First on-line isotopic characterization of N 2 O above intensively managed grassland

. The analysis of the four main isotopic N 2 O species ( 14 N 14 N 16 O, 14 N 15 N 16 O, 15 N 14 N 16 O, 14 N 14 N 18 O) and especially the intramolecular distribution of 15 N (“site prefer-ence”, SP) has been suggested as a tool to distinguish source processes and to help constrain the global N 2 O budget. How-ever, current studies suffer from limited spatial and temporal resolution capabilities due to the combination of discrete ﬂask sampling with subsequent laboratory-based mass-spectrometric analysis. Quantum cascade laser absorption spectroscopy (QCLAS) allows the selective high-precision analysis of N 2 O isotopic species at trace levels and is suit-able for in situ measurements. Here, we present results from the ﬁrst ﬁeld campaign, conducted on an intensively managed grassland site in central Switzerland. N 2 O mole fractions and isotopic composition were determined in the atmospheric surface layer (at 2.2 m height) at a high temporal resolution with a modiﬁed state-of-the-art laser spectrometer connected to an automated N 2 O preconcentration unit. The analytical performance was determined from repeated measurements of a compressed air tank and resulted in measurement repeatability of 0.20, 0.12 and 0.11 ‰ for δ 15 N α , δ 15 N β and δ 18 O, respectively. Simulta-neous eddy-covariance N 2 O ﬂux measurements were used to determine the ﬂux-averaged isotopic signature of soil-emitted N 2 O.

velopment of adequate mitigation strategies is pertinent and requires a better understanding of the processes driving N 2 O fluxes. To date, nitrification, nitrifier-denitrification and denitrification are considered to constitute the dominant N 2 Oproducing processes, especially in agricultural soils (Wrage et al., 2001). Other N 2 O source processes such as abiotic N 2 O production, co-denitrification and heterotrophic nitrification have also been observed; a concise overview of observed processes is given elsewhere (Butterbach-Bahl et al., 2013). This complexity inherent in the N cycle and associated transformation processes is a major challenge in developing mitigation strategies, as the attribution of N 2 O production to the respective processes is required to tailor targetoriented actions (Baggs, 2008). Approaches for apportioning N 2 O emissions to nitrification and denitrification and N 2 O reduction to N 2 (source partitioning) have mostly relied on acetylene (C 2 H 2 ) inhibition and isotope labeling (Groffman et al., 2006), but denitrification rates are underestimated by the C 2 H 2 method (Butterbach-Bahl et al., 2013;Groffman et al., 2006;Watts and Seitzinger, 2000). Isotope labeling approaches are vulnerable to incomplete diffusion of the tracer and to the stimulation of process rates by the addition of the labeled substrates themselves (Groffman et al., 2006). Changes in the natural abundance of 15 N and 18 O in N 2 O have been explored to investigate N 2 O production processes, but the δ 15 N and δ 18 O determined depend on both fractionation factors and the isotopic composition of precursors, which in turn exhibit strong variations (Baggs, 2008;Bedard-Haughn et al., 2003;Heil et al., 2014;Toyoda et al., 2011). N 2 O is a linear molecule and four main isotopic species can be discerned: 14 N 14 N 16 O, 14 N 15 N 16 O, 15 N 14 N 16 O and 14 N 14 N 18 O. The isotopic species 14 N 14 N 16 O, 14 N 14 N 18 O and 14 N 15 N 16 O (or 15 N 14 N 16 O) are isotopologues, while 14 N 15 N 16 O and 15 N 14 N 16 O are isotopomers and will be termed 15 N α -N 2 O and 15 N β -N 2 O (Toyoda and Yoshida, 1999). The umbrella term isotopocule is used for both isotopomers and isotopologues. The intramolecular distribution of 15 N in N 2 O ("site preference"; SP = δ 15 N α − δ 15 N β ) has been reported to be independent of the substrate's isotopic composition, as SP in the N 2 O produced de novo remained constant even though δ 15 N and δ 18 O values of both N 2 O and substrates changed markedly during experiments with pure cultures (Heil et al., 2014;Sutka et al., 2003Sutka et al., , 2006Sutka et al., , 2008Toyoda et al., 2005). Therefore, SP can be considered as a tracer conserving the source process information . The SP of different processes has been characterized in a number of pure-culture, mixedculture (Ostrom et al., 2007;Sutka et al., 2003Sutka et al., , 2006Toyoda et al., 2005;Wunderlin et al., 2012Wunderlin et al., , 2013 and soil-incubation studies (Köster et al., 2011(Köster et al., , 2013aLewicka-Szczebak et al., 2014;Well et al., 2006Well et al., , 2008, with a compilation of data in Toyoda et al. (2011). A recent review on source partitioning and SP (Decock and Six, 2013b) concluded that SP is capable of distinguishing between the process groups N 2 O N (NH 2 OH oxidation, fungal denitrification and abiotic N 2 O production; SP = 32.8 ± 4.0 ‰) and N 2 O D (nitrifierdenitrification and denitrification; SP = −1.6 ± 3.8 ‰). In addition, N 2 O isotopocules can be used as an independent validation of the global, measurement-based bottom-up N 2 O budget and have already confirmed that the isotopically light sources such as agriculture and industry contribute to the increase in atmospheric N 2 O (Toyoda et al., 2013;Yoshida and Toyoda, 2000). Owing to the temporal and spatial variability of isotopomer ratios, it is indispensable to derive flux-weighted average values from different sources (such as ecosystems) for later use in budget analysis using box models (Kim and Craig, 1993;Perez et al., 2001;Yoshida and Toyoda, 2000).
N 2 O isotopomers can be measured by mass spectrometry, but this requires discrete flask sampling with subsequent laboratory analysis. Hence, this approach is limited in temporal resolution and spatial representation of a given site. Additionally, it is indirect, as information on the site-specific isotopic composition is derived from the analysis of the NO + fragment and N 2 O + molecular ion. Recently, a quantum cascade laser absorption spectrometer (QCLAS) capable of selective analysis of the three most abundant N 2 O isotopocules has been presented (Waechter et al., 2008), and its potential for in situ measurements in conjunction with an automated pre-concentration unit has been shown (Mohn et al., 2010. Here we present the results obtained from a, to our knowledge first, campaign in which the isotopic composition of N 2 O (SP, δ 15 N, δ 18 O) in the atmospheric surface layer was determined on-line by using an optimized state-of-the-art laser spectrometer. Using the combination of N 2 O isotopic analysis by QCLAS, accompanying eddy-covariance-based N 2 O flux measurements as well as the monitoring of environmental conditions and inorganic nitrogen concentrations, our specific objectives for this study were (i) to demonstrate the capability of QCLAS systems for carrying out high-precision isotopic analysis of (soil-emitted) N 2 O in ambient air; (ii) to investigate management and weather effects on isotopic composition and source processes; and (iii) to characterize the flux-averaged isotopic composition of N 2 O emitted from an intensively managed grassland site.

Study site
The agricultural research station Chamau (CHA) is located in central Switzerland at an elevation of 400 m a.s.l. The experiment was conducted on intensively managed grassland belonging to CHA which is primarily used for fodder production and occasional winter grazing by sheep (Zeeman et al., 2010). The soil type is a Cambisol with a bulk density of 0.97 g cm −3 , 30.6 % sand, 47.7 % silt and 21.8 % clay in the top 10 cm and pH of 5.7-6.2. Soil carbon and nitrogen con- Table 1. Reference gas and compressed air tanks used during the campaign. S1 and S2 represent the anchor and calibration standard. C1 and C2 are the target gases used for determination of system performance. The reported precision is 1σ . tent in the top 10 cm was 37.9 and 4.1 g kg −1 (Roth, 2006). Mean annual temperature and annual precipitation are 9.1 • C and 1151 mm, respectively (Zeeman et al., 2010). Management practices aim at fodder production and consist of mowing followed by slurry application, with up to six mowing and slurry applications per year and the occasional grazing of sheep and cattle in October and November. During the campaign in summer 2013, three management cycles were carried out. Harvest dates were 6 June, 11 July and 21 August, and slurry was applied within 10 days after each mowing event. Nitrogen input was calculated from the amount of slurry applied to the field and the N concentration determined (Labor für Boden-und Umweltanalytik, Eric Schweizer AG, Thun, Switzerland) in a sample drawn from the supply to the trailing hose applicator. The applied N amounted to 30, 40 and 43.3 kg N ha −1 for the first, second and third application, respectively. Grassland is reestablished via ploughing and resowing approximately every 10 years. The last reestablishment event took place in 2012 (Merbold et al., 2014).

Instrumental setup for analysis of N 2 O isotopocule ratios
The four most abundant N 2 O isotopic species were quantified using a modified QCLAS (Aerodyne Research Inc., Billerica MA, USA) equipped with a continuous-wave quantum cascade laser (cw-QCL) with spectral emission at 2203 cm −1 , an astigmatic Herriott multi-pass absorption cell (204 m path length, AMAC-200) and a reference path with a short (5 cm) N 2 O-filled cell to lock the laser emission frequency . During the campaign, the QCLAS was operated in an air-conditioned trailer located 60 m west of the eddy-covariance (EC) tower. This trailer position contributes < 20 % to the main flux and is at the far side of the prevailing wind direction (Zeeman et al., 2010). The sample air inlet was installed next to the inlet of the EC tower (2.2 m height). Sample air was drawn through a PTFE tube (4 mm ID) by a membrane pump (PM 25032-022, KNF Neuberger, Switzerland). Upstream of the pump, the sample air was pre-dried with a permeation drier (MD-050-72S-1, PermaPure Inc., USA). Following the pump, the pressure was maintained at 4 bar overpressure using a pres-sure relief valve. Humidity as well as CO 2 were quantitatively removed from the gas flow by applying a chemical trap filled with Ascarite (7 g, 10-35 mesh, Fluka, Switzerland) bracketed by Mg(ClO 4 ) 2 (2 × 1.5 g, Fluka, Switzerland). Finally, the sample gas was passed through a sintered metal filter (SS-6F-MM-2, Swagelok, USA) and directed to a preconcentration unit described in detail previously (Mohn et al., 2010. For an increase in N 2 O mixing ratios from ambient level to around 50 ppm N 2 O, approx. 8 L of ambient air were preconcentrated. Afterwards, the preconcentrated N 2 O was introduced into the evacuated multi-pass cell of the QCLAS. Isotopic fractionation during preconcentration (increase by 0.31 ± 0.10, 0.34 ± 0.16 and 0.29 ± 0.07 ‰ for δ 15 N α , δ 15 N β and δ 18 O, respectively) was quantified by the preconcentration of N 2 O with a known isotopic composition and subsequently corrected. The compatibility of N 2 O isotopomer analysis by QCLAS with isotope ratio mass spectrometry (IRMS) laboratories was recently demonstrated in an inter-laboratory comparison campaign .

Measurement and calibration strategy
To ensure high accuracy and repeatability of the analytical system, a measurement and calibration strategy similar to the one presented by Mohn et al. (2012)  moles of trace gas per mole of dry air). Both standards were calibrated against primary standards which were previously measured by the Tokyo Institute of Technology (TIT, Toyoda and Yoshida) to anchor δ values to the international isotopic standard scales. The first standard (S1, Table 1) was used as an anchor point to the international δ scale and used as input data for data analysis algorithms (see data processing). Therefore, the N 2 O isotopic composition of S1 was tar- geted to closely resemble background air. As the N 2 O isotopic composition of surface layer air is mainly a mixture of soil-derived and background composition, the second standard (S2, Table 2) used for span correction was depleted in δ 15 N α , δ 15 N β and δ 18 O compared to background air in accordance with the expected terrestrial source signatures. The measurement protocol started with the injection of S1, dynamically diluted to 50 ppm, the mole fraction of ambient N 2 O after preconcentration. After flushing the absorption cell with synthetic air, S2 was injected and also diluted to 50 ppm. For the determination of the slight concentration dependence already reported , S1 was injected again but at a higher mole fraction of 67 ppm (later referred to as S1 h ). This mole fraction represents the mole fraction expected after the preconcentration of highconcentration surface layer air. Subsequently, S1 was injected again and diluted to 50 ppm, before the cell was filled with preconcentrated ambient N 2 O (A). This subroutine (S1 + A) of injection of S1 and preconcentrated ambient N 2 O took 35 min and was repeated three times. For an independent determination of repeatability, the fourth sample was preconcentrated compressed air (target gas). During the campaign, two compressed air cylinders (C1 and C2, referred to as target gas) were used. The isotopic composition and N 2 O mixing ratio of both cylinders were determined in the laboratory prior to the campaign start (Table 1). N 2 O mole fractions and isotopic composition analyzed in the laboratory and at the field site agreed within their analytical uncertainty. Following target gas analysis, S1 and S1 h were analyzed again. Another set of three subroutines S1+A completed one run. One complete cycle of 6 ambient air samples and one compressed air sample took 340 min, leading to approx. 25 ambient air samples being analyzed during 24 h. N 2 O mole fractions were determined according to Mohn et al. (2012).

Data processing
Data processing is based on individual mixing ratios of the four main N 2 O isotopic species and spectrometer characteristics as recorded by the instruments's software (TDLWintel, Aerodyne Research Inc., Billerica, MA, USA). In the first step, variations in the isotope ratios induced by drifts in the instrument working parameters during the field operation were corrected. A linear additive model explaining the deviation of isotope ratios R α , R β and R 18 O for repeated measurements of standard S1 from their mean value by absorption cell temperature (T 1), laser temperature (T 2), line position (LP) and pressure (p) was calibrated based on S1 injections. For isotope ratios of S1, S1 h , S2, sample air and compressed air, these systematic deviations were corrected based on the respective values of T 1, T 2, LP and p. In a second step, concentration dependence of isotope ratios, determined using the measurements of S1 and S1 h , was addressed with corrections (0.013, 0.028 and 0.004 ‰ ppb −1 for δ 15 N α , δ 15 N β and δ 18 O) being in the same range as described earlier . Subsequently, remaining drifts were corrected based on the analysis of S1. Finally, isotope ratios were converted to δ values using a two-point calibration derived from corrected values of S1 and S2.

Determination of soil-emitted N 2 O isotopic composition
The isotopic composition of the source process "soil N 2 O emission" was derived using the Keeling plot approach (Keeling, 1958), where δ values measured (here at 2.2 m height) are plotted versus the inverse of N 2 O mole fractions. The intercept of the linear regression line can be interpreted as the isotopic composition of soil-emitted N 2 O (Pataki et al., 2003). Therefore, the determination of soil N 2 O isotopic composition requires an increase in N 2 O mole fraction. During the day, turbulence mixes surface layer air into the at- Figure 1. Long-term stability (standard deviation σ ) derived by target gas injections (n = 331) over a 3-month period. As two target gas tanks were used, histograms show deviation of respective tank means,x, for δ 15 N α , δ 15 N β , δ 18 O, δ 15 N bulk and SP. mospheric background. At night, the surface layer becomes more stable and the N 2 O mole fraction increases, shifting isotopic composition towards its source composition. As a consequence, Keeling plots were based on noon-to-noon periods. This approach is discussed in Sect. 4.6.

N 2 O flux measurement
At CHA, greenhouse gas mole fractions, including N 2 O, have been measured continuously since 2012 by means of the EC method (Baldocchi and Meyers, 1998). The system consists of a three-dimensional sonic anemometer to measure wind speed and direction (2.41 m height, Solent R3, Gill Instruments, Lymington, UK) and a QCLAS (mini-QCLAS, Aerodyne Research Inc., Billerica, MA, USA) to determine N 2 O mole fractions at a temporal resolution of 10 Hz. Both data streams are merged in near real time within a data acquisition system (MOXA embedded Linux computer; Moxa, Brea, CA, USA) via an RS-232 serial data link . The setup has been described in detail previously (Merbold et al., 2014). Post-processing of N 2 O fluxes included screening for obvious out-of-range values (±100 nmol m −2 s −1 ). N 2 O fluxes were further aggregated to noon-to-noon daily averages to smoothen the large variability in the 30 min flux averages. Daily averages were calculated for days where more than 30 half-hour values were available, with this filter excluding 3 days from analysis.

Soil inorganic N, dissolved organic C and environmental conditions
Ammonium (NH + 4 ) and nitrate (NO − 3 ) concentrations were determined from soil (0-20 cm depth) sampled at 10 positions along a transect within the footprint of the EC measurements following the predominant wind direction. Samples were taken weekly throughout the campaign or daily during mowing and slurry application events. Per sample, ∼ 15 g of fresh soil were added to specimen vessels containing 50 mL 1 M KCl. After 1 h on a shaker, the supernatant was filtered (Whatman no. 42 ashless filter paper, 150 mm diameter) and analyzed colorimetrically for NH + 4 and NO − 3 . For a subset of extracts, we determined dissolved organic carbon (DOC) concentrations by combustion of KCl extracts using a to-tal organic C analyzer (Shimdazu TOC-V, Columbia, MD, USA).
Soil temperatures and volumetric soil moisture contents at 10 cm depth were measured at the same 10 locations along the transect (5TM-sensors, Decagon Devices Ltd., Pullman, USA). Data were stored as 10 min averages on a data logger (EM50, Decagon Devices Ltd., Pullman, USA). The volumetric water content was converted to water-filled pore space (wfps) using a bulk density of 1.09 g cm −3 . Precipitation was measured with a tipping bucket rain gauge (Type 10116, Toss GmbH, Potsdam, Germany) and stored as 10 min averages on a data logger (CR10X-2M, Campbell Scientific Inc., Logan, USA).

Long-term precision for target gas analysis
System performance for N 2 O mole fractions and isotopic composition was determined based on the repeated analysis of compressed air from target gas tanks (C1, C2). There was no significant drift in the δ values and N 2 O mole fractions, indicating the stability of the applied measurement technique. Repeatability, calculated as the standard deviation (σ ) of 331 target gas measurements, amounted to 0.20, 0.12, 0.10, 0.12 and 0.22 ‰ for δ 15 N α , δ 15 N β , δ 18 O, δ 15 N bulk and SP, respectively (Fig. 1). The standard deviation for the N 2 O mole fraction of the target gas was 0.25 ppb.

N 2 O mole fractions and isotopic composition at 2.2 m height
Air samples were taken at 2.2 m height, which is within the lowest 10 % of the atmospheric boundary layer (ABL) where mechanical generation of turbulence exceeds buoyant generation or consumption. This part of the ABL is called surface layer; hence, corresponding air samples are referred to as surface layer air samples. N 2 O isotopic composition of the surface layer air samples (n = 2130) ranged from 2.5 to 16.1 ‰, −11.9 to −2.4 ‰, 37.6 to 44.6 ‰, −4.6 to 6.6 ‰ and 14.3 to 19.3 ‰ for δ 15 N α , δ 15 N β , δ 18 O, δ 15 N bulk and SP, respectively (Fig. 2). Surface layer N 2 O mole fractions varied between 325 and 469 ppb and followed a diurnal cycle with highest values during the night, when the boundary layer became more stable. Increasing N 2 O mole fractions were associated with decreasing δ values, indicating that soil-emitted N 2 O that was mixed into the surface layer was depleted in 15 N as compared to N 2 O in the atmospheric background.

Auxiliary measurements
Half-hourly N 2 O fluxes were averaged from noon-to-noon (f N 2 O ) and ranged from −1 to 5 nmol m −2 s −1 . Maximum N 2 O fluxes coincided with an overnight build up in N 2 O mole fractions ( N 2 O) as analyzed by QCLAS and could not be attributed to slurry application events alone (Fig. 3).
Among the correlations of f N 2 O and auxiliary variables, only the one with a nitrate concentration (r 2 = 0.18) was significant (p < 0.01). Soil water content (wfps) was modulated by precipitation, and two clear states could be identified. During the "wet" part of the campaign, lasting until 7 July, average wfps was 62 ± 4 % and thus was significantly (t test, p < 0.001) higher than the average of 37 ± 4 % calculated for the remainder of the campaign (referred to as the "dry" part). Soil temperature did not show such a clear two-phase pattern; however, temperatures during the first, wet part were 16.7 ± 4 • C and thus significantly (p < 0.001) lower than during the dry phase (21.2 ± 2 • C). Background NH + 4 and NO − 3 concentrations were smaller than 3 µg g −1 soil and clearly responded to mowing and slurry application in the second and third management events. The NO − 3 concentration was higher than the NH + 4 concentration and peaked at 16 and 50 µg g −1 soil , while NH + 4 concentration peaked at 9 and 15 µg g −1 soil for these two management events. In contrast, dissolved organic carbon concentrations (DOC) did not respond to management events but were higher during the dry phase of the campaign (p < 0.001).

Isotopic composition of soil-emitted N 2 O
The uncertainty of the determined source isotopic composition was estimated based on the standard error of the Keeling plot intercept and depends on the degree to which soil air accumulates in the surface layer ( N 2 O, Fig. 4). For instance, the intercept (source) standard error ranged from 0.3 to 82 ‰ for SP. To apply the Keeling plot approach only to situations in which soil air accumulated in the surface layer, only source isotopic compositions for overnight increases in N 2 O mole fractions of more than 12 ppb were considered in this study. This filter led to a maximum and average (µ) standard error of 6.8 (µ = 2.2) ‰, 4.5 (µ = 1.4) ‰ and 2.2 (µ = 1) ‰ for SP, δ 15 N bulk and δ 18 O isotopic source signatures, respectively.
During the field campaign, the Keeling-plot-derived isotopic composition of soil-emitted N 2 O ranged from 1.4 to 17.3 ‰, −29 to −3 ‰ and 22.6 to 34.8 ‰ for SP, δ 15 N bulk and δ 18 O, respectively. All explanatory variables except NH + 4 and NO − 3 were found to significantly correlate with SP (Table 2). For δ 15 N bulk , correlations with N 2 O, wfps, soil temperature, DOC and NO − 3 and for δ 18 O correlations of f N 2 O , N 2 O, precipitation, soil temperature and NO − 3 were significant. However, the adjusted r 2 for all regressions was below 0.4; in addition, multiple explanatory variables such as NH + 4 and NO − 3 or wfps and temperature (Fig. 5) did not increase the explained variance above this value.

Event-based data aggregation
As already described in the section "Auxiliary measurements", there was a wet phase (n = 27 Keeling-plot-derived N 2 O isotopic compositions) in the beginning of the campaign, which lasted about 1 month, and a dry phase lasting about 2 months (n = 38). Therefore, the data set was split into two corresponding parts with averages of 7.4 ± 3.6 vs. 11.1 ± 4.2 ‰ for SP, −19 ± 3.8 vs. −12.5 ± 5.9 ‰ for δ 15 N bulk and 28.7 ± 2.2 vs. 29.7 ± 3.4 ‰ for δ 18 O in the wet vs. the dry phase, respectively. Averages of SP and δ 15 N bulk were significantly different (p < 0.001), but δ 18 O averages were not. Based on this simple classification, the dry phase contains rewetting events. A rewetting event was defined as a 2-day period starting at the day for which wfps increased. Exclusion of these rewetting events during the dry phase increased average δ values (n = 30) and decreased standard deviations for SP, δ 15 N bulk and δ 18 O to 12.5 ± 3.4, −10.8 ± 4.5 and 30.7 ± 2.8 ‰ . Moreover, the difference in δ 18 O was significant (p < 0.001).
In addition to the dry/wet classification, we also defined three subsets representing the N 2 O emission associated with the management events of mowing followed by fertilization ("Mana I"-"Mana III"), one subset representing a rewetting event between Mana II and III ("Rewetting") and one subset representing background ("BG", all remaining measure- ments). There were two distinct rewetting events between management events II and III, but N 2 O isotopic composition is only available for the first one (29-31 July 2014). Isotopic compositions of soil-emitted N 2 O were assigned to subsets of management or rewetting if the associated flux or nutrient concentration was elevated. This classification scheme led to 3 to 7 measurements for management and rewetting events (Fig. 3, underlaid in blue), while 47 measurements were assigned to class BG. Boxplots for SP, δ 15 N bulk , δ 18 O and wfps (Fig. 6) showed characteristic δ values and wfps for manage-ment and rewetting but not for subset BG. Measurements assigned to BG covered practically the whole range of values observed across all the other classes. Therefore, standard deviations for class BG were 1 order of magnitude larger than for the four other classes.
Statistical analysis is confounded by low and unequal sample size, so we compared exclusively the subsets management and rewetting using multiple nonparametric Wilcoxon tests after having checked the homogeneity of variances us- ing the Bartlett test. For all investigated δ values, only differences between groups Mana II and Mana III were significant.

Averages of N 2 O isotopic signature for intensively managed grassland
Simple averages of the daily isotopic composition of soilemitted N 2 O were 9.6 ± 4.4, −15.2 ± 6.0 and 29.3 ± 3 ‰ for SP, δ 15 N bulk and δ 18 O, respectively (n = 62). Representative isotopic composition of N 2 O emitted from a given site or treatment can be estimated based on flux-weighted averages of daily isotopic composition. For some noonto-noon periods included in the above average, thus including an overnight increase in N 2 O mole fractions of at least 12 ppb, negative N 2 O fluxes were detected by the EC system (−0.17 ± 2.1 nmol m −2 s −1 ; n = 14). This might be due to the uncertainty of N 2 O flux measurements, temporal averaging over positive and negative fluxes in a noonto-noon period, or different footprint regions for N 2 O flux and isotopic analysis (flux vs. concentration footprint). To avoid bias in the flux-weighted average of emitted N 2 O due to either one of the abovementioned possible reasons, the weighted averages were calculated for positive flux events only. Flux-weighted averages were 6.9 ± 4.3, −17.4 ± 6.2 and 27.4 ± 3.6 ‰ for SP, δ 15 N bulk and δ 18 O, respectively (n = 48).  In both studies, in accordance with the present study, ana-lytical performance was determined based on the repeated analysis of compressed air samples. Thereby, the analytical precision reached in the present study was distinctly higher for δ 15 N β and δ 18 O and similar for δ 15 N α compared to these two previous studies, even though the measurements were done under field conditions and over a much longer, 3-month, period. This confirms the high level of precision associated with the QCLAS-based determination of N 2 O isotopic composition. Standard errors for Keeling plot intercepts (Fig. 4) confirm that this precision is sufficient to resolve the variability of atmospheric N 2 O sampled close to the ground. As our instrument was located directly at the field site and measurements were conducted over a period of more than 3 months, our study indicates that this level of repeatability can be achieved both on long timescales and in the field.

N 2 O isotopic composition in the atmospheric surface layer (2.2 m height)
In our study, δ values of single preconcentrated air samples were between atmospheric background and 14.3 ‰ (SP) and −4.7 ‰ (δ 15 N bulk ). Mohn et al. (2012) reported similar values between atmospheric background and 12 ‰ (SP) and −4 ‰ (δ 15 N bulk ). Therefore, the variation observed in both studies is much higher compared to the measurements by Harris et al. (2014) (Harris et al., 2015;Ogawa and Yoshida, 2005).

Isotopic composition of soil-emitted N 2 O
The SP of soil-emitted N 2 O observed in our study (1 to 17 ‰) is within the ranges expected for a mixture of the two process groups N 2 O N and N 2 O D and does not necessarily indicate a significant contribution of N 2 O reduction, an effect which is discussed further below. Isotopic composition of soil-emitted N 2 O has been predominately determined in laboratory incubation studies (Köster et al., 2013a, b;Perez et al., 2006;Well and Flessa, 2009b;Well et al., 2006Well et al., , 2008. Additionally, results from field experiments us-ing static chambers (Opdyke et al., 2009;Ostrom et al., 2010;Toyoda et al., 2011;Yamulki et al., 2001) and N 2 O accumulation below a snowpack have been published . Based on pure-culture studies, SP values from 19.7 to 40 and −8.7 to 8.5 ‰, were observed for N 2 O N and N 2 O D , respectively (Decock and Six, 2013b). In field experiments SP was found to range between −1 and 32 ‰ (Opdyke et al., 2009), −3 and 18 ‰ (Yamulki et al., 2001), −14 and 90 ‰ (Toyoda et al., 2011) and 0 and 13 ‰ (Ostrom et al., 2010). The very high SP values detected by Toyoda et al. (2011) may have resulted from extensive N 2 O reduction to N 2 , a process increasing SP, δ 15 N bulk and δ 18 O (Ostrom et al., 2007). For δ 15 N bulk and δ 18 O, a much wider variation as compared to SP is expected because these variables depend both on fractionation factors, which vary among different microbial communities and depend on reaction conditions as well as on the isotopic composition of the substrate (Baggs, 2008). Under field conditions, δ 15 N bulk was reported to range between −17 and 9 ‰ (Opdyke et al., 2009)

Changes in N 2 O source signatures induced by N 2 O reduction to N 2
Quantitative source partitioning between process groups N 2 O N and N 2 O D based on SP is possible only when no other processes except those contained in the process groups have an influence on the site-specific N 2 O isotopic composition. However, in the terminal step of denitrification, namely the reduction of N 2 O to N 2 , N-O bonds between lighter isotopes are cleaved preferentially, leading to an increase in SP, δ 15 N bulk and δ 18 O in the remaining N 2 O. Consequently, part of the N 2 O originating from a combination of the two process groups, i.e., N 2 O N and N 2 O D , may have been consumed by N 2 O-to-N 2 reduction prior to emission. For the identification of processes determining N 2 O isotopic composition, isotopocule maps were suggested on which site preference is plotted versus the difference in substrate and product isotopic composition (Koba et al., 2009). The determination of isotopic composition in the substrates is time consuming and additionally confounded in our study by the large and varying footprint area. Therefore, we present a modified isotope map of SP versus δ 15 N bulk (Fig. 7, left panel) instead of δ 15 N, the δ 15 N differences between substrate and product (i.e., N 2 O gas). Rectangles for the process groups N 2 O N and N 2 O D are defined by SP values given by Decock and Six (2013b) and by δ 15 N bulk values calculated based on process fractionation factors and substrate isotopic composition. For nitrification and denitrification, minimum and maximum fractionation factors of −90 to −40 ‰ and −40 to −15 ‰ were assumed (Baggs, 2008); for the isotopic compositions of the N 2 O precursors (i.e., NH + 4 and NO − 3 ) a range of −20 to +10 and −25 to 15 ‰ were assumed. Koba et al. (2009) attributed a concurrent decrease in δ 15 N bulk with increasing SP values as indicative of an increasing contribution of N 2 O N . In contrast, an increase in δ 15 N bulk in parallel with increasing SP values (enrichment of 15 N in the α position relative to the β position), as observed in the present study, was attributed to a substantial increase in N 2 O reduction to N 2 . Our results (Fig. 7, left panel) indicate that N 2 O is predominately formed by bacterial denitrification and that deviations in the isotope values from denitrification may have been caused by variations in the extent to which N 2 O was reduced to N 2 . It is noteworthy that, based on such modified isotope maps, systematic changes in δ 15 N bulk induced by systematic changes in the N isotopic composition of one of the precursors NH + 4 or NO − 3 could be misinterpreted as reduction events .
In addition to the SP vs. δ 15 N bulk maps, SP vs. δ 18 O maps have been suggested to trace N 2 O reduction to N 2 (Lewicka-Szczebak et al., 2014Well et al., 2012). While δ 15 N bulk depends on the isotopic composition of the precursor (e.g., NO − 3 ) and, thus, may vary considerably, δ 18 O-N 2 O is expected to be more stable as, during both nitrification and denitrification, oxygen (O) later found in N 2 O may almost completely originate from water (Kool et al., 2009). Due to this almost complete O exchange with water, relatively stable δ 18 O in soil water, and the observed constant ratio of fractionation factors for SP and δ 18 O-N 2 O (r sp-o ), variation in the share of N 2 O reduced to N 2 should be reflected in a linear re-lationship between SP and δ 18 O-N 2 O with a slope of 0.2-0.5 (Jinuntuya-Nortman et al., 2008;Ostrom et al., 2007;Well and Flessa, 2009a). In this study, a linear relationship with a slope of 1.02 was found (Fig. 7, right panel). Tracking the management events (Mana I to Mana III) and the rewetting event in SP vs. δ 18 O space revealed that the onset of such an event is associated with a decrease in both SP and δ 18 O, gradually increasing back to approximately initial values, except for Mana II. During Mana II, no significant change in SP vs. δ 18 O occurred (Fig. 7, right panel, red trace). The gradual increase in isotopic composition supports the conclusion from the SP / δ 15 N bulk map that N 2 O was mainly produced by bacterial denitrification and that variations in isotopic composition may have been caused predominately by N 2 O reduction to N 2 . This interpretation is in agreement with observations of the isotopic composition of N 2 O, NO − 3 and NH + 4 during a rewetting event in an agricultural field (Decock and Six, 2013a). Additionally, δ 18 O was found to be positively correlated with δ 15 N bulk , which enforces the interpretation that varying shares of N 2 O reduction occurred because it acts on both N and O isotopic composition (Koehler et al., 2012).
As stated above, the ratios of fractionation factors for δ 18 O and δ 15 N bulk (r o-n ) and SP and δ 18 O (r sp-o ) during N 2 O reduction were 2.5 and 0.2 to 0.5 in laboratory incubation experiments (Jinuntuya-Nortman et al., 2008;Ostrom et al., 2007;Well and Flessa, 2009a). In our study, r o-n and r sp-o were 0.5 and 1, respectively, for the whole data set. We also calculated these ratios for a subset of data for which all δ values (SP, δ 15 N bulk and δ 18 O) increased for 2 consecutive days, indicating that N 2 O reduction may have occurred. Such events were observed on eight occasions. If source processes (N 2 O D , N 2 O N ) contributed constantly over 2 consecutive measuring days, changes in the isotopic composition of emitted N 2 O were solely attributed to changes in the fraction of N 2 O reduction. Under such conditions one would expect that the ratio of the changes in δ 18 O and δ 15 N bulk (r o-n ) is around 2.5 and that the ratio of the changes in SP and δ 18 O (r sp-o ) is between 0.2 and 0.5. The mean (median) ratios for r o-n and r sp-o for these selected events were 0.69 (0.44) and 2.1 (1.16), respectively. While the high values of r sp-o indicate that, for instance, changing physical conditions such as soil moisture may play a role in field measurements, the deviation of r o-n from the value of 2.5 could either indicate that the fractionation factor for 18 O might be smaller than the one for 15 N or that there is no correlation of fractionation factors in natural environments. This is in line with recent findings showing that apparent isotope effects associated with N 2 O reduction are sensitive to experimental conditions which influenced diffusive isotope effects (Lewicka-Szczebak et al., 2014. The same study also showed that fractionation factors during N 2 O reduction for 15 N and 18 O were variable (from −11 to +12 ‰ and from −18 to +4 ‰, respectively) and not predictable for field conditions yet. Therefore, to date, the amount of N 2 O reduction prior to emission cannot be inferred with sufficient robustness from field measure-ments alone without the knowledge of the isotopic composition of the substrates.

Controls on isotopic composition and event-based data aggregation
The high temporal resolution of N 2 O isotopic and auxiliary measurements allowed us to investigate the controls on N 2 O isotopic composition over the 3-month campaign period. Correlations with isotopic composition were highest and positive for DOC and soil temperature ( Table 2). The significant correlation with temperature for the whole campaign was due to a significant correlation during the dry part of the campaign. If the increase in SP was due to an increased contribution of nitrification, δ 15 N bulk should decrease due to the higher isotopic fractionation during this process. The simultaneous increase in SP, δ 15 N bulk and δ 18 O revealed in Fig. 7, however, indicates an increased share of N 2 O reduction to N 2 , which might have been triggered by increased substrate availability (DOC) for heterotrophic denitrification. The reported effect of temperature on the N 2 O : N 2 ratio is not entirely certain, but a decrease has been observed with increasing temperature, supporting the hypothesis that N 2 O reduction increased as temperature rose throughout the measurement period (Saggar et al., 2013). Though substrate availability has been identified as a major control on N 2 O source processes (see references in Saggar et al., 2013), correlations between N 2 O isotopic composition and NO − 3 and NH + 4 concentrations were low, except for the correlation with δ 15 N bulk . The reason might be both the number of measurement points for substrate concentrations being lower compared to other explanatory variables and substrate concentrations not necessarily reflecting process or turnover rates (Wu et al., 2012).
The low explanatory power of all linear regressions underlines that drivers of N 2 O emissions are highly variable and may even change from event to event. In the absence of management or rewetting events (group BG), isotopic composition covered the whole range of measured values, while management or rewetting events were characterized by lower variability in isotopic composition. Values for SP, δ 15 N bulk and δ 18 O were low for Mana I, rewetting and Mana III, whereas Mana II showed increased SP, δ 15 N bulk and δ 18 O. This indicates that processes must have been different for Mana II, although management was almost identical.

Short-term variation in isotopic composition
The Keeling plot approach is based on the conservation of mass and assumes that the atmospheric concentration of a gas in the surface layer is a mixture of background atmospheric concentration and a variable amount of gas added by a source, raising the atmospheric concentration above background. The source's isotope value can be determined, given that its isotope value remains constant during the observation period. In this study, we used noon-to-noon data in the Keeling plots to determine isotope values of soil-derived N 2 O for the respective noon-to-noon period. Hence, the source processes underlying these N 2 O emissions have to be constant on this timescale. Currently, little is known about the rate of change of N 2 O source processes over time steps of minutes to hours. However, changing relative contributions of source processes, which change the isotopic composition in soil-emitted N 2 O, would be reflected in deviations from a linear relation between inverse concentration and isotopic composition. As the Keeling plots showed no obvious deviations from a linear relation within our measurement precision (see Supplement S1), we conclude (1) that the use of the Keeling plot approach was valid in our study and (2) that changes in N 2 O source processes in our study site occurred at a time step of 1 day or more. While our data suggest that there are little or no changes in source processes underlying N 2 O emissions within a noon-to-noon period, clear and distinct day-to-day variation in isotope values of soilderived N 2 O, especially in SP, were observed. Such changes were often strong and abrupt following management events (Mana I and III, Rewetting), indicating a significant response of microbial processes to the imposed disturbance. Larger than expected variability in isotope values was observed between management events (class BG) when no obvious variation in environmental drivers occurred. Since noon-to-noon concentration increases were very small during these periods, part of this variability may be attributed to increased uncertainty around the intercept of the Keeling plot. This is also reflected in the relatively large error bars around isotope values on days when overnight N 2 O concentration increases was low (Fig. 3). Alternatively, the variation in isotope values associated with small overnight concentration increase may result from other land use or land cover. The EC fluxes are calculated from the turbulent fluctuation of concentration and vertical wind speed (i.e., the covariance of the concentration and wind speed deviations from the half-hourly mean) and therefore account for the modulation of concentration around a short-term (30 min) mean caused by locally emitted N 2 O. Isotopic composition based on Keeling plots, however, is determined from total N 2 O accumulated in the nocturnal boundary layer, and, thus, this approach also contains molecules that were emitted outside the flux footprint, which almost exclusively comprised our grassland site (Zeeman et al., 2010), within the larger concentration footprint (Griffis et al., 2007). However, two facts indicate a major influence of the studied grassland on the determined N 2 O isotopic composition: First, the N 2 O isotopic composition is very stable for a noon-to-noon period as indicated by a linear relationship between individual measurements (Supplement S1). This relationship persists even though wind speed and direction change and, therefore, individual N 2 O isotope measurements integrate over a 16-min sampling interval originate from different source areas. Secondly, the CHA grassland can be characterized as a site with vigor-B. Wolf et al.: Real-time grassland N 2 O isotopic signature ous N 2 O emission, and therefore may dominate the determined N 2 O isotopic composition, as the influence of a source area increment scales with the source strength. The grassland site was restored in 2012, which led to an extraordinarily high N 2 O-N emission of 29.1 kg ha −1 year −1 (Merbold et al., 2014). In the following year 2.5 kg N 2 O-N ha −1 were released. This value is still in the range of maximum emissions reported for a different intensively managed Swiss grassland site, emitting 1.5-2.6 kg N ha −1 year −1 , and at least a factor of 5 greater compared to an extensively managed grassland site with less than 0.5 kg N ha −1 year −1 (Ammann et al., 2009). With regard to distant land use and land cover, the 2.5 kg N 2 O-N is also more than double the median (between the 70th and 75th percentile) of all reported values for cultivated temperate sites and is higher than the highest value reported for forests presented in a study containing 1008 N 2 O emission measurements from agricultural fields (Stehfest and Bouwman, 2006). However, the possibility cannot be excluded that the N 2 O isotopic signatures analyzed above the grassland site were influenced by adjacent ecosystems.

Flux-weighted averages of source isotopic compositions
N 2 O isotopic composition can be used to calculate and further constrain the global N 2 O budget (Kim and Craig, 1993;Yoshida and Toyoda, 2000). The analysis of emissions from different sources such as agricultural soils or managed grassland based on box models and isotopic composition is complicated by the distinct temporal and spatial variability of isotopic composition (Kim and Craig, 1993;Toyoda et al., 2011;Yoshida and Toyoda, 2000); hence, flux-weighted averages are required to obtain representative values for agricultural N 2 O (Perez et al., 2001). Our flux-weighted averages of 6.9 ± 4.3, −17.4 ± 6.2 and 27.4 ± 3.6 ‰ for SP, δ 15 N bulk and δ 18 O are well within the range of the values of 2.9 to 36.6, −41.5 to −1.9 and 23.2 to 51.7 ‰ for agricultural soils (Park et al., 2011;Toyoda et al., 2011). The comparison with other grassland soils (Opdyke et al., 2009;Park et al., 2011) indicates that the variability of isotopic composition within a group, such as grassland, may be considerable (for SP: 2.2 to 11.1 ‰). One has to keep in mind, however, that part of the observed variability may be attributed to the fact that the footprint area of the N 2 O isotopic composition includes areas with other land use or land cover. Another part of the variability might also be explained by a limited compatibility of different laboratory results, as recently demonstrated in an inter-laboratory comparison campaign . The uncertainty in budgets derived by isotopic composition depends on the uncertainty of the representative isotopic composition for a single source, which can be reduced by a quasi-continuous measurement approach, as shown in this study.

Conclusions
Our field observations indicate that nitrifier-denitrification and denitrification (process group N 2 O D ) dominated throughout the measurement period and that variation in isotopic composition was more likely due to variation in the extent of N 2 O reduction rather than contributions of NH 2 OH oxidation or fungal denitrification. A high temporal resolution of isotopic composition in soil-emitted N 2 O showed that, at the beginning of the growing season, medium wfps and low temperature induced low isotope values (representative for process group N 2 O D ), whereas in the second part of the measurement period, higher temperature and DOC stimulated N 2 O reduction to N 2 , although wfps was lower. Management or rewetting events were mostly characterized by low SP, δ 15 N bulk and δ 18 O, but the event Mana II indicated that processes underlying N 2 O emissions can vary even under similar management conditions. With this study, a new method is available that can provide real-time data sets for various individual N 2 O-emitting (eco)systems, such as grassland or agricultural soils, which will help in further constraining the global N 2 O budget based on box model calculations. However, future campaigns should be accompanied by footprint modeling in order to optimize of the inlet height and associated concentration footprint size.
The Supplement related to this article is available online at doi:10.5194/bg-12-2517-2015-supplement.