reduces the accumulation of water on surfaces, which can reduce the surface reaction of NO2 (Finlayson-Pitts et al., 2003; Jenkin et al., 1988; Stutz et al., 2004) and analytical bias in the measurement of trace gases like HONO, especially when instrument gas sampling inlets do not take this into account (Crilley et al., 2019; von der Heyden et al., 2022).

The replacement of the brass-lined push-to-connect bulkhead fittings with PTFE led to a similar decrease in fNO2, which was reduced by 17 % to a final value of less than 0.05 ± 0.02 (Fig. S7). These fitting surfaces act as the largest surface-driven NO2 loss despite their surface area being very small compared to that of the entire chamber configuration and with a very small contact time against the gas sample (0.012 s per fitting at a flow rate of 2 L min⁻¹).

The loss of NO2 in the commercially available system is challenging to attribute solely to the heterogeneous hydrolysis reaction. During the characterization experiments, the conditions inside the chamber were matched to those reported by previous lab studies, which have shown that high RH, presence of NO2, and surface adsorbed water on surfaces favour this loss mechanism (Jenkin et al., 1988; Stutz et al., 2004). The reaction is known to occur on surfaces such as Pyrex (Jenkin et al., 1988) and borosilicate glass (Finlayson-Pitts et al., 2003), but no prior studies to date, nor this study, have demonstrated metallic surfaces as facilitating this mechanism.

The inert PTFE fittings dramatically minimized transformations, while PFA film lining the inner chamber surfaces was also effective, but less so. Our results indicate that water-adsorbed and metallic surfaces, such as brass, facilitate substantial loss and/or transformations of NO2. Further investigation is required to confirm the mechanism(s) at play and is beyond the scope of this work.

A complete characterization of fNO2 and the amount of HONO in the fully modified chamber was determined across a range of environmentally relevant RHs and NO2 concentrations. We found that the absolute and fractional NO2 losses were highest under the highest RH conditions (85 %; Table 2). However, the fNO2 does not appear to follow a concentration-dependent trend across the additions made at lower RHs, with at most 0.4 ppbv NO2 lost across the remainder of the tests, a value which is equivalent to the LOD of the NOx analyzer used. This would generate 0.2 ppbv of HONO according to the disproportionation of the hydrolysis mechanism, which is well below the analyzer detection limits. The modifications successfully reduced NO2 losses below 10 % across all environmentally relevant conditions the chambers are expected to encounter, with our findings here suggesting that the mass lost is nearly constant and independent of NO2 mixing ratio at RHs below 85 %, while being marginally higher at and above this value.

Quantifying fNO2 and the amount of HONO made in the chamber is required for the correction of field datasets. The dual chamber system, via the RC, can also quantify any changes in these processes over time if standard additions to the headspace are conducted. Consequently, important parameters such as NO2 deposition fluxes on surfaces can be better estimated (Pape et al., 2009).

Since the inferred HONO mixing ratios from the chamber surfaces across various environmental RHs are nearly invariant at 0.2 ± 0.1 ppbv, it is simple to background correct any observational datasets by subtracting this amount from the total HONO measured in the chamber. In addition, as the NO2 values expected in most atmospheric gas samples during field measurements are well into the ppbv range (> 3.3 ppbv per 30 min flux measurement for a 0.08 µgNm-2h-1 emission), the corrections would be easy to implement in post-processing of datasets and have minimal impact on the technical aspects of the analytical determinations.

It should be noted that the amount of HONO in the chamber was below the LOD of the NOx analyzer for HONO (1.1 ppbv), meaning that the upper limit of HONO inferred may perhaps, in fact, be negligible. Therefore, future experiments that wish to detect small Nr fluxes accurately will need to focus on reproducing these experiments with a higher performance instrument, such as a time-of-flight chemical ionization mass spectrometer (ToF-MS) or long-path absorption photometer (LOPAP), which have lower detection limits (Crilley et al., 2019; Lee et al., 2014; Neuman et al., 2016; Reed et al., 2016).

Ozone loss was observed in both modified and unmodified chamber configurations, with 18 % lost to clean 15 m PFA lines alone when transferring 30 ppbv. The unmodified chambers lost 45 % across delivered mixing ratios spanning 150–250 ppbv. This was reduced to 35 % when the fluoropolymer modifications were implemented. When the PFA film was aged by 15 d of ambient sampling and 2 years of exposure to lab air, the losses were substantially exacerbated, reaching 80 %. Such outcomes are expected and can be attributed to several factors discussed in detail in Sect. S5, primarily involving surface reactions with built-up films of deposited organics, adsorption, and material interactions (Burkholder et al., 2015; Ebnesajjad, 2017; George et al., 2015; Plake et al., 2015b). We recommend regular replacement of the PFA film as part of the Nr system maintenance, coupled with quality control procedures to characterize material performance for target gases.

Emission fluxes were measured from two pooled and two individual soil samples collected from a single agricultural field (Table 3; Sect. S6). The average and integrated fluxes of N2O, NH3, NO, NO2, and HONO were assessed under controlled, environmentally realistic (65 % RH), drying conditions (Table 3).

As the soils dried, NO and NO2 emissions increased, with NO fluxes highest across all replicates and reaching up to 2.50 µgNm-2h-1. This trend is consistent with the prior work of other researchers, showing peak NO emission potentials when VWC drops below 25 % during soil drying, which is when microbial nitrification and denitrification processes are suggested to become more active (Bao et al., 2022; Oswald et al., 2013). The soil VWC at which these maxima occur can vary depending on soil type, texture, and microbial diversity therein (Ludwig et al., 2001; Schindlbacher et al., 2009). Plot-level replicates from our field had a higher integrated NO flux (i.e., > 2600 µgNm-2), compared to the pooled replicates (< 1000 µgNm-2), likely indicating real differences in preserved microbial hotspots, intact plot-level soil aggregates, true spatial variability, and plot-specific N availability (Table 3). Soil texture and aggregate size, for example, play an important role in building the porous structure of soil, which has implications for the release of gases (Mangalassery et al., 2013). Soil aggregates, therefore, govern the release of gaseous Nr analytes like NO based on the aerobic or anaerobic state of the soil. Here, our low level of soil manipulation (i.e. not ground, no sieving) will drive some of the variability by preserving these features, which exist across and within real soil systems (Lipiec et al., 2007). Individual plot samples also retain plot-specific microbial communities when working with intact soil, whereas soil grinding can temporarily inhibit microbial activity. While we tried to minimize soil handling and processing extremes in these experiments, a measure of homogeneity was also pursued, and fully intact soil cores were not assessed.

Integrated NO2 fluxes showed the same trend, with more sample-to-sample variability. For example, one pooled replicate (R2) produced over three times the emissions of (R1), despite both experiments being conducted across identical moisture content ranges (Table 3). Given the limited studies directly measuring NO2, such as Purchase et al. (2023), this variability is difficult to interpret and highlights the need for more assessments of its production pathways and controls, which our developed chambers show promise for.

The observed average HONO fluxes remained low across all of the samples, ranging from 0.05 to 0.25 µgNm-2h-1 (Table 3). These values are lower by more than an order of magnitude compared to those reported in other controlled laboratory studies, where HONO fluxes exceed 900 µgNm-2h-1 (Oswald et al., 2013; Su et al., 2011; Wang et al., 2021). These discrepancies are concerning, given recent emphasis from the scientific community on the atmospheric impacts of soil-derived HONO on air quality. Here, the results from our agricultural soil samples may reflect the differences in our methodology, such as the soil handling and preparation steps prior to and during experiments. Many prior reports prepare their samples in ways that strongly deviate from real-world conditions (e.g. initial soil drying temperatures above those occurring under ambient conditions, extreme storage conditions, grinding, sieving, use of dry zero air to flush chambers). Further drivers of variability within the category of heavily altered soil samples from the literature include pH, NO2- availability, and NH4+ or NO3- content, all of which are known to influence biotic and abiotic HONO formation pathways (Wu et al., 2019).

Our HONO fluxes from the agricultural soil samples studied here are consistent with field observations under ambient conditions, where average emissions have been reported to largely remain below 7.2 µgNm-2h-1 (Tang et al., 2019; Xue et al., 2024). This does suggest that greater care in sample preparation, and likely also a widely agreed-upon standard procedure, is needed to study soil HONO emissions relevant to atmospheric models.

The integrated HONO fluxes for the pooled replicates yielded 185 and 146 µgNm-2, respectively. From the individual sample replicates, which were slightly wetter than the pooled, the integrated HONO fluxes were 690 and 739 µgNm-2, which was unexpected. The drier soils would have been expected to yield greater integrated HONO emissions (Oswald et al., 2013), yet this was not the case. Additional replicates and experimental controls, while beyond the scope of this study, would allow further attribution of the controls over the observed HONO variability.

These findings demonstrate the utility of the modified custom-built dynamic chambers for accurately capturing Nr fluxes under controlled laboratory settings, but they also highlight the need for more such systems to be implemented across the scientific community to better consider both biogeochemical soil properties and environmental context when interpreting the impacts of Nr fluxes obtained in the lab and scaling them to real soils. There seems to be potential for skewing the atmospheric impacts as a result, in particular for HONO, as the standard approaches have been designed to replicate NO fluxes (Behrendt et al., 2014). Most global models do not consider the effect of soil HONO on air quality through O3 production and oxidation chemistry. Several modelling studies, like Ha et al. (2023) and Tian et al. (2024), have incorporated the order of magnitude or higher HONO fluxes reported from lab studies, like those by Su et al. (2011), Wang et al. (2021), Oswald et al. (2013), and Meusel et al. (2018). They estimated significant HONO production with maximum flux potentials of 830, 95, 70, and 55 µgNm-2h-1, respectively. In contrast, the field observations that do exist suggest that real HONO fluxes are much smaller at 2–17.5 µgNm-2h-1 (Song et al., 2023; Tang et al., 2019). Similarly, Wu et al. (2022) have used the regional WRF-Chem model to explore the impact of soil HONO emissions on the concentrations of atmospheric HONO, OH, and O3.

Agricultural soil HONO emissions have been suggested to significantly contribute to OH radical production, accounting for approximately 10 % to 60 % of total OH formation in rural areas before noon (Oswald et al., 2013; Su et al., 2011), which often exceed the contributions from O3 photolysis. Additionally, high HONO emissions from agricultural soils have been reported to increase local O3 concentrations by ∼ 0.5–1.0 ppb in low-NOx rural environments where VOCs are not limiting (Zhang et al., 2021), with even greater impacts suggested during fertilization periods (Wu et al., 2022). Modelling studies, using GEOS-Chem and CMAQ, for example, claim that incorporating soil HONO emissions improves the agreement between observed and simulated O3 levels, particularly during the morning (Zhang et al., 2021).

Only a few studies have conducted estimates of soil NOx emissions under reasonable conditions, like Bao et al. (2022) and Wu et al. (2022), where the researchers tried to mimic field conditions. Even in such an area that has been long studied, large uncertainty still exists around soil sources of NOx, particularly for agricultural activities, where uncertainty is still at least ±30 % due to limitations in lab characterizations and field experiments (Gong et al., 2025). It is not surprising, then, that a similar issue exists for the very recent work on HONO soil emissions. The NOx uncertainty range results from intricate soil biogeochemical processes and varies with crop types, soil texture, fertilizer types and application rate (Gong et al., 2025). This longstanding and established difficulty in predicting soil NOx for use in global chemical models means that doing so for HONO without careful ground truthing of real-world emissions could lead to substantial inflation of the impacts on atmospheric chemistry and air quality. Care should be taken in using HONO emissions from lab studies in global models, as it seems they pose a risk of overestimating their atmospheric impacts until a more representative experimental design can be obtained with chamber systems like the one used here.

Agricultural soils amended with chemical fertilizers are expected to be hotspots for NH3, N2O, HONO, and NOx emissions. Here, we demonstrate the use of our developed chambers to measure these analytes under controlled lab conditions using our lightly processed pooled soil samples and four fertilizers: urea, ammonium nitrate (AN), ammonium bicarbonate (ABC), and ammonium carbonate (AC) with the temperature maintained at 23 °C, VWC between 25 % and 29 %, and headspace RH held at 65 % for three days, simulating realistic atmospheric and environmental Nr flux exchange conditions following farm field fertilization (Fig. 3). In all experiments conducted with the Picarro G2509, the mixing ratio of N2O began to rise approximately 4 h after the experiment started. In the example shown for urea, it peaked at 0.79 ppm after 12 h, which was followed by a gradual decline (Fig. 3a). This pattern likely reflects the incubation period of nitrifying and denitrifying bacteria that leads to the subsequent release of gases like HONO, as depicted in Wang et al. (2021), and N2O in Liu et al. (2022). In contrast, the mixing ratio of NO remained relatively constant throughout the three days (Fig. 3b), with NO2 increasing as the N2O emissions decreased (Fig. 3c). In this example experiment, no measurable emissions of HONO were detected despite the substantial presence of urea and evidence of active microbial nitrification and denitrification from the other emitted gases (Fig. 3d). Lastly, the urea application example in Fig. 3e shows the expected significant NH3 emissions, with the integrated amount reaching 22 % of the applied N over the 3 d incubation period. These findings are consistent with our existing knowledge that NH3 volatilization as an N loss mechanism dominates early Nr losses from fertilizers.

Volatilization of NH3 is well-characterized as a major pathway for N loss from fertilizers (Behera et al., 2013; Govoni Brondi et al., 2024; Liu et al., 2020; Moravek et al., 2019; Pan et al., 2016, 2022; Paulot et al., 2014). Besides agronomic concerns due to N loss and reduced fertilizer efficiency related to NH3 emissions from fertilized soil (Anas et al., 2020), it is a key precursor to secondary inorganic aerosols in the atmosphere with impacts on respiratory and ecosystem health, visibility, and climate (Dennis et al., 2010; Edwards et al., 2024; Fowler et al., 2013; González Ortiz et al., 2020; Jang et al., 2025; Seinfeld and Pandis, 2006).

Emissions of NH3 and N2O were observed to be far greater in terms of integrated amounts from the fertilized samples (Fig. 4). Integrated flux for NH3 produced by soil treated with ABC accounted for 77 % of total Nr flux (i.e., 85 900 µgNm-2), followed by AC, which was 63 700 µgNm-2. This is not surprising, since the use of chemical fertilizers increases the concentration of NH4+ in the soil that can deprotonate to emit neutral NH3 and, in the presence of ammonia-oxidizing microorganisms, increase the production of N2O (Luo et al., 2025). Nitrous oxide released from the addition of urea accounted for the highest integrated flux of 32 400 µgNm-2 observed, representing 89 % of total Nr released, followed by 25 300 and 9600 µgNm-2 for ABC and AC, respectively. One way that has been proposed to reduce these large N2O emissions from inorganic fertilizers is to change the application form to organic fertilizer, as the NH4+ in soil is produced more slowly (Luo et al., 2025). Due to a limited duration of access to the G2509 to conduct this work, we were unable to measure the N2O and NH3 emitted from unamended soils or those treated with AN. Regardless, based on the obtained data, the values found here are similar to those observed under real environmental conditions (Fig. 4). For our sample without N amendment, the integrated flux of NO was the largest (2400 µgNm-2; 87 % of total NOy), followed by comparable levels of NO2 (160 µgNm-2) and HONO (190 µgNm-2). The fluxes of NOx and HONO were below 1 µgNm-2h-1 for all the nutrient addition treatments, suggesting a similar trend as those observed under field conditions and from our lab results with unamended soils (Fig. 4, Table S1, Sect. S6). The exact mechanism behind the HONO release, being due to nitrification and/or denitrification, cannot be definitively assigned based on flux data alone, and many factors drive these emissions. There are discrepancies still observed between HONO flux measured from the treated soil samples in the laboratory and similar measurements in literature (Oswald et al., 2013; Su et al., 2011), some of which report fluxes of up to ∼ 3600 µgNm-2h-1 and are likely overestimating the soil Nr fluxes found in the real world. In contrast, our results are in close agreement with the field-based measurements of Tang et al. (2020), which also used a dynamic chamber flux method.

These in-lab experiments show that simultaneous speciated Nr emissions directed towards mass balance analysis in a controlled environment can be conducted using a single chamber and potentially applied to an array of chambers, as others have done for a subset of gaseous Nr (Scharko et al., 2015; Tang et al., 2019). With the limited replicates we explored here, our results raise a question for researchers who have been using lab studies as a standard to predict and incorporate HONO soil emission values, particularly for regional and global models.

A pilot scale field campaign was carried out to demonstrate the application of our dual soil flux chambers in capturing Nr gas exchange processes. A paired MC-RC setup was deployed in the same field a year after the soil samples were collected for our lab experiments. During a period of stimulated Nr emission from an in situ experimental application of urea, the mixing ratios of NO, N2O, and NH3 were impacted compared to the unfertilized state. The changes within both the MC and RC were measured and used to calculate fluxes (Fig. 5). The purpose of Fig. 5 is methodological to demonstrate how rate, dilution, and reaction terms combine in the observed rates of concentration change during chamber cycles. The selected data for NH3 correspond to measurements taken after the fertilization event, while the selected NO and N2O data segments are examples for separate observation times which best demonstrated the contributions of all terms before the fertilization event. Taken together, these three separate examples for the mathematical terms contributing to the net flux in Eqs. (6) and (7) can be considered more easily – they are entirely ascribed to the rate term otherwise. Each set of observations spans 3 consecutive hours of dynamic changes in gas concentrations within the chambers, which allow fluxes to be calculated. For the 0.5 Hz measurement rate of the Picarro, it is clear from the accumulation and depletion of target gases in Fig. 5 that a shorter observation period than 30 min could be used when high time resolution instrumentation across all target species is available. The benefit of this would be to reduce both the alteration of the composition of the chamber headspace and diverging physical conditions between the chamber and ambient environment, ultimately obtaining better flux estimates. However, for this pilot study, the 1 min time resolution of the NOx analyzer and the method for determining HONO by difference with an annular denuder every 5 min required a 30 min interval. A shorter closure period could also have the drawback of worse flux detection limits when fluxes are small, and more variability due to a less robust regression of the accumulation or depletion rate. For example, this would increase the value of λ for NH3 (Fig. S11; Sect. S7) and its relative error (4 % for a clean system, Sect. S7.3) as well as other surface-interacting gases.

The breakdown of the flux components for NH3, N2O, and NO can allow the contributions of the experimental setup (e.g. dilution) and environmental factors (e.g. reaction) to be considered independently (Fig. 6). We consider these across a consecutive triplicate of flux determinations from the pilot field deployment, to provide meaningful examples. The uncertainty in each term is estimated from the variance between the triplicate of consecutive MC and RC observations, through the measured slopes (i.e. using Eq. 6 or 7) and assumes that ambient atmospheric composition did not change substantially during this error derivation period.

The rate term (dC/dt) is the dominant contributor to the determined flux for all three gases (Fig. 6), which are all positive and indicate emissions despite their clearly differing temporal trends. Each contribution term is defined and discussed in further detail in Sect. S7. Accurate quantification of trace gas fluxes using dynamic chamber systems requires correction for attenuation caused by surface interactions. These effects can significantly suppress the measured accumulation rate of reactive species within the closed MC (Fig. 5). To account for this, the attenuation factor (λ) was introduced in this work (Eq. S14 in the Supplement), defined as the ratio of the theoretical concentration signal for an inert gas (e.g. as in our fill/empty experiments) to the measured signal over the chamber closure period. The value of λ is gas-specific and time-dependent, reflecting wall affinity and kinetics. This correction reduces flux bias, independent of chamber-specific losses. The highest attenuation was observed for NH3, for which a λ of 5.40 was determined empirically for the 30 min closure period. Comparative values for NO2 and HONO are also provided in Sect. S7.3.

For N2O, the first cycle exhibits the highest flux (1.07 ± 0.12 nmol m⁻² s⁻¹), while the second cycle showed uptake by the soil (-0.18 ± 0.12 nmol m⁻² s⁻¹), and the final cycle showed no net flux within error (0.08 ± 0.12 nmol m⁻² s⁻¹). This occurs for a non-reactive greenhouse gas like N2O despite the concentration decreasing with time for both the measurement and reference periods, as the rate of decrease during the measurement cycle is slower than that from dilution in the RC alone (Fig. 6). Across the three cycles, the rate of concentration change term contributed 84 ± 15 % to the measured flux in the first cycle, 109 ± 98 % in the second cycle, and 50 ± 171 % in the third cycle. The negative rate of concentration change in all of the observation periods arises due to the use of N2O-free purge gas delivered to the chamber headspace, which is required for the gas sample to be destructively quantified, while not introducing ambient air into the chamber (Fig. 5). This simultaneously prevents sudden changes in local outdoor air composition from making small fluxes difficult to detect, as well as reduces the uncertainty in the fluxes assigned. Here, the dilution terms range from 8 % to 50 % of the net flux, as would be expected from the N2O exchange switching from emission to deposition during the three-cycle example (i.e. 3 h). In the final measurement cycle, the rate of concentration change transitions from a loss to steady state during the observation period, suggesting that an instant of “hot spot, hot moment” emission of N2O was likely occurring. As such, the transition leads to no net flux for the observation period, which is a limitation of our dual-chamber approach compared to one with recirculating headspace, or that of an EC approach that can capture higher temporal resolution changes in concentration. The N2O fluxes observed here are consistent with those reported for European arable soils using both dynamic and static chambers, where event-driven N2O peaks after fertilization commonly range between 0.2 and 2.4 nmol m⁻² s⁻¹, and background periods can be as low as 0.08 nmol m⁻² s⁻¹ (Kong et al., 2025; Murphy et al., 2022; Maier et al., 2022; Manco et al., 2025). Field-scale EC studies, such as Maier et al. (2022), have captured post-harvest pulses up to 1.6 nmol m⁻² s⁻¹, highlighting the capacity of EC to resolve large, rapid emission events that single-point chamber systems may miss. However, dynamic chamber systems provide high-precision, temporally resolved flux data under controlled conditions, enabling direct attribution of emissions to specific soil management or environmental factors (Butterbach-Bahl et al., 2013; Kong et al., 2025). This makes dynamic chambers, like those demonstrated here, especially valuable for mechanistic and process-level studies.

For NO, the only gas we consider in the examples here with a reaction term, the reaction contribution is the smallest among the three flux components. Over all three cycles, the reaction term opposed the net flux by less than 13 %. The magnitude of the reaction term is always negligible, remaining within the ±0.01 nmol m⁻² s⁻¹ variability caused by background correction from the RC. Instead, the time rate of change in concentration term drove 74 %–81 % of the total flux for all three cycles and dilution accounted for the remaining 30 %–33 %. The reaction term falling within the uncertainty range of no contribution in all three cycles (e.g. -0.01 ± 0.01 nmol m⁻² s⁻¹ in the second cycle, where it had the greatest potential), indicates that its contribution is less certain than the other flux components as it is driven by the presence of O3, which is rapidly lost upon chamber closure. Overall, the low contribution of the reaction term suggests this process has a minor role in measured NO fluxes. In other, more polluted regions, such as the North China Plain and the Pearl River Delta, where summertime ambient O3 concentrations range from 60 to 275 ppbv, the exceedances above 200 ppbv during pollution episodes (Wang et al., 2017) could make this term very important. As noted above, the PFA film on the chamber surface will change its properties with respect to O3 transmission over time, highlighting the utility of the RC in tracking this because it is designed to accumulate or lose the same atmospheric compounds as the MC on all sampling surfaces.

The resulting emission fluxes of NO observed in these three cycles were 0.056–0.078 nmol m⁻² s⁻¹, which are well within the range reported for agricultural soils in North America and Europe, such as Taylor et al. (1999) who observed -0.07 to 4.2 nmol m⁻² s⁻¹ in Canadian fertilized cropland, and Pape et al. (2009) and Almand-Hunter et al. (2015) who reported values of 0.05–4.0 nmol m⁻² s⁻¹, using dynamic chambers in grass and cropland soils. Therefore, in the case of NO for this example, the variability in the fluxes is largely driven by real fluctuations in the rate term. This highlights the sensitivity of the total flux to changes in the rate of concentration change, and the precision of the method, as the uncertainty in the final fluxes here is on the order of 14 %. Compared to the dynamic chambers reported by these prior studies, the RC in our dual-chamber system offers a clear advantage by directly correcting for baseline fluctuations and environmental drift that can confound single-chamber approaches. This is especially important for reactive gases such as NO, which are susceptible to rapid loss to O3 and short-term background variability (Taylor et al., 1999). In comparison to EC or flux-gradient techniques, which integrate over larger areas but may underestimate true NO fluxes due to post-emission chemistry (Taylor et al., 1999; Plake et al., 2015b), our approach yields high-frequency, process-resolving data ideal for mechanistic and plot-scale studies.

For NH3, the rate of change term dominates the total flux, contributing 103 ± 10 %, 97 ± 16 %, and 101 ± 17 % of the flux in cycles one, two, and three, respectively, while the dilution term contributes inconsequentially at -3 ± 7 %, 3 ± 12 %, and -1 ± 12 % in the same three cycles, respectively. The resulting emission fluxes of NH3 observed across these cycles ranged from 0.43 ± 0.03 nmol m⁻² s⁻¹ in cycle one down to 0.25 ± 0.03 nmol m⁻² s⁻¹ in cycle three. The relatively small contribution of the dilution term is consistent with the expectation that the RC and MC exhibit similar dilution effects. The relative uncertainty in the final NH3 fluxes is 7 % for cycle one, and 12 % for cycles two and three. By comparison, relative uncertainties were 14 % for NO and ranged from 15 % to 171 % for N2O, reflecting the greater variability and lower precision associated with the smaller flux magnitudes for those gases.

Our observed NH3 fluxes (0.25–0.43 nmol m⁻² s⁻¹) are consistent with literature values for managed grass and croplands. For instance, Milford et al. (2009) reported bi-directional background fluxes from -3.8 to 2.5 nmol m⁻² s⁻¹ before cutting intensively managed grassland, with larger diurnal emissions up to 42 nmol m⁻² s⁻¹ after cutting and maxima up to 224 nmol m⁻² s⁻¹ following fertilizer application. Abdulwahab et al. (2025) observed highly variable fluxes in intensively grazed French grassland, ranging from -6.6 to 188 nmol m⁻² s⁻¹, with short-lived maxima above 300 nmol m⁻² s⁻¹ after slurry application, though most measurements were at much lower magnitudes. Notably, accurate quantification of NH3 fluxes in this system critically depends on the application of the chamber-specific attenuation factor λ (Sect. S7.3; Fig. S11). Without this correction, true NH3 fluxes would be underestimated by more than fivefold. While the RC correction plays a role in the flux calculations, its impact on NH3 is less pronounced than on N2O, where the correction significantly alters the net flux direction (Fig. 5; first cycle). Chamber-based approaches such as ours provide a key advantage for NH3 over micrometeorological methods like EC, which are especially prone to high-frequency attenuation and chemical interferences for reactive gases. As shown in Moravek et al. (2019), even state-of-the-art closed-path EC systems may recover less than half (as little as 46 %) of true NH3 fluxes due to instrument limitations and turbulence losses. Challenges were also evident for REA, where Xu et al. (2011) found that turbulence and surface effects complicated flux interpretation in cropland. Related methodological considerations have also been noted in other contexts. Schlossberg et al. (2017) highlighted how airflow and canopy structure can influence chamber NH3 fluxes in turf systems, underscoring the need for chamber methods that minimize such artifacts. In contrast, our dynamic chamber system, with a chamber-specific λ correction and RC, enables robust, bias-corrected quantification of both low and episodic NH3 fluxes, as well as clear partitioning of emission and dilution terms, even under highly variable field conditions.