We built gas-permeable soil probes from microporous tubes of sPTFE (Fig. 1a). sPTFE is hydrophobic and has uniform pore distribution, which improves gas diffusion (Dhanumalayan and Joshi, 2018). The material is structurally stable and non-reactive, properties that make this material a good candidate as long-term soil gas probes. We selected four probes with different pore sizes and dimensions (Table 1) to evaluate their equilibration properties. Probes were machined (White Industries, Inc., Petaluma, CA, USA) from solid sPTFE blocks (Berghof GmbH, Eningen, Germany). We constructed probe prototype assemblies to connect probes to inlet and outlet transport lines of 1/8′′ fluorinated ethylene propylene (FEP; Versilon™, Saint-Gobain, Malvern, PA, USA) using stainless-steel reducing unions (Swagelok, Solon, OH, USA). In some cases, probes were assembled from two pieces (Table 1) using perfluoroalkoxy (PFA) unions (Swagelok, Solon, OH, USA). After assembly, probe assembly leak tightness at the fittings was tested by submersion under water while flowing ultra zero air through the probe.

We used soil columns to evaluate probe performance under controlled soil gas in a non-reactive matrix (silica sand). Silica sand (Granusil 4095, high-purity industrial quartz; Covia Corporation, Emmett, ID, USA) was used as the non-reactive matrix, which is a low-alkaline-oxide matrix with a characterized particle size distribution (Table S1). We designed the column to allow a gas of controlled composition (control gas) to be advectively forced through the silica matrix from below (Fig. 1) to evaluate probe performance (System 1 tests at University of Arizona, UA, and System 2 tests at Aerodyne Research Inc., ARI; Sect. 2.3.1). We also used the columns to measure in situ gas dynamics in response to environmental manipulation (e.g., wetting, redox state) in a complex matrix (soil) (System 2 tests at ARI, Sect. 2.3.2).

The lower column section (Fig. 1b) supported drainage and buffered delivery of control gas, and the upper section contained the matrix (silica or soil), with a headspace layer for uniform column outflow. Together, the two column sections had a 20.3 cm inner diameter, 87.6 cm length (including base and cover), and 28 L volume. The probe was positioned centrally in the upper section to allow sufficient distance from column walls (10 cm) and the soil–gas interface (15.2 cm) to avoid edge effects (Fig. 1c). The upper and lower column sections were separated by a layer of perforated PVC (3.17 mm thickness, staggered 3.17 mm holes, 4.67 mm center to center, 40 % open area) and a type 304 stainless-steel wire cloth mesh (325 × 325 mesh (44 µm), 0.051 mm opening size) to allow the passage of control gas and drainage of water (sealed during sampling) while retaining matrix integrity in the upper section. Column sections were joined using schedule-80 PVC pipes, flanges, bolts, and rubber gasket seals, allowing columns to be modular and easy to disassemble, transport, and refill. Additionally, PTFE and polyetheretherketone (PEEK) bulkhead fittings (IDEX Health and Science LLC., Oak Harbor, WA, USA) and washers provided airtight and watertight connections for gas tubing. Soil sensors (e.g., moisture, temperature) flanked the soil probes (Fig. 1c).

The soil probe sampling system operated in a continuous-flow mode whereby carrier gas (ultra zero air, UZA; Airgas Inc.) flowed through the soil probe to equilibrate with soil gas (probe flow), the outflow was diluted online (dilution flow), and the combined flow (total flow) was sent to the gas analyzer for real-time measurement. The gas sampling system consisted of a controlled soil gas transfer system, sampling probes, and a measurement and data acquisition system that coordinated sampling in three gas columns (Fig. 2). Nearly identical sampling systems were built at UA (System 1) and Aerodyne (System 2) and differed in the specific TILDAS and gas control components deployed at each location (Table 2). To prevent bulk gas advection in the soil, it was critical to ensure that flows into and out of the probe were matched such that the sum of the probe and dilution flows were equal to the total flow at the instrument intake. This depended on precise flow control by digital mass flow controllers (MFCs; Alicat Scientific, Tucson, AZ, USA). Dilution flow (Fig. 2) was important to reduce risk of condensation, avoid exceeding the optimal detection range, and increase the gas analyzer cell response time. The control gas system allowed us to stipulate the specific mole fractions and relative isotope mixtures at the column inlet. Two streams of UZA controlled by MFCs (probe and dilution) were delivered in tandem through a stream selector 16×2 port valve (VICI – Valco Instruments Co. Inc., Houston, TX, USA) with the total flow directed to the analyzer (Fig. 2) by a separate multiport selector (VICI – Valco Instruments Co. Inc., Houston, TX, USA). The custom control gas composition added to soil columns was mixed from UZA and concentrated gas cylinders (e.g., 5 % CO2; Table 3). A bypass line was installed to independently verify the control gas composition entering the column, while the column outflow line was used to measure column headspace concentrations (Fig. 2). In System 1, we used a custom LabVIEW (National Instruments, Austin, TX, USA) program to execute scripts generated in MATLAB (2018; The MathWorks Inc., Natick, MA, USA) for the timing and control of MFC gas flow rates and VICI valve switching. The custom LabVIEW program was also used to query and log all the MFC parameters via USB multi-drop box (BB9 RS-232, Alicat Scientific, Tucson, AZ, USA) and all SDI-12 (serial digital interface at 1200 Bd) sensors with a SDI-12-to-RS-232 converter (Vegetronix, Inc., Riverton, UT, USA). In System 2, TDLWintel, the TILDAS measurement and data acquisition program, controlled the multi-valves on a schedule for continuous unattended operation.

To evaluate the probe and the column performance, we corrected observed concentrations (Cobs) using the ratio of the dilution and total flows to obtain true probe sample, column/headspace, and control gas concentrations (C). For example, for soil probe sample concentrations we used the ratio of the total flow (Ft, probe plus dilution flow) to the probe flow (Fp) as shown in Eq. (1): 1C=Cobs⋅Ft/Fp.

In System 1 we integrated two TILDAS trace gas analyzers (Aerodyne Research Inc., Billerica, MA, USA) with the soil probe system to evaluate the feasibility of coupling with the sintered PTFE probes and evaluate performance under controlled conditions. TILDAS-1 was a dual-laser instrument configured for measurement of water isotopes at 3765 cm-1 and 12C16O16O, 13C16O16O, 12C16O17O, and 12C16O18O at 2310 cm-1 with an 18 m absorption cell. TILDAS-2 was a compact “mini” single-laser instrument configured to quantify carbonyl sulfide (OCS), carbon monoxide (CO), water (H2O), and CO2 at 2050.4–2051.3 cm-1 with a 76 m absorption cell. The dual and mini TILDAS analyzers had a 500 and 300 cm3 sample cell volume, respectively. The TILDAS platforms draw air samples through an absorption cell at low pressure where laser light is transmitted in a multipass configuration for long effective absorption pathlengths. The laser is scanned at kilohertz rates over the rovibrational absorptions of the molecule(s) of interest. Transient light absorptions were fit to known Voigt profiles to determine molecular concentrations on the fly using Aerodyne's proprietary acquisition and analysis software, TDLWintel. For this experiment, we connected the two TILDAS analyzers at a controlled flow rate (500–250 sccm, MC-1SLPM-D, Alicat) in series, and cell pressure was dynamically controlled to 40 Torr (PCS-EXTSEN-D-ISC/5P, Alicat) between the two analyzer sample cells and vacuum pump (MPU 2134-N920-2.08, KNF Neuberger, Trenton, NJ, USA). The TILDAS optical tables were each purged with 100 sccm zero air.

In System 1, CO2 concentrations varied linearly with controlled dilutions of 10 % CO2 tanks (Fig. S1, dual CO2 cal), and absolute CO2 concentrations were calibrated with a linear curve. We calibrated the δ13C CO2 from the concentration-dependent relationship of δ13C CO2 vs. observed [CO2] (Fig. S2); specifically, we fit a Gaussian equation to the relationship between (δ13C CO2 observed – δ13C CO2 true ∼-39.2 ‰ vs. Vienna Pee Dee Belemnite (VPDB)) and CO2 concentration (accounting for standard deviation in δ13C-CO2 measurements). We applied this CO2-dependent correction to all reported δ13C-CO2 values.

System 2 integrated a second and nearly identical (Table 2) gas sampling system with a novel dual TILDAS analyzer for isotopomers of methane (CH4) and nitrous oxide (N2O) (Aerodyne Research Inc., Billerica, MA, USA) to test instrument modifications that help integrate soil gas sampling probes with laser spectrometry.

In this study, we identified and selected the best spectral region and laser technology for continuous high-precision measurements of isotopomers of CH4 (12CH4 and 13CH4), and N2O (14N14N16O (“446”), 14N15N16O (“456”), 15N14N16O (“546”), and 14N14N18O (“448”)). The regions near 2196 cm-1 (4.56 µm) and 1295 cm-1 (7.72 µm) provide interference-free measurements of N2O and CH4, respectively, and their rare isotopes. The 2196 cm-1 region is also capable of measuring CO2 at soil-relevant concentrations (parts-per-thousand levels). The CH4 and N2O TILDAS system was optimized with respect to optical alignment, laser operating parameters (i.e., scan length, laser current, and temperature settings), and fit parameters. Short-term (seconds) and long-term (minutes–hours) noise was determined by sampling from a compressed air cylinder as a constant gas source, followed by Allan–Werle variance analysis (Werle et al., 1993). We chose 30 Torr as the optimum cell pressure to minimize both noise and spectral crosstalk between isotopomer absorptions. To reduce sample volume we designed a new cell insert and a compact 76 m pathlength multipass sampling cell. The novel volume-reducing insert for the 76 m cell has interior walls that match the contour of the multipass pattern and was 3D-printed using PA2200 nylon. After printing, the interior and exterior surfaces of the insert were sealed with urushi lacquer – a stable, durable, inert lacquer (McSharry et al., 2007). The turnover time of the cell volume with insert was evaluated in continuous sampling mode.

The concentration dependence of isotope δ values derived from infrared isotopic measurements is an analytical challenge that is instrument dependent. To minimize the concentration dependence we followed two steps. (i) We used frequent spectral backgrounds to minimize offsets (i.e., immediately prior to each sample measurement). A sample spectrum is recorded with the instrument sample cell filled with UZA. This spectrum is used to normalize sample spectra, improving accuracy and sensitivity by accounting for changing instrument conditions and possible drift. (ii) We identified best-fitting parameters for each spectral region and application. During System 2 operation, we automated script schedules using an external command language (ECL) within TDLWintel that ran backgrounds, calibrations, and controlled valves.

Alcohols (e.g., methanol and ethanol) have weak features in the methane spectral window (1295 cm-1), at levels typically below that of the isotopic precision. We tested whether VOCs would cause infrared spectral interferences with TILDAS analysis by exposing the instrument to artificially elevated part-per-thousand levels of methanol, ethanol, and formaldehyde – three species that may be common in soil. We found potential for interference near the 13CH4 absorption at elevated alcohol levels but did not observe this interference in the spectra collected from probes in the soil tested.

System 2 calibration used online mass flow control to dilute concentrated N2O or CH4 calibration gases into UZA. We used pure samples of N2O from the Massachusetts Institute of Technology (MIT Ref I and Ref II). The isotopic ratios of N2O were determined by isotope ratio mass spectrometry (IRMS) and TILDAS measurements and externally verified by Sakae Toyoda at the Tokyo Institute of Technology (McClellan, 2018). For calibration of the soil matrix tests discussed below, we used MIT Ref II to make a surveillance standard of 1000 ppm N2O. After calibrating N2O isotopes against the reference gas, observed lab air N2O isotopic ratios were within 3 ‰ of the relatively stable isotopic ratios of ambient tropospheric N2O (Snider et al., 2015): bulk 15N value of 6.3 ‰–6.7 ‰, site preference of 18.7 ‰ (Mohn et al., 2014), and 18O value of 44.4 ‰ (Snider et al., 2015). For CH4 concentrations, a CH4 surveillance tank served as a stable isotopic source to identify changes in isotopic composition. Measured instrumental precisions with an averaging time of 2 min were 0.9 ‰ and 1.6 ‰ for N2O bulk 15N and the site preference, respectively, at 325 ppb N2O, and 0.2 ‰ for 13CH4.

In System 2 experiments, we integrated a PTR-TOF-MS instrument (Vocus; Aerodyne Research Inc., Billerica, MA, USA) (Krechmer et al., 2018) into the sampling system in parallel with the N2O–CH4 TILDAS to detect soil VOCs such as monoterpenes, isoprene, and pyruvic acid (Gonzalez-Meler et al., 2014; Guenther et al., 1995). The Vocus technology contains a corona discharge reagent-ion source and focusing ion molecule reactor (fIMR) that has low limits of detection (less than parts per trillion by volume) and a fast time response, acquiring the entire mass-to-charge spectrum on the order of microseconds. A TOF instrument also has high resolving power in the mass dimension, enabling separation of isobaric signals (occurring at the same nominal mass-to-charge ratio). The TOF instrument employed in this work consisted of a 1.2 m flight tube enabling a resolving power > 10 000 m / Δm. A sample flow of 100 sccm was injected continuously into the Vocus source, with no extra overblow or carrier flow in the inlet line.

Data were processed using the Tofware (Aerodyne and TOFWERK A.G.) software package in Igor Pro (WaveMetrics). For these experiments the PTR-TOF-MS instrument was not quantitatively calibrated for the signals reported below, as we were only interested in relative concentration responses to wetting. Thus, signals are reported in non-normalized counts per second (Hz).

Silica sand was used to limit trace gas production or consumption from the matrix for controlled evaluation of the probe. Three columns were filled with a dry silica matrix (Table S1) and closed hermetically. Gas concentrations and isotopic signatures of the inlet, soil probe, and column headspace samples were quantified while the gases flowed continuously through the column and dilution rates were varied (Table 3).

We evaluated the effect of probe sampling on the column (Experiment 1) by changing the probe flow rate with constant control gas concentration and dilution. With System 1 and a single column, we alternated measurement of CO2 concentration in headspace gas (1 h) and the probe (15 min) to determine the impact of probe sampling on soil column outflow concentrations. Next, we tested the flow conditions that support the probe delivering fully equilibrated and representative samples by varying flow and dilution at constant column concentrations (Experiment 2). We evaluated 42 combinations of set points for total flow (from 50 to 300 sccm, at 50 sccm intervals) and dilution (from 90 % to 9 %, at 15 % intervals). Each measurement cycle lasted 25 min (15 min probe, 10 min column headspace) using one probe in System 1 and System 2.

We scaled up the sampling systems to three probes to evaluate multiple probes (Experiment 3). We measured probe and headspace gas at a constant dilution (75 %) of a 2000 ppm CO2 control gas for a target observation concentration of 500 ppm and probe flow rates of 5, 10, 20, 30, 40, 50, and 100 sccm (System 1). System 2 was similarly evaluated with N2O and CH4 control gases in the silica matrix (Table 3).

We replaced the silica matrix with soil in the columns to understand (1) probe behavior and response when monitoring soil gases in a complex and dynamic soil matrix and (2) soil processes that drive dynamic changes in subsurface soil gases. We measured N2O and CH4 concentrations and isotopic signatures with the improved TILDAS instrument in System 2 (Fig. 2) in a series of experiments (Table 4). For soil experiments, headspace measurements can be used to track surface gas fluxes but do not represent control gas concentrations as in the silica experiments. We evaluated how measured soil gas concentrations changed in response to the following: probe sample flow rate (Experiment 4), environmental manipulation of the soil matrix (e.g., increased soil moisture with 5.1 cm of simulated rainfall) (Experiment 5), and forced changes to the soil redox state (e.g., forced N2 and UZA through the columns to shift from anoxic to oxic soil environments) (Experiment 6). In this last experiment, we integrated the Vocus PTR-TOF-MS instrument into the system to measure soil VOCs (Fig. 2).

We selected optimal spectra windows and laser technologies for detection of the isotopomers of both CH4 and N2O using fundamental rovibrational transitions (Fig. 3). We used Aerodyne-developed simulation programs that utilize the HITRAN database (Rothman et al., 2013) to perform spectral simulations to identify potential measurement regions. Based on these simulations, we obtained appropriate lasers and detectors for the selected spectral regions. Simulations assumed an N2O mixing ratio of 1 ppm (lower end of expected values; Rock et al., 2007) in a mixture with 1.3 % H2O, 1 % CO2, 220 ppb CO, and 1.9 ppm CH4, at 30 Torr in a 76.4 m pathlength sample cell. This resulted in the selection of a spectral region (Fig. 3a) where all four N2O isotopomers of interest, 14N14N16O (446), 14N15N16O (456), 15N14N16O (546), and 14N14N18O (448), have absorptions in close spectral proximity (< 1 cm-1) but without overlap of absorptions of each other or other trace gases such as CO2. The 2196 cm-1 region was used to monitor the N2O isotopologues and CO2 in the soil gas matrix using a quantum cascade laser (QCL) (Alpes Lasers, Switzerland). We selected a second QCL (Alpes Lasers) based on simulations of methane isotopes in the 1294 cm-1 region to monitor 12CH4 and 13CH4 isotopomers (Fig. 3b). This region also provided measurement of H2O content in the soil gas via a water spectral feature at ∼ 1294.0 cm-1.

TILDAS operational parameters were optimized to increase isotope ratio precision. For example, we monitored the slightly weaker doublet at 2196.2 cm-1 that had lower concentration dependence than the stronger absorber singlet at 2195.6 cm-1 that would produce nonlinear dependence at high mixing ratios. In addition, we modified fitting parameters to minimize the impact of baseline variability on measurement precision (fit shown in Fig. S3). These improvements in spectral fitting helped minimize the dependency of N2O and CH4 isotopic ratios on concentration. Specifically, we reduced the slope of δ vs. the mole fraction to 0.7 ‰ ppm-1 N2O (for N2O < 8 ppm) and 0.5 ‰ ppm-1 CH4 (for CH4 < 14 ppm). The online dilution approach was critical for avoiding N2O and CH4 concentrations in soil exceeding these linear ranges. We quantified the precision of the isotopic ratios (Table S2) using Allan–Werle plots (Werle et al., 1993) (Fig. S3).

We improved measurement response time by reducing the TILDAS sample cell volume while maintaining the spectroscopic pathlength. Unnecessary “dead” volume in the sample cell was eliminated through two approaches. First, we reduced the cell volume (port to port) by 20 % (610 to 485 cm3) by shortening the cell by 4.2 cm, eliminating dead volume behind the mirrors. Second, the insert reduced the cell volume by ∼ 50 % (485 to 245 cm3) by filling volume between the mirrors but in the region outside of the multipass laser path. Overall, these changes reduced cell volume from 610 cm3 (previous ARI 76 m astigmatic multipass absorption cell (AMAC)) to 245 cm3, which improved the cell response time by 40 %, here defined as the time to observe 75 % of a full transition in concentration (Fig. S4) (i.e., from 1.13 (0.005) to 0.76 (0.01) s; 30 Torr and 1 slpm). At the cell pressure of 30 Torr used here, this 245 cm3 absorption cell volume corresponds to 9.7 cm3 of sample gas at ambient pressure.

Soil probes sample subsurface gases by diffusion across the probe membrane into a UZA stream flowing through the probe. In our balanced mass flow approach, an equal proportion of UZA molecules diffuse out of the probe relative to soil gas diffusing in, which can affect (i.e., dilute) concentrations in the subsurface environment. To quantify the impact of probe sampling on soil column concentrations, we set control gas to 1000 ppm CO2 and varied the probe flow rate from 5 to 300 sccm and back, at a constant dilution (50 %). We evaluated the impact of a 15 min soil probe measurement on subsequent 1 h measurements of the column headspace. We found that column CO2 concentrations were depleted directly following probe sampling (from 0.6 % to 1.6 % depletion) and took > 1 h to fully stabilize. Column CO2 was most depleted after higher probe flow rates (Fig. 4) due to increased CO2-free UZA diffusion through the probe membrane. Low probe flow rates helped minimize these sampling artifacts on subsurface concentrations.

Compared to the controlled soil gas concentrations (Fig. 5), the probe-sampled concentrations were lower. When probe carrier gas is not flowing, the volume inside the probe is fully equilibrated with soil gas. This resulted in the observed initial “pulse” of high gas concentrations when a probe was first selected and measured. During sampling, probe gas concentrations drop to a steady-state value that represents a balance between the probe flow rate and the diffusion rate of soil gas molecules into the probe.

Gas samples obtained by probes at low probe flow rates were most representative of soil gas as the slower flow rates allow more complete diffusive equilibration. We evaluated the impact of combinations of different total flow rates (from 50 to 300 sccm at 50 sccm increments) with sample dilution ratios (from 0 % to 90 % dilution at 15 % increments) resulting in probe sampling flow rates of between 5 and 300 sccm. These tests were conducted in the silica matrix with controlled soil gas composition (1000 ppm CO2) (Experiment 2). We calculated the residence time of carrier gas in the soil probe by considering the internal volume of the probes (V is 2.6–4.6 mL) and the range of flow rates evaluated (F is 5–300 sccm). This indicates that the residence time (V/F) could range from < 1 s for high flow rates to 55 s for the lowest flow rates and larger volume (5 sccm in probes P5, P8, P10). We found that observed soil probe concentrations decreased with increases in the probe flow rate (Figs. 6, 7), with no systematic influence of the dilution ratio. For the probe tested (Table 4), flow rates below 24.5 sccm produced representative samples (within 90 % of true concentration). We did not observe any clear drawbacks to sampling CO2 at flow rates < 50 sccm (Fig. 7).

Probe flow rates affected gases unequally and based on their diffusivity. Probe recovery was lower for CO2 with lower diffusivity than CO (molecular diffusion coefficients in air at 20 ∘C (CO2 0.14, CO 0.18) (Bzowski et al., 1990; Massman, 1998) (Fig. 7). The fractional recovery of true soil gas concentrations by probe gas sampling (i.e., probe : column headspace ratios) was higher (0.65) for CO than CO2 (0.2) at high flow rates (300 sccm). Additionally, the recovery ratios at specific flow rates were more scattered at a higher flow rate for CO. Regardless of the diffusion coefficient, both CO2 and CO reached equilibrium at low probe flow rates, but CO was well equilibrated over a 4× wider range (5–100 sccm) than CO2 (5–25 sccm). Moreover, for molecular isotopologues (e.g., 12CO2 vs. 13CO2), at increasing probe flow rates, the sampled CO2 δ13C appears to be lighter than the headspace control by ca. -6 ‰ (Fig. 8) at the highest probe flow rates. That this fractionation was observed relative to the headspace measurements implies it is derived from the probe rather than the rest of the sampling system (tubing, multiport valves, MFCs). These concentration and isotopic fractionation results underscore the need to ensure that the probe flow rate is sufficiently low to ensure full diffusive exchange between zero air and soil gas before the gas sample exits the probe.

We upscaled the online diffusive probe sampling method in both System 1 and System 2 to automatically control multiple probes using flow rates (< 100 sccm) to measure soil gas concentrations and isotopic ratios. To fully constrain probe measurements in the silica matrix (Table 3), each probe was evaluated repeatedly over a full sampling cycle (∼ 25 m) to measure headspace–probe–headspace. In both systems, we could scale to sequential measurements of multiple probes with good sample recovery (e.g., minimal concentration loss, isotope fractionation). In particular, probe recovery of N2O isotopomers was within 3 ‰ of true headspace values, and equilibration of all trace gas species generally was near or above 85 % (Fig. 9). Multiprobe tests showed that the system has a high potential for scalable spatial resolutions and scalability.

We used the multiprobe system to determine whether probes with different properties would exhibit the same flow dependency and, in particular, the effect of the characteristic pore size of an sPTFE probe on concentration recovery. The flow rate dependence of the different probes was determined with CO2 in silica sand (Fig. 10). We found that the flow rate dependency for one pore size (P1) predicted the general behavior of others (P2–P3) across a 5–10 µm pore size range. Unexpectedly, we did not find a clear link between the pore size and the fractional recovery of true soil CO2 concentrations for any given flow rate. For example, we might expect that a pore size of 10 µm would permit greater diffusion and favor probe equilibration; instead, the 8 µm probe produced a more equilibrated sample than either the 5 µm or the 10 µm (Fig. 10).

In System 2, at low probe flow rates the concentration measured from the probe was similar to the concentration in the headspace in the silica matrix. Probe flow rates above 25 sccm decreased probe concentration for both the 10 and the 25 µm pore sizes (Fig. 11). Similarly to System 1 (Fig. 10), the fractional recovery did not increase with pore size, and we did not find that the 25 µm pore size transferred more gas into the carrier flow. In tests at a higher probe flow in the silica matrix, the fraction of CH4 recovered in the probe was higher than for N2O, consistent with System 1 results (Fig. 7) and the known molecular diffusion rates of N2O and CH4 through soil, 0.14 and 0.19 cm2 s-1, respectively (Wang et al., 2014). Thus CH4 diffuses into the probe and replenishes the area around the probe more quickly during sampling than N2O.

In System 2, even in soil where controlled soil gas conditions were lacking (i.e., cannot constrain with headspace measurement), we observed a decline in measured soil gas concentrations with flow rate, similarly to the silica matrix experiments (Table 3).

We used soil trace gas sampling and nitrogen isotopic mapping to identify real-time, in situ changes in N2O production pathways in response to soil wetting. Soil wetting induced a strong pulse in subsurface N2O concentrations, isotopic signatures, and site preference that was captured in detail with the N2O and CH4 TILDAS and real-time in situ soil gas probe sampling. We found that the isotopic ratios of all three N2O isotopomers (δ448, δ546, δ456), site preference, and N2O concentration responded to the wetting over the subsequent 36 h period. N2O rose from approximately 3 ppm to over 40 ppm, with a corresponding and slightly delayed response in isotopic signatures (Fig. 12). The dramatic increase in N2O required additional dilution at concentrations above the expected range of the TILDAS (> 20 ppm). The response of the two 15N-N2O isotopomers diverged enough to drive a shift in the site preference (SP) upward by approximately 4 ‰–6 ‰ before falling back down toward 2 ‰. After the peak, the decline in concentration and isotopic signatures was not explained by soil moisture, which was a relatively steady 25 %–30 % volumetric water content (VWC) throughout the period. N2O isotopes point to pathways such as hydroxylamine decomposition, chemodenitrification, nitrifier denitrification, or denitrifier denitrification. When mapped into a 3-dimensional isotope space (Fig. 12b) that is based upon previous observations of the SP, 15Nbulk, and 18O for a variety of different processes (Toyoda et al., 2017; Wei et al., 2019), the observed isotopic signature falls between chemodenitrification and bacterial denitrification. While the 15Nbulk, and 18O signals are dependent upon the substrate 15N and 18O compositions, the shift over the course of the rewetting measurement indicates a period of more denitrification (at higher SPs), then decreasing back to bacterial denitrification. Importantly, the observed range of SP values is well below the expected range for bacterial and archaeal nitrification (AOB, AOA), which are > 20 (off the scale in Fig. 12b).

In contrast to the dynamic response in N2O, soil CH4 concentrations remained low, leading to low signal-to-noise ratios in the detected 13C-CH4 isotopologue, and did not respond to wetting (data not shown). The dilution rate of the sample was increased by 1.9× at hour 18, resulting in a 1.9× reduction in N2O concentration measured by TILDAS (accounted for in Fig. 12). Despite the large change in concentration, the isotopic signatures barely changed, even after readjusting the dilution rate at hour 42, indicating that their concentration dependence had been well accounted for.

We measured a diverse suite of soil trace gases, including VOCs, to test whether we would observe consistent responses in real-time, in situ changes in multiple compounds to shifts in redox from anoxic to oxic conditions in soil. Shifting the soil redox environment from anoxic to oxic conditions induced a cascade of subsurface gas pulses in CO2, N2O, and VOCs that we measured by integrating TILDAS and Vocus analyzers with the real-time in situ soil gas probe sampling (Fig. 13). Before this experiment, the soil column was forced into anoxic conditions by advectively flushing with N2 through the control gas ports for 3.5 h; subsequently, conditions were driven oxic by flushing the system with UZA for a short time at time zero. Conversion to oxic conditions drove a pulse in N2O concentrations that was slow and considerably weaker (reaching 1.6 ppm after 72 h) than the wetting response (Experiment 5). The onset of oxic conditions brought a strong CO2 increase from 0.1 %–0.4 %, suggesting an increase in microbial respiration. Along with CO2 and N2O, we measured a cascade of responses in masses corresponding to different VOCs. As respiration and nitrogen processing increase, the larger VOCs exhibit either immediate loss (C9H18O, C11H20O, e.g., nonanal, methylborneol) or delayed loss (C10H16 (monoterpenes), C12H22O, e.g., geosmin) in the soil. In contrast, after 5 h, the sulfur-containing compounds methanethiol (CH4S) and dimethyl sulfide (C2H6SH) exhibited a surge in production. The approach captured different sensitivities and temporal responses to a shift in soil redox across a suite of soil gases that reflect different biochemical processes and their sensitivity to redox conditions.