Dynamics of nitrous oxide in groundwater at the aquatic–terrestrial interface

Introduction

The atmospheric concentration of nitrous oxide (N₂O) has increased since preindustrial times and continues to do so. The United Nations Framework Convention on Climate Change (UNFCCC) established the Kyoto protocol, which requires participating countries to either reduce or take responsibility for their excess greenhouse gas (GHG) emissions. This necessitates signatories of the Kyoto protocol to develop and publish national inventories of anthropogenic GHG emissions on an annual basis. Guidelines for constructing national inventories have been prescribed by the Intergovernmental Panel on Climate Change (IPCC).

In the IPCC methodology, agricultural sources of N₂O are partitioned into three categories: direct emissions from agricultural land, emissions from animal waste systems, and indirect emissions associated with nitrogen (N) that is removed in biomass, volatilized, leached, or exported from agricultural land (Mosier et al., 1998). These categories are thought to contribute an equal 1/3 share of the total estimated agricultural N₂O source but with 2/3 of the uncertainty in the total agricultural source due to the wide range of estimates for indirect emissions (Nevison, 2000). Globally, the total agricultural N₂O source equals 6.3 Tg N yr⁻¹ (Mosier et al., 1998) with 2.1 Tg N yr⁻¹ coming from indirect emissions. The predominate sources of the indirect emissions, over 75%, are associated with N leaching and runoff (Mosier et al., 1998; Nevison, 2000). N in leachate and runoff enters groundwater, riparian zones, wetlands, rivers, and oceans.

The IPCC methodology provides estimates of the amount of N leached (NLEACH) based on the amount of N input, assuming a certain fraction (FRACLEACH) of these inputs are lost to leaching and runoff (Mosier et al., 1998). The emissions of N₂O arising from leached N (i.e. N₂O(L) are calculated as follows:

The N₂O emissions resulting from NLEACH are assumed to evolve from: (1) groundwater and surface drainage (EF5-g), (2) rivers (EF5-r), and (3) coastal marine areas (EF5-e). Combined these factors represent the emission factor EF5, thus

The default value for EF5 is 0.025 kg N₂O-N kg⁻¹ NLEACH (Mosier et al., 1998), with component values of 0.015 for EF5-g, 0.0075 for EF5-r and 0.0025 for EF5-e. The value for EF5- is based on observations of supersaturated concentrations of N₂O in drainage waters due to leaching of N₂O from soil toward groundwaters, or production of N₂O in the groundwater via nitrification or denitrification, and the idea that this N₂O eventually degasses to the atmosphere. In contrast with direct emissions, there are very few data available to validate the IPCC emission factors for indirect emissions (Groffman et al., 2002). The IPCC model for estimating N₂O emissions from groundwater is highly uncertain because the controls on both N₂O production and consumption in groundwater are not well understood. Nevison (2000) noted that the default value for EF5-g was based on a literature review of only six studies and concluded that the EF5-g default factor should perhaps be reduced to a value of 0.001. More recent studies have also highlighted uncertainties with the magnitude of the default factors that comprise the EF5. (Reay et al., 2003; Clough et al., 2006). We are not aware of any published studies that have tried to determine the potential for reduction of N₂O in groundwater. It is clear that further studies are needed to examine the fate of N₂O in groundwater.

In situ push–pull methods have been used to study the fate of nitrate (NO₃⁻) and its potential denitrification rate in the subsoil–groundwater matrix (Istok et al., 1997; Addy et al., 2002), yet there has been no analysis of the fate of intermediate denitrification end-products such as N₂O. Addy et al. (2002) used a tracer gas (sulfur hexafluoride, SF₆) as a conservative tracer in their push–pull studies, suggesting that gases could be successfully traced with these methods. Few studies have directly measured N₂O consumption rates in soils (Kroeze et al., 1989; Hénault et al., 2001; Mei et al., 2004) let alone the vadose zone. The use of ¹⁵N labeled NO₃⁻ can be used to determine the rate of N₂O production. However, the determination of ¹⁵N₂O consumption rates is hampered by ongoing production of ¹⁵N₂O from the ¹⁵NO₃⁻ pool. Thus, only a net measure of N₂O is achieved i.e. the difference between production and consumption of N₂O. To accurately determine ¹⁵N₂O consumption rates there is a need to decouple the production of ¹⁵N₂O from the ¹⁵NO₃⁻ pool. Clough et al. (2005) recommended the use of ¹⁵N₂O as a starting substrate in order to better identify the fate of the N₂O molecule. The introduction of ¹⁵N₂O into the subsoil–groundwater matrix, along with a conservative tracer, could potentially provide data on the fate and production of N₂O belowground. Thus, the objectives of this study were to adapt the in situ push–pull method of Addy et al. (2002) to allow the introduction of dissolved ¹⁵N₂O into the subsoil–groundwater matrix and to assess the potential for ¹⁵N labeled N₂O consumption at two sites with differing denitrification potentials.

Materials and methods

Groundwater extraction, ¹⁵N₂O labeling, and dosing

Groundwater was extracted from the piezometers at sites A and B using a Masterflex L/S portable peristaltic pump (Cole Parmer, Vernon Hills, IL, USA). This water (10 L aliquots) was brought from the field in plastic carboys and stored overnight at 4 °C until it was labeled with SF₆. Labeling with SF₆ was achieved by purging the water with an SF₆ gas mixture (100 μL L⁻¹ SF₆ in He; Med-Tech, Medford, MA, USA) via a sparge stone for 15 min. This also served the purpose of lowering dissolved oxygen (DO) concentrations to ambient levels as noted by Addy et al. (2002). Immediately after this, 2 L aliquots of the SF₆-dosed water were decanted into 2 L volumetric flasks and sealed with a Suba-seal. A 5 mL volume of potassium bromide (KBr) solution was then injected into the dose water within each 2 L volumetric flask to achieve a Br⁻ tracer concentration of 33 mg L⁻¹.

Water was then labeled with ¹⁵N₂O that had been previously prepared in the laboratory by adding ¹⁵N labeled ammonium nitrate (NH₄¹⁵NO₃, min. 98 + atom%¹⁵N, Sigma Aldrich, Auckland, New Zealand, Category No. 366536) to unenriched NH₄NO₃ and distilled water to form a well-mixed uniform salt solution containing NH₄¹⁵NO₃ (34 atom%¹⁵N excess relative to N₂ in air). This solution was then dried at 80 °C to an NH₄¹⁵NO₃ salt. The NH₄¹⁵NO₃ was then weighed out and placed in glass tubes, sealed and baked at 300 °C according to Friedman & Bigeleisen (1950) generating N₂O. The mass of NH₄¹⁵NO₃ used was adjusted so that the internal pressure inside the glass tubes did not exceed 2 atm. One glass tube was prepared for each dosing event per peizometer. The resulting N₂O used in dosing the groundwater was enriched in ¹⁵N, 34.0 atom%¹⁵N excess relative to N₂ in air [standard error of the mean (SEM) equaled 0.3, n= 9].

Labeling of the water with ¹⁵N₂O was achieved by scoring the glass tube that contained the ¹⁵N₂O with a metal file and placing it inside a silicone tube (12.8 mm ID, 2.4 mm wall thickness, Cole Parmer, Vernon Hills, IL, USA: EW-06411-19) that had been stored in a freezer such that one end of the silicone tube was already plugged with a 3 cm length of previously frozen deionized water. The open end of the silicone tube was then capped with a silicone turn-over flange stopper (12.7 mm diameter; Saint-Gobain Verneret, Charny, France: 467013-50). Working rapidly, the Suba-seal was then removed from the 2 L flask, the sealed silicone tube was then bent so that the internal glass tube snapped at the scored mark, releasing ¹⁵N₂O inside the silicone tube. The silicone tube was then placed into the 2 L flask and the Suba-seal repositioned to seal the flask. Within minutes the ice plug inside the silicone tube melted and released the ¹⁵N₂O into the dose water. The flask was gently shaken by hand for 2–3 min and then left overnight (12 h) at 4 °C for gases to equilibrate.

The 2 L flasks containing the dose water were transported to the field sites in coolers. At the field sites, the water table depth was recorded prior to pumping 1 L of water from the well. This was discarded before a further 150 mL ambient predose water sample was taken and placed in a nalgene screw-top bottle. These predose samples were transported back to the laboratory in a cooler and stored at 4 °C before chemical analysis. DO in the groundwater was measured on site using a portable DO kit (La Motte, Maryland, BA, USA). The dose water was then readied to be pumped into the peizometers. This was achieved by quickly swapping the Suba-seal on the flask with a rubber stopper fitted with stainless steel (SS) (6 mm OD, 4 mm ID) and copper (3 mm OD, 2 mm ID) tubes. The SS tube was the outlet port for the dose water and was connected to C-Flex tubing (4.8 mm ID; Cole Parmer HV-06424-15) that was then routed through the peristaltic pump to the air tight SS-sampling apparatus (Addy et al., 2002) that was in turn connected to the Teflon tube of the peizometer head (Fig. 1). The copper tubing was attached to a 10 L Tedlar bag filled with He at atmospheric pressure (Fig. 1). Helium entered the flask as the dose water was pumped out, thus preventing a vacuum being created inside the flask, which could lead to rapid degassing of the gases of interest from the dose water and/or the possible implosion of the flask. As the dose water was pumped (8 L h⁻¹) into the well a subsample was taken of the dose water as follows. The first 20 mL of the dose water pumped was discarded to waste with the following 20 mL collected in a gas tight syringe fitted with a two-way stop-cock using the gas tight SS-sampling apparatus. Using this apparatus the pumped water flow could be directed either toward a luer-lok fitting attached to a gas syringe or to the well head. This water sample was then injected into a pre-evacuated (− 0.93 atm.) 160 mL serum bottle sealed with a rubber suba-seal. The serum bottle was inverted so that the water sample covered the septa. Then He gas from a Tedlar bag, was released into the inverted serum bottle via a hypodermic needle until atmospheric pressure was attained, as indicated by the cessation of bubbles coming from the hypodermic needle connected to the He supply. Water samples were then transported back to the laboratory in a cooler and kept at 4 °C until analysis, within 24 h.

After the dose water had been pumped into the soil-groundwater matrix it was left to incubate for a total time of 1.5 and 4.3 h at sites A and B, respectively. These incubation times were based on previously measured denitrification rates and tracer recoveries at these sites (Addy et al., 2002; Kellogg et al., 2005). Following the incubation period the dose water was slowly ‘pulled’ out of each peizometer, using the peristaltic pump (6 L h⁻¹). Samples of the extracted groundwater were taken and treated in an identical manner to the water samples previously described. Six samples were taken from each piezometer when the cumulative volume of water extracted reached 0.22, 0.46, 0.70, 0.94, 1.48, and 2.02 L. Site A was dosed and sampled once, on August 19, 2005, while site B was dosed and sampled on two occasions, August 22 and 29, 2005. These sampling dates at site B are subsequently referred to as site B test-1 and site B test-2, respectively.

Results

Ambient groundwater concentrations of NO₃⁻, NH₄⁺, and Br⁻ were low or undetectable at both sites (Table 1). The groundwater at site B contained more DOC (P < 0.01) and lower DO (P < 0.01) than site A. Dissolved SO₄²⁻ concentrations were higher at site A (P < 0.01) than site B, while Cl⁻ concentrations did not differ between sites (Table 1). Salinity at site A was four times higher than at site B (Table 1), a reflection of groundwater mixing with sea water that had infiltrated the aquifer in the transition zone, producing groundwater of intermediate salinity (Moore, 1999). Ambient N₂O concentrations in the groundwater averaged 4 μg L⁻¹ (Table 1) and were above saturation (0.6 μg L⁻¹), with ambient ¹⁵N₂O enrichments at site A and B of 0.203 and 0.295 atom%¹⁵N excess (relative to ambient N₂ in air), respectively. Ambient dissolved CH₄ concentrations were significantly higher (P < 0.01) at site B (Table 1).

When averaged across all nine push–pull tests that were performed the mean (SEM, n= 9) Br⁻ and SF₆ concentrations in the initial dose solutions were 33.2 (0.6) mg L⁻¹ and 1.5 (0.1) μg L⁻¹, respectively. Initial concentrations of dissolved N₂O in the dose water were well in excess of ambient concentrations and averaged (SEM, n= 3) 150 (44), 65 (12), and 70 (44) μg L⁻¹ for site A, site B test-1, and site B test-2, respectively. The mean ¹⁵N enrichment of the N₂O gas in the initial dose solutions was as noted above 34.0 atom%¹⁵N excess relative to N₂ in air.

Concentrations of CH₄ in the ambient water at sites A, site B test-1, and site B test-2 were 5 (1), 415 (120), and 333 (154) μg L⁻¹, respectively, (SEM, n= 3). At site B the CH₄ concentrations in replicate three were much lower than in replicates one and two, leading to high variability at this site. Immediately before ‘pushing’ the mean (SEM, n= 3) dose water contained CH₄ concentrations of 3 (1), 12 (1), and 10 (4) μg L⁻¹ at sites A, site B test-1, and site B test-2, respectively.

After incubation of the pushed dose water in the soil–groundwater matrix the dose-water was withdrawn. There was one particular well where equal volumes of gas were withdrawn along with the water sample during the pull phase of the test. This was at site B, replicate 3, during both tests.

The relationship between C/C_o, where C is the pulled groundwater concentration and C_o is the original pushed groundwater concentration, and V/V_t, where V is the cumulative volume pulled and V_t is the total volume of water pushed, was plotted for Br⁻ at site A (Fig. 2) and site B (Fig. 3a–c). Recovery of the conservative Br⁻ tracer, averaged (SEM, n= 3) over the three peizometers, was 64 (4)%, 67 (6)%, and 67 (3)% for the push–pull tests at site A, site B test-1 and site B test-2, respectively (SEM in brackets, n= 3). The concentration of Br⁻ in the recovered pull samples had decreased to an average (SEM) 10.7 (0.6) mg L⁻¹ in the final aliquot of the pull phase.

The conservative tracer SF₆ behaved in a similar manner to the Br⁻ anion (2, 3) but with lower average (SEM, n= 3) recoveries, 60 (1)%, 58 (11)%, and 65 (5)% for the push–pull tests at site A, site B test-1, and site B test-2, respectively. Of note however was the divergence between the two conservative tracers at site B, replicates 2 and 3, which was in contrast to site A and replicate 1 at site B. The recovery of N₂O, based upon C/C_o, was extremely variable and averaged (SEM, n= 3) 55 (3)%, 63 (36)%, and 57 (17)% for the push–pull tests at site A, site B test-1 and site B test-2, respectively. This variability was due to both intersite variability, between sites A and B, and intrasite variability at site B. At site A the mean N₂O C/C_o values closely tracked those of the mean C/C_o values for the conservative tracers (Fig. 2). At site B, replicate 1 behaved in a conservative manner, similar to site A with N₂O C/C_o values closely tracking the conservative tracers Br⁻ and SF₆ (Fig. 3a). However, in replicate 2 at site B the N₂O C/C_o values decreased rapidly with respect to the conservative tracers during both tests (Fig. 3b) while for replicate 3 the N₂O C/C_o values increased over and above those of the conservative tracers for both tests. N₂O concentrations in the final water sample withdrawn averaged (SEM, n= 3) 46 (13), 25 (15), and 32 (10) μg L⁻¹ for sites A, B test-1 and B test-2, respectively.

There was no significant decrease in the ¹⁵N enrichment of the N₂O during incubation for all replicates at site A and for replicates 1 and 3 at site B (Fig. 4). However, for replicate 2 at site B the ¹⁵N enrichment of the N₂O decreased exponentially in both dosings, as V/V_t increased, to a mean of 1.634 atom%¹⁵N excess relative to N₂ in air (Fig. 4).

Denitrification rates, i.e. the N₂O reduction rates, were only determined where N₂O concentrations decreased significantly, (i.e. for site B, replicate 2, and equated to 8 and 3 μg N₂O-¹⁵N kg⁻¹ soil day⁻¹ for tests 1 and 2, respectively).

Dissolved CH₄ concentrations in the pulled samples varied between sites (Fig. 5) with concentrations in the final aliquots sampled of 3 (1), 256 (37), and 199 (33) μg L⁻¹ for sites A, site B test-1, and site B test-2, respectively. Thus, there was no significant change in the C/C_o ratio for dissolved CH₄ at Site A. However, at site B the dissolved CH₄ concentrations increased significantly as the cumulative volume of water pulled increased (Fig. 5), although these concentrations were still below ambient levels recorded at the start of the experiment.

Discussion

There are several possible fates for N₂O injected and incubated in the subsoil–groundwater matrix in terms of both concentration and ¹⁵N enrichment (Table 2). These fates range from conservation of the added N₂O, [i.e. C/C_o behaves in a similar manner as the conservative tracers over time, with constant N₂O ¹⁵N enrichment (scenario A, Table 2), to a decrease in both the N₂O C/C_o ratio and its ¹⁵N enrichment (scenario E, Table 2)].

The data from sites A and site B, replicate 1, show that the N₂O concentration behaved in a conservative manner with no significant decrease in the ¹⁵N enrichment of the N₂O (Fig. 4), (i.e. scenario A in Table 2). We know that the ambient groundwater contained some antecedent N₂O but the N₂O concentration of this groundwater was insignificant when compared with the added ¹⁵N labeled N₂O. Therefore, hydrodynamic dispersion or advective groundwater flow that occurred, either during the actual dosing event or during the incubation period, did not significantly affect the N₂O ¹⁵N enrichment. Had we used a lower concentration of N₂O or ¹⁵N enrichment in our original dose water then we may have observed a significant decrease in ¹⁵N enrichment.

For site B, replicate 2, we observed a different result with both the N₂O concentration and ¹⁵N enrichment decreasing more rapidly than the conservative tracers. This leads us to consider scenario E in Table 2 where a decrease in ¹⁵N enrichment must be due to either an advective influx or in situ production of N₂O. While a higher advective flux is possible at this replicate, a comparison of the recovery of the conservative tracers at site B replicates 1 and 2 suggests that groundwater conditions were similar between these two replicates and that in situ denitrification was responsible for both the depletion of the ¹⁵N₂O enrichment as well as the overall decline in the N₂O concentration. Site B, replicate 2, appears to be a relative ‘hot-spot’ for denitrification. We assume N₂O production did not occur via nitrification due to the low DO concentrations and the lack of any measured NH₄⁺ substrate. This assumption could easily be verified by using ¹⁵NH₄⁺ as a substrate in future studies.

At site, B replicate 3, we observed marked increases in N₂O concentration, (i.e. C/C_o, increases, but the ¹⁵N enrichment of the N₂O remained constant). The increase in N₂O concentration would suggest that active net production of N₂O was occurring, but the constant ¹⁵N enrichment is difficult to explain. It is possible that production of low enrichment N₂O was balanced by isotopic discrimination in N₂O reduction, resulting in no net change in the ¹⁵N enrichment of N₂O in the incubation. A further compounding factor at this peizometer was the occurrence of significant gas bubbles, during the extraction of the dose water after its incubation. This gas had a negligible N₂O content but a considerable CH₄ concentration. In theory it is possible that the gas bubble presence caused an error in the calculations of V, but had this been the case the results for the tracer (e.g. Br⁻ would have been erroneous and this was not the case). We do not believe that this particular result draws the method into question since another eight piezometers were successfully sampled providing data that was interpreted in a sensible and logical fashion. However, further field work is required to fully understand the processes at site B, replicate 3.

Our CH₄ data suggest that conditions are sufficiently anaerobic at site B to support production of this important GHG. At site A, ambient groundwater concentrations of CH₄ were relatively low and there was no change over the course of the incubation. However, at site B, the process of SF₆ labeling stripped out the high ambient CH₄ concentrations in the dose water so that when the dose water was injected at the start of the incubation there was a large differential between the CH₄ concentration in the dose water and the ambient groundwater. The increase in dissolved CH₄ concentration in the pulled water samples (Fig. 5) was likely driven by diffusion of CH₄ from the surrounding groundwater matrix. The differences in CH₄ dynamics between sites A and B are consistent with the differences in ambient DO levels between the sites and support the results showing that some peizometers at site B are located in denitrification hotspots (Yu & Patrick, 2004).

A previous study in the fringe area of site A recorded a denitrification rate of 2 μg NO₃⁻N kg⁻¹ soil day⁻¹ following the addition of NO₃⁻ (Addy et al., 2002), while a previous study at site B has measured much higher but more variable rates of NO₃⁻ denitrification (61–40 μg NO₃⁻N kg⁻¹ soil day⁻¹ at 65 cm depth, (Kellogg et al., 2005). However, these previous studies did not present information on the relative production of N₂O and N₂ production during denitrification. While the relative N₂O reduction rates in the present study are consistent with these previous studies (site B higher than site A), the rate of N₂O reduction that we measured at site B is much lower than the total denitrification rates measured in the previous study. The magnitude of the N₂O reduction rates that we measured are also considerably lower than rates measured in surface soils. Hénault et al. (2001) reported a lag phase of 48 h in a gley soil before N₂O was reduced, with measured N₂O reduction rates from a variety of soils that were > 3360 μg N kg⁻¹ soil day⁻¹ while Blackmer & Bremner (1976) measured N₂O consumption rates > 570 μg N kg⁻¹ soil day⁻¹. A possible reason for the low N₂O reduction rates that we observed could be the time required for the denitrifying community to generate N₂O reductase. Although there were ambient levels of N₂O present in our sites, the higher concentrations injected with the dose water may not have been able to be immediately processed by the denitrifier community. Other studies have shown the denitrification enzymes and communities to be highly responsive to factors such as the temperature and water regime and carbon availability (Chèneby et al., 1998; Hénault et al., 2001) and in some instances denitrification may reduce NO₃⁻ in preference to N₂O. Previous studies have observed significant lag times between addition of NO₃⁻ and denitrification activity (Aelion & Shaw, 2000; Addy et al., 2002). A similar lag may also occur for N₂O. If so, it may be necessary to expose the peizometers to elevated N₂O to condition the microbes to record true potential N₂O reduction rates.

The divergence of the conservative tracers at site B, replicates 2 and 3, could possibly have been due to the relative physical states of the tracers and their respective interactions with the soil–groundwater matrix. It is possible that the peizometers at replicates 2 and 3 were in a soil matrix that was less dense or denser than the other peizometers. Thus, the resulting physical turbulence or mixing of the dose water with the groundwater may have resulted in the gas tracer behaving differently to the anion tracer as a result of varying pressure during injection.

Further modifications of this method are possible to facilitate measurement of N₂O dynamics at multiple sites. While we used highly enriched ¹⁵N₂O, the use of N₂O that is closer to levels of natural abundance ¹⁵N enrichment could be used if there was a sufficient difference between the ¹⁵N enrichment of any dissolved ambient N₂O and the N₂O supplied in the dose water. When ambient N₂O concentrations are low, it may even be possible to use commercially manufactured N₂O, with a sufficiently different ¹⁵N signature from that of the ambient dissolved N₂O, so that the dose water could be simultaneously labeled with SF₆ and N₂O by bubbling a tank-gas mixture of these gases through the dose water for a suitable period. While we have used a 2 L dose water volume there is the potential for a greater volume of N₂O labeled dose water to be used (e.g. 10 L as used by Addy et al., 1999). This would allow the integration of N₂O dynamics to occur over a greater volume of soil and reduce the potential impact of ambient groundwater diffusing into the dose plume. This could be achieved by replacing the flask holding the dose water with a gas impermeable bag. Then there would also be no requirement for the He gas, as the gas impermeable bag would deflate as the dose water was injected into the groundwater. A further modification could be the inclusion of a pressure gauge to note the pressure of the dose water as it is injected into the groundwater. This could indicate the relative densities of the soil matrix. Modified methods that allow for collection of data at multiple sites could allow for information on N₂O dynamics that could be scaled to address questions about the importance of these dynamics to IPCC inventories.