Nitrous oxide generation, denitrification, and nitrate removal in a seepage wetland intercepting surface and subsurface flows from a grazed dairy catchment

Additional keywords: bromide, denitrification, ¹⁵N, NO₃⁻ removal, N₂O, N₂, wetland, SF₆.

Nitrogen (N) losses from applied chemical fertilisers, dairy effluent irrigation, and grazing animal excreta potentially cause eutrophication in receiving waters (Phipps and Crumpton 1994; Naiman et al. 1995; Carpenter et al. 1998). The export of N via surface and subsurface runoff to water bodies can either be minimised by best farm management practices: applying N fertiliser with N inhibitors (Zaman et al. 2008a), and using riparian zones along river and stream banks (Cooper 1990; Ambus and Lowrance 1991; Haycock and Burt 1993; Carpenter et al. 1998; Burt et al. 1999), seepage wetlands (Blackwell et al. 1999), or permanently wet swales located at gullies of agricultural hillslopes (Burns and Nguyen 2002; Rutherford and Nguyen 2004).

Nitrate moving through riparian and seepage areas is subject to denitrification, plant uptake, dissimilatory reduction of NO₃⁻ to ammonium (NH₄⁺) (DNRA), and microbial immobilisation (Bowman et al. 1989a, 1989b; Hefting et al. 2003; Matheson et al. 2003). Denitrification is considered to be the major NO₃⁻ removal process, in which NO₃⁻ is enzymatically reduced to N₂O and N₂ (Hoffmann et al. 2000). Nitrous oxide is a long-lasting greenhouse and potential ozone (O₃) depleting gas, whose generation rates from highly organic enriched wetlands/seepage zones are likely to be substantial, particularly if they receive high NO₃⁻ pulses in seepage and runoff from adjacent, intensively grazed dairy farming systems (Fig. 1).

Nitrous oxide has been found to be an important end product of denitrification in riparian soils receiving high NO₃⁻ loadings (Hefting et al. 2003). Nitrous oxide in wetlands is produced not only by denitrification but also by nitrification and DNRA (Stevens and Laughlin 1998; Hefting et al. 2003; Smith et al. 2003). Nitrous oxide produced by various processes might form 1 pool before being reduced to N₂ by nitrous oxide reductase, the enzyme involved in reduction of N₂O to N₂ (Stevens and Laughlin 1998), which suggests that wetlands can act as a source or a sink of N₂O (Blicher-Mathiesen and Hoffmann 1999; Well et al. 2001). A wide range of N₂O : N₂ ratios has therefore been reported for wetlands, depending on factors (e.g. soil pH, carbon, anaerobicity, NO₃⁻ concentrations) that govern the nitrification–denitrification–DNRA processes (Groffman et al. 2000, 2002; Smith et al. 2003). Thus, the use of wetlands in decreasing diffuse N pollution of streams and rivers may shift the potential environmental problem from water pollution to greenhouse gas emission (Well et al. 2001; Hefting et al. 2003).

Several techniques have been used to investigate NO₃⁻ removal in wetlands. Many studies have focused on quantifying NO₃⁻ removal by measuring the rates of processes such as denitrification in laboratory incubation and microcosm studies (Seitzinger 1994; Groffman and Hanson 1997; Zaman et al. 2008b). These studies suffer from the effects of soil disturbance on oxygen (O₂) status and hence N transformation processes, and from the difficulty of obtaining sediments from depths below the water table for microcosms (Jacinthe et al. 1998; Addy et al. 2002).

In-situ studies conducted by Burns and Nguyen (2002) have compared NO₃⁻ movement in wetland groundwater to that of a more conservatively transported ion such as bromide (Br⁻). Nitrate removal was then estimated from the changes in NO₃⁻ : Br⁻ ratios and NO₃⁻ and Br⁻ mass with time, assuming that Br⁻ is not taken up by plants or processed by soil microorganisms (Schnabel et al. 1996; Whitmer et al. 2000). Addy et al. (2002) successfully demonstrated the use of the push–pull method to determine in-situ subsurface water denitrification and NO₃⁻ removal rates in riparian zones. In that study, collected wetland subsurface water was spiked with Br⁻, sulfur hexafluoride (SF₆), and ¹⁵N-enriched NO₃⁻ (potassium nitrate). The amended subsurface water was pushed (i.e. injected) into a mini-piezometer (diameter 18 mm with screen length 20 mm) and then pulled (i.e. extracted) from the same piezometer after an incubation period ranging from 6 to 48 h. Sulfur hexafluoride, an inert and slightly water-soluble gas with an extremely low atmospheric background concentration of 3 parts per trillion by volume, was found to behave as conservatively as Br⁻, indicating negligible degassing, thus allowing the authors to confidently use the sum of ¹⁵N dissolved gas (¹⁵N₂O and ¹⁵N₂) generation and changes in NO₃⁻ concentration (NO₃⁻ consumption) as estimates of the denitrification rate and NO₃⁻ removal.

The objectives of this study were to (i) quantify in-situ NO₃⁻ removal, N₂O and N₂ emissions, and denitrification rates in a wetland swale in a dairy landscape using the push–pull technique; (ii) compare in situ denitrification rates as estimated by either the ¹⁵N tracer technique or the decrease in added NO₃⁻ concentrations with time; and (iii) provide information on N₂O : N₂ ratios in emitted gas from wetlands under non-limiting and limiting NO₃⁻ conditions.

The studied seepage wetland (6817 m²) was located at a foot-slope of 3 pasture paddocks of a grazed dairy farm (3 cows/ha) in a dairy catchment at Kiwitahi, about 32 km from Hamilton (37°44′S, 175°35′E), New Zealand. The climate is humid–temperate with mean annual temperature of 15°C and annual rainfall of 1150 mm.

The wetland comprises a broad, low gradient (<1° slope) area that receives water from a spring (approximately 1 m from the wetland inlet), natural seepage, and discharges from shallow channels that intercept surface runoff and overland flow from the adjacent pasture paddocks. Grazing animals have been excluded from the wetland since 1999. An artificial swale that is now filled with sediment, organic floc, and wetland plants runs through the wetland and carries most of the flow. The entire wetland complex developed due to a flow constriction at the end of the permanently wet swale which creates partially flooded conditions within the area surrounding the swale (with a slope of <1° across the wetland).

There are 2 major soil types in the catchment, the Topehaehae silt loam and the Kiwitahi silt loam. The Topehaehae silt loam, which is derived from volcanic ash alluvium, is a gley recent soil (Aeric Haplaquent; USDA Soil Taxonomy) with silt loam topsoil and blocky clay loam at 0.3–0.75 m depth. It is a poorly drained soil with very slow subsoil permeability (<0.5 cm/h; Wilson 1980). The Kiwitahi silt loam is a yellow-brown loam (Typic Andept) with a brown silt loam soil texture and moderately permeable subsoil. Its parent material is volcanic ash over late Pleistocene terrace deposits. The soil is well drained and friable with well-developed fine crumb structure. Our study focused on the wetland swale. The soil within the wetland swale is saturated to the surface and is composed of very loose organic material approximately 0.2 m deep. The top 0.1 m depth comprises mainly a thick root mat of wetland plants plus unconsolidated organic mucks. It is followed by a 0.1 m layer of unconsolidated organic flocs and decayed plant materials. Beyond the 0.2 m depth, sediment is more condensed with a mixture of organic matter, silt, and clay, probably originating from the eroded soil materials that have been washed in from the adjacent pasture paddocks plus organic matter decay from wetland vegetation. At a depth of 0.5 m, there is a transition to bluish grey silty clay that increases in its firmness and density with depth. A dense silty clay layer of low permeability particularly at 0.7–0.9 m depth acts as an aquiclude, restricting water movement to a deeper subsurface water. Beyond 0.9 m depth, sediments consisted of a sand and clay mixture. Using tracer tests in a similar seepage wetland, Rutherford and Nguyen (2004) found rapid dilution and pore water velocities of approximately 0.5 m/day within the upper layers of soil. Wetland vegetation consists mainly of soft brome (Bromus hordaceus L.) with some floating glaucous sweet grasses (Glyceria declinata Breb.) and soft rush (Juncus effuses L.) and wiwi (Juncus edgariae L.) in areas around the wetland channel and the remaining wetland area. The herbage in pasture paddocks was a mixture of ryegrass (Lolium perenne L.) and white clover (Trifolium repens L.).

Push–pull studies are poorly suited to wetlands with high advection and dispersion that can carry the introduced plume away from the dosing piezometer. Because seepage wetlands have high advective and dispersive flux, we installed 4 large lysimeters made up of polyvinyl chloride (PVC) (0.5 m inner diameter and 1.2 m length with a sharpened bevel at the bottom end) to create a confined control volume for push–pull experiments in January 2003. This lysimeter set-up allowed us to investigate denitrification and to follow changes in N₂O and NO₃⁻ with time. To facilitate the installation of each lysimeter into the wetland, a serrated knife and 0.6-m machete were used to cut through the wetland plant roots and sediments around the lysimeter perimeter and to remove above ground wetland herbage within each lysimeter to 0.05 m height above the wetland surface. The lysimeter was then gently pushed through the wetland media to the underlying silt-clay layer and then slowly pounded into the aquiclude layer. Each lysimeter extended 0.25 m above the wetland surface and at least 0.25 m into the consolidated aquiclude layer of silty clay. Thus, in each lysimeter the depth of wetland sediment above the aquiclude material was approximately 0.7 m.

In the centre of each of the 4 lysimeters, a PVC piezometer (PVC pipe 0.03 m internal diameter and 1.25 m long screened with 0.25-mm slots at 6-mm intervals over the 0.15-m length between 0.20 and 0.35 m depth below the surface of the wetland subsurface water) was installed for the purpose of dosing tracers and water sampling. An impermeable barrier (PVC sleeve) was created within the piezometer at 0.35 m from the wetland ground surface. The estimated well volume of the piezometer over the top 0.35 m depth of the wetland was 247 mL (thereafter a well volume was designated as 250 mL). The piezometer was installed into a PVC sleeve, 0.08 m diameter and 0.40 m deep, hollowed out from the wetland sediment with an auger. The piezometer was inserted into this PVC sleeve and pounded into the clay aquiclude for stability. Each piezometer extended 0.35 m above the wetland ground surface and over 0.2 m deeep into the aquiclude. After installation, the lower 0.2 m of space between the 0.08-m PVC sleeve and the piezometer was backfilled with quartz drilling sand (36% pore space), and the upper top 0.2 m was filled with bentonite before removal of the PVC sleeve. Added bentonite acted as a seal to minimise water flow along the side of each piezometer. Two additional piezometers, which were not confined within lysimeters, were also installed at the site to obtain ambient subsurface water to supply the amended dosing volumes used in the push–pull test. The screened depth, construction, and installation of these piezometers were identical to the piezometers in the lysimeters.

A preliminary study was conducted about 6 weeks before the initiation of the ¹⁵N-tracer push–pull study. The objective of this study was to provide background information on the physical behaviour of an introduced tracer plume and the approximate rate of NO₃⁻ removal within the studied lysimeters. We dosed 4 lysimeters through their piezometers with 10 L of wetland subsurface water at a rate of 0.2 L/min., using a peristaltic pump. This subsurface water was amended with 200 mg/L of chloride (Cl⁻) and 8 mg/L of NO₃⁻-N). Surface and subsurface water samples were then obtained 1, 2, 3, 4, 24, 48, 72, and 96 h after dosing and analysed for NH₄⁺, NO₃⁻, and Cl⁻ concentrations. During and immediately following this study we extracted a total of 15 L from each lysimeter to remove most of the introduced plume.

To prepare a dosing solution for 4 lysimeters (10 L per lysimeter), about 42 L of clear wetland water was extracted from the 2 additional piezometers outside the lysimeters. To achieve clear wetland water, the dead volume (muddy water) was pumped until clear water was visible. The collected water was brought back to the laboratory in insulated boxes with ice cubes and stored <4°C until used for preparation of the dosing solution. The dosing solution contained 30 mg/L of Br⁻ (as LiBr) and 12 mg/L of ¹⁵N-labelled NO₃⁻ as KNO₃ (99 atom% ¹⁵N). The level of Br⁻ was higher than the background level (<1 mg/L) typically found in wetlands in the studied area while the NO₃-N level was comparable to that found in the studied wetland inflow (Nguyen et al. 2002). The dosing solution for each lysimeter was then transferred to a polypropylene carboy with a 3-port lid for sparging with 1 ppmv SF₆ through a sparge stone for 5 min to lower the dissolved oxygen (DO) to 3 mg/L. After SF₆ sparging, the headspace was sparged with helium (He) gas to lower the DO to <2 mg/L, to minimise the potential for artificially aerating the soil solution during dosing. The lid on the carboy was then securely closed to avoid any gas leakage during the transportation of the dosing solution to the field.

The push–pull techniques of Addy et al. (2002) was modified to measure NO₃⁻ removal and N₂O and N₂ generation during late summer (March 2003), the time of the year when no major inflows to and outflows from the studied wetland occurred. Surface water from each lysimeter was collected for ion (NH₄⁺, NO₃⁻, and Br⁻) and gas (N₂O and SF₆) analyses prior to injecting dosing solution into each piezometer. The dosing solution was again sparged with SF₆ to bring the DO <1 mg/L in the field. Two subsamples of dosing solution were then taken to determine the concentration of SF₆. Dosing solution was injected between 1100 and 1200 hours. into the piezometer of each lysimeter at the rate of 250 mL/min using a peristaltic pump. This injection was found to increase the surface water level in each lysimeter to about 0.03–0.05 m. Replicated samples of surface water and subsurface water (piezometer) were removed from each lysimeter after 1, 2, 3, 4, 24, and 48 h of dosing for analysis of dissolved gases (¹⁵N₂O, ¹⁵N₂, N₂O, SF₆) and ions (NH₄⁺, NO₃⁻, and Br⁻). Before subsurface water was taken from each piezometer for gas and ion analyses, 2 well volumes of water (500 mL) at the first sampling (i.e. 1 h after dosing) and 1 well volume at subsequent samplings (i.e. 2, 3, 4, 24, and 48 h after dosing) were pulled from each lysimeter using a peristaltic pump and discarded. Another 200 mL was then taken from each piezometer and stored in two 100-mL plastic bottles for ion analyses. For gas analyses, duplicate 20-mL subsurface water samples were collected from each piezometer using a 60-mL syringe through a closed system to avoid exposure to air, and injected into 2 separate 120-mL pre-evacuated gas bottles. All samples were brought back to the laboratory in insulated boxes with ice cubes and stored at 4°C before analyses.

Four samples of mixed ryegrass–white clover pasture herbage (standing pasture herbage of 0.1–0.15 m height) were randomly taken from adjacent paddocks by cutting to 20 mm height. Similarly, grab samples of vegetation in the wetland swale (mainly sweet grass) were also collected. Pasture herbage and wetland vegetation were dried at 60°C for 7 days, ground, sieved <2 mm, and analysed for total N. Four sediment samples per replicate (85 mm diameter) were taken from 0–0.1, 0.1–0.2, 0.2–0.4, and 0.4–0.7 m depths outside each lysimeter for bulk density (BD), porosity, and saturated hydraulic conductivity (Ks). Estimates of Ks for each horizon were obtained from undisturbed cores using a constant head laboratory method described in Rutherford and Nguyen (2004). Soils were then extruded and oven-dried at 105°C for 48 h to determine BD and porosity. The unconsolidated nature of the top 0–0.1 m sediment depth made it difficult to use the soil core technique for measuring Ks, and hence the pump test technique was used (Klute 1986). Four additional sediment samples (0.05 m diam.) were taken from the same depths and analysed for denitrification enzyme activity (DEA) (Tiedje et al. 1989), pH, organic matter, moisture, and NH₄⁺ and NO₃⁻ contents.

Water samples collected from the preliminary test were analysed for Cl⁻ and Br⁻ by ion chromatography (American Public Health Association 1998), and for total Kjeldahl-N, NO₃⁻-N, and NH₄⁺-N with a flow injection analyser (FIA). The DO and temperature of surface and subsurface waters were recorded at every sampling event using a WP-82Y Model DO/temperature meter (TPS). Water samples for ion analyses during the main experiment were filtered through GFC filters (1 μm pore size) to remove suspended materials and subsequently analysed for Br⁻ by inductively coupled plasma emission spectroscopy-mass spectrometry (American Public Health Association 1998), and for NH₄⁺, NO₂⁻, and NO₃⁻ by FIA.

Immediately after arrival at the laboratory, He gas was injected into the 120-mL gas bottle samples collected for dissolved gas analyses, to bring them to atmospheric pressure. After storage overnight at 4°C, the sample bottles were shaken for 30 s to equilibrate SF₆, N₂O, and N₂ between the aqueous (subsurface water) and gaseous phases (headspace). Two gas samples of 12 mL each were then collected from each bottle headspace and stored in pre-evacuated 12-mL glass vials fitted with a screw cap and a rubber septum (Exetainers; Labco, High Wycombe, UK), one for ¹⁵N₂O and ¹⁵N₂ analyses and another for N₂O and SF₆ analyses. The ¹⁵N analyses of N₂O and N₂ were performed by automated continuous flow isotope ratio mass spectrometry (CF-IRMS) and N₂O and SF₆ concentrations were determined using a gas chromatograph (GC) (Shimadzu GC-17A, Japan) equipped with a ⁶³Ni-electron capture detector operating at column, injector, and detector temperature of 55, 75, and 330°C, respectively. Correction for dissolved N₂O and N₂ in the water phase at 5°C was achieved by using the Bunsen solubility coefficients of 1.06 and 0.021 for N₂O and N₂, respectively (Weiss and Price 1980).

Sediment water content was determined gravimetrically by oven drying at 105°C for 48 h. Sediment NH₄⁺ and NO₃⁻ contents were determined by shaking freshly collected sediment (10 g on oven-dried basis) with 20 mL of 2 m KCl solution for 1 h, followed by centrifugation at 4000G, filtration through Whatman no. 42 filters, and analysis by FIA. Sediment organic matter content was determined as the proportion of the weight of oven-dried soil based on loss on ignition (LOI) at 550°C for 4 h. Sediment pH was determined using a combination glass electrode after equilibrating freshly collected samples (10 g on oven-dried basis) with 10 mL of deionised water for 30 min. All sediment analyses were corrected for sediment water content.

Wetland sediment was bulked on a soil depth basis and 3 replicate samples from each depth were analysed for DEA using a Bunsen coefficient of 0.632 for 20°C (temperature of the laboratory) to account for N₂O in aqueous solution (Tiedje et al. 1989). Sediment (5 g fresh weight) was amended with 5 mL solution containing NO₃⁻ (KNO₃ 0.1 g/L) and glucose (0.2 g/L). The headspace was flushed with N₂ gas and 10 mL acetylene (C₂H₂) to inhibit the reduction of N₂O to N₂. After 15 and 60 min of incubation at laboratory ambient temperature (20°C), duplicate headspace samples were transferred to 12-mL exetainers. The samples from exetainers were then analysed for N₂O by the Shimadzu GC as described above.

Plant samples (0.1 g per sample) were digested with 5 mL of Kjeldahl mixture for 3 h at 350°C and N content was determined by automated analysis with a Technicon Auto Analyzer II (Technicon Instruments Corp., Tarytown, NY). Total S and total P content in plant samples were determined using the technique of Quin and Woods (1976) in which plant samples were digested with nitric and perchloric acid mixture for 1 h at 200°C.

Measured concentrations and ¹⁵N atom% in ¹⁵N¹⁵N¹⁶O (⁴⁶N₂O) and ¹⁴N¹⁵N¹⁶O (⁴⁵N₂O) in dissolved gases extracted from surface and subsurface water samples were used to determine denitrification rates after correcting for the ¹⁵N background (natural abundance) of 0.3663 atom%. The dissolved N₂ concentration in water samples (obtained from CF-IRMS) was converted to the amount (μg/L) as described in Tiedje (1982) taking into account the volume of an exetainer, the volume of water samples collected from each lysimeter for N₂ analyses, and the volume of a headspace in a 0.12-L bottle (i.e. 0.012, 0.020, and 0.10 L, respectively). The amount of N₂ produced was then converted to a rate of N₂ generation by multiplying it by the ratio of applied ¹⁵N (99% atom) present as ¹⁵N atom % in dissolved N₂. The rate was expressed as μg/L.day after taking into account the time interval between each sampling event.

Similarly, the N₂O concentration in water samples as measured by GC was converted to the amount of N₂O produced (μg/L), which was then converted to the rate of N₂O generation (i.e. the amount of N₂O that was derived from the denitrification of labelled NO₃⁻) by multiplying it by the ratio of applied ¹⁵N (99% atom) present as ¹⁵N atom% in dissolved N₂O. The data on ¹⁵N atom% in ⁴⁶N₂O was used in this calculation since it was significantly correlated (r² = 0.86; P < 0.001) with the ¹⁵N atom% in ⁴⁵N₂O of the same samples.

Nitrate removal attributed to denitrification and other biological processes was calculated as the difference between measured and estimated ‘conserved’ NO₃⁻ concentrations in subsurface water at a sampling time (t_s). The ‘conserved’ NO₃⁻ estimate at t_s was based on concentration reductions associated with the introduced Br⁻ tracer and represented an estimate for NO₃⁻ concentration reductions resulting from physical processes such as dilution and not denitrification or other biological processes. Based on this assumption, the estimated NO₃⁻ concentration was calculated by multiplying NO₃⁻ concentration at the time of dosing (t₀) by the ratio of Br⁻ concentration at t_s and t₀ (Burns and Nguyen 2002).

Standard deviation and standard error were calculated for different parameters using data collected from the 4 lysimeters. The first-order kinetic equation was used to fit the curve in the surface and subsurface data of NO₃⁻ concentration and ratio of NO₃⁻ over Br⁻ concentrations.

The studied wetland sediment had low BD and high organic matter (LOI) content (Tables 1 and 2). Its high porosity and high hydraulic conductivity in the top 0.1 m depth (Table 1) made it difficult to use the push–pull technique of Addy et al. (2002) without the use of a confined lysimeter to investigate NO₃⁻ removal and gaseous emissions of N₂O and N₂ due to rapid dispersion of the introduced tracer plume. Using a confined lysimeter minimised advection–dispersion of the introduced Br⁻ tracer plume in the subsurface water. The major form of N in surface and subsurface depths of wetland sediments was NH₄⁺-N (Table 2) because anaerobicity inhibits nitrification (Zaman et al. 2007).

Table 1.  Sediment physical properties
Values are means ± standard deviations

Table 2.  Sediment chemical properties
Values are means ± standard deviations

Both surface water and subsurface water had low DO levels (<1 mg O₂/L; Table 3), suggesting that anaerobic conditions favourable for denitrification (Smith et al. 2003; Zaman et al. 2008b) were prevalent in this wetland. Various workers (Tiedje 1988; Patrick and Jugsujinda 1992; Achtnich et al. 1995; Blicher-Mathiesen and Hoffmann 1999) have reported that NO₃⁻ reduction occurs mainly when subsurface water DO falls below 0.5–1.6 mg O₂/L.

Table 3.  Surface and subsurface water characteristics before the commencement of the study
Values are means ± standard deviations of 4 measurements

The low level of NO₃⁻ in wetland waters (Table 3) suggests that any NO₃⁻ that enters the wetland via seepage springs and runoff is readily removed by denitrification (Zaman et al. 2008b) and other biological processes (Seitzinger 1994; Hill 1996; Fennessy and Cronk 1997; Matheson et al. 2002, 2003). Ammonium and organic N were the predominant N fractions in both surface and subsurface waters (Table 3), likely due to inputs from farmland runoff and/or the incomplete breakdown of organic matter originating from wetland organic sediments and/or plant vegetation under anaerobic conditions (Nguyen 2000).

Denitrification enzyme activity in the upper 0.4 m of wetland sediment ranged from 53 to 205 mg N₂O-N/kg soil.day (Table 4). Highest DEA was measured in the top 0.1 m and decreased sharply with depth, probably due to a reduction in the level of organic matter and denitrifier populations with depth. Since NO₃⁻ was unlikely to reach the lower depths because of active denitrification in the top sediment layer, DEA at the lower depths was also likely limited by low NO₃⁻ concentrations (Xue et al. 1999; Hoffmann et al. 2000; Well et al. 2001).

Table 4.  Denitrification enzyme activities at different depths in the studied wetland sediment
Values are means ± standard deviations (n = 3)

At the time of the experiment, nutrient status of the sweet grass collected from the wetland channel was lower than that in pasture herbage of the adjacent paddocks (Table 5). These results were similar to those obtained in the following spring (6 months later in September) when wetland plants were at their most active growing stage. The lower N (and also P and S) status in wetland vegetation was likely to be due to the difference in plant species of wetland and pasture soils. Perennial ryegrass (Lolium perenne L.) is a highly nitrophyllous species, selected for high production potential, whereas wetland plants do not require as much N as pastures; therefore, their roots systems are less adventitious in accessing N. The high N concentrations (Table 5) of pasture herbage with a mixture of ryegrass and white clover (Trifolium repens L.) suggest a high level of soil fertility; the paddock results are consistent with those reported in New Zealand conditions (Machado et al. 2005; Blennerhassett et al. 2006; Zaman et al. 2008a).

Table 5.  Nutrient status (g/100 g dry matter) of wetland sweet grass and pasture herbage in the adjacent paddocks
Values are means ± standard deviations

The Cl⁻ concentration of the introduced plume changed <25% over a 72-h period (data not shown), suggesting that the introduced subsurface water plume in the in-situ wetland lysimeters was subject to minimal dilution. The low hydraulic conductivity of the wetland sediment at depths beyond the top 0.4 m (Table 1) may limit any water transfer out of, or into, this sediment zone. This was further confirmed by the slow drop in the level of the ponded water that occurred after injection of the dosing solution. No water was observed emerging around the piezometers or the bentonite seals, arguing against any significant short-circuiting of flow between surface and subsurface water in the piezometer–lysimeter system. Thus, the modification of the push–pull technique of Addy et al. (2002) with the use of 0.03-m internal diameter piezometers, instead of mini-piezometers in conjunction with in-situ lysimeters, was appropriate in our wetland environment.

Nitrate in both surface and subsurface water returned to background levels (<0.015 mg/L) within 24 h after dosing (data not shown). As a result of this rapid removal rate, we chose to use a higher initial concentration (12 v. 8 mg N/L) during the denitrification study with ¹⁵N-enriched NO₃⁻.

Following dosing with ¹⁵N-enriched NO₃⁻, Br⁻, and SF₆, NO₃⁻ concentrations (Fig. 2a, b; Table 6) and NO₃⁻/Br⁻ ratios (Fig. 3a, b) in both surface and subsurface waters significantly (P < 0.001) decreased with time, indicating an active NO₃⁻ removal from surface water and particularly from subsurface water. Bromide is a conservative tracer and its fate over a short period of time (48 h) is probably governed by hydrologic conditions and not by transformations from microbial processes or plant uptake (Xue et al. 1999; Addy et al. 2002; Burns and Nguyen 2002). If decreases in NO₃⁻ concentrations were strictly due to physical processes, changes in NO₃⁻ concentration with time would mirror the reduction in Br⁻ concentration with time, yielding a constant NO₃⁻/Br⁻ ratio over the sampling period (Addy et al. 2002).

Table 6.  Measured and theoretical (based on the changes in bromide concentrations with time) concentrations of NO₃⁻-N (g/cm³) in wetland waters sampled over a 48-h study period
Values are means ± standard deviations

Changes in the ratio of Br⁻ (C_i/C₀) concentration between t_i and t₀ can be used to estimate the expected concentrations at time t_i of ‘conserved’ NO₃⁻ (the concentration resulting from physical processes such as dilution and dispersion). The conserved concentration estimates based on changes in the Br⁻ (C_i/C₀) ratios were higher than measured NO₃⁻-N concentrations within each lysimeter during virtually all (46 out of 48) of the sampling periods (Table 6). Measured NO₃⁻-N concentrations remained >2 mg/L, the level at which NO₃⁻-N availability begins to limit denitrification (Schipper and Vojodic-Vukovic 1998), for the first 3–4 h following dosing in all subsurface water samples and in the surface water of 2 of the 4 lysimeters. The surface water was a mixture of the dosing solution and displaced water and exhibited considerable dilution as it moved from the 0.2–0.35 m dosing depth to the surface, with relative Br⁻ concentrations ranging from 0.1 to 0.3 (C_i/C₀) 1 h after dosing. All samples from all locations obtained after 24 and 48 h had low NO₃⁻-N concentrations.

The average NO₃⁻ removal rates during non-limiting conditions (i.e. NO₃⁻-N > 2.0 mg/L) were 3.96 and 15.7 mg/L.day for surface and subsurface water, respectively (Table 7). These rates were substantially higher than the average rates of ¹⁵N-enriched denitrification gas production (N₂O-N plus N₂) over the same period (0.25 and 1.1 mg/L.day in the surface and subsurface water, respectively; Table 7). Denitrification only accounted for 6–7% of NO₃⁻ removal during non-limiting conditions, suggesting that other transformation processes were responsible for most of the NO₃⁻ removal. Because many studies expressed the rates on a real basis, we transformed the rates by assuming that the observed groundwater rates reflect conditions within the upper 0.4 m of the wetland soil. Below that depth, organic matter and Ks decline precipitously. Given the measured porosity of 0.65 (Table 1), the upper 0.4 m of a 1-m² area contains 260 L of water, so areal subsurface water denitrification and NO₃⁻ removal rates were 289 and 4094 mg N/m².day.

Table 7.  Nitrate-N removal rates (μg/L.day) and corresponding flux (μg/L.day) of ¹⁵N-enriched denitrification gases (N₂O-N and N₂) during non-limiting phase of push–pull experiments
Nitrate-N removal rates calculated from changes in Br⁻ : NO₃⁻-N ratios. Non-limiting phase defined as NO₃⁻-N concentrations >2.0 mg/L. Values are means ± standard deviations. n.d., No values obtained for NO₃⁻-limiting conditions

The observed groundwater denitrification rates are comparable to those (quoted below in mg N/m².day) reported for other wetland sediments (76.8–115.2, Christensen and Sorensen 1986; 33.6–336, Lindau et al. 1990, Lowrance et al. 1995; 48–283, Xue et al. 1999; 20–80, Hefting et al. 2003; 50–741, Pinay and Decamps 1988, Cooper 1990, Haycock and Burt 1993). The elevated rates we observed reflect environmental conditions conducive to denitrification, i.e. low DO (<1 mg O₂/L; Table 3) plus a highly enriched organic matter sediment (Table 2) and optimum sediment–water contact time within the confined lysimeter environment (Hill 1996; Fennessy and Cronk 1997; Hoffmann et al. 2000; Smith et al. 2003). The observed high denitrification rates could be attributed to the elevated DEA values (Table 4) in the wetland soils (Groffman et al. 1999).

The amount of N₂O and N₂ generation (and hence denitrification) did not account for most of the NO₃⁻ removal; however, it is unlikely that we have underestimated the amount of N₂O and N₂ generation. The SF₆ concentration in wetland waters (data not shown) did not significantly change with time after its dosing; the transport of N₂O and N₂ to atmosphere via stem and aerenchyma of wetland vegetation (Well et al. 2001; Hefting et al. 2003), or through the interface between the sediment and the edge of piezometers or channels in the sediment or around the dead roots (Blicher-Mathiesen et al. 1998), is assumed to be insignificant in our study.

Given that denitrification was a minor pathway for NO₃⁻ transformation, the rapid NO₃⁻ removal observed in the first several hours following dosing may be attributed to several processes that we did not measure; specifically DNRA (Silver et al. 2001; Matheson et al. 2002, 2003), abiotic immobilisation (Davidson et al. 2003), and microbial immobilisation and plant uptake (Hill 1996; Fennessy and Cronk 1997). Because the wetland soils likely included eroded materials from the adjacent Allophanic volcanic soils of the paddocks, there is potential for some portion of the dosed NO₃⁻ to be subject to anionic sorption (Magesan et al. 1998). However, considerable NO₃⁻ leaching has been observed in the paddock soils (Wilcock et al. 1999), suggesting that any sorption is likely to account for a limited proportion of the NO₃⁻ removal. Uptake of NO₃⁻ by wetland plants in our study appears to be plausible even over a short period of 48 h, particularly because the vegetation appeared to be N-deficient at the time of the experiment. Bowman et al. (1989a, 1989b) found extremely high, short-term N uptake rates by moderately N-deficient turfgrass. In particular, they noted that N uptake rates by N-deficient ryegrass during the first 6 h of exposure to NO₃⁻-enriched solution was >4-fold greater than the rates that occurred after 96 h of exposure.

The pasture growth rate in the studied Waikato region during February–March is at least 45–50 kg dry matter (DM)/ha.day and the standing biomass in our wetland site was estimated to be 3–4 times that of the adjacent pastoral land (data not shown). Assuming a daily wetland plant above-ground growth rate of 135–200 kg DM/ha.day (assuming that production rate is proportional to standing biomass), the amount of N that can potentially be taken up by wetland vegetation with N concentration of 2.6–2.7% (Table 5) is 3.5–5.4 kg N/ha.day (351–540 mg N/m².day). In our study approximately 680 mg NO₃⁻-N/m² was depleted over a 4-h period, suggesting that short-term plant uptake by N-deficient plants could account for a substantial portion of the NO₃⁻ removal.

Although considerable increases in ¹⁵N-enriched N₂O-N levels were initially observed following NO₃⁻ dosing, no net emissions were generated over the 48-h study (Fig. 4a, b). The wetland served as a source of N₂O during the non-NO₃⁻-limiting phase of the experiment, but functioned as a sink for N₂O during the NO₃⁻-limited phase of the study. Concentrations of N₂O rapidly increased with time, reaching a peak 4 h after dosing and declining to background by 24 h after dosing (Fig. 4a, b). This was observed for both surface and subsurface waters and was related to the decline in NO₃⁻ concentration with time (r = 0.41–0.46; P < 0.05). Nitrous oxide generation was more predominant than isotopically enriched N₂ generation (Fig. 4a, b v. Fig. 5a, b) but this predominance declined with time (after 3–4 h of tracer dosing; Table 8), suggesting that the generation of N₂O and N₂ is dependent on the level of NO₃⁻ in sediments. With a decline in NO₃⁻ concentration with time (Fig. 2a, b), the N₂O : N₂ ratio decreased, and more N₂ instead of N₂O was emitted (Fig. 5a, b), probably because of the reduction of N₂O to N₂. Our results are in agreement with several studies (Swerts et al. 1996; Cho et al. 1997; Dendooven et al. 1997; Blicher-Mathiesen and Hoffmann 1999; Well et al. 2001; Zaman et al. 2008b) which found that during the initial stages of denitrification and in the presence of significant NO₃⁻ inputs, NO₃⁻ may be denitrified to N₂O and not fully reduced to N₂. However, when NO₃⁻ becomes limited, N₂O that is dissolved in wetland water is reduced to N₂ by microbes due to the demand for electron acceptors when NO₃⁻ concentrations are low.

Table 8.  Nitrous oxide (N₂O-N) fluxes, their ¹⁵N atom%, and ratios of N₂O-N : N₂ in wetland waters sampled over a 48-h study period
Flux is computed as the difference in concentration of dissolved gases between each sampling period per time interval during this period. Values are means ± standard deviations. n.a., N₂O concentrations declined while negligible N₂ generated during sampling interval

The extent of N₂O and N₂ generation in wetland waters was likely influenced not only by the concentration of NO₃⁻ in wetland waters but also by the depth from which N₂O was produced. If N₂O is produced at sites just below the top sediment layer, it may readily diffuse into sediment–air water interface as N₂O, instead of being reduced to N₂. In contrast, N₂O produced in subsoils may be entrapped in subsurface waters and subsequently be reduced to N₂ by sediment denitrifiers (Hefting et al. 2003; Smith et al. 2003). In addition to NO₃⁻ concentration and depth, sediment pH, C, and O₂ content and temperature may also affect the generation rates of N₂O and N₂ (Del Grosso et al. 2000; Dobbie and Smith 2001; Smith et al. 2003; Zaman et al. 2004, 2007, 2008b). The influence of this variety of sediment and environmental factors as discussed above could explain why a range of N₂O : N₂ ratios (0–20) has been reported in the literature (Rolston et al. 1978; Weier et al. 1993; Maag and Vinther 1996; Cho et al. 1997; Well et al. 2001; Rochester 2003) and in our study (0.33–188; Table 8) and in other associated laboratory studies using pastoral and wetland sediments collected from areas within the same catchment (0.9–1.4) (Zaman et al. 2007).

Since the potential contribution of a wetland to greenhouse emissions is dependent on the amount and fraction of N emitted as N₂O, and these parameters are likely to vary depending on sediment and environmental factors as outlined above, the net annual production (net balance between source and sink) of N₂O from the studied wetland with temporal variation in NO₃⁻ inputs is not known. This aspect needs to be evaluated in future studies. Various studies have shown that the extent and duration of contact time between wetland inflows and wetland sediment (Hill 1996; Cirmo and McDonnell 1997; Devito et al. 2000; Hill et al. 2000) can affect the extent of denitrification and hence the wetland capacity to remove NO₃⁻. During high flow events, low removal is expected because high NO₃⁻-containing water bypasses microbially active wetland sediments by flowing across the top of wetlands (Gold et al. 2001; Burns and Nguyen 2002; Rutherford and Nguyen 2004). The use of engineered bypass flow designs to regulate NO₃⁻ loading may enhance sediment–water contact time, thus promoting NO₃⁻ removal and the potential for NO₃⁻-limiting conditions. Our results suggest that the studied wetland may be a source of N₂O emissions when NO₃⁻ concentrations are elevated (non-limited), but can readily remove N₂O (function as a N₂O sink) when NO₃⁻ levels are low. Although the proposed bypass flow design may short-circuit some of wetland NO₃⁻ inflows and hence reduce the full potential for wetland NO₃⁻ removal, it provides a balance between water quality goals and greenhouse gas emissions. Future research is therefore required to investigate this balance issue and assess temporal variations in NO₃⁻ removal rates, denitrification, and N₂O generation in responses to changes in NO₃⁻ inputs and water-sediment contact time.

The studied seepage wetland can remove substantial amount of NO₃⁻ under conditions where sediment water contact time is optimum (confined lysimeter) for denitrification. The amount of NO₃⁻ removal was much higher than the amount of denitrification gases (N₂O and N₂) produced, suggesting that additional processes (e.g. plant uptake) were responsible for NO₃⁻ removal. The push–pull method, in combination with confined lysimeters to minimise advection and dispersion of water, is a promising tool for quantifying NO₃⁻ removal and N₂O and N₂ generation rates under a variety of conditions.