Water, Air, & Soil Pollution

, 224:1693

Quantification of In Situ Denitrification Rates in Groundwater Below an Arable and a Grassland System


  • M. M. R. Jahangir
    • Teagasc Environment Research CentreJohnstown Castle Estate
    • Department of Civil, Structural and Environmental EngineeringTrinity College
    • Department of Soil ScienceBangladesh Agricultural University
  • P. Johnston
    • Department of Civil, Structural and Environmental EngineeringTrinity College
  • K. Addy
    • Department of Natural Resources ScienceUniversity of Rhode Island
  • M. I. Khalil
    • Environmental Protection AgencyJohnstown Castle Estate
  • P. M. Groffman
    • Cary Institute of Ecosystem Studies
    • Teagasc Environment Research CentreJohnstown Castle Estate

DOI: 10.1007/s11270-013-1693-z

Cite this article as:
Jahangir, M.M.R., Johnston, P., Addy, K. et al. Water Air Soil Pollut (2013) 224: 1693. doi:10.1007/s11270-013-1693-z


Understanding denitrification rates in groundwater ecosystems can help predict where agricultural reactive nitrogen (N) contributes to environmental degradation. In situ groundwater denitrification rates were determined in subsoil, at the bedrock interface and in bedrock at two sites, grassland and arable, using an in situ ‘push–pull’ method with 15N-labelled nitrate (NO3–N). Measured groundwater denitrification rates ranged from 1.3 to 469.5 μg N kg−1 day−1. Exceptionally high denitrification rates observed at the bedrock interface at grassland site (470 ± 152 μg N kg−1 day−1; SE, standard error) suggest that deep groundwater can serve as substantial hotspots for NO3–N removal. However, denitrification rates at the other locations were low and may not substantially reduce NO3–N delivery to surface waters. Denitrification rates were negatively correlated with ambient dissolved oxygen, redox potential, ks and NO3 (all p values, p < 0.01) and positively correlated with SO42− (p < 0.05). Higher mean N2O/(N2O + N2) ratios at an arable (0.28) site than the grassland (0.10) revealed that the arable site has higher potential to indirect N2O emissions. Identification of areas with high and low denitrification and related site parameters can be a tool to manage agricultural N to safeguard the environment.


Denitrification15N-enrichment15N–N2O15N–N2GroundwaterN2O mole fraction

1 Introduction

The nitrogen (N) cascade is an increasingly important global issue with multiple impacts on terrestrial, aquatic and atmospheric environments (Galloway et al. 2008). The high rates of N deposition result in N saturation in agricultural land causing high nitrate (NO3–N) delivery to groundwater which is of concern with result to global environment and human health (Organisation of Economic Co-operation and Development 2009). In Ireland, groundwater beneath some agricultural systems is contaminated with NO3, and this also contributed to the eutrophication of estuarine and near coastal waters (McGarrigle et al. 2010). The OECD (2009) urged Ireland to strengthen measures to achieve ‘good ecological status’ for Irish waters by 2015, paying special attention to eutrophication. The requirement for good ecological status for Irish waters is a requirement of the EU Water Framework Directive (WFD; EC 2002).

The biogeochemical process, denitrification, is the principal process which converts the NO3–N to nitrous oxide (N2O) and dinitrogen (N2) gas (Rivett et al. 2008). The intermediate product N2O is a potent greenhouse gas with global warming potential of 298 over a 100-year time period. Indirect N2O emissions resulting from N leaching into associated groundwater are an important but poorly understood component of global N2O budget (Clough et al. 2007). The quantity of the end product of denitrification process, N2, is by far the largest uncertainty of the N cycle at all scales (Galloway et al. 2004). Therefore, narrowing this uncertainty is critical if improvements are to be made in global N2O and N2 budgets. Quantification of N2O/(N2O + N2) ratios in groundwater would help refine greenhouse gas inventories and provide insights into the relative contribution of denitrification to environmentally benign N2 production.

As denitrifiers are reported to be ubiquitous in shallow to deep groundwaters (Linne von Berg and Bothe 1992; Francis et al. 1989), the availability of energy sources and suitability of the hydrogeological environments for denitrifiers need to be investigated. Barrett et al. (2013) quantified denitrification genes in four Irish aquifers (up to 50 m), including the two sites in the current study. They found similar concentrations of denitrification genes across sites and piezometer depth. Therefore, optimum hydrogeochemical conditions for microbial denitrification can help the biodegradation of NO3–N (ITRC 2002). Analysis of dissolved N2O and N2 in groundwaters from subsoil (5 m), bedrock interface (12 m) and bedrock (22 m) in Ireland underlined that denitrification can be an important NO3 removal pathway across shallow to deep groundwaters (Jahangir et al. 2012a). However, in groundwater denitrification studies, it is often unclear if the denitrification products are produced in situ or if they have been leached from surface soils (Groffman et al. 1998). Application of in situ remediation to any contaminant and site is gaining wide acceptance as viable and economic technology (ITRC 2002). However, the denitrification process in groundwater is very difficult to measure, and existing methods used to measure denitrification are problematic for a variety of reasons (e.g., high background N2, degassing of samples and physical attenuation) (Groffman et al. 2006). The in situ NO3 push–pull method has been used to determine the denitrification in shallow groundwater (<3 m) (Addy et al. 2002; Kellogg et al. 2005). Istok et al. (1997) used the push–pull method for measuring groundwater denitrification in a sand and gravel aquifer at a depth of approximately 10 m. However, in the deep groundwater zones, it can be more challenging due to the complex hydrogeological settings, e.g., high permeability or preferential flow through fracture in bedrock resulting in high physical attenuation (Buss et al. 2005). In this study, the push–pull method was extended from shallow to deep groundwaters (up to 22 m) to quantify denitrification rates. The objectives of this study were to (a) assess the application of the ‘push–pull’ method in deep groundwaters, (b) determine in situ denitrification rates in shallow to deep groundwaters, (c) quantify the N2O mole fractions, N2O/(N2O + N2) and (d) identify factors controlling the observed spatial trends of denitrification rates.

2 Materials and Methods

2.1 Experimental Site Characteristics

The in situ NO3 push–pull method was used at two groundwater monitoring sites in southeastern Ireland (Fig. 1). The sites were the Johnstown Castle (52° 17′ 30″ N, 6° 29′ 50″ W), a poorly drained intensively managed grazed (35 years) grassland, and the Oak Park (52° 51′ 43″ N, 6° 54′ 53″ W), a well-drained arable land with spring barley cover crop rotation (10 years). Both sites receive approximately 312 and 150 kg N ha−1 as organic and inorganic forms of N, resulting in N surpluses of 243 and 75 kg N ha−1, respectively. The grassland site comprises poorly drained top soils overlying clayey subsoils intermixed with sands and gravels followed by Ordovician sediments, sandstone and shale at 10 m. At the arable site, soil profile comprises well-drained top soil overlying subsoils of sands, gravel and interbedded clay band followed by grey limestone at 10 m (Fig. 2). Three distinct water tables were encountered on each of the sites, and these were specifically targeted with piezometers (Fig. 2). The aquifer beneath the grassland site is poorly productive, with a shallow perched water table but has had elevated NO3–N concentrations reported (Fenton et al. 2009). At the arable site, there had a productive sand and gravel aquifer overlying a productive limestone aquifer, both of which were vulnerable to NO3-N pollution as previously described by Premrov et al. (2012). The hydrologic and geochemical properties of the sites were presented in Table 1. The grass and arable sites represent approximately 62 and 37 % of Irish soil types and 21 and 71 % of bedrock types, respectively.
Fig. 1

Experimental sites and multilevel well locations, grassland at Johnstown Castle and arable land at Oak Park in southeastern Ireland. Receptors are carrying groundwater to the nearby rivers (river ‘Kildavin’ at grassland and river ‘Barrow’ at arable land)

Fig. 2

Borehole installation cross sections from sites: Johnstown castle (JC) and Oak Park (OP) with average water table and ks values. Wells installation depths, geochemical properties, details of water table depths and ks values were summarised in Table 1

Table 1

Ambient hydrologic and hydrochemical properties




Fe (II)








mg L−1


(m V)

(m, bgl)

(m day−1)


 Subsoil 5 m bgl

4.7 ± 1.6b

1.0 ± 0.1b

12 ± 4b

20.0 ± 1.6b

0.26 ± 0.04b

1.9 ± 0.1b

6.9 ± 0.1b

94 ± 28b

1.8 ± 0.1b

0.008 ± 0.002b

 Interface 12 m bgl

2.0 ± 1.8c

3.5 ± 2.3b

48 ± 27c

19.2 ± 1.6b

0.21 ± 0.06b

1.3 ± 0.4c

6.8 ± 0.1b

25 ± 62c

2.9 ± 0.9c

0.024 ± 0.004b

 Bedrock 22 m bgl

2.9 ± 1.3c

3.4 ± 2.7b

14 ± 13b

16.4 ± 1.6b

0.24 ± 0.04b

1.6 ± 0.1c

6.8 ± 0.1b

47 ± 43c

3.4 ± 1.0c

0.030 ± 0.005b

Arable land

 Subsoil 5 m bgl

12.8 ± 2.6b

1.1 ± 0.2b

4.4 ± 1.1b

27.2 ± 1.0b

0.17 ± 0.01b

9.5 ± 1.4b

7.8 ± 1.3c

178 ± 60b

4.2 ± 0.2b

0.033 ± 0.006b

 Interface 12 m bgl

10.4 ± 0.3b

0.8 ± 0.2b

4.8 ± 0.7b

23.3 ± 1.1b

0.24 ± 0.08b

6.2 ± 1.6c

8.9 ± 1.2b

163 ± 50b

4.6 ± 0.1b

0.053 ± 0.003b

 Bedrock 22 m bgl

12.6 ± 2.5b

0.7 ± 0.2b

2.7 ± 1.0b

27.3 ± 0.7b

0.18 ± 0.05b

4.1 ± 1.4c

7.5 ± 0.1c

107 ± 39c

5.1 ± 0.1b

0.123 ± 0.003c

Values are means ± SE, n = 3. The same letter within each site does not differ significantly between depths (p > 0.05)

DOC dissolved organic carbon, GWT groundwater table, ks saturated hydraulic conductivity, bgl below ground level

aDO dissolved oxygen

2.2 In Situ Push–Pull Method

We adapted the in situ push–pull method of Addy et al. (2002) and Kellogg et al. (2005) to estimate the denitrification rates in shallow (5 m below ground level, bgl) to deep (12–22 m bgl) groundwaters. Groundwater wells (PVC with 0.05 m i. d.; 2-m screen section) were placed along groundwater flow paths at three depths to target samples in (S) subsoil (5 m bgl), (I) bedrock interface (12 m bgl) and (B) bedrock (22 m bgl). The push–pull method comprised two steps: (1) the push–pull pre-test and (2) the NO3 push–pull test. The study (pre-test and NO3 push–pull test) was conducted during October–December 2010.

2.3 In Situ Push–Pull Pre-Test

In situ push–pull pre-test was conducted to gain insights into balancing high recovery of the plume with sufficient time in situ for microbial denitrification to occur at detectable levels. Twenty litres of groundwater was collected from each well, amended with a conservative tracer bromide (Br, 20 mg L−1) and pushed into the same well (at least one well per depth per site) using a peristaltic pump (Model 410, Solinst Canada Ltd.). The dosing solution amended with Br was sampled during the push phase to obtain the undiluted concentration of Br. The push–pull pre-test was conducted repeatedly with initial incubation for a 12-h period and then lowered to 3 h with the estimation of corresponding recoveries of Br. An incubation period less than 3 h was not attempted because of the concern that there would be low detection of denitrification gases in the subsequent NO3 push–pull test, particularly in deep bedrock. After the incubation period, groundwater (twice the dosing volume), pulled up using a Grundfos pump (Model MP1, Grundfos, Fresno, CA, USA) taking samples at 2-L intervals, was analysed for the Br recovery at each sample intervals. A peristaltic pump was not used to pump water because of its inability to pump water from depths greater than 6 m bgl. Groundwater injection and pumping back were conducted slowly to prevent changes in hydraulic gradient around the well. After 1 week, groundwater in the pre-tested wells was resampled and analysed for Br to ensure that the tracer concentration was at ambient level before conducting another pre-test with a shorter incubation period or before conducting the in situ NO3 push–pull test.

The injected volume of water was sufficient to fill approximately 270 to 1,000 kg of aquifer materials (bulk density = 1,650–2,500 kg m−3, porosity = 0.03–0.12) after correcting for the sand and gravel pack around the well. The total amount of aquifer materials covered by the solution was calculated using Eq. 1 below:
$$ \mathrm{Mt}=\left[\frac{\left(\mathrm{Vt}-\mathrm{Vg}\right)}{\mathrm{Porosity}\;\mathrm{of}\;\mathrm{aquifer}}\right]\kern0.37em \times \mathrm{Bd} $$
where Mt is the total mass of aquifer materials (kg), Vt is the total volume of solution (m3), Vg is the volume of gravel pack (m3) and Bd is the bulk density (kg m−3).

2.4 In Situ 15N–NO3 Push–Pull Experiment

In situ NO3 push–pull tests were conducted in S, I and B with three replications per depth. Twenty litres of groundwater was collected in a carboy from each well and stored in a cold room at 4 °C for a maximum of 2 days. To adjust the dissolved oxygen (DO) back to ambient conditions, groundwater solution was bubbled with sulphur hexafluoride (SF6, 98.2 %, Cryoservice Ltd., Worcester WR4 9RH, UK) while the DO concentration was monitored using a DO probe (Multi 340i/SET, WTW, Germany). The SF6 can also serve as a conservative tracer. The dosing solution was prepared with ambient groundwater, 20 mg L−1 Br as KBr and 20 mg N L−1 as isotopically enriched KNO3 (50 at.% 15N–KNO3, purity 99 %). The carboy with dosing solution was capped, and its headspace was filled with the SF6 gas. The SF6 headspace was maintained with the same pressure while connected to a SF6 gas cylinder (carried to field) during the injection of the dosing solution. The dosing solution was injected into the respective well over the course of 1–2 h (depending on the permeability, Table 1, and hydraulic gradient, data not shown) with a peristaltic pump with a Teflon outlet at a very low rate (10 to 15 L h−1). Samples were collected for DO, SF6, Br and other dissolved gases and hydrochemistry during the middle of the injection phase.

The incubation period was defined as the length of time between the end of the push phase and the start of the pull phase since the plume core would consist mostly of the later injected groundwater. The incubation period for the dosing solution was set at 6 h, based on pre-test results so that there was substantial plume recovery and sufficient incubation time. After the incubation period, groundwater was pumped back from the well slowly (10 to 15 L h−1) using a Grundfos pump with a Teflon outlet. As the injected volume was pumped, such samples were taken using a syringe attached to an air-tight sampling apparatus made of stainless steel tubing connected to the outlet of the Grundfos pump. Groundwater samples (120 ml) were injected into an evacuated serum bottle (160 ml), and the headspace (40 ml) was filled with high-purity helium gas (He/water ratio, 1:3, v/v) and then submerged under water in a polystyrene box and stored at 4 °C. For each well, conservative tracer (Br and SF6) recoveries were estimated as C/Co; where C was the tracer’s concentrations in the pulled groundwater following incubation and Co was the tracer’s concentrations in the original pushed groundwater (Freeze and Cherry 1979).

2.5 Dissolved Gas Analysis

Groundwater dissolved gases (N2O, N2 and SF6) in ambient, pushed and pulled samples were extracted using the phase equilibration headspace extraction technique, with He filling the headspace (Lemon 1981; Davidson and Firestone 1988) in the lab on the same day of sample collection. Groundwater samples collected in the serum bottles were shaken for 5 min on a Gyrotory shaker (Model G-10, New Brunswick Scientific Co., USA) and left for a standing period of 30 min. Headspace samples were then taken for the analysis of SF6, N2O and N2 concentrations and the 15N enrichment of N2O and N2 in 12 ml exetainers (Labco Inc. Wycomb, UK) after injecting additional 12 ml high purity He. The N2O and SF6 gases were analysed on a gas chromatograph (CP-3800 GC, Varian, Inc. USA/CTC Analytics combi PAL Auto Sampler, Switzerland) equipped with an electron capture detector using Ar as a carrier gas. The GC had a Porapak-Q column (80–100 MESH), 3.7 m × 1/8″ × 2.0 mm. Concentrations and 15N enrichment of N2O and N2 were determined on a dual-inlet isotope ratio mass spectrometer (Stable Isotope Facility, UC Davis, Davis, CA) as described by Mosier and Schimel (1993).

2.6 Calculations of Denitrification Rate

Dissolved N2O and N2 concentrations were calculated using the three highest recovery values within sample replicates (Harrison et al. 2011). The masses of dissolved N2O–N and N2 gases (μg) were calculated from the headspace extraction samples using equations and constants provided by Tiedje (1982) and Mosier and Klemedtsson (1994). The total mass of N2O–N or N2 was then transformed to the mass of 15N2O–N or 15N2 by multiplying it by the respective 15N sample enrichment proportion (ratio of pulled at.% of the dissolved N2O–N and N2 to pushed NO3–N at.%, both corrected for ambient at.%). Gas production rates for 15N2O–N and 15N2–N were expressed as μg N kg−1 soil day−1 as below:
$$ {\mathrm{Rates}\ \upmu \mathrm{g}\ \mathrm{N}\ \mathrm{kg}}^{-1}{\mathrm{day}}^{-1}=\frac{\mathrm{Total}\;\mathrm{mass}\;\mathrm{of}{}^{15}\mathrm{N}2\mathrm{O}\hbox{--} \mathrm{N}\;\mathrm{and}{\;}^{15}\mathrm{N}2\hbox{--} \mathrm{N}\;\mathrm{per}\;\mathrm{volume}\;\mathrm{of}\;\mathrm{water}\;\mathrm{pulled}}{\mathrm{Dry}\;\mathrm{mass}\;\mathrm{of}\;\mathrm{soil}\;\mathrm{per}\;\mathrm{volume}\;\mathrm{of}\;\mathrm{water}\kern0.37em \times \kern0.37em \mathrm{incubation}\;\mathrm{per}\mathrm{iod}\;\mathrm{pulled}} $$

Mass of aquifer materials was calculated for individual depths at each site. Total denitrification rates were the sum of 15N2O–N and 15N2 generation rates. All samples used in denitrification calculations contained at least 8 mg L−1 NO3–N to ensure that calculated denitrification rate estimates were not limited by the amount of NO3–N available (Schipper and Vojvodic-Vukovic 1998).

2.7 Hydrological and Geochemical Analyses

Groundwater permeability (ks) was estimated using the slug test method (Bouwer and Rice 1976) with 20 s for the initial linear point to eliminate the drainage in the gravel pack. Groundwater table (GWT) depth was measured using an electrical dip metre. Samples for DO were collected in a 12-ml exetainer (Labco Ltd., Wycombe, UK), after slowly overflowing of approximately 10 ml excess water and closed immediately using double septum (butyl rubber + Teflon) stopper. Samples were submerged underwater in a polystyrene box, stored at 4 °C and analysed within 1 week. DO was measured by membrane inlet mass spectrometry (Kana et al. 1994). Groundwater pH, electrical conductivity and redox potential (Eh) were measured using a multiparameter probe (Troll 19500, In Situ Inc., USA). Groundwater was analysed for NO3–N and Br on DX-120 ion chromatography (Metrohm UK Ltd.). The dissolved organic carbon (DOC) was analysed using total organic carbon analyser (TOC-V cph/cpn, Shimadzu Corporation, Kyoto, Japan). Groundwater non-metallic ions, e.g., total oxidised N, nitrite, NH4+ and P, and reduced metals, e.g., Fe2+, Mn2+ and S2− were analysed with an Aquakem 600 discrete analyser (Aquakem 600A, Vantaa, Finland). Groundwater SO42− concentration was measured with a turbimetric method (Askew and Smith 2005).

2.8 Statistical Analyses

The measured denitrification rates were approximately log-normally distributed. Therefore, non-parametric Kruskal–Wallis H tests were performed to determine significant differences in groundwater denitrification rates among depths within each site. After significant differences were observed among depths, Mann–Whitney U tests (Ott 1993) were performed as a post hoc test to determine which depths were significantly different. Paired t tests (Ott 1993) were performed to determine significant differences in recovery (C/Co) between Br and SF6. Spearman rank order correlations were performed to determine significant correlations between groundwater denitrification rates and ambient DO, Eh, NO3–N, DOC and ks. All statistical analyses were performed on VSN International Ltd. (2011). All statistical differences were considered significant at p < 0.05 level.

3 Results

3.1 Groundwater Physico-Chemical Properties

Groundwater ambient physico-chemical properties related to denitrification differed among sites (Table 1). Mean NO3–N concentrations were significantly different between sites (p < 0.001). Considering the within site differences among various depths, NO3–N concentrations were significantly higher (p < 0.01) in S than at the I and B at grassland site but were similar at the arable site. Mean NH4+ concentrations were low at both sites, with being 0.14 and 0.02 mg L−1 at grassland and arable sites, respectively. Groundwater pH was near neutral in all depths at grassland but was higher at I compared with B at the arable site. Reduced Fe (Fe II) concentrations were higher at grassland than that at arable site (Table 1). Mean groundwater SO42− concentrations were significantly higher at arable site than at the grassland (p < 0.05) but were similar between depths at each site. Mean S2− concentrations were similar across sites and depths. Mean DO concentrations were significantly lower at the grassland site than at the arable site (Table 1). Mean DOC concentrations were significantly higher at grassland (2.6 ± 0.8 mg L−1), than at the arable site (0.9 ± 0.1 mg L−1) (p < 0.05). Interestingly, DOC was similar between depths at each site, whereas DO significantly decreased (p < 0.05) with depth at both sites (Table 1). The C/N ratios were significantly higher at grassland than at the arable site (data not shown). Irrespective of depths, C/N ratios ranged 1.2–20.5 and 0.10–0.14 at grassland and arable sites, respectively. Phosphorous (orthophosphate, PO43−) concentrations were below the detection limit in groundwater at both of these study sites (<0.005 mg L−1). The Eh at grassland (25–94 mV) site was lower compared with that at the arable site (107–178 mV) (Table 1). The arable site had a higher aquifer saturated hydraulic conductivity coupled with a deeper groundwater table than at the grassland (Table 1). Saturated hydraulic conductivity (ks) increased with the increase in groundwater depth (Table 1).

3.2 Assessment of Push–Pull Method for Deep Groundwaters

The predetermined ks value in each piezometer (mean 0.009 m day−1 ± 0.002 (standard error, SE) at grassland, mean 0.049 m day−1 ± 0.008 at arable) provided an insight into the potential incubation times for push–pull pre-test. However, push–pull pre-test at both sites revealed a significant influence of incubation time on the recovery of tracer (Br) injected into the piezometer (p < 0.001). Reducing the incubation time increased the tracer recovery from 9–30 % for the 12-h incubation to 30–80 % for the 3-h incubation. In the NO3 push–pull test, the percentage recovery of the two tracers used (Br and SF6) were similar (p > 0.05) to each other. Mean recovery of the Br and SF6 tracers did not differ significantly among groundwater depths within each site but differed between the two sites. Mean Br recoveries in the core plume (the first 2–4 L of the pull where recovery is the highest) after a 6-h incubation ranged from 43 % in B to 59 % in S at grassland and 39 % in B to 55 % in S at the arable site.

3.3 In Situ Denitrification Rates

Over the short incubation period (6 h), NO3 removal via denitrification was detected at both sites. Denitrification rates at grassland site (mean = 163 μg N kg−1 day−1 ± 153 (SE)) were significantly higher than that at the arable site (mean = 3.9 μg N kg−1 day−1 ± 2.0). Among depths within the grassland site (Fig. 3a), significantly higher denitrification rates were measured at I (mean = 470 μg N kg−1 day−1 ± 111); than S (mean = 10.9 μg N kg−1 day−1 ± 3.5) or B (mean = 9.2 μg N kg−1 day−1 ± 2.8). Similarly, denitrification rates in the three different depths at the arable site were significantly higher (p < 0.05) at I (6.4 μg N kg−1 day−1 ± 1.8) than S (3.8 μg N kg−1 day−1 ± 0.7) or B (1.4 μg N kg−1 day−1 ± 0.4) (Fig. 3b). Mean denitrification rates were equivalent to a weighted average of 3.92 and 0.09 mg NO3–N L−1 day−1, respectively, at the grassland and arable sites, which accounted for 24.5 and 0.33 % of the N input to the land. Denitrification rates individually in the S, I and B at grassland were equivalent to 0.2, 10.3 and 0.3 mg N L−1 d−1 which accounted for 1, 65 and 2 % of the N input, respectively. The coefficient of variations (CV) for denitrification rates between wells was 55, 115 and 109 % in the S, I and B, respectively, at the grassland and 117, 60 and 47 % in S, I and B at the arable site, respectively.
Fig. 3

Mean denitrification rates (N2O + N2) in a three different depths of groundwater at grassland (n = 3) and b at arable land (n = 3)

3.4 N2O Mole Fraction

The N2O/(N2O + N2) ratios were significantly higher at the arable site (mean = 0.28 ± 0.04) than at the grassland site (mean = 0.10 ± 0.02) (Fig. 4). Among the three depths, N2O/(N2O + N2) ratios were significantly higher in S and I than in B at the arable site. In contrast, they were lower in S than I and B at the grassland (Fig. 4a, b). In situ production of environmentally benign N2 was the dominant end product of denitrification and ranged from 89–93 % of the total denitrification gases at the grassland site, whereas at the arable site, it ranged from 62–85 % of the total denitrification gases.
Fig. 4

Mean N2O mole fraction, N2O/(N2O + N2) in a three different depths of groundwater at grassland (n = 3) and b at arable land (n = 3)

3.5 Relationships Between Denitrification Rates and Ambient Hydrogeochemical Conditions

Spearman rank order correlation between denitrification rates and ambient geochemical properties showed significantly negative correlations between denitrification rates and ambient DO (r = −0.52, p < 0.05), Eh (r = −0.52, p < 0.05), NO3–N concentrations (r = −0.69, p < 0.01) and saturated hydraulic conductivity (r = −0.50, p < 0.05). There was no significant correlation observed between denitrification rates and ambient DOC concentrations in groundwater. In addition, denitrification rates showed a positive correlation with reduced Fe (Fe II, r = 0.39, p < 0.05), SO42− (r = 0.32, p < 0.05) and NH4+ (r = 0.33, p < 0.05). A conceptual model showing the site hydrogeochemistry, groundwater denitrification and NO3–N pollution potential was presented in Fig. 5.
Fig. 5

A conceptual model showing site hydrogeochemistry vs. denitrification and nitrate (NO3–N) pollution potential

4 Discussion

4.1 Assessment of Push–Pull Method for Deep Groundwaters

Estimation of tracer recovery is very important for quantifying groundwater denitrification rates and to understand the decline in concentrations of denitrification end products by physical processes like advection, dispersion and diffusion. Both Br and SF6, being used in these sites, had similar rates of recovery in the NO3 push–pull test and indicated that there was no degassing loss of SF6 during the incubation and sampling. The similarities in the recovery of both tracers also enhance the confidence of estimating groundwater dissolved gas concentrations produced via denitrification during the incubation period. Bromide has been used as a tracer because in groundwater, it does not come into contact with vegetation, thus uptake by plant is minimised (Richards et al. 2005). However, either of the tracers can be used for investigating groundwater denitrification using the push–pull test. Only Br has been used as the conservative tracer in many riparian groundwater NO3 studies (Simmons et al. 1992; Nelson et al. 1995; Starr et al. 1996), and other in situ riparian studies (Addy et al. 2002; Clough et al. 2007; Kellogg et al. 2005) have used both Br and SF6 as conservative tracers.

Recovery rates in this study (5–22 m bgl) were relatively lower than the push–pull studies by Addy et al. (2002) and Kellogg et al. (2005). Both studies incubated the dosing solutions for variable times, e.g., 4 to 24 h (Kellogg et al. 2005) and 5 to 72 h (Addy et al. 2002). Their higher tracer recoveries were found at shallower depths, i.e., in 0.65 to 1.25 m and 0.65 to 3 m that provided a maximum recovery of 80 and 70 %, respectively. Our tracer recoveries were within the range found by Harrison et al. (2011); 42–54 % recovery in summer and 20–26 % in winter in two alluvial wetlands with minipiezometer to a depth of 0.5 m and incubated for 4 h. Low tracer recovery in our study is likely due to high advective dispersion and diffusion and low residence time in these aquifers which have sediments with larger and more connected secondary pores or preferential flow path via fracture/fissure (Buss et al. 2005; Misstear et al. 2009). Sedimentary rocks, e.g., Ordovician sediments, sandstones in the grassland site and limestones at the arable sites showed increased hydraulic conductivity with depth of aquifers. Solute movement follows piston flow model in subsoil, but in bedrock, it follows complex pattern of movement because bedrock might have both vertical and horizontal flow paths via fractures developed by glacial movement.

4.2 Variations in Groundwater Denitrification Rates

Denitrification rates were highest at I of the grassland site, higher than observed in the S. Our lower denitrification rates were within the range of shallow groundwater denitrification rates reported by Kellogg et al. (2005) (<1 to 330 μg N kg−1 day−1), Addy et al. (2002) (2.1 to 123.2 μg N kg−1 day−1) and Harrison et al. (2011) (<0.1 to 193 μg N kg−1 day−1), but our grassland I high value was higher than reported by these other in situ push–pull papers. Higher denitrification rates at I (10 m bgl) are in line with the findings of Weymann et al. (2010) who, from a laboratory incubation experiment, observed that NO3 removal in the autotrophic zone (6.5 to 7.0 m bgl) is much more intensive than shallow zone (1.5 to 4.0 m bgl).

Our results suggest that while denitrification is not ubiquitous in deep groundwaters, it can serve as substantial hotspots for groundwater N removal before its delivery to surface waters. Higher denitrification rates at the I indicate that denitrification is not limited to shallow groundwater; rather, it can occur in deep groundwaters. This notion is in contrast to the assumption of Van Drecht et al. (2003) who developed an empirical model with an assumption that denitrification is zero in deep groundwater. However, underestimation of denitrification rates may also occur because NO2 and NO production rates are not included in the calculation (Bollmann and Conrad 1997; Harrison et al. 2011; Istok et al. 1997).

Denitrification rates showed high spatial variability because groundwater hydrogeological properties that control denitrification are heterogeneous. The coefficients of variation of N2O concentrations between wells within each site ranged from 55–115 and 47–60 % at grassland and arable sites, respectively, and were similar to the coefficients of variation of N2O production found by other workers, in surface soils, e.g., 71–139 % (Mathieu et al. 2006), 78–122 % (Jahangir et al. 2011), 14–132 % (Ishizuka et al. 2005) and in shallow groundwater, e.g., 219 % (Von der Heide et al. 2008). This variation indicates that denitrification is likely to be an active process, as it is in top soil, of natural NO3 reduction in shallow to deep groundwaters. Moreover, high spatial variability of N2O production is consistent with the high spatial variability of groundwater DO (CV 120 %), Eh (CV 219 %) and DOC (CV 98 %), suggesting that NO3 in groundwater is being processed, and these properties can be the key indicators of groundwater denitrification. The in situ push–pull tests were only conducted during one season because dissolved N2O and N2 at the sites were previously observed to be similar throughout the year (Jahangir et al. 2012a).

4.3 Variations in N2O Mole Fractions

Higher N2O mole fractions at the arable site than that at the grassland might have occurred due to low N2O reduction rates at this site because high DO at this site might have reduced N2O reduction and thus increased its accumulation. Mean N2O mole fractions in the in situ measurements were comparable with those measured in a laboratory incubation of subsoil from the grassland site with values of 0.25 to 0.42 in 0–10 cm; 0.06 to 0.36 in 45–55 cm and 0.04 to 0.24 in 120–130 cm depths (Jahangir et al. 2012b). The N2O mole fraction in this study (0.07–0.38) was comparable with Harrison et al. (2011) who measured N2O/(N2O + N2) ratios of 0.02–0.21 in 0.5 m bgl in alluvial wetlands using the in situ push–pull method. Mean N2O mole fraction, calculated at each site, implies two possibilities: (1) the groundwater could be an important source of atmospheric N2O when it discharges to surface streams and rivers (Deurer et al. 2008) or diffused upwardly from water table to the atmosphere (Ueda et al. 1993) or (2) N2O can further be reduced to N2 (Weymann et al. 2008). Mean mole fractions 0.02 at grassland to 0.09 at the arable site from monthly measurements over 2 years (2009–2010) in these wells (Jahangir et al. 2012a) were lower than that of the measurements by in situ push–pull test, possibly because N2O might have been further reduced to N2 while passing through and from the sediments to the streams due to its longer residence times. However, another possible reason for higher N2O/(N2O + N2) ratios in the in situ study than that of the monitoring results of Jahangir et al. (2012a) could be the addition of NO3–N to groundwater by at least two times of the ambient concentration, as high NO3–N concentration can accelerate N2O production (Scholefield et al. 1997; Blackmer and Bremner 1978), inhibit N2O reduction (Simek and Cooper 2002) and eventually increase the N2O mole fraction. The monitoring results suggest that denitrification is more complete, resulting in lower N2O mole fractions, taking into consideration the travel time through aquifers which can take from months to years at these sites (Fenton et al. 2011).

4.4 Relationships Between Denitrification and Ambient Hydrogeochemical Conditions

The differences in denitrification rates between sites and depths may be explained by their contrasting hydrologic and geochemical conditions (Table 1). The ITRC (2002) highlighted that in situ hydrologic conditions (e.g., groundwater table, ks and hydraulic gradient), geochemistry (e.g., Eh, Fe II, DO and TOC) and microorganisms are important factors for bioremediation. The lower ks at grassland site favoured denitrification. In comparable study, Fenton et al. (2009) found that subsoil ks was negatively related to groundwater N2/Ar ratio. Fenton et al. (2009) measured saturated hydraulic conductivity in 17 wells in subsoil at grassland site by slug which ranged from 0.001 to 0.016 m day−1. These hydraulic conductivity values were comparable with the range of the present study. Fitzsimon and Misstear (2006) reported the hydraulic conductivity values of some low to moderate permeable tills in Ireland ranging from 0.0004 to 0.009 m day−1 which was within the range of the current study at the grassland site. The DO, being comparable in all depths at the grassland site, was lower than the arable site. The low DO and low Eh indicate the higher anaerobiocity of groundwater that could foster denitrification. Rivett et al. (2008) identified DO and electron donor concentrations and availability as the primary factors governing denitrification in groundwater. Böhlke et al. (2007) observed that <1.6 mg L−1 of DO was required for complete denitrification of NO3–N to N2. The higher DO and Eh at the arable site suggests that in situ denitrification may be either very low or zero under these conditions. The observed denitrification rates, though small at the arable site, could be attributed to either deriving from aerobic denitrification (Robertson et al. 1995) or through denitrification occurring in anaerobic microsites (Seitzinger et al. 2006). From groundwater monitoring results of hydrochemistry and dissolved gases (N2O and excess N2, called denitrified N2), higher NO3–N and lower N2O and N2 concentrations were previously observed at the arable site (Jahangir et al. 2012a) supporting this theory. On the same sites, Barrett et al. (2013) observed nir and nosZ abundance (these are the functional genes associated with nitrite and nitrous oxide reductase) of 13.5–4.6 × 103 and 9.8–18.3 × 102 (gene copy conc. L−1), indicating that microbial occurrence is unlikely to be a limiting factor for groundwater denitrification.

DOC enhances denitrification by reducing DO through aerobic respiration, releasing CO2 and as an electron donor for denitrifier community. Moreover, DOC was available to shallow groundwater and also the deep groundwater as there was no significant decline in DOC with depth from 5 to 22 m bgl. The lack of any significant correlation between DOC and denitrification rates may be due to the high spatial variabilities in DOC concentration (<1 to >10 mg L−1). In deep groundwaters, however, other electron donors, such as Fe minerals, can be of importance as denitrification rates showed positive correlation with reduced Fe, which was the highest at the I at grassland site. The oxidation of sulphide compounds (bound with Fe) under anaerobic conditions may release Fe (II) or Mn (Kolle et al. 1985). Negative correlations between denitrification and ambient NO3 concentration implies that low ambient NO3 existed in groundwater wells due to the occurrence of natural denitrification process that substantially reduced NO3 (Konrad 2007; Vogel et al. 1981; Weymann et al. 2008). In denitrification process, if reduced S is the electron donor, SO42− is formed (Rivett et al. 2008). The positive correlation between groundwater SO42− and denitrification rates might be contributed to the oxidation of sulphur in anaerobic environment where S2− (reduced S or metal bound S) might be an important electron donor (autotrophic denitrification). The NH4+, being observed mainly at grassland, showed positive correlation with denitrification rates because NH4+ generation via dissimilatory nitrate reduction to ammonium might have occurred in the anaerobic environment which is a requirement for denitrification.

5 Conclusions

The results of this study show that the push–pull method for groundwater denitrification rates using 15N-enriched NO3–N can be used in the deep groundwater systems. Low conservative tracer recovery may have underestimated denitrification estimates. The bedrock interface at the grassland site with low DO, Eh and high DOC demonstrates that deep groundwater can serve as a ‘hot spot’ for NO3 removal. Even where we observed low denitrification rates at the arable site with high DO, Eh and low DOC, its contribution to indirect N2O emissions should still be accounted for in global N2O budgets. The strong correlations between denitrification rates and hydrogeologic conditions suggest that modelling within geographical information systems may help to predict locations with substantial subsurface denitrification rates. These findings show important implications about the natural NO3–N attenuation capacity of groundwater beneath intensively managed grassland that reduces the risk of NO3–N delivery to the surface waters. In addition, N2O mole fractions from in situ measurements indicated that groundwater denitrification can reduce indirect N2O emissions to the atmosphere. Therefore, NO3–N reduction to N2O and to N2, while transported through groundwater to the receptors are simultaneous processes which balance net NO3–N delivery to surface waters and indirect N2O emissions to atmosphere.


The study was funded by the Department of Agriculture and Food, Ireland, through the Research Stimulus Fund Programme (grant RSF 06383) in collaboration with the Department of Civil, Structural and Environmental Engineering, The University of Dublin, Trinity College. The authors sincerely acknowledge the contribution of Mr. John Murphy in the field work.

Copyright information

© Springer Science+Business Media Dordrecht 2013