Our chambers were located at eight plots on 17 m intervals along a topographic gradient in a conventionally managed corn–soybean (Zea mays–Glycine max) agricultural field in central Iowa, USA (41.98∘ N, 93.69∘ W). The transect spanned a distance of 120 m (Fig. 1a), 2.25 m elevation, and included very poorly to moderately poorly drained soils (Mollisols classified as Okoboji to Clarion series under the US Department of Agriculture taxonomy). Chambers were placed immediately adjacent to crop plants; due to frequent tillage and herbicide application, recruitment of other plants inside the chamber collars was uncommon, but any plants were removed from the chamber interiors as soon as they were observed. Roots from crop plants were not excluded and likely grew beneath chambers. The lower half of the transect often experienced flooding after large rain events (Logsdon and James, 2014), and chambers were occasionally completely inundated. The foreground of Fig. 1b shows one open and one closed chamber located in the lowest topographic position. The open chambers in the background are positioned along the topographic transect.

The chambers we utilized were constructed in-house, and various aspects were modified from previously published methods. The chamber lid was first described by Rayment and Jarvis (1997), and Riggs et al. (2009) pioneered a pneumatic piston and stainless-steel frame that opened and closed a chamber lid relative to a collar installed in the soil. Bowling et al. (2015) implemented a similar chamber design to measure CO2 and δ13C fluxes from a forest but did not include extensive details on chamber design, construction, or operation.

The dimensions of many of the materials used were commercially specified with imperial units but are reported here in metric equivalents for consistency. A table providing the instrument part names in the order that they are described, along with use, supplier, part number, and total cost, is supplied in the Supplement. Small, unspecified items (e.g., bolts) which do not require exact dimensions are not listed. Approximately 130–260 h of labor was required to construct nine chambers and assemble the associated control system. Figure 3 shows the chamber design. Here we define the chamber base as the rigid, rectangular polyethylene structure (Fig. 3a) and the chamber frame as the metal structure superior to the base which allows for movement of the chamber lid (Fig. 3). The chamber collar is defined as the length of polyvinyl chloride (PVC) pipe that forms the interface between the chamber lid and the soil. Chamber bases were constructed from 2.54cm×7.62cm high-density polyethylene (HDPE) plastic (Fig. 3a). Custom L brackets cut from 5.08 cm aluminum angle stock and bolted to the plastic base provided two horizontal platforms to attach female spherical rod ends that served as the pivot point for opening and closing the chamber (Fig. 3b). By routing vertical slots in the L brackets, we provided a means to adjust the lateral orientation of the pivot rod on each chamber after installation in the field (Fig. 3b). This was useful to ensure that the chamber lid sealed against the collar given the inherent variability of soil microtopography. A 0.64 cm diameter threaded rod between the rod ends provided an axle to attach the chamber frame (Fig. 3c). Most of the chamber frame was constructed from 0.95 cm diameter stainless-steel tubing; dimensions are listed in the caption and correspond to the numbered labels in Fig. 3. To drill holes in the stainless-steel tubing, we flattened the ends of each piece of tubing to a length of 1 cm in a bench vise and then drilled holes through the flattened portion to accommodate attachment bolts. The stainless-steel tubing was attached to the threaded rod described above or to aluminum angle brackets bolted to the chamber lid, noted by yellow or red circles respectively in Fig. 3. Two lengths of 1.27 cm diameter stainless-steel tubing surrounding a second 0.64 cm diameter threaded rod and inserted into a 5.08cm×2.54cm HDPE bar with a slot for a spherical rod end were attached to the end of a pneumatic cylinder rod piston (Clippard, UDR-17-6) (Fig. 3d). Extension of the piston moved the chamber lid open or closed, and the HDPE bar and stainless-steel tubing were used to prevent the threaded rod from flexing during movement of the chamber lid. The three spherical rod ends, two located on the pivot point and one at the end of the cylinder piston, served as rotational degrees of motion (Fig. 3, yellow circles). All other connection points were rigid (Fig. 3, red circles).

The chamber lid followed a previous design which was shown to minimize the pressure differential between the inside and outside of the chamber (<0.2 Pa at flow rates of 4.5 L min-1) (Moyes et al., 2010b; Rayment and Jarvis, 1997). The circular chamber lid (38 cm diameter) was cut from an HDPE panel (1.27 cm thick). A 2.54 cm diameter hole cut into the center of the lid allowed a vertical gas inlet tube (Fig. 2a) to be fixed to the lid via custom-machined threads and a nut on the bottom of the tube. The inlet tube (15 cm length) was machined from aluminum bar stock and had internal and external diameters of 2.54 and 3.81 cm, respectively, and a taper (2.54 cm length) at the superior end (Fig. 2a). The inlet tube was covered by a PVC cap (10.16 cm diameter and 16 cm length; Fig. 2b) attached to the lid surface with three bolts, each with 1 cm spacers to create an air gap between the cap and the lid surface (Fig. 2). The gap created by the spacers allowed air to flow to the inlet while preventing the direct horizontal flow of wind over the inlet tube opening. On the lower surface of the lid, a D-shaped rubber seal (EPDM foam, 2.54 cm width) was affixed with silicone caulk in a ring where the lid contacted the collar to create an airtight seal when pressure was applied to the piston that closed the chamber (Fig. 1b). Early in our study, we observed that high pressure (>550 kPa) was needed to ensure a tight seal between the collar and chamber lid. To minimize the piston air pressure required to seal the chamber lid against the collar, and thus conserve power, we bolted two nested, 26 cm sections of slotted steel construction strut to the top of the chamber lid to provide additional mass (Fig. 3). Gas from the inside of the chamber was sampled via a circular outlet manifold consisting of polyethylene tubing (6.4 mm o.d., 3.2 mm i.d.) perforated by drilling 2 mm diameter holes through the tubing at 2 cm intervals, and it was held in place approximately 3 cm below the lower surface of the lid with three stainless-steel eyebolts. All tubing connections in our chamber and instrument manifolds were made using 0.64 cm brass Swagelok compression fittings. A threaded bulkhead union and tee fitting were used to connect to the outlet manifold to external tubing above the chamber lid.

Chamber collars were made from PVC pipe segments (20 cm length, 30.48 cm i.d.) with the lower edge beveled with a belt sander to facilitate insertion into the soil. The beveled edge was pounded 10 cm into the soil for a total collar height of 10 cm and volume of approximately 7.3 L. The volume of air inside the longest length of tubing (120 m) connecting the chamber lid to the gas analyzers was <1.8 L. To hold the chamber base in place relative to the collar, we initially used a ratchet strap. However, we found that pressure exerted by the pneumatic arm when opening or closing the chamber occasionally shifted the position of the chamber base or collar and prevented a seal between the chamber lid, collar, and soil. This occasionally occurred following tillage or when soils were extremely dry, given that these soils contained swelling clays. To address this problem, we anchored the chamber base using two steel rebar rods (60 cm length, 1.27 cm diameter) pounded 45 cm into the ground on either side of the chamber base and affixed to the outside of the chamber base with U bolts positioned along the central axis of the collar (Fig. 3). We periodically checked that the chamber lids were effectively sealing against the collars. Application of this method to true Vertisols, with even greater shrink/swell behavior, could likely be achieved using similar use of rebar to anchor the chamber.

Chambers were opened and sealed by alternatively applying 550 kPa pressurized air to either side of the pneumatic cylinder described above via two lengths of tubing connecting each chamber and the instrument shed (Fig. 4a). We used 0.64 cm o.d., 0.43 cm i.d. low-density polyethylene (LDPE) plastic tubing. We initially used aluminum composite tubing (Synflex 1300), which has been commonly used in other field trace gas measurement studies (e.g., Bowling et al., 2015), but we found this to be impractical for our application given its vulnerability to kinking during chamber installation and removal through dense vegetation. Pressurized gas tubing was connected to the pneumatic cylinder via national pipe thread (NPT) to Swagelok connections (Fig. 4a). Needle valves (Clippard JFC-2A) located between the pressurized tubing and either side of the pneumatic piston were used to manually adjust the rate of chamber opening and closing to prevent damage to the frame. Pressurized gas was initially supplied by a pressurized cylinder and regulator as described in Riggs et al. (2009). However, we found that cylinders were impractical to supply the volume of gas necessary to pressurize the ∼100 m lengths of tubing between the cylinder and chambers with frequent opening and closing. To provide a less-labor-intensive source of pressurized air, we installed a Gast 12 VDC oil-less air compressor regulated by an air compressor switch (Condor MDR 3) with cut-in pressure set to 450 kPa and cut-out pressure set to 550 kPa (Fig. 4b). It was important to remove excess moisture from the pressurized air to maintain downstream metal components and valves. A 15 m coil of copper tubing immediately downstream of the compressor allowed the pressurized air to cool and water to condense. Excess moisture was removed by a water trap (Speedaire no. 4ZL49) connected to an additional 1 L reservoir made from PVC pipe and Swagelok fittings, which was periodically drained via a needle valve to the exterior of the instrument enclosure (Fig. 4c). From the water trap, the pressurized air flowed to a manifold of four-channel, two-way valves (Clippard MME-41PEEC-W012) which controlled the open/sealed position of each chamber by supplying pressure to either of two lengths of tubing extending to each chamber (Fig. 4a, d). Each valve was wired to one channel of a 12 V data-logger-controlled relay controller (Campbell Scientific SDM-CD16AC) such that pressurized air maintained the chamber in an open position when the relay was closed.

Figure 5 outlines the movement of sample gas between chamber and analyzers. Air was pulled through two separate 0.64 cm o.d., 0.43 cm i.d. LDPE plastic tubes. One tube sampled gas adjacent to the chamber inlet tube (Fig. 2), while the second pulled air from the perforated tubing manifold inside the chamber (Fig. 2). The second sampling tube is referred to here as the chamber outlet, as it served to pull ambient air from the chamber inlet tube through the chamber. Both sampling tubes were filtered through 1 µm Teflon (i.e., hydrophobic) filters (Pall Corporation) affixed to LDPE tubing via Swagelok connections immediately outside the chamber to prevent any particulates and liquid water from being pulled through the tubing (Fig. 5a).

Chamber sample selection was achieved by two sets of eight normally closed solenoid valves, one for inlet and one for outlet selection (Clippard, DV-2M-12-L, Fig. 5b). Solenoid valves were controlled by a second Campbell relay controller. Downstream of the chamber selection manifolds, both inlet and outlet gases flowed through additional 1 µm filters. Inlet and outlet flow rates were set independently by two mass flow controllers (Aalborg, GFCS-010201) upstream of two 12V diaphragm gas pumps (KNF Neuberger UNMP830; Fig. 5c). Both flow rates were set to 2 L min-1 by the mass flow controllers, and actual flow rates were recorded on the data logger (which was important for diagnosing potential problems during operation, as discussed later). To mediate selection of the gas sample that flowed to the analyzers, a third sample selection manifold with four normally closed solenoid valves selected among gas sources: chamber inlet, chamber outlet, high concentration standard, or low concentration standard (Fig. 5d). To operate 16 chambers without reducing measurement period or frequency, separate parallel selection manifolds, additional mass flow controllers for chamber inlet/outlet, and diaphragm pumps were added. Two additional solenoid valves on the sample selection manifold allowed selection between each of the two inlet and outlet manifolds. To maintain a constant flow rate through the inlet and outlet sampling tubes when the sample was not being routed to the analyzers, needle valves vented excess flow between the gas pumps and the selection manifold (Fig. 5). The selected sample gas flowed through a common sample gas mass flow controller set to 0.9 L min-1 (Fig. 5e). An internal pump in the N2O analyzer sampled gas at 0.8 L min-1, and this pump also served to pull sample through the CO2 analyzer which had no internal pump. The remaining 0.1 L min-1 was vented through a final needle valve placed upstream of the CO2 analyzer (Fig. 5e).

Two instruments in series were used to analyze CO2 and N2O, respectively (Fig. 5e). The CO2 analyzer was placed upstream of the N2O analyzer to avoid artifacts from the high oven temperature and Nafion drying column in the latter. We used a LI-COR 830 (or, subsequently, a LI-COR 850) analyzer to measure CO2 concentrations by infrared absorbance. Downstream, a Teledyne 320U gas filter correlation analyzer measured N2O concentration via infrared absorbance by frequently comparing the sample to a reference gas in a rotating filter (Fassbinder et al., 2013). Instantaneous gas concentrations, as well as the air temperature, inlet flow, outlet flow, and sample flow, were measured every 10 s and recorded on a data logger (Campbell CR3000).

Each chamber flux measurement was conducted over the course of a half-hour cycle. When 16 chambers were deployed, a new chamber was closed every 15 min and two chambers were closed simultaneously with the sample gases vented during a 15 min equilibration period prior to a 15 min measurement period. Here we describe the eight-chamber arrangement. To reduce possible conflation between measurement time and plot topographic position, we chose a consistent but staggered plot measurement sequence for each 4 h period (1, 5, 3, 7, 2, 6, 4, 8), where plot one was the lowest topographic position. When 16 chambers were deployed, the plot sequence was maintained so paired chambers at each plot were measured in a single half-hour cycle. At the beginning of each half-hour cycle when a new chamber was going to be measured, a chamber lid was closed by triggering a relay to apply pneumatic pressure to the piston, and the inlet and outlet sampling tubes of the respective chamber began to be sampled at 2 L min-1. Both inlet and outlet tubes were sampled continuously at a constant rate during the half-hour cycle while a downstream selection manifold alternated which gas was routed to the instruments, with residual flow vented to the instrument shed through needle valves (Fig. 5). All pneumatic and sample selection valves were controlled by the data logger. Calibration gases (standards) were measured every 2 h (Fig. 5d). If standards were measured during a given chamber measurement sequence, this was conducted at the beginning of the half-hour period: each standard was measured for 3 min by opening a valve on the gas selection manifold while chamber inlet and outlet flows were vented (Fig. 6a, b). During measurement periods where standards were not measured, the inlet sample was opened first on the selection manifold (Fig. 6c). After 11 min, the inlet was vented while the outlet sample was routed to the instruments until the 16th minute of the half hour (Fig. 6d). The first inlet and outlet gas concentration values from a given chamber measurement cycle (Figs. 6c and 7d respectively) were not used to calculate fluxes, as the chamber headspace concentrations of CO2 and N2O were often not at a steady state during this time. These values, however, were useful for troubleshooting and assessing temporal trends in chamber gas concentrations. Between 16–21 and 21–29 min, the inlet and outlet were respectively measured for a second time (Fig. 6e, f). The minimum 5 min measurement period for inlet and outlet samples was chosen to overcome a lagged response in the N2O analyzer following a switch in sample gas composition, which was as long as 2 min when there were large concentration differences between the inlet and outlet samples; the CO2 analyzer typically stabilized much faster (tens of seconds). The differences between the inlet and outlet gas concentrations averaged over the last 2 min of their second respective measurement periods (Fig. 6e, f) were used to calculate soil gas fluxes (units of µmolm-2s-1) using Eq. (1). The last 10 s of data from each period were excluded because of transient values during valve switching. 1Flux=P⋅FConcOut-ConcInR⋅T⋅A, where P is equal to mean atmospheric pressure at our study site (atm), F is the outflow rate (L s-1), ConcOut is the standard-corrected second outlet measurement period gas concentration (µmolmol-1) (Fig. 6f), ConcIn is the standard-corrected second inlet gas concentration (µmolmol-1) (Fig. 6e), R is the ideal gas constant (L atm K-1 mol-1), T is temperature (K), and A is the area covered by the chamber (m2). Following the end of the measurement period (29 min total), the chamber was opened by applying pneumatic pressure to the opposite end of the piston via the open/sealed manifold (Fig. 4a, d) and would remain open prior to the next measurement sequence.

Corrected gas concentration values were obtained by applying two-point linear standard corrections updated every 2 h (e.g., Fig. 6a, b). The instrument output during the last minute of each standard measurement, again excluding the last 10 s, was averaged for calibration. Corrected gas concentrations were obtained by regressing measured standard values against known values to obtain a linear slope and intercept used to correct raw values. Working standards were prepared by filling two 50 L gas cylinders with higher and lower concentrations of analytes by mixing CO2- and N2O-free air (zero air) with a concentrated standard gas to achieve values that approximately spanned the range of CO2 and N2O concentrations observed in the field. The mole fractions of each standard gas were verified in our laboratory by analyzing five replicates each on a gas chromatograph (Shimadzu 2014A) with thermal conductivity and electron capture detectors, which were calibrated according to additional NIST-traceable standards using a four-point curve. Gas cylinders filled to 140 MPa lasted approximately 9 months.

All data cleaning, flux calculation, and data analysis were conducted with R statistical software version 3.6.1 (R Core Team, 2019). Cleaning and calibration required R packages lubridate, nlme, and reshape (Grolemund and Wickham, 2011; Pinheiro et al., 2020; Wickham, 2007). The CR3000 data logger code we used to operate the chambers and record data, along with an example dataset and R script for data cleaning and flux calculations, is provided as archived files associated with this publication.

At our field site, six 265 W solar panels (Kyocera) with 16 deep cell marine batteries (Trojan J305E-AC 6V) were able to power the analysis system for much of the year. Figure 7 illustrates the solar charging and battery storage system. Two sets of three solar panels each were wired in series through parallel 15 A circuit breakers within a combiner box. The positive lead flows through a 30 A circuit breaker with a second combiner box before joining the negative at a charge controller (Morningstar TS-MPPT-60, Fig. 7). Indicator lights on the charge controller were used to assess the remaining battery charge, and we occasionally shut the entire system down during prolonged periods of low sunlight to avoid completely discharging the batteries. The charge controller positive output flowed through a 63 A circuit breaker (Fig. 7) to the final positive lead of a battery bank consisting of four sets of four serially wired batteries, each connected in parallel (Fig. 7). The negative output from the charge controller flowed to the negative lead at the opposite end of the battery bank. A 24 V DC output connected to a 60 A breaker (Fig. 7) and a DC/AC converter provided power for the 110 V AC N2O analyzer. A subset of two batteries provided 12 V DC power to the other components (data logger, CO2 analyzer, switches, valves, and additional sensors).

While often no maintenance was required, we typically checked the measurement system every several days to prevent data gaps if a failure occurred. Under ideal conditions (permanent chamber installation, ample sunlight, no flooding), the analysis system may be able to operate over periods of weeks to months without maintenance. However, we found that problems related to chamber submergence, component failure, or unintended faunal interactions occurred on occasion. This section highlights some common issues and practices that we found helpful for addressing them.

A key principle of steady-state chamber operation is that the gas concentration inside the chamber headspace is approximately at equilibrium (gas flux from the soil is balanced with gas removed via the chamber outlet) when the flux measurement is made. The time to achieve steady-state conditions is a balance between the soil flux rate and the flow of gas through the chamber. Here, to enable the use of smaller pumps and conserve power we employed lower flow rates (2 L min-1) than often employed previously in dynamic chambers (e.g., 4 L min-1; Bowling et al., 2015). Initial tests revealed that use of larger pumps needed to achieve 4 L min-1 flow rates over >100 m tubing runs was not sustainable from the perspective of power supply. To validate the steady-state assumption at 2 L min-1, we analyzed the slope of a linear regression between concentration of CO2 and N2O and time over the final outlet measurement period (Fig. 6f, approximately 27–29 min) using data from three separate periods chosen to cover a broad range of fluxes and spanning 2 weeks in total. We found an average increase of 0.18±10.51 ppm CO2 min-1 (mean and SD) and 0.57±8.40 ppb N2O min-1, respectively, indicating that both gases were approximately at a steady state at the end of the measurement period (relative to mean chamber outlet values of 684 ppm and 494 ppb for CO2 and N2O, respectively). We repeated this analysis for the final inlet measurement period (Fig. 6e, approximately 19–21 min) and found a change of less than 1 ppm or ppb min-1 CO2 and N2O relative to mean chamber inlet concentrations of 539 ppm and 331 ppb, respectively.

To assess temperature sensitivity of both gas analyzers under field conditions we examined the slope and intercept of standard curves measured during a 20 d period when air temperature ranged from -4 to 21 ∘C and during which the instruments ran continuously. There was no significant directional trend in air temperature over this period to avoid conflating temperature-related drift and drift of the instrument over time unrelated to temperature. All four metrics examined (slope and intercept of CO2 and N2O calibration curves) displayed correlations with temperature. However, the impact of temperature on the slope of the CO2 and N2O calibrations was less than 10-3 ppm ∘C-1 for both values. These values correspond to <1 % difference in instrument output between the highest and lowest observed temperature values at CO2 and N2O concentrations of 400 and 0.3 ppm, respectively. The intercept values showed greater sensitivity (0.02 and 0.003 ppm ∘C-1 for CO2 and N2O, respectively). These values correspond to a ∼0.5 ppm difference in CO2 and ∼0.08 ppm difference in N2O at the high and low temperature range observed. Taken together, we found that the N2O instrument had a -0.006 ppm ∘C-1 sensitivity, in close agreement to the -0.009 ppm ∘C-1 found by Fassbinder et al. (2013) for a similar instrument from the same manufacturer. As detailed above, standards were measured every 2 h to account for instrument sensitivity to environmental conditions. Additionally, because gas flux was calculated as the difference between and inlet and outlet concentration the intercept values canceled mathematically, thereby removing any additional bias due to temperature-related intercept drift between standard measurements. Therefore, temperature variation between measurements had negligible impact on the final flux calculation.

Optical trace gas measurements may be affected by a number of interacting factors including temperature, pressure, and water vapor pressure (McDermitt et al., 1993). Water vapor can be removed through chemical traps. However, the high gas flow in our system (2 L min-1) made reagent replacement in chemical traps impractical, and preliminary work showed that membrane-based driers did not always completely remove water vapor in our operating environment, where relative humidity often reached 100 %. The N2O analyzer we utilized removed moisture through a multitube Nafion dryer (Model NMP850KNDCB, KNF Neuberger Inc.). Water vapor was not removed prior to measuring CO2 concentration. As we calculated the soil CO2 flux as proportional to the concentration difference between the inlet and outlet gases, we were primarily concerned with a change in water vapor between the inlet and outlet measurement (Fig. 6e, f). In 2019, measurements were made with a LI-COR 850 that included a water vapor correction and measurement, which we used to constrain the potential impact of water vapor on our previous CO2 measurements. McDermitt et al. (1993) found that the required water vapor correction using a similar analysis was <10 ppm CO2 at water vapor pressure of 25.3 mmol mol-1 and CO2 concentration up to 1000 ppm. Water vapor pressure in the gases we measured spanned 1.0–53.6 mmol mol-1, with an average difference between the inlet and outlet gas of 1.8 mmol mol-1 and a maximum of 36.4 mmol mol-1. These small observed changes in water vapor between the inlet and outlet measurements indicate a minor impact on measured CO2 fluxes: if the water vapor difference between the inlet and outlet caused a <10 ppm bias in the measured CO2 concentration (as expected in >99.9 % of our observations), this would impact the average measured CO2 flux (3.47 µmolm-2s-1) by <5.2 % (0.18 µmolm-2s-1), which is within the typical range of measurement uncertainty for reproducing a known flux value under controlled conditions (Pumpanen et al., 2004). The correction under a more moderate water vapor difference between the inlet and outlet (<12.6 mmol mol-1) that spans >97 % of observed differences is approximately half the impact of this extreme example (0.09 µmolm-2s-1). Unrelated to its impacts on instrument performance, water vapor can also impact flux measurements by dilution (Harazono et al., 2015). Given an average water vapor difference between the inlet and outlet of 1.8 mmol mol-1 and maximum of 36.4 mmol mol-1, impacts of dilution on measured fluxes would also be small: typically <0.18 % and as much as 3.6 %.

To constrain the potential impacts of water vapor on measured N2O concentrations, we conducted a laboratory experiment comparing the N2O instrument output between a high and low moisture measurement on a three-point standard curve. Water vapor was measured with a LI-COR 850 installed in-line and upstream of the N2O sensor. To quantify the impact of water vapor on instrument output, we compared the standard curve created from dry standards to a curve created after bubbling the gas through a jar of deionized water. The bubbling technique added a mean of 25.4 mmol mol-1 of water vapor, spanning >99.9 % of observations of the difference between water vapor at the inlet and outlet in the field. Standard gases ranged up to 9.96 ppm N2O, greater than all differences between the inlet and outlet observed in the field. No difference was noted in N2O instrument output due to the presence of water vapor, which suggested the drying column was effective at removing water vapor or that the gas filter correlation method corrected for any impacts of residual vapor.

Manual trace gas sampling by field crews is generally accomplished during normal daytime work hours. In contrast, automated measurements can be scheduled throughout the 24 h diel period. Figure 8 displays boxplots of N2O emission from days when chambers were measured at each 4 h interval during 2017 and 2019 (the years of Zea mays cultivation). Though infrequent, we observed occasional instantaneous negative N2O flux values, as observed in other ecosystems (Schlesinger, 2013; Wu et al., 2013). Figure 9 shows the N2O and CO2 emissions from two typical 1-week periods from September 2017 and August 2018. A diel trend is visible for most chambers in August and some chambers and time periods in September. In general agreement with previously published automated chamber N2O studies from agricultural soils, we found the lowest rates of emission during early morning (04:00–08:00 LT) and highest emissions during early afternoon (12:00–16:00 LT) (Akiyama et al., 2000; Alves et al., 2012; Bai et al., 2019; Flessa et al., 2002; Savage et al., 2014). Early afternoon measurements were on average 28 % greater than the daily average from each chamber (4.04 vs. 3.15 nmol m-2 s-1), but this difference varied from -13.9 to 110 nmol m-2 s-1 among all chambers/days that were compared. The relative difference between average and peak daily emissions was in reasonable agreement with previous data from agricultural fields in the United Kingdom, Australia, and the United States (approximately 31, 47, and 33 % respectively; Alves et al., 2012; Bai et al., 2019; Savage et al., 2014). Although CO2 fluxes were highest and lowest during the same time periods as N2O, early afternoon CO2 fluxes were 22 % greater than the daily mean, on average (4.38 and 3.60 µmolm-2s-1), and this difference varied between -7.72 and 21.1 µmolm-2s-1 among all chambers/days that were compared.

N2O emissions pulses have often been observed following rain events (Savage et al., 2014; Sehy et al., 2003). To assess the length of the delay between rainfall and peak emissions, we analyzed the number of hours between heavy rainfall (defined as >2 cm total over 24 h) and subsequent peak N2O emission rate averaged over all chambers. A rain gauge located on-site recorded precipitation data that were obtained through the Iowa Flood Center (2017). There were 45 d with total rainfall >2 cm. To avoid conflating more than one rain event, we chose isolated events without rainfall >4 mm d-1 in the preceding or the following 2 d. Of the 15 isolated rain events observed, 4 were analyzed that did not span data gaps (Fig. 10). The rain-to-peak-emission delay varied from 12 to 26 h among precipitation events which varied from 2.4 to 4.4 cm.