Application of the 15N gas-flux method for measuring in situ N2 and N2O fluxes due to denitrification in natural and semi-natural terrestrial ecosystems and comparison with the acetylene inhibition technique

. Soil denitriﬁcation is considered the most un-constrained process in the global N cycle due to uncertain in situ N 2 ﬂux measurements, particularly in natural and semi-natural terrestrial ecosystems. 15 N tracer approaches can provide in situ measurements of both N 2 and N 2 O simul-taneously, but their use has been limited to fertilized agro-ecosystems due to the need for large 15 N additions in order to detect 15 N 2 production against the high atmospheric N 2 . For 15 N–N 2 analyses, we have used an “in-house” laboratory designed and manufactured N 2 preparation instrument which can be interfaced to any commercial continuous ﬂow isotope ratio mass spectrometer (CF-IRMS). The N 2 prep unit has gas puriﬁcation steps and a copper-based reduction furnace, and allows the analysis of small gas injection volumes (4 µL) for 15 N–N 2 analysis. For the analysis of N 2 O, an automated Tracegas Preconcentrator (Isoprime Ltd) coupled to an tracer rate kg 15 N ha − 1 detectable ﬂux h

Abstract.Soil denitrification is considered the most unconstrained process in the global N cycle due to uncertain in situ N 2 flux measurements, particularly in natural and semi-natural terrestrial ecosystems. 15N tracer approaches can provide in situ measurements of both N 2 and N 2 O simultaneously, but their use has been limited to fertilized agroecosystems due to the need for large 15 N additions in order to detect 15 N 2 production against the high atmospheric N 2 .For 15 N-N 2 analyses, we have used an "in-house" laboratory designed and manufactured N 2 preparation instrument which can be interfaced to any commercial continuous flow isotope ratio mass spectrometer (CF-IRMS).The N 2 prep unit has gas purification steps and a copper-based reduction furnace, and allows the analysis of small gas injection volumes (4 µL) for 15 N-N 2 analysis.For the analysis of N 2 O, an automated Tracegas Preconcentrator (Isoprime Ltd) coupled to an IRMS was used to measure the 15 N-N 2 O (4 mL gas injection volume).Consequently, the coefficient of variation for the determination of isotope ratios for N 2 in air and in standard N 2 O (0.5 ppm) was better than 0.5 %.The 15 N gas-flux method was adapted for application in natural and semi-natural land use types (peatlands, forests, and grasslands) by lowering the 15 N tracer application rate to 0.04-0.5 kg 15 N ha −1 .The minimum detectable flux rates were 4 µg N m −2 h −1 and 0.2 ng N m −2 h −1 for the N 2 and N 2 O fluxes respectively.Total denitrification rates measured by the acetylene inhibition technique in the same land use types correlated (r = 0.58) with the denitrification rates measured under the 15 N gas-flux method, but were underestimated by a factor of 4, and this was partially attributed to the incomplete inhibition of N 2 O reduction to N 2 , under a relatively high soil moisture content, and/or the catalytic NO decomposition in the presence of acetylene.Even though relatively robust for in situ denitrification measurements, methodological uncertainties still exist in the estimation of N 2 and N 2 O fluxes with the 15 N gas-flux method due to issues related to non-homogenous distribution of the added tracer and subsoil gas diffusion using open-bottom chambers, particularly during longer incubation duration.Despite these uncertainties, the 15 N gas-flux method constitutes a more reliable field technique for large-scale quantification of N 2 and N 2 O fluxes in natural terrestrial ecosystems, thus significantly improving our ability to constrain ecosystem N budgets.

Introduction
There has been a renewed interest recently in developing new or enhancing existing measurement approaches for improving our ability to constrain dinitrogen (N 2 ) fluxes due to denitrification in terrestrial ecosystems (Kulkarni et al., 2014;Lewicka-Szczebak et al., 2013;Wang et al., 2011;Yang et Published by Copernicus Publications on behalf of the European Geosciences Union.F. Sgouridis et al.: Application of the 15 N gas-flux method al., 2014).Denitrification, the reduction within soils of nitrogen oxides (NO − 3 and NO − 2 ) to NO, N 2 O, and ultimately N 2 gas, constitutes the most important mechanism for the removal of reactive nitrogen (N r ) in terrestrial ecosystems (Galloway et al., 2008;Groffman, 2012).Despite its importance, denitrification is considered the most un-constrained process in the global N cycle (Groffman, 2012;Kulkarni et al., 2008) due to uncertainties in N 2 flux estimations that are likely leading to underestimations of denitrification rates at multiple scales (Butterbach-Bahl et al., 2013).Considering contemporary atmospheric N deposition rates globally including the UK (Dore et al., 2012;Galloway et al., 2008;Payne, 2014), the available N r pool in soils may be greater than the capacity of denitrification for its removal, with important consequences of chronic N enrichment of natural terrestrial ecosystems (Galloway et al., 2008;Limpens et al., 2003).Moreover, nitrous oxide (N 2 O), an obligate intermediate of denitrification, is a potent greenhouse gas involved in the breakdown of stratospheric ozone (Ravishankara et al., 2009).Therefore, a reliable estimation of the relative magnitude of the major denitrification end products (N 2 + N 2 O) in soils is crucial in evaluating the role of denitrification as an N r sink (Kulkarni et al., 2008).
N 2 comprises ∼ 78 % of the atmosphere and thus it is extremely difficult to measure small N 2 fluxes from soil against this high background, particularly in natural terrestrial ecosystems (Groffman et al., 2006).Available methods for measuring both N 2 and N 2 O are limited and can be categorized into the direct flux and 15 N isotope tracer methods (Kulkarni et al., 2014), whilst micrometeorological approaches (Eddy covariance) are impossible in the N 2 -rich atmosphere (Felber et al., 2012).The gas-flow soil core method (Burgin and Groffman, 2012;Butterbach-Bahl et al., 2002;Scholefield et al., 1997;Wang et al., 2011) allows the direct measurement of N 2 flux (without the addition of any substrate such as nitrate) from intact soil cores where the soil atmosphere is replaced by a mixture of He / O 2 .However, despite the high precision of the technique, cores still need to be extracted from the field and conditioned over lengthy periods of time for the complete removal of N 2 from the soil atmosphere.This method is therefore time and resource intensive, which precludes its application to intensive temporal and large spatial scales (Kulkarni et al., 2014).Moreover, the gas-flow soil core method cannot discriminate between sources of N 2 O, thus overestimating the denitrification product ratio N 2 O / (N 2 + N 2 O; Butterbach-Bahl et al., 2013;Morse et al., 2015).The acetylene inhibition technique (AIT) is also a direct flux method that exploits the ability of acetylene (C 2 H 2 ) at high concentrations (10 % v/v) to inhibit the reduction of N 2 O to N 2 (Tiedje et al., 1989); thus, total denitrification (N 2 + N 2 O) is measured in C 2 H 2 amended soil cores in situ, whilst N 2 flux is estimated indirectly by difference from un-amended soil cores.Despite its simplicity and cost-effectiveness, the AIT is becoming increasingly unpopular due its several limitations (Groffman et al., 2006), of which the catalytic decomposition of NO in the presence of C 2 H 2 under oxic or suboxic conditions in the field (Bollmann and Conrad, 1997;Nadeem et al., 2013), in particular, precludes its use for reliable estimates of in situ denitrification rates (Felber et al., 2012).
The 15 N gas-flux method (Mosier and Klemedtsson, 1994) has the advantage of providing in situ measurements of both N 2 and N 2 O simultaneously, thus allowing its application over large temporal and spatial scales.It requires the addition of a 15 N-labelled tracer in a soil enclosure in the field which is subsequently covered by a chamber while the chamber headspace is progressively enriched with 15 N-N 2 and 15 N-N 2 O produced by denitrification (Stevens and Laughlin, 1998).Assuming that both N 2 and N 2 O originate from the same uniformly labelled soil NO − 3 pool (Stevens and Laughlin, 2001), the true denitrification product ratio can be more accurately estimated as opposed to the direct flux approaches (Bergsma et al., 2001).Field applications of the 15 N gas-flux method so far have been limited to fertilized agro-ecosystems (Baily et al., 2012;Cuhel et al., 2010;Graham et al., 2013) and more recently restored peatland soils (Tauchnitz et al., 2015) with high 15 N tracer application rates (between 10 and 200 kg N ha −1 ), with the exception of Kulkarni et al. (2014), who have measured denitrification rates in northern hardwood forests of the USA by adding tracer amounts of 15 Nlabelled nitrate, and Morse and Bernhardt (2013), who applied the same technique in intact soil cores collected from mature and restored forested wetlands in North Carolina, USA.These recent studies hold much promise that the 15 N gas-flux method can be applied to a range of natural and semi-natural terrestrial ecosystems, allowing the quantification of the relative magnitude of N 2 and N 2 O fluxes due to denitrification from these under-represented ecosystems.
Natural and semi-natural terrestrial ecosystems in the UK (i.e.peatlands, heathlands, acid grasslands, deciduous and coniferous forests), where there is no fertilizer use and the impact from grazing and commercial forestry is minimal (Mills et al., 2013), along with improved and unimproved grasslands (grazed and/or fertilized) constitute approximately 49 and 85 % of rural land use cover in England and Wales respectively (Morton et al., 2011).Unlike arable agriculture, these land use types have been poorly investigated for their role in N r loss through denitrification.
The major challenge in measuring 15 N-N 2 at near-natural abundance levels is the possibility of interference at m/z 30 ( 30 N 2 ) due to the reaction of oxygen in the ion source with N and the formation of NO + ions that also have m/z 30 (Stevens et al., 1993).Commonly, this issue is addressed in a continuous flow isotope ratio mass spectrometer (CF-IRMS) with the inclusion of a copper (Cu) oven for reducing O 2 in the gas sample (Russow et al., 1996).Recently, it has been suggested that the interference at m/z 30 can be further reduced by including a molecular sieve column in gas chromatograph IRMS (GC-IRMS) systems to not only separate N 2 and O 2 in the gas sample, but also to quantitatively remove O 2 and other trace gases such as carbon monoxide (Lewicka-Szczebak et al., 2013;Yang et al., 2014).We hypothesize that the precision for m/z 30 determination can be greatly improved by using a custom-built preparative unit for the removal of H 2 O, CO 2 , N 2 O, NO + and CO, a device which also permits the micro-scale injection of volumes of < 5 µL.These injection volumes are much smaller than have previously been reported in the literature.
Studies that have compared the 15 N gas-flux method with the AIT in the field are rare and have exclusively focused on highly fertilized agro-ecosystems with moderate to low soil moisture contents (Aulakh et al., 1991;Mosier et al., 1986;Rolston et al., 1982).These studies have measured comparable denitrification rates with both field techniques, although the relatively low soil moisture contents have probably allowed greater diffusion of C 2 H 2 to the anaerobic microsites where denitrification occurs (Malone et al., 1998), whilst the high nitrate application rates have probably favoured nitrate reduction over N 2 O reduction (Dendooven and Anderson, 1995), resulting in high denitrification rates from the AIT.Conversely, laboratory studies have shown that the AIT significantly underestimates total denitrification compared to the 15 N tracer approach (Yu et al., 2010) and the direct N 2 flux approach (Qin et al., 2012) due to the incomplete inhibition of N 2 O reduction to N 2 by C 2 H 2 in wet soils (Yu et al., 2010) or in soils with low nitrate content, where N 2 O reduction is more energetically favourable (Qin et al., 2013(Qin et al., , 2014)).A comparison of the 15 N gas-flux method with the AIT under in situ conditions across a range of natural and semi-natural terrestrial ecosystems has not been attempted before.It can provide valuable insights in terms of the validity and applicability of the two field techniques for measuring denitrification rates across broad spatial and temporal scales.
The objectives of the present study were (1) to determine the precision and suitability of our preparative-IRMS instrumentation for measuring 15 N-N 2 and 15 N-N 2 O at low enrichment levels, (2) to adapt the 15 N gas-flux method for application across natural and semi-natural terrestrial ecosystems, and (3) to compare the validity and applicability of the 15 N gas-flux method with the AIT for measuring in situ denitrification rates.

IRMS system
For N 2 gas isotopic analysis we used an Isoprime isotope ratio mass spectrometer (Isoprime Ltd, UK, Wythenshawe) coupled to an in-house built N 2 preparative interface (Fig. 1).Headspace gas (4 µL) was manually injected with a gas-tight syringe (SGE Analytical science) into the preparative interface via an open split.Prior to its introduction into the IRMS, the sample was treated as follows: (a) dried by passing through Mg(ClO 4 ) 2 (Elemental Microanalysis Ltd, Devon, UK), (b) CO 2 removed with 0.7-1.2mm Carbosorb (Elemental Microanalysis Ltd, Devon, UK), (c) N 2 O cryogenically trapped under liquid nitrogen, and (d) O 2 removed over a copper-packed reduction furnace heated at 600 • C. The N 2 was then directed towards the triple collectors of the isotope ratio mass spectrometer where m/z 28, m/z 29 and m/z 30 mass ions were measured.Mass / charge ratios for the m/z 28, m/z 29, and m/z 30 nitrogen ( 28 N 2 , 29 N 2 , and 30 N 2 ) were recorded for each sample at a trap current of 300 µA.Instrument stability checks were performed prior to each analysis by running a series of 10 reference pulses of N 2 (BOC special gases) until a standard deviation of δ 15 N better than 0.05 ‰ was achieved.Additionally, 10 consecutive injections (4 µL) of atmospheric air were analysed prior to the analysis of actual samples.The precision of the instrument was better than δ 15 N 0.08 ‰ in all quality control tests.
Nitrous oxide was analysed using modified headspace methods described for the analysis of nitrogen gas above.Headspace gas (ca. 4 mL) was injected into a TraceGas ™ Preconcentrator coupled to an Isoprime ™ IRMS (GV instruments Ltd, UK) whereupon the sample was directed through a series of chemical traps designed to remove H 2 O and CO 2 .The N 2 O was cryogenically trapped under liquid nitrogen.analysis, trace gas N 2 O measurements were made on three 100 mL flasks containing atmospheric air collected from outside the stable isotope laboratory.δ 15 N precisions using the TraceGas Preconcentrator and Isoprime IRMS were better than 0.3 ‰ respectively at the 600 µA trap current.

Field application of the 15 N gas-flux and AIT methods
In situ measurements of N 2 and N 2 O were made using static chambers according to the 15 N gas-flux method (Mosier and Klemedtsson, 1994).Five plots were randomly established in June 2013 in each of four study sites in the Ribble-Wyre River catchments (area 1145 km 2 ; north-western England; 53 • 59 99 N, 2 • 41 79 W).The study sites were a heathland (R-HL), a deciduous woodland (R-DW), an unimproved grassland (R-UG), and an improved grassland (R-IG).In August 2013, four more study sites were tested in the Conwy River catchment (area 345 km 2 ; northern Wales; 52 • 59 82 N, 3 • 46 06 W) following a similar sampling design.These sites were an acid grassland (C-UG), an ombrotrophic peat bog (C-PB), a mixed deciduous and coniferous woodland (C-MW), and an improved grassland (C-IG).Further details on the location, land management status, and major soil properties for all study sites can be found in Sgouridis and Ullah (2014).
In each plot a round PVC collar (basal area 0.05 m 2 ; chamber volume 4 L) was inserted into the soil at ca. 10 cm depth (15 cm for the R-HL and C-PB plots) 2-4 weeks before the measurement date.The collars were open at the bottom to maintain natural drainage and root growth during the measurements.The natural vegetation cover at the soil surface of each installed collar remained unchanged.The PVC collars were fitted with a circular groove of 25 mm depth to fit in an acrylic cylindrical cover (chamber) providing a gastight seal when filled with water (Ullah and Moore, 2011).The gas leak rate from the chamber was determined in the laboratory by placing the sealed collar and chamber over a tray of water, injecting CH 4 (10 ppm) and determining the change in CH 4 concentration within the chamber headspace over time (Yang et al., 2011).The CH 4 concentration change within 24 h was negligible, with the relative standard deviation (RSD) being < 5 %.We did not use a vent tube for pressure equilibration, as suggested by Hutchinson and Mosier (1981), in our chamber design, which could have diluted the chamber headspace with atmospheric N 2 , as part of our effort to increase the probability of a detectable 15 N-N 2 signal in the chamber headspace.Instead chambers were covered with reflective foil for minimizing temperature increase within the chamber headspace during the incubation period (Ullah and Moore, 2011).Labelled K 15 NO − 3 (98 at % 15 N, Sigma-Aldrich) was applied in each plot via 10 injections of equal volume through a grid (4 × 6 cm) using custom-made 10 cm long lumber needles (15 cm for the R-HL and C-PB plots) attached to a plastic syringe (Ruetting et al., 2011).
The 15 N tracer was delivered as the needle was pushed into the soil from the surface up to 10 or 15 cm depth, aiming to achieve as uniform as possible labelling of the soil volume enclosed by the collar, as required by the 15 N gas-flux method (Mosier and Klemedtsson, 1994).The volume and concentration of the labelled K 15 NO − 3 tracer solution was determined from measurements of soil nitrate and moisture content, as well as bulk density adjacent to each plot made during the installation of the collars (Morse and Bernhardt, 2013).Lower application rates (< 0.1 kg N ha −1 ) were administered to natural study sites (e.g.peat bog, heathland) and higher rates (< 1 kg N ha −1 ) administered to semi-natural ones (e.g.unimproved and improved grasslands).The tracer solution (50-200 mL) was adjusted between 3 and 5 % of the ambient volumetric water content (see Supplement Table S1 for detailed data from each sampling plot).It should be noted that no time was allowed for the equilibration of the added tracer solution in the soil enclosure to avoid significant loss of the low amount of added nitrate via plant uptake.
Following the 15 N tracer application, the collars were covered with the acrylic chamber fitted with a rubber septum for gas sampling.Two sets of gas samples (20 mL each) were collected with a gas-tight syringe (SGE Analytical Science) through the septum of the chamber cover at T = 1 h, T = 2 h, and T ≈ 20 h after the tracer injection, while a T = 0 h sample was collected immediately after tracer injection above the plot surface before fitting the chamber cover.The gas samples were transferred into pre-evacuated (< 100 Pa) 12 mL borosilicate glass vials with butyl rubber septa (Exetainer vial; Labco Ltd., High Wycombe, United Kingdom) for storage under positive pressure and were analysed within 8 weeks from collection without any significant change in the gas concentration (Laughlin and Stevens, 2003).
Adjacent to each PVC collar in each plot, two intact soil cores (50 mm I.D., 15 cm long) were extracted from 10 cm depth, leaving the top 5 cm void as a headspace volume.The cores were capped on both ends with the top cap fitted with a rubber septum for gas sampling.One set of cores was amended with pure C 2 H 2 with 5 mL injected through the septum directly in the middle of the soil core before 10 % of the headspace was also replaced with pure C 2 H 2 .The second set of cores was not amended with C 2 H 2 and both cores were placed back in the ground where they came from.Gas samples (5 mL) were collected with a gas-tight syringe (SGE Analytical Science) through the septa of the cores at T = 1 h and T = 2 h after amendment with acetylene.The gas samples were transferred into pre-evacuated (< 100 Pa) 3 mL borosilicate glass vials with butyl rubber septa (Exetainer vial; Labco Ltd., High Wycombe, UK) for storage under positive pressure.

Flux calculations
The 15 N content of the N 2 in each 12 mL vial was determined using the IRMS system described above and the ra-Table 1. Measured ratios of R29 and R30 for N 2 in ambient air (n = 10), ratios of R45 and R46 in standard N 2 O gas (0.5 ppm concentration, n = 15), and 15 N at % abundance calculated from the respective ratios for both gases.SD: standard deviation; CV: coefficient of variation.
For calculating the total N 2 flux from a uniformly labelled soil nitrate pool when both R29 and R30 are measured, the "non-equilibrium" equations were applied as described by Mulvaney (1984) for estimating first the 15 N fraction in the soil NO − 3 denitrifying pool ( 15 X N ) as 15 where R29 and R30 are the difference between R29 and R30 respectively between enriched (T = 1, 2, and 20 h) and reference samples (T = 0 h).Subsequently, the 15 X N allows the quantification of the fraction of the N 2 evolved from the 15 N-labelled pool (d) using either R30 or R29: Using d and the concentration of [N 2 ] (µg N) in the chamber headspace, the evolved N 2 from the soil pool was calculated: The N 2 flux was then calculated using linear regression between the maximum evolved N 2 and the respective incubation time per plot surface area, and was expressed in µg N m −2 h −1 representing the total N 2 flux from the mixture of the 15 N-labelled tracer and the soil N at natural abundance (Stevens and Laughlin, 1998).
The 15 N content of the N 2 O in the same 12 mL vials as well as the ratios R45 ( 45 N 2 O / 44 N 2 O) and R46 ( 46 N 2 O / 44 N 2 O) were measured in both enriched (T = 1, 2, and 20 h) and reference samples (T = 0 h).The application of the "non-equilibrium" equations to N 2 O is analogous to N 2 after correcting for the naturally occurring oxygen isotopes (Bergsma et al., 2001).Therefore, the ratios R45 and R46 were converted to ratios of R29 and R30 respectively by applying the following equations: where for R17 ( 17 O / 16 O) the value 0.000373 was used and for R18 ( 18 O / 16 O) the value 0.0020052 was used (Bergsma et al., 2001).There was no significant instrumental drift for the ratios R45 and R46 over time.The MDC was defined, for the converted R29 and R30, with repeated automatic analyses of 0.5 ppm N 2 O standards (n = 15) as 3.4 × 10 −5 and 2.9 × 10 −5 respectively.The second set of gas samples collected at the same time in the field was analysed for total N 2 O on a GC-µECD (7890A GC Agilent Technologies Ltd., Cheshire, UK) and the concentration of [N 2 O] (µg N) was used in Eq. ( 5) to calculate the N 2 O flux due to denitrification of the mixture of the 15 N-labelled tracer and the soil N and expressed in µg N-N 2 O m −2 h −1 .Assuming that the N 2 O originates from the same uniformly labelled pool as N 2 , the 15 X N from N 2 O was used to estimate d for N 2 using either R30 (Eq. 3) or R29 (Eq.4), thus lowering the limit of detection for N 2 (Stevens and Laughlin, 2001) and allowing measurement of N 2 gas flux from natural terrestrial ecosystems at low 15 N-tracer application rates.Gas samples collected from the intact soil cores with or without acetylene amendment were analysed for N 2 O on a GC-µECD (7890A GC Agilent Technologies Ltd., Cheshire, UK) and for CO 2 on a GC-FID (7890A GC Agilent Technologies Ltd., Cheshire, UK) and flux rates were determined by linear regression between 0 and 2 h.The instrument precision was determined from repeated analyses of 6 ppm N 2 O F. Sgouridis et al.: Application of the 15 N gas-flux method and 200 ppm CO 2 standards respectively (n = 8), and the RSD was < 1%.

Statistical analysis
Using factor analysis on selected soil physico-chemical properties, the samples from the eight field sites were grouped into three broad land use types: organic soils (C-PB, C-UG, R-HL), forest soils (C-MW, R-DW), and grassland soils (C-IG, R-UG, R-IG) according to Sgouridis and Ullah (2014).All subsequent statistical analyses were performed on the broad land use types rather than individual field sites.The data were analysed for normality and homogeneity of variance with the Kolmogorov-Smirnov test and the Levene statistic respectively and logarithmic transformations were applied as necessary.One-way ANOVA combined with Hochberg's GT2 post hoc test for unequal sample sizes or the Games-Howell post hoc test for unequal variances was performed for comparing the variance of the means between land use types for all gas fluxes.The nonparametric Kruskal-Wallis test was used to compare mean flux rates between incubation time intervals.Pearson correlation was used between log-transformed flux rates.Comparisons between the 15 N gas-flux and AIT techniques were made with an independent samples t test.All statistical analyses were performed using SPSS ® 21.0 for Windows (IBM Corp., 2012, Armonk, NY).

IRMS system evaluation
The precision of the IRMS systems was evaluated using repeated analyses of ambient air samples for N 2 (n = 10) injected manually in one batch and repeated analyses of N 2 O gas standard at natural abundance and 0.5 ppm concentration (n = 15) using automated injections.The mean measured ratios of R29 and R30 for N 2 and of R45 and R46 for N 2 O are shown in Table 1.Measurement precision was defined as the coefficient of variation (%), and it was lower for R29 compared to R30 and lower for R45 compared to R46, but still less than 0.5 % for all four measured ratios.We estimated the 15 N atom% abundance for both gases as per Yang et al. (2014), and the precision was less than 0.01 % for N 2 in air and 0.26 % for standard N 2 O at natural abundance.The mean measured R30 (5.16 × 10 −5 ) was higher than the theoretical value of 1.35 × 10 −5 for N 2 in ambient air, suggesting some interference at m/z 30 potentially due to the formation of NO + ions in the ion source of the mass spectrometer despite the inclusion of the Cu reduction oven.The contribution of NO + ions (R30 measured-R30 theoretical) was 3.81 × 10 −5 , whilst the ratio of R30 theoretical / R30 measured was 0.26.Correcting the R30 ratio for the contribution of NO + ions results in a lower "true" precision for the R30 (CV = 1.67 %).

Field application of the 15 N gas-flux method
The 15 N tracer application rate was variable between land use types and ranged between 0.03 and 1 kg 15 N ha −1 , while it was lower in the case of the organic soils and higher for the woodland and grassland soils (Table 2).Based on the soil nitrate content on the day of the tracer amendments (Table 2), the estimated enrichment of the total soil nitrate pool was on average between 13 and 25 15 N at % (detailed data on the 15 N tracer application per field site are shown in Table S2).
The 15 N fraction in the denitrifying pool ( 15 X N ), as calculated from the measured isotopic ratios of the N 2 O after 1 h of incubation using Eq. ( 2), ranged between 65 and 93 15 N at %.The average change in the 15 X N with incubation time, indicated by the slope shown in Table 2, was not different from 0 in the case of the organic (t test; t = 0.520, df = 18, p > 0.05) and grassland soils (t test; t = 0.047, df = 28, p > 0.05), whilst it was significantly below 0 for the woodland soils (t test; t = 2.917, df = 18, p < 0.05).Separating the woodland soils into C-MW and R-DW sites, only the former displayed a significant negative slope of 15 X N with incubation time (t test; t = 3.306, df = 8, p < 0.05), suggesting N 2 O production from a second nitrate pool, possibly nitrate produced from the oxidation of NH + 4 via nitrification, in the C-MW.In cases where the 15 X N could be calculated from the N 2 isotope ratio data (woodland and grassland soils; data shown in Table S3), this was not significantly different from their respective 15 X N calculated from the N 2 O isotope ratio data (t test; t-WL = 0.929, df = 12, p > 0.05; t-GL = 1.511, df = 20, p > 0.05).
The mean evolved amount of N 2 and N 2 O gases due to denitrification in each land use type increased with increasing incubation time (Fig. 2).The increase in the evolved N 2 was statistically significant after 20 h incubation in GL (ANOVA; F = 19.8,p < 0.01), whilst due to the high variability among plots, shown by the large error bars at 20 h incubation in Fig. 2a, it was not significant for the OS and WL soils.The amount of N 2 O accumulated after 20 h (Fig. 2b) was significantly higher than at the previous time points for all land use types (ANOVA; F OS = 4.6, F WL = 5.1, F GL = 14.7, p < 0.05).However, this pattern was not consistent in every sampling plot (data presented in Tables S4 and S5); for example, in C-MW, the highest N 2 accumulations were observed after the first or second hour of incubation, whilst in most cases the increase in N 2 and N 2 O concentrations was not linear throughout the incubation period (Tables S4 and S5).This suggested a complex temporal sequence of events, which was not consistent between plots among the different land use types, probably as a result of complex interactions between environmental controls of denitrification and the length of the incubation period (details below).Consequently, the N 2 flux rate decreased with increasing incubation time (Fig. 3a), and this decrease was significant between each time interval in the OS (Kruskal-Wallis; χ 2 = 11.35,p = 0.003), between 1 and 20 h in the WL (Kruskal-Wallis; χ 2 = 10.78,p = 0.005), and between 1 and 2 h in the GL (Kruskal-Wallis; χ 2 = 10.10,p = 0.006).Conversely, the N 2 O flux rates increased between the first and second hour of incubation (Fig. 3b), followed by a decrease after 20 h, albeit the mean differences between time intervals were not sta- 11.2 < 0.001 CO 2 production (C 2 H 2 amended cores) 11.7 < 0.001 tistically significant in any land use type (Kruskal-Wallis; χ 2 OS = 3.58, χ 2 WL = 3.47, χ 2 GL = 3.01, p > 0.05).The N 2 flux ranged between 2.4 and 416.6 µg N m −2 h −1 and was significantly different among land use types based on 20 h incubation duration for comparison purposes (Table 3).The grassland soils showed on average 3 and 14 times higher denitrification rates than the woodland and organic soils respectively (Fig. 4a).A similar pattern was observed for the N 2 O flux due to denitrification (range: 0.003-20.8µg N m −2 h −1 ), with the grassland soils emitting on average 14 and 120 times more N 2 O than the woodland and organic soils respectively (Fig. 4b), whilst the N 2 O flux was on average 20 to 200 times lower than the N 2 flux among land use types.Consequently, the denitrification product ratio N 2 O / (N 2 + N 2 O) was low, ranging between 0.03 and 13 %, and was highest in the GL and similar between the WL and OS (Fig. 4c).The change in the denitrification product ratio with incubation time was evaluated in each sampling plot where both N 2 and N 2 O fluxes were available (data shown in Table S6).Generally, there was no consistent pattern between individual sampling plots, with the exception of the grassland soils, where the maximum product ratio was observed after 2 h of incubation (ANOVA; F = 6.11, p < 0.05).This was www.biogeosciences.net/13/1821/2016/Biogeosciences, 13, 1821-1835, 2016 an indication of some reduction of the denitrification-derived N 2 O to N 2 during the extended closure period (up to 20 h) in the grassland soils.

Comparison with the AIT
The total denitrification rate measured from the C 2 H 2 amended intact soil cores in the same land use types ranged between 0.5 and 325.2 µg N m −2 h −1 and correlated positively with the total denitrification rate (N 2 and N 2 O fluxes combined) measured with the 15 N gas-flux method (Pearson; r = 0.581, n = 25, p < 0.01) following a similar trend among land use types, albeit with only the OS being significantly lower than the grassland and woodland soils (Table 3).The AIT denitrification rates were between 3 and 5 times lower than the total denitrification from the 15 N gas-flux (Fig. 5a), with the difference being significant in woodland (t test; t = 3.914, df = 18, p < 0.01) and grassland (t test; t = 3.521, df = 25, p < 0.01) soils.The total N 2 O flux measured from the un-amended intact soil cores ranged between 0.15 and 86.6 µg N m −2 h −1 and was between 1 and 3 times lower than the total denitrification rate from the C 2 H 2 amended cores.There were no significant differences between bulk N 2 O fluxes measured with the static chambers and the un-amended intact soil cores (Fig. 5b), which indicated that total N 2 O emissions were comparable between the two field techniques.Consequently, estimating the denitrification product ratio from the un-amended and C 2 H 2 amended intact soil cores resulted in significantly higher ratios compared to the 15 N gas-flux approach (Fig. 5c), which were on average between 50 and 60 % and not significantly different among land use types (Table 3).
The mean CO 2 production rate was similar irrespective of whether it was measured in static chambers, in C 2 H 2 amended or un-amended intact soil cores (Fig. 6), indicating that soil respiration (including both microbial and plant respiration) was not affected by the measurement technique.

IRMS system evaluation
The precision of our trace gas isotope ratio mass spectrometer (TG-IRMS) for manual analysis of 15 N-N 2 in gas samples was comparable for both the R29 and R30 ratios to the recently developed gas chromatograph-IRMS (GC-IRMS) systems that included a combination of a copper reduction oven and a molecular sieve (Lewicka-Szczebak et al., 2013) or only a molecular sieve (Yang et al., 2014) for the removal of O 2 from the samples.This was achieved while injecting a trace amount of headspace gas sample (4 µL), which is less than half of what is used by Lewicka-Szczebak et al. (2013) and 10 times less than the required sample volume by Yang et al. (2014).Furthermore, the interference at m/z 30 by NO + ions was reduced by an order of magnitude (3.81 × 10 −5 ) compared to the value (1.6 × 10 −4 ) reported by Lewicka-Szczebak et al. (2013).Consequently, correcting the R30 ratio for the NO + ion interference led to a CV value of < 2 %, which was significantly lower than the precision reported for natural abundance samples in previous studies (Lewicka-Szczebak et al., 2013;Russow et al., 1996;Stevens et al., 1993), thus constituting a significant improvement in m/z 30 determination in N 2 gas samples with low 15 N enrichment.However, the correction of the R30 ratio is only useful for estimating the "true" instrument precision for m/z 30 and is not necessary for calculating N 2 fluxes as shown by Lewicka-Szczebak et al. (2013), unless using the mathematical formulations of Spott and Stange (2007).
The TraceGas ™ Preconcentrator IRMS system used for 15 N-N 2 O analysis displayed similar precision for the deter-  mination of R45 and R46 in standard N 2 O gas at circa ambient concentration to a similar system used by Bergsma et al. (2001) while injecting only 4 mL of gas sample as opposed to 0.5 L used by Bergsma et al. (2001).When expressed in delta values (δ 15 N), the precision of our system was better than 0.05 ‰, which is significantly better than the respective precisions reported in Lewicka-Szczebak et al. (2013) and Yang et al. (2014), but comparable to Well et al. (1998).Therefore, the analytical precision achieved for both the 15 N-N 2 and 15 N-N 2 O analyses, using smaller gas sample volumes than previously reported, allowed us to quantify in situ N 2 and N 2 O fluxes with low tracer addition under field conditions.

Field application of the 15 N gas-flux method
The average 15 N tracer application rate (0.04-0.5 kg 15 N ha −1 or 0.4-1.2mg 15 N kg −1 dry soil) across land use types was 1 to 2 orders of magnitude lower than previous applications of the 15 N gas-flux method in highly fertilized agricultural systems (Baily et al., 2012;Bergsma et al., 2001;Cuhel et al., 2010;Graham et al., 2013) and in restored peatland soils (Tauchnitz et al., 2015).The estimated enrichment of the total soil NO − 3 pool was variable (2-40 15 N at %, Table S2) and this wide range was due to the fact that the tracer concentration was calculated based on the previous campaign's soil nitrate data, which in some cases did not reflect the soil nitrate content on the day of the tracer application a month later.It should be noted that the soil nitrate enrichment levels reported in this study correspond to the high end of the average soil NO − 3 pool enrichment (10-15 15 N at %, Table S2) for the period April 2013 to October 2014, which is presented in a separate publication (Sgouridis and Ullah, 2015).To our knowledge, only Kulkarni et al. (2014) have applied the 15 N gas-flux method in the field with soil nitrate enrichment levels (5 15 N at %) lower than in our study, but this had as a consequence poorly detected 15 N-N 2 fluxes.Nevertheless, for the organic soils the average tracer application rate corresponded to current estimates of daily atmospheric N deposition (0.05 kg N ha −1 d −1 ) in the UK (∼ 15-20 kg N ha −1 yr −1 ; Dore et al., 2012;Payne, 2014), whilst for the grassland soils the tracer application mimicked a daily fertilizer application rate of 0.5 kg N ha −1 d −1 .Due to the inclusion of the NO − 3 -rich C-MW site in the woodland soils, tracer application rates were higher than the daily atmospheric N deposition rates, but also reflected internal N cycling processes (e.g.nitrification) as an additional source of nitrate in these well-drained forest soils.Therefore, the application of the 15 N tracer at these low rates should not be expected to enrich the soil nitrate pool significantly, and potentially enhance the denitrification activity, in excess of the amount of nitrogen normally deposited via natural processes and common management practices.
The major assumptions of the 15 N gas-flux method and the associated "non-equilibrium equations" are that the denitrifying soil NO − 3 pool is uniformly labelled with 15 N and that the N 2 and N 2 O originate from the same denitrifying pool (Stevens and Laughlin, 1998).The 15 N fraction in the denitrifying pool ( 15 X N ), calculated non-destructively from the measured isotope ratios, ranged between 65 and 93 % and was well above the 10 % threshold for the correct application of the "non-equilibrium equations" (Lewicka-Szczebak et al., 2013).However, the calculated 15 X N was higher than the estimated total soil NO − 3 pool enrichment (range: 2-40 15 N at %), suggesting non-homogeneous mixing of the added tracer (98 15 N at %) with the ambient soil nitrate at natural abundance despite our effort for uniform tracer application with multiple injections across the investigated soil depth (Ruetting et al., 2011).Wu et al. (2011) have optimized the number of injections and the volume of tracer needed to achieve homogeneous labelling of a soil core (diameter 15 cm; height 20 cm) and reported that 38 injections of 4 mL volume each were necessary.We have used only 10 injections of 5-20 mL volume (depending on the soil water content of each land use type) to minimize the disturbance of the soil matrix, particularly in the highly porous media such as peatland soils, and this was clearly sub-optimal for the homogenous labelling of the soil enclosure but probably a necessary compromise for large-scale intensive measurements.We were not able to sample the soil within the chamber collars to directly estimate the 15 NO − 3 content of the soil pool due to time and budget constraints.However, in cases where destructive soil sampling was used to measure the soil nitrate pool enrichment (Kulkarni et al., 2014), the results were significantly different from the estimated enrichment due to sampling bias of the volume of soil affected by the tracer application.Non-uniform mixing of the 15 N label may lead to overestimation of the 15 X N and underestimation of the denitrification flux rates (Boast et al., 1988).However, under field conditions, it is unlikely to achieve complete mixing of the added tracer with the ambient nitrate pool, and experimental studies (Mulvaney, 1988;Mulvaney and Van den Heuvel, 1988) have shown that the associated error is well constrained and that accurate measurements can be made even with a less-uniformly labelled denitrifying pool.
The larger volume of tracer per injection (> 4 mL) in combination with the lower number of injections compared to Wu et al. (2011) may have created localized saturation effects (saturated soil cylinders around the injection holes), even if the total soil moisture content of the enclosure was not increased by more than 5 %, which would require several hours to equilibrate with the ambient soil moisture.We did not allow time for this soil moisture equilibration to occur following the tracer injection to avoid significant loss of the added nitrate via plant uptake (measurements occurring during the growth season).Therefore, it is likely that in plots where denitrification activity may have been limited by soil moisture (e.g.C-MW with mean WFPS 42 ± SE 0.76 %), the flux rates after 1 and 2 h of incubation may be overestimated due to moisture-induced denitrification events.
Most studies using 15 N tracers and static chambers in highly fertilized systems typically deploy their chambers between 1 and 2 h (Baily et al., 2012;Cuhel et al., 2010;Tauchnitz et al., 2015), but it has been shown that longer incubation periods (up to 24 or 48 h) may be needed in case of low 15 N enrichment applications in intact soil cores (Morse and Bernhardt, 2013) and laboratory incubations (Yang et al., 2014) for a more precise and accurate detectable 15 N-N 2 signal.However, it should be noted that in these cases the soil cores or slurries were incubated in fully enclosed systems and were thus not affected by potential bias from diffusion of evolved N 2 and N 2 O to the subsoil (Clough et al., 2005).The openbottom, un-vented static chamber design used in this study in combination with the extended incubation period up to 20 h may have potentially allowed some loss of the evolved N 2 and N 2 O through downward subsoil diffusion and/or reduction of gas exchanges at the soil-atmosphere interface due to decreasing concentration gradients (Healy et al., 1996).This could partly explain the non-linear increase in the evolved N 2 and N 2 O in the chamber headspace (Fig. 2a and b) and also the decrease in the N 2 flux rate with increasing incubation time (Fig. 3a).The N 2 O flux rate increased up to 2 h incubation followed by a decrease after 20 h consistently across land use types (Fig. 3b), indicating that the extended enclosure period lowered N 2 O fluxes due to subsoil diffusion and enhanced N 2 O reduction to N 2 .However, due to the high spatial heterogeneity within each land use type, the mean N 2 O flux rate was not significantly different between the different incubation intervals.In other words, the non-linearity of N 2 O evolution had less effect on the flux rate estimation than the inherent spatial variability within each land use type, which is in agreement with the findings of Chadwick et al. (2014), who suggested that the spatial variability of N 2 O fluxes far exceeds the bias due to assumed linearity of fluxes.
The lack of a consistent pattern of N 2 flux rate change with incubation time among the different land use types suggested a more complex temporal variability of N 2 fluxes that apart from the duration of incubation could have also been affected by the distribution of the added nitrate tracer.In the OS sites with the lowest average nitrate content (Table 2) and the highest water filled pore space (mean WFPS: C-PB = 70 ± SE 3.21 %; C-UG = 66 ± SE 1.58 %; R-HL = 69 ± SE 2.00 %), non-homogeneous tracer distribution ( 15 X N = 90 15 N at %) could have led to the creation of hotspots of denitrification activity due to substrate availability resulting in potentially overestimated flux rates in the first or even the second hour of incubation.However, analytical uncertainty due to fluxes being close to the limit of detection could not be ruled out.Conversely, in the soil moisture-limited forest site (C-MW), the injection of even 50 mL of tracer solution could have led to an increased moisture-induced denitrification activity within the first 1-2 h of incubation, until the added water started to equilibrate with the ambient soil moisture.Therefore the N 2 flux rate in C-MW may be significantly overestimated after 1 h of incubation.In the grassland sites and the R-DW forest site with intermediate soil moisture (mean WFPS: R-DW = 65 ± SE 1.79 %; R-UG = 64 ± SE 1.41 %; C-IG = 60 ± SE 1.45 %; R-IG = 61 ± SE 2.46 %) and nitrate content, the tracer injection is unlikely to have significantly affected the denitrification rate when all the conditions (i.e.soil moisture and substrate availability) were favourable, and therefore flux rates estimated after 1 h of incubation should be more reliable as long as the bias from analytical uncertainty was low.In these sites denitrification rates estimated after 1 or 20 h of incubation were not significantly different (Fig. 3a), suggesting a quasi-linear N 2 evolution throughout the incubation period (at least in 37.5 % of the sampling plots; see Table S4).However, the N 2 flux rates were significantly lower after 2 h of incubation, whereas the N 2 O flux rates were maximum at 2 h of incubation, consequently leading to an increased product ratio N 2 O / (N 2 + N 2 O) (Table S6).This observation could potentially be explained by a delay in the de novo synthesis of denitrification enzymes and the fact that the N 2 O reductase is known to have a slower expression than the preceding reduction enzymes (Knowles, 1982), leading to N 2 O accumulation and lower N 2 production after 2 h of incubation.After 20 h incubation, the decrease in the product ratio could be explained by a higher reduction rate of N 2 O to N 2 due to probably higher N 2 O reductase activity but also slower soil-atmosphere exchange of N 2 O due to the decreasing concentration gradient (Healy et al., 1996).It has been shown that the N 2 flux estimation with the 15 N gas-flux method is sensitive to the incubation time interval and the homogeneity of the tracer distribution due to the combination of several antagonistic effects such as decreasing gas diffusion gradients and soil moisture and substrate availability effects due to the added tracer solution.The uncertainty in the estimated in situ N 2 fluxes can be significantly reduced if additional effort is made to increase the homogeneity of the tracer application by increasing the number of injections while reducing the volume of the applied tracer (particularly in soils where denitrification is limited by moisture).Moreover, allowing the equilibration of the added tracer solution with the ambient soil water before gas sampling commences and by closely monitoring the linear evolution of the produced gases with more frequent gas sampling at shorter equal incubation intervals could help in identifying the appropriate length of incubation, thus avoiding potential overestimation of denitrification in nitrate and moisturelimited ecosystems and potential underestimation due to subsoil diffusion of evolved gases.The detailed uncertainty analysis of the N 2 and N 2 O flux estimation presented in this study complements the large-scale application of the 15 N gas-flux method in the same land use types between April 2013 and October 2014 for estimating annual rates of denitrification and N 2 O emission, which is presented in Sgouridis and Ullah (2015).
The minimum detectable N 2 and N 2 O fluxes depend on the precision of the IRMS systems, the soil NO − 3 pool enrichment, and the incubation parameters, such as the dimensions of the static chamber and the incubation time (Bergsma et al., 2001;Stevens and Laughlin 2001).For our chamber design, an incubation time of up to 20 h (which integrates the equilibration of the added tracer solution within the soil enclosure), and using the estimated MDC values (for both N 2 and N 2 O) for calculating a 15 X N value of 60 15 N at %, the minimum detectable flux rates were 4 µg N m −2 h −1 and 0.2 ng N m −2 h −1 for the N 2 and N 2 O fluxes respectively.These were significantly better than the minimum rates (175-900 µg N 2 -N m −2 h −1 and 0.04-0.21µg N 2 O-N m −2 h −1 ) reported by Bergsma et al. (2001), Kulkarni et al. (2014), andTauchnitz et al. (2015), using similar field 15 N tracer approaches, and comparable to the minimum rates measured by a high-precision 15 N gas flux approach in a laboratory soil incubation (Yang et al., 2014) and the gas-flow soil core method (8 µg N 2 -N m −2 h −1 and < 1 µg N 2 O-N m −2 h −1 ) by Wang et al. (2011).Our N 2 fluxes from woodland soils compare well with the rates reported in the literature for restored forested wetlands in North America (Morse and Bernhardt, 2013) and with the rates from northern hardwood forests in the USA (Kulkarni et al., 2014), using 15 N tracers at application rates similar to or lower than ours.Our results are also comparable to the rates reported from central European forests, under similar atmospheric N deposition rates, using the gas-flow soil core method (Butterbach-Bahl et al., 2002).For the grassland soils, the N 2 fluxes measured in the present study were significantly lower than previous applications of the 15 N gas-flux method at high fertilizer application rates (Baily et al., 2012;Cuhel et al., 2010;Graham et al., 2013), whilst for the organic soils our rates were significantly lower than the ones reported by Tauchnitz et al. (2015) since their 15 N tracer application rate (30 kg N ha −1 ) was 300 times higher than ours.The N 2 O fluxes were up to 200 times lower than the N 2 fluxes leading to low denitrification product ratios in all land use types, a result which is in line with the N 2 O yields reported from 15 N tracer studies in forest (Kulkarni et al., 2014;Morse and Bernhardt 2013) and grassland soils (Baily et al., 2012;Bergsma et al., 2001).In the present study we have compared the in situ denitrification rates between three major land use types using an extended field incubation period to increase the probability of detecting a reliable 15 N-N 2 signal, particularly under conditions of low denitrifier activity due to seasonality of denitrification and/or inherent capacity of soils (for example, organic and deciduous forest soils).However, these rates should be considered conservative since confounding issues such as subsoil diffusion and non-homogeneous labelling of the soil nitrate pool may in some cases have led to underestimations of the in situ denitrification rates.

Comparison with the AIT
The total denitrification rates measured with the C 2 H 2 amended intact soil cores followed the same trend as the total denitrification (N 2 and N 2 O fluxes combined) from the 15 N gas-flux measurements, while they were on average 168 times lower than the denitrification potential measured in the same land use types in anaerobic soil slurries amended with acetylene and nitrate in a previous study (Sgouridis and Ullah, 2014), thus reflecting lower in situ rates.The AIT denitrification rates were between 3 and 5 times lower than the 15 N gas-flux rates despite the fact that the AIT intact soil cores were capped at the bottom, thus not allowing any subsoil diffusion of the evolved gases due to denitrification.Therefore, the AIT rates should have been higher than the 15 N gas-flux rates if serious underestimation was occurring due to subsoil diffusion in the open-bottom static chambers, which was not the case.Adding nitrate to the C 2 H 2 amended cores would have been desirable for directly evaluating the priming effect of the added substrate on denitrification rates.The 15 N tracer addition to the static chambers corresponded to the amounts of N naturally deposited in these land use types either via management practices and/or atmospheric deposition, thus avoiding excessive N fertilization of the sampling plots.However, it cannot be conclusively argued that the same amount of applied nitrate would not have led to similar denitrification rates between the AIT and the 15 N gas-flux method.Previous comparisons between the AIT and the 15 N tracer method in field studies showed no significant difference between the two methods in measuring in situ total denitrification rates when the tracer is applied at high fertiliza-tion rates (50-200 kg N ha −1 ) and relatively low soil moisture contents (WFPS: 40-60 %; Aulakh et al., 1991;Mosier et al., 1986).Conversely, in laboratory incubations it was shown that the AIT significantly underestimated total denitrification compared to the 15 N tracer approach (Yu et al., 2010) and the direct N 2 flux approach (Qin et al., 2012) due to the incomplete inhibition of N 2 O reduction to N 2 by C 2 H 2 in wet soils (Yu et al., 2010) or in soils with low nitrate content (Qin et al., 2013(Qin et al., , 2014)).In our study, the soil WFPS ranged between 60 and 70 % in all land use types, with the exception of the C-MW site (mean WFPS 42 %), whilst the 15 N-NO − 3 tracer application rate was low (< 1 kg N ha −1 ).Moreover, the disturbance of the soil structure during the extraction of the soil cores and the effect of the acetylene addition to microbial activity were not significant, as was suggested by the similar CO 2 production rates (Aulakh et al., 1991), representing soil respiration (Felber et al., 2012), in the static chambers and the C 2 H 2 amended and un-amended intact soil cores.Therefore, we could argue that it is possible that the AIT underestimated total denitrification rates compared to the 15 N gas-flux method due to the likely incomplete inhibition of N 2 O reduction to N 2 under relatively high soil moisture contents, although the shorter incubation time (2 h for the intact cores) may have limited the ability of C 2 H 2 to fully equilibrate within soil pore spaces.Other confounding factors such as the catalytic decomposition of NO in the presence of C 2 H 2 (Bollmann and Conrad, 1997;Nadeem et al., 2013) may have also contributed to the lower denitrification rates measured by the AIT.This study has confirmed some of the drawbacks of the AIT as a quantification method of in situ denitrification rates compared to the 15 N gas-flux.
The estimation of the denitrification product ratio using the AIT method, from the un-amended cores (N 2 O only) and the C 2 H 2 amended cores (N 2 + N 2 O), is usually overestimated since the source of N 2 O cannot be discriminated with the AIT, whilst the N 2 flux is underestimated due to the incomplete inhibition of N 2 O reduction (Butterbach-Bahl et al., 2013).This was confirmed in the present study for all the land use types, and even the maximum denitrification product ratio after 2 h incubation in the case of the grassland soils (23 %) was still significantly lower than the respective ratio from the AIT (50 %).Therefore, the much lower denitrification product ratio estimated from the 15 N gas-flux measurements is significantly more reliable, and the wider application of this field technique across a range of land use types can have important implications for evaluating the role of denitrification as a reactive nitrogen sink and as a source of N 2 O emissions (Butterbach-Bahl et al., 2013;Kulkarni et al., 2008).

Conclusions
The improved analytical precision for both 15 N-N 2 and 15 N-N 2 O analyses allowed us to quantify in situ N 2 and N 2 O fluxes with low 15 N tracer addition under field conditions in natural and semi-natural land use types for the first time.The estimation of N 2 fluxes was sensitive to the incubation time interval and the homogeneity of the tracer distribution due to the combination of several antagonistic effects such as decreasing gas diffusion gradients over time and soil moisture and substrate priming effects due to the added nitrate tracer solution.The spatial variability of N 2 O fluxes superseded any bias associated with non-linear fluxes due to the extended incubation period.The uncertainty in the estimated N 2 and N 2 O fluxes can be significantly reduced by increasing the homogeneity of the tracer application and by closely monitoring the linear evolution of the produced gases with more frequent gas sampling at shorter equal incubation intervals to avoid underestimation or overestimation of denitrification.Comparing the 15 N gas-flux method with the AIT confirmed the drawbacks of the AIT as a reliable quantification method of in situ denitrification rates.Moreover, the AIT method overestimated the denitrification product ratio compared to the 15 N gas-flux method.The 15 N gas-flux method holds much promise as a more reliable field technique for measuring in situ denitrification rates and its wider application across a range of terrestrial ecosystems can lead to its refinement and improvement and in the long term can significantly contribute to our understanding of the role of denitrification as a reactive nitrogen sink.
The Supplement related to this article is available online at doi:10.5194/bg-13-1821-2016-supplement.

Figure 1 .
Figure 1.Schematic of the 15 N-N 2 analysis system.
The waste was flushed out of the instrument.The N 2 O was further cryofocused in a second liquid nitrogen trap prior to being introduced onto a 25 m × 0.32 mm Poraplot Q gas chromatography column (Chrompack column, Varian, Surrey, UK).The column separated N 2 O from any residual CO 2 , and both entered the IRMS via an open split.The retention time between the first eluting CO 2 (< 2 × 10 −10 amplitude) and second eluting N 2 O peak typically fell in the range between 60 and 70 s to avoid isobaric interference of the CO 2 with the calculated 15 N.The N 2 O was directed towards the triple collectors of the isotope ratio mass spectrometer where m/z 44, m/z 45, and m/z 46 mass ions were measured and recorded.Instrument stability checks were performed prior to each analysis by running a series of 10 reference pulses of N 2 O (BOC special gases) until a standard deviation of δ 15 N better than 0.05 ‰ was achieved.Prior to each sample batch wwwet al.: Application of the 15 N gas-flux method

Figure 2 .
Figure 2. Evolved (a) N 2 and (b) N 2 O gas measured between 1, 2, and 20 h incubation time intervals using the 15 N gas-flux method in the organic soil (OS), woodland (WL), and grassland (GL) land use types.Data points are means and the error bars represent standard errors.

Figure 3 .
Figure 3. Mean rates of (a) N 2 flux and (b) N 2 O flux due to denitrification at the three incubation time intervals in the three land use types (OS: organic soils; WL: woodland; and GL: grassland).Same lower-case letters indicate no significant differences (p > 0.05) between incubation time intervals according to the non-parametric Kruskal-Wallis test.Error bars represent standard errors.

Figure 4 .
Figure 4. Mean rates of (a) N 2 flux, (b) N 2 O emission due to denitrification, and (c) the denitrification product ratio N 2 O / (N 2 + N 2 O) in the three land use types (OS: organic soils; WL: woodland; and GL: grassland).Same lower-case letters indicate no significant differences (p > 0.05) between land use types according to one-way ANOVA and the Games-Howell post hoc test.The sample size (n) is given in parentheses for each land use type on the x axis.Error bars represent standard errors.

Figure 5 .
Figure 5. (a) Mean total denitrification measured with the 15 N gas-flux method and the AIT, (b) mean bulk N 2 O emission measured in the static chambers of the 15 N gas-flux method and in unamended intact soil cores, and (c) the denitrification product ratio N 2 O / (N 2 + N 2 O) with the 15 N gas-flux method and the AIT in the three land use types (OS: organic soils; WL: woodland; and GL: grassland).Same lower-case letters indicate no significant differences (p > 0.05) between measurement methods according to an independent samples t test.The sample size (n) is given in parentheses for each land use type and each method on the x axis.Error bars represent standard errors.

Figure 6 .
Figure 6.Mean CO 2 production measured in the static chambers of the 15 N gas-flux method, in un-amended and C 2 H 2 amended intact soil cores in the three land use types (OS: organic soils; WL: woodland; and GL: grassland).Same lower-case letters indicate no significant differences (p > 0.05) between measurement methods according to an independent samples t test.The sample size (n) is given in parentheses for each land use type on the x axis.Error bars represent standard errors.
et al.: Application of the 15 N gas-flux method

Table 2 .
The ambient soil nitrate pool, the 15 N tracer application rate, the estimated enrichment of the total soil nitrate pool, the calculated 15 X N value from N 2 O, and the slope of the 15 X N change with incubation time in the three land use types.Data are means with standard errors in parentheses.

Table 3 .
Comparison of mean flux rates and ratios between land use types for the two field methods using one-way ANOVA.All variables are log-transformed.F : F statistic; P : probability level.