Evidence for the predominance of denitrification as a source of N2O in temperate agricultural soils based on isotopologue measurements



[1] Isotopologue values were used to quantify the contribution of denitrification to nitrous oxide (N2O) flux in agricultural and early successional fields in southwest Michigan. Nitrous oxide-δ15N and δ18O values were poor estimators of microbial origins compared to site preference (SP) (difference in δ15N between the outer and central N atoms of N2O). Site preference was used to evaluate the importance of denitrification (including nitrifier denitrification) in N2O production. Average flux-weighted SP values for each field ranged between 2.9 and 14.6‰ and, on the basis of SP values for N2O production from denitrification (0‰) and collectively for nitrification and fungal denitrification (37‰), these values indicate that between 61 and 92% of N2O originated from denitrification. Reduction of N2O ranged from undetectable to as much as 50% of production; and because reduction increases SP, our results underestimate the percentage of N2O from bacterial denitrification. The SP values in our study clearly indicate that denitrification is the predominant source of soil-derived N2O.

1. Introduction

[2] Nitrous oxide is an important greenhouse gas that reached an atmospheric concentration of 319 ppb in 2005 and continues to increase at an annual rate of 0.26% [Intergovernmental Panel on Climate Change (IPCC), 2007]. Considering these attributes and its long residence time (118 ± 25 years), N2O is recognized as an important driver of climate change [Intergovernmental Panel on Climate Change (IPCC), 2006, 2007]. Major sources of N2O include emissions from soils and oceans, with agricultural soils supplying ∼50% of total anthropogenic output of N2O [Mosier et al., 1998]. Enhanced N2O emissions from agricultural soils result from human induced alteration of soil microbial activity in response to increased nitrogen (N) inputs (e.g., fertilizers, cultivation of N fixing plants, and atmospheric N deposition) [IPCC, 2006, 2007; Mosier et al., 1998].

[3] Agricultural soils can be managed to mitigate emissions of greenhouse gases, however, this requires an understanding of the specific microbial process responsible for N2O production and reduction, and how the relative importance of different production pathways change with time and across ecosystems. The primary pathways of N2O production are nitrification and denitrification (Figure 1). Nitrification is the microbial oxidation of ammonium (NH4+) or ammonia (NH3) to nitrite (NO2) or nitrate (NO3) and occurs in the presence of oxygen. During this process, nitrifying bacteria produce nitrous oxide, which is formed as a byproduct of hydroxylamine (NH2OH) oxidation. Under oxygen (O2) limiting conditions, nitrifiers may also engage in denitrification, whereby NO2 is reduced to N2O; often termed nitrifier denitrification [Wrage et al., 2001, 2005]. Denitrification involves the successive reduction of NO3 to NO2, nitric oxide (NO), N2O and finally dinitrogen gas (N2), and predominantly occurs under anaerobic conditions [Firestone and Davidson, 1989]. During the successive reduction of NO3 by denitrification, N2O may leak out of cells. While the emphasis of most efforts has been on net N2O production in soils, reports of net negative N2O fluxes from soils indicate that net reduction also occurs [Cavigelli and Robertson, 2001; Chapuis-Lardy et al., 2007; Kellman and Kavanaugh, 2008].

Figure 1.

Production of N2O from hydroxylamine oxidation, nitrifier denitrification and denitrification (modified from Wrage et al. [2005]).

[4] A definitive means for evaluating the relative importance of nitrification and denitrification on the production of N2O and the influence of reduction has been elusive. This is, in part, because some approaches for apportioning N2O production between nitrification and denitrification rely on chemical inhibitors or isotope labeling approaches. The former inherently alters soil conditions and microbial activity while the latter frequently requires the incubation of soil samples rather than in situ experimentation [Bateman and Baggs, 2005; Tiedje et al., 1989; Wrage et al., 2005]. Alternative approaches for evaluating the relative proportion of N2O emitted from nitrification and denitrification which solely rely on bulk δ15N and δ18O analysis of N2O evolving from soils avoid biases imposed by inhibitors [Bol et al., 2003, 2004; Mandernack et al., 2000; Perez et al., 2000, 2001; Tilsner et al., 2003; Webster and Hopkins, 1996; Wrage et al., 2004a; Yamulki et al., 2000, 2001]. The natural isotopic abundance of N2O, however, is controlled by many factors in addition to its microbial origin; hence, this approach is marked by uncertainty. Signatures of δ18O in N2O, for example, are influenced by exchange of oxygen between intermediates and water during N2O production [Casciotti et al., 2002; Kool et al., 2009]. Methods based on δ15N rely on estimates of discrimination against 15N during a reaction and are frequently described by a net isotope effect (NIE), which is the difference in δ15N between N2O and the substrate of nitrification and denitrification (NH4+ or NO3, respectively). The difference in the NIE between nitrification (−45 to −66‰) and denitrification (−13 to −27‰) is the basis for apportionment [Perez et al., 2000]. The application of NIE for apportioning N2O is problematic because, unlike fractionation factors, the NIE can vary temporally, spatially and with substrate availability [Ostrom et al., 2002; Sutka et al., 2008]. Furthermore, the influence of reduction on bulk stable isotope values of N2O is not accounted for.

[5] A more recent approach for identifying microbial sources of soil-derived N2O, fragmentation ion compound specific isotope analysis (FICSIA), is based on the intramolecular distribution of 15N in N2O (isotopologues) [Sutka et al., 2003, 2006]. Isotopologue distributions for N2O are expressed as site preference (SP): SP = δ15Nαδ15Nβ. Here, δ15Nα and δ15Nβ reflect the abundance of 15N in the central (α) and terminal (β) atom of the linear N2O molecule, respectively. The isotope value for δ15Nα is determined from the fragment ion, NO, after correcting for rearrangement occurring within the ion source [Brenninkmeijer and Rockmann, 1999; Toyoda and Yoshida, 1999; Yoshida and Toyoda, 2000].

[6] Promising evidence of SP as a conservative tracer of N2O production was presented in pure culture studies [Sutka et al., 2003, 2006; Toyoda et al., 2005]. In monospecific cultures, SP values during N2O production from nitrification (∼33‰) and denitrification (∼0‰) are distinct, are constant during the course of N2O production and are not influenced by the nitrogen and oxygen isotope composition of the substrates [Sutka et al., 2003, 2004, 2006; Toyoda et al., 2005]. Note that in the work of Sutka et al. [2006] the SP of 0‰ associated with denitrification was found in N2O production from both denitrification and nitrifier denitrification. Therefore, in this manuscript we refer to denitrification as nitrate or nitrite reduction regardless of whether or not reduction is carried out by nitrifying or denitrifying bacteria, whereas, nitrification refers specifically to the oxidation of ammonia.

[7] In a tropical field study, SP values associated with nitrification and denitrification differed from those obtained in the culture studies [Perez et al., 2006]. However, SP values from this field study are confounded by the use of the inhibitor acetylene, calibration [Westley et al., 2006] and the potential for N2O reduction. Acetylene reacts with nitric oxide, inhibits the production of nitrate from nitrification, does not completely inhibit N2O production via nitrification, alters microbial activity and may not completely inhibit fungal production [Bateman and Baggs, 2005; Groffman et al., 2006; McKenney et al., 2001; Tilsner et al., 2003; Wrage et al., 2004a, 2004b]. Moreover, the affects of these perturbations on SP are unknown. Isotope effects associated with N2O reduction may influence SP and, to date, no field study has dealt with this issue [Jinuntuya-Nortman et al., 2008; Ostrom et al., 2007]. Thus, at present, the SP values from pure culture studies appear to be the most promising approach for resolving sources of N2O in the field and additional efforts on assessing the influence of reduction on SP are needed.

[8] Until recently, the influence of fungal N2O production on SP has not been considered [Sutka et al., 2008]. Fungal denitrification produces N2O with an average SP of 37‰ [Sutka et al., 2008], quite similar to the SP value for bacterial nitrification (33‰). Thus, while fungal denitrification and bacterial nitrification may not be distinguished on the basis of SP no process has yet been identified with a SP similar to that of N2O from denitrification. Therefore, in this study we estimate the percent N2O derived from bacterial denitrification relative to other sources.

[9] In this manuscript, SP is used to estimate the proportion of soil-derived N2O from denitrification on the basis of a simple mass balance equation with two end-members, and the influence of reduction on SP and associated apportionment estimates is considered retrospectively. The two end-members include bacterial denitrification with a SP of 0‰ and nitrification and fungal denitrification, collectively, with a SP of 37‰. By using the maximum SP value for N2O production from nitrification and fungal denitrification the percent N2O derived from bacterial denitrification will be slightly underestimated.

[10] This paper presents N2O flux, bulk isotope and SP results from two complimentary field studies in southwest Michigan. The first part of this study addresses the affect of treatment history on isotopologue values by comparison of data from three agricultural fields with unique treatment histories at four time periods between May and November 2006. The treatments include (1) an unfertilized, tilled field grown in soybean; (2) a tilled and fertilized (chemical and manure) field grown in corn; and (3) an unfertilized early successional field abandoned from agriculture in 1989. Samples were collected within three different plots with the same treatment history (soybean and successional) or three different locations within the same treatment (corn). A second part of this study addresses temporal variation in isotopologue values via a higher-resolution data set from a single chamber collected within a soybean field and consists of 12 time periods between May and November 2006. In both cases we evaluate the use of bulk isotopes for source apportionment, apply SP to estimate the percentage of N2O derived from bacterial denitrification and constrain the affect of N2O reduction on SP using the approach of Ostrom et al. [2007]. Thus, our study addresses previously unaccounted for issues in the application of SP to the field and provides insight into the enigmatic microbial origins of soil-derived N2O.

2. Methods

2.1. Research Site

[11] Isotopologue measurements were taken at Michigan State University's W. K. Kellogg Biological Station (KBS) in southwest Michigan between May and November 2006. This region of the U.S. corn belt (42°24′N, 85°24′W) receives on average 890 mm yr−1 of precipitation and has a mean annual temperature of 9.7°C. Percent carbon, percent nitrogen, pH and bulk density, as determined from the KBS Long-term Ecological Research Web site (http://lter.kbs.msu.edu/datasets), were similar across sampling sites, averaging 0.9%, 0.09%, 6.3 and 1.6 g/cm3, respectively. Climate, soil and management are described in more detail at the KBS Long-term Ecological Research Web site (http://lter.kbs.msu.edu/datasets).

[12] Soils at KBS are developed from glacial outwash deposited at the end of the Wisconsin glaciation and are Typic Hapludalfs (fine-loamy, mixed, mesic) of moderate fertility. The agro-ecosystem treatments sampled included (1) an unfertilized, conventionally tilled agricultural field planted in soybean as part of a corn-soybean-wheat rotation (AGsoy); (2) a conventionally tilled agricultural field planted in corn, which received manure fertilizer in February and September of 2006 (AGcorn); and (3) an unfertilized, early successional community dominated by herbaceous perennials (SUCC) abandoned from agricultural activities in 1989.

[13] The AGsoy and SUCC fields are located at the Long-term Ecological Research (LTER) site within KBS (http://www.kbs.msu.edu). The LTER site is subdivided into seven 1-ha agro-ecosystem treatments, each segmented into smaller replicate field plots. Three replicate field plots were randomly chosen for sampling within the AGsoy and SUCC treatments. The AGsoy treatments have been in a corn-soybean-wheat rotation since 1988 and the SUCC fields were abandoned in 1989, having previously been treated as conventionally tilled agricultural fields. Sampling within the AGcorn field was performed within a subplot of a larger agricultural field (∼10 ha) approximately 1 km from the boundaries of the LTER site, but still part of KBS. Historically, the agricultural field had been planted in a corn-soybean-alfalfa rotation, with inorganic and manure fertilizer applied when required.

2.2. Field Sampling

[14] Soil temperature and moisture were recorded using a soil temperature recorder buried 5 cm beside the flux chambers. Collection of N2O samples from flux chambers placed in the three agro-ecosystem treatments occurred four times between May and November 2006. One flux chamber was placed in each of three replicate field plots for both the AGsoy and SUCC agro-ecosystem treatments, and within the AGcorn treatment, one flux chamber was placed at three locations within a single field plot (96 m2) chosen to facilitate sampling and avoid conflict with agricultural activities. At each sampling time, three consecutive samples were taken from a single flux chamber. Additionally, to better constrain temporal variation in isotopologue values a higher-resolution data set was obtained within the AGsoy treatment, but only from one of the three replicate field plots in this treatment. Higher-resolution samples were collected once in May, June, and November, with biweekly sampling in July, August, and September. During each sampling, samples were collected in triplicate.

[15] Two weeks prior to the first sampling period in May, the AGsoy and AGcorn fields were plowed, with the AGcorn fields receiving approximately 1.0 ton manure per hectare between January and February. Manure was not added to the AGcorn fields again until the second week of September at a similar loading. During July, soybean and corn crops were in early growth stages and herbaceous vegetation covered approximately 100% of the ground surface at the SUCC fields. By the September sampling period, soybean was in late growth stage and corn had been harvested 1 week prior. Soybean was harvested 1 week prior to the November sampling period and at this time vegetation was senescing in the SUCC fields.

[16] Samples of air were collected from within cylindrical polyvinyl chloride (PVC) soil gas flux chambers (surface area = 500 cm2; headspace volume = 8.4 L) buried 5 cm deep in soil and covered with an airtight PVC lid sealed with a viton o-ring. Flux chambers were maintained at atmospheric pressure by a stainless steel tubing (0.5 m × 0.3 cm OD) “pigtail” extending through the wall of the chamber into the interior that was open to the atmosphere. A leak test was conducted to ensure that gases were not escaping the chambers through lids or tube extensions. A 30 mL sample of 100% CO2 was injected into a sealed chamber through a side-sampling port. The top and bottom of the chamber was sealed with clamped lids while the pigtail remained open. Percent recovery was measured at 2, 6 and 24 h. At all time intervals, recovery was approximately 100% and constant with time, confirming our ability to evaluate flux during the extended closing periods. Closing periods for the field chambers ranged between 6 and 48 h to assure a sufficient amount of N2O for precise isotopologue analyses. Samples were not collected with time to assure linear trends in N2O fluxes but we have observed linear fluxes at comparable concentrations in previous studies at KBS (long-term N2O flux data since 1989). Three gas samples were collected at the end of each closing from flux chambers in preevacuated 60 or 250 mL glass bottles (depending on sample size requirement), which were capped with Geo-Microbial® butyl rubber septa. During sampling, a volume of atmospheric air equivalent to the sample bottle volume was passively introduced via the pigtail. For this reason, concentration and isotopologue values were corrected for the introduction of atmospheric N2O on the basis of reported values for tropospheric N2O [Yoshida and Toyoda, 2000]. In all cases, this correction was minor and within the limits of analytical uncertainty. All sample bottles contained magnesium perchlorate pellets (3 mL in glass) to absorb moisture. Laboratory tests confirmed that magnesium perchlorate improved accuracy and precision of isotopologue measurements. Gas samples were transferred from the chamber to the bottles through peek tubing (0.5 m × 0.6 cm OD) that extended through the wall of the chambers and was externally connected to an 18-gauge needle. Concentrations of N2O were determined during isotopic analysis on the basis of the m/z 44 ion beam signal within the mass spectrometer following cryogenic purification and chromatographic separation on a GV Instruments Trace Gas Inlet System [Sutka et al., 2003]. The N2O fluxes were, therefore, specific to the period of measurement.

2.3. Isotopologue Analysis

[17] Analyses of the isotopologue compositions of N2O were based on FICSIA. Nitrous oxide fluxes and FICSIA measurements were determined within 1 month of collection on a multicollector IsoPrime mass spectrometer (GV Instruments) interfaced with a continuous flow trace gas inlet system. As described by Sutka et al. [2003], this instrument simultaneously monitors five masses for N2O isotopologues and the fragment ion NO: 30, 31, 44, 45 and 46. The ion beam ratios of 45:44, 46:44 and 31:30 provide the raw data for the calculation of δ15Nbulk, δ18O and δ15Nα. The δ15Nbulk is an average of the nitrogen isotope values of the central (α) and terminal (β) N atoms and allows for the determination of δ15Nβ:

equation image

Raw isotope ratios are corrected for the contribution of δ17O to masses 31 and 45 and rearrangement of δ15N between the central (α) and terminal (β) N atom positions [Toyoda and Yoshida, 1999]. The internal reference laboratory tank standard used for all analyses was calibrated for isotopologue abundances in collaboration with the laboratory of Naohiro Yoshida at the Tokyo Institute of Technology as recommended by Westley et al. [2007]. All δ15N and δ18O values are expressed with respect to Air and Vienna Standard Mean Ocean Water international standards and defined as

equation image

where R is 15N/14N or 18O/16O. Our instrumental precision for bulk δ15N and δ18O is 0.5‰ and for SP is 1.3‰.

[18] The N2O collected in flux chambers is a mixture of atmospheric and soil-derived N2O. Consequently, the values we report have been corrected for the contribution of N2O from the atmosphere assuming N2O concentration, δ15Nbulk, δ15Nα, δ15Nβ, and δ18O values of 320 ppb-v, 7.0, 16.9, −1.80, and 43.7‰, respectively [Toyoda and Yoshida, 1999], and the following mass balance equation:

equation image

where δmeasCmeas, δatmCatm, and δSDCSD are the product of the isotope values and concentrations of measured, atmospheric, and soil-derived N2O, respectively. We do not report isotopologue values when soil-derived N2O was less than 30% of the ambient atmospheric N2O concentration. This is because the error associated with the calculation of soil-derived isotopologue values by equation (3) increases with a decrease in the proportion of soil derived to atmospheric N2O [Ostrom et al., 2007].

2.4. Apportioning N2O Flux

[19] SP is used to estimate the proportion of soil-derived N2O from denitrification on the basis of a simple mass balance equation with two end-members: denitrification with SP of 0‰ and nitrification and fungal denitrification collectively with a SP of 37‰. The influence of N2O reduction on SP and associated apportionment estimates is evaluated on the basis of the methods of Ostrom et al. [2007]. Note that in referring to N2O reduction, we are not referring to the production ratio of N2 to N2O from denitrification that can be obtained by 15N labeling studies [Bergsma et al., 2001, 2002]. Rather, we are referring to gross N2O reduction; both atmospheric and that which is produced microbially. The approach of Ostrom et al. [2007] shows that in the absence of reduction, in a plot of N2O δ18O versus δ15N values, data fall on a “mixing line” between the isotopic composition of atmospheric N2O and soil-derived N2O. As soil-derived N2O is commonly depleted in 15N and 18O relative to atmospheric N2O the slope of the mixing line is commonly less than 1 [Ostrom et al., 2007]. When reduction in the absence of production occurs the relationship between δ18O and δ15N is constant with a slope of 2.6 [Ostrom et al., 2007; Jinuntuya-Nortman et al., 2008]. Data that are influenced by reduction simultaneous with production fall between the slopes of these two lines. The magnitude of the shift from the mixing line is proportional to the extent of reduction and provides a means to estimate the shift in SP when reduction occurs [Ostrom et al., 2007] (see Text S1 and Figure S1).

2.5. Statistical Analysis

[20] Differences in flux-weighted average δ15N, δ18O, and SP among the three agro-ecosystems with different treatment histories were assessed by one-way analysis of variance followed by post hoc pair-wise comparisons (Bonferroni). A Bonferroni's adjustment was used since it is more conservative and controls the family wise type II error rate. A Kruskal-Wallis test was used to assess differences in flux among the three agro-ecosystem treatments. Homogeneity of variance was evaluated with Levene's test, and normality was evaluated by a normality plot and companion statistics (Lilliefors, Shapiro-Wilk). Alpha was set at 5% and all significance tests were performed in SYSTAT (version 12).

3. Results

3.1. Characteristics and Management of Treatment Field Plots

[21] Throughout the study, precipitation was highly variable with mean daily air temperatures lowest in November and highest between July and August. Accumulative precipitation 7 days before each sampling period was 83.6, 17.8, 0.0 and 1.9 mm in May, July, September, and November, respectively (http://lter.kbs.msu.edu/datasets). Daily mean soil temperatures at 5 cm depths during the four sampling dates were 16.8, 23.9, 18.7 and 2.5°C for May, July, September, and November, respectively. Percent soil moisture was measured in September and November, averaging 8.6 and 24.9%, respectively, among all fields. Data from the KBS LTER indicate that soil moisture is greatest in spring and fall and least in late summer (http://lter.kbs.msu.edu/datasets).

3.2. Comparison of Flux and Isotopologue Values Among Treatments

[22] Analysis of variance indicated that N2O flux did not differ significantly among sites (Kruskall-Wallis test, H = 2.08, p = 0.353). A comparison of flux between sites by month shows that soil-derived N2O fluxes were greater at the AGcorn fields than the AGsoy and SUCC fields in May, July, and September. In May, fluxes averaged 46.9, 3.5 and 3.1 g N2O-N ha−1 d−1 for AGcorn, AGsoy, and SUCC, respectively (Figure 2). In July and September, fluxes at AGcorn averaged 2.5 g N2O-N ha−1 d−1 and were <0.7 g N2O-N ha−1 d−1 for the AGsoy and SUCC treatments. The average flux for all treatments in November was 0.3 g N2O-N ha−1 d−1, with an outlier in the SUCC field being 4.8 g N2O-N ha−1 d−1.

Figure 2.

Soil-derived N2O fluxes in the early successional (SUCC) and agriculture planted in soybean (AGsoy) and corn (AGcorn) treatment plots at the W. K. Kellogg Biological Station between May and November 2006. Each bar represents flux chamber data from one of three replicate plots per treatment. The labels above the bars are in percent N2O flux derived from denitrification averaged over treatment and sampling period. Asterisks indicate fluxes below detection limits of 0.05 g N2O ha−1 d−1. Error bars represent the standard deviation about the mean of three consecutive samples taken from a single flux chamber.

[23] Soil-derived δ15N-N2O values were variable between treatments and sample periods throughout the study (Figure 3a). Nitrogen isotope values ranged from −18 to 11‰. To provide temporally integrated values for each field we report flux-weighted isotope values. Here, each isotope measurement is weighted by the flux of N2O present at that time, summed for all time points, and divided by the total flux during our study. Even though the δ15N flux-weighted averages for the entire study ranged between −16.1 and 9.7‰, with the lowest value in AGsoy and the highest value in SUCC (Table 1), an analysis of variance showed no difference in flux-weighted average δ15N-N2O values among sites (F2,19 = 2.723, P = 0.091). Temporally, δ15N flux-weighted averages increased from May to November at AGsoy (−16.1 to 0.6‰) and decreased in AGcorn (−3.7 to −13.3‰).

Figure 3.

(a) Soil-derived δ15N-N2O, (b) δ18O-N2O, and (c) site preference at the early succession (SUCC), agriculture planted in soybean (AGsoy), and agriculture planted in corn (AGcorn) treatment plots at the W. K. Kellogg Biological Station between May and November 2006. Each bar represents flux chamber data from one of three replicate plots per treatment. Asterisks indicate that N2O fluxes were below detection limit for isotopologue analyses (detection limits of the isotopic measurements were dependent on the length of the incubation period, averaging 0.2 g N2O ha−1 d−1). Error bars represent the standard deviation about the mean of three consecutive samples taken from a single flux chamber.

Table 1. Flux-Weighted Averages for Soil-Derived δ15N-N2O, δ18O-N2O, and Site Preference at Three Treatment Fieldsa
 δ15N-N2O (‰)δ18O-N2O (‰)Site Preference (‰)
  • a

    Treatment fields are succession (SUCC), agriculture planted in soybean (AGsoy), and agriculture planted in corn (AGcorn) at the W. K. Kellogg Biological Station between May and November 2006.

16 May 2006
10 July 2006
8 September 2006
5 November 2006
Mean For Entire Study Period
Mean of all treatments−5.332.04.7

[24] Over time and treatment, soil-derived δ18O-N2O values exhibited a large range of 22 to 63‰ (Figure 3b). The flux-weighted averages for δ18O ranged between 22.2 and 57.3‰ for the entire study (Table 1). Analysis of variance indicated significant differences in flux-weighted δ18O values among sites (F2,19 = 7.020, P = 0.005). Results from pair-wise comparisons indicated that flux-weighted δ18O values from AGcorn differed from SUCC (P = 0.005) and that AGsoy also differed from SUCC (P = 0.005). At each sampling period, flux-weighted δ18O values were higher in AGsoy compared to the other fields; 33.0 and 35.9‰ in November and May, respectively, and between 47.9 and 57.3‰ in July and September. Among the fields, SUCC had the lowest flux-weighted average δ18O values (22.2 to 28.5‰).

[25] In May, July, and September the SP values for soil-derived N2O were lower in AGcorn than AGsoy and SUCC (Figure 3c). In AGcorn, the low SP values in May coincided with the highest N2O flux. In AGsoy, SP decreased from May to September as shown in Figure 3c. The flux-weighted average SP calculated for May through November was 2.9, 11.1 and 14.6‰ for soils from AGcorn, SUCC and AGsoy, respectively (Table 1) and the analysis of variance indicated that differences in flux-weighted average SP values among sites were significant (F2,19 = 13.134, P = 0.046). Results from pair-wise comparisons indicated that AGcorn and AGsoy differed (P = 0.043).

[26] To quantify the percentage of N2O derived from bacterial denitrification we use an isotope mixing model. We found that 92, 61 and 70% of N2O production originated from denitrification in AGcorn, AGsoy, and SUCC, respectively. Temporally, the contribution of denitrification to N2O production increased from May to November at the SUCC field (57 to 90%), was least during July and September in AGsoy (47 and 49%) and greatest in May and September in AGcorn (94 and 95%, respectively) (Figure 2).

3.3. Higher-Resolution Data in AGsoy

[27] To better constrain temporal variation in fluxes and isotopomer values, samples were collected from one field plot in AGsoy at a higher frequency than in the treatment history study. Soil-derived N2O fluxes ranged from less than 0.2 g N2O-N ha−1 d−1 in August, when atmospheric temperatures were at their greatest and precipitation was minimal, to 2.4 g N2O-N ha−1 d−1 in September, 1 week after harvest (Figure 4). There was a general increase in δ15N values between June and November (−13.3 to −2.0‰), with a flux-weighted average of −7.0‰ for the entire study period (Figure 5a). The δ18O values, were greatest in the summer (June and July: 50.5 to 66.1‰, respectively) and least in the fall (September and November: 30.5 and 28.6‰, respectively). The flux-weighted average for the entire study period was 46.5‰ (Figure 5b). Site preference decreased between June and July (range of 19.8 to 32.5‰) and September and November (10.4 and −1.2‰, respectively) (Figure 5c). The flux-weighted SP for the four sampling periods was 18.0‰, suggesting that denitrification contributed to 51% of the N2O production during this time.

Figure 4.

Temporal trends of soil-derived N2O fluxes at the agriculture planted in soybean (AGsoy) treatment plots at the W. K. Kellogg Biological Station between May and November 2006. Each point represents the mean of three samples taken consecutively from a single chamber in the same plot. Error bars represent the standard deviation about the mean of the three consecutive samples.

Figure 5.

Temporal trends of (a) soil-derived δ15N-N2O, (b) δ18O-N2O, and (c) site preference at the agriculture planted in soybean (AGsoy) treatment plots at the W. K. Kellogg Biological Station between May and November 2006. Each point represents the mean of three samples taken consecutively from a single chamber in the same plot, and the error bars represent the standard deviation about the mean. The labels above the points in Figure 5c are mean percent of the proportion of N2O from denitrification for that sampling period.

3.4. Influence of N2O Reduction on SP and Apportionment Estimates

[28] On the basis of the approach of Ostrom et al. [2007] we saw little or no evidence of reduction for some of our data (see Text S1 and Figure S1). Other cases could have experienced as much as 50% reduction (upper limit). Because reduction serves to increase SP, our apportionment estimates to denitrification are conservative. For example, if 50% reduction had occurred, the SP that we report of 4.7‰ for the overall flux of N2O from KBS soils would have been 0.7‰ in the absence of reduction. In this case, the proportion of N2O from denitrification would increase to 98%.

4. Discussion

[29] Nitrogen isotope values of N2O have been used to apportion production to nitrification and denitrification on the basis of the difference between the δ15N of N2O and the substrates of these reactions (NH4+ and NO3) [Bol et al., 2003; Perez et al., 2001]. This approach relies on the observation that production of N2O from denitrification induces a NIE of −13 to −28‰, which is of a smaller magnitude to that of production from nitrification, −45 to −66‰ [Barford et al., 1999; Ueda et al., 1999; Yoshida, 1988]. Using the flux-weighted average δ15N for soil NO3 at KBS of 2.5‰ [Ostrom et al., 1998] to represent the nitrogen isotope value of the substrate, the NIE for our data, 7 to −19‰, falls within or above the range of values indicative of N2O production from denitrification (−13 to −28‰). However, like Perez et al. [2001], our estimate of the NIE relies on a single value for the δ15N of the substrate and, thus, does not account for isotopic variation over the course of a reaction or season. Feigin et al. [1974], for example, found that the δ15N of ammonium increased by ∼13‰ upon fertilizer application. Such a shift would result in an increase in the δ15N of N2O and an apparent decrease in NIE. In this case, estimates of NIE would underestimate the magnitude of isotopic discrimination and overemphasize the influence of denitrification.

[30] Apportionment approaches based on NIE associated with δ15N do not account for variation in fractionation factors associated with N2O production nor for enrichment in 15N during N2O reduction [Menyailo and Hungate, 2006; Ostrom et al., 2007; Sutka et al., 2008; Vieten et al., 2007; Jinuntuya-Nortman et al., 2008]. Variation in a fractionation factor can result from changes in the relative importance of the reaction rate, constants for diffusion and enzymatic reduction that occur during an experiment [Jinuntuya-Nortman et al., 2008; Sutka et al., 2008]. When a reaction is controlled by the rate constant for diffusion, the resulting fractionation factor will be small and the opposite is true if the rate constant for the enzyme predominates. Any variation in the fractionation factor will also influence the NIE and this has implications for apportionment. If the rate constant for diffusion begins to prevail, the NIE associated with N2O production from nitrification will decrease and obscure our ability to distinguish nitrification and denitrification on the basis of the NIE. Thus, although the range of NIE values we observed for δ15N (7 to −19‰) are consistent with the range of values reported for denitrification (−13 to −28‰), N2O production from nitrification cannot be eliminated on this basis.

[31] Because the light isotope is preferentially transferred to the product during kinetic reactions, the NIE should not exceed 0‰. Thus, the positive NIE values observed at KBS could indicate that (1) the 15N content of NO3 was enriched above average KBS values; (2) rate constants for diffusion control the NIE, and the NIE approaches 0‰; (3) nitrate was enriched in 15N due to denitrification; and/or (4) N2O reduction is operative in addition to production. These issues challenge the use of NIE to define microbial sources of N2O and, thus, δ15N and NIE are not convincing source indicators.

[32] Similar to δ15N, variation in the oxygen isotope of N2O presents a problem for their application in source apportionment. Oxygen isotope values reflect substrate isotopic compositions and fractionation associated with production pathways. During the microbial production of NO3 by nitrification, the first O atom added is from atmospheric O2 with a δ18O value of 23.5‰, whereas, the second and third are derived from water with isotope values commonly less than 0‰ [Ostrom et al., 2000]. Consequently, δ18O-N2O values lower than 23.5‰ must be influenced by water either through reduction of NO2 or NO3 or via exchange reactions. Kool et al. [2009] recently showed that an average of 85% of O-N2O was derived from exchange with water rather than from NO3. Our δ18O-N2O values (21.7 to 63.0‰) were rarely less than that expected for atmospheric O2, which suggests that isotope fractionation effects must be important. The majority of our values were well above that of atmospheric O2, which could reflect fractionation during a number of processes including microbial respiration of O2, production of N2O and reduction of N2O. The overall complexity of isotope effects during N2O production prohibits definitive source identification on the basis of δ18O.

[33] When we neglect reduction in the consideration of our data, the flux-weighted average SP values for soil-derived N2O during the study periods indicate that 71, 60 and 92% of N2O production was derived from denitrification in SUCC, AGsoy, and AGcorn fields, respectively. The high proportion of N2O from denitrification in the AGcorn field was likely enhanced by the addition of fertilizer and carbon associated with organic manure; inputs that favor denitrification. The proportion of N2O produced from the AGsoy fields was approximately balanced between nitrification and denitrification, with a greater proportion of nitrification compared to the other fields. While the AGsoy fields were not fertilized, tillage and the presence of N-fixing soybean plants may favor production of N2O from nitrification. The proportion of N2O from denitrification averaged over all treatments and time periods was 87% (or 74% if the large N2O flux occurring in May at AGcorn is excluded). Thus, our results support the idea that the majority of N2O production in temperate agricultural soils may be from bacterial denitrification and nitrifier denitrification.

[34] We used the approach of Ostrom et al. [2007] to assess the impact of N2O reduction on our estimates of SP and associated apportionment estimates (see Text S1 and Figure S1). While reduction had little or no influence on some of our data, in other cases there was ample evidence of reduction. The upper limit of N2O reduction was approximately 50%. However, reduction does not compromise our conclusion that denitrification is an important source, and in most cases the dominant production pathway in KBS soils. This is because reduction serves to increase SP and leads to an underestimate of the percentage of N2O derived from denitrification. Thus, our determination that 87% of N2O production was via bacterial denitrification is likely an underestimate.

[35] The flux-weighted average for SP within the AGsoy higher-resolution data set determined over the study period was 18.0‰, suggesting that the proportion of N2O from denitrification was 51%. The range in SP for June and July was 20.0 to 32.5‰, dropping to less than 10.4‰ in September and November. On the basis of SP values of 0‰ for denitrification and 37‰ collectively for nitrification and fungal denitrification [Sutka et al., 2006], the origin of N2O was predominately from nitrification and/or fungal denitrification in summer shifting to denitrification later in the year. The high δ18O values for N2O from the higher-resolution AGsoy study strongly suggest that N2O reduction has altered the isotope signals in this data set by as much as 80% (on the basis of the model of Ostrom et al. [2007]). Consequently, because reduction serves to increase estimates of production from dentrification, we believe that our overall estimate of 51% of N2O production from denitrification in the higher-resolution AGsoy data set is conservative. Thus, even in a treatment that favors nitrification during certain months (AGsoy during June and July), denitrification appears to be the predominant source of N2O production over the growing season.

5. Conclusions

[36] The use of SP is a powerful tool for differentiating microbial sources of N2O in soil ecosystems [Bol et al., 2003; Sutka et al., 2006, 2008; Well et al., 2006]. SP is unaffected by O exchange with substrates, water and atmospheric O2, and provides a conservative tracer of production [Sutka et al., 2006; Toyoda et al., 2005]. Furthermore, the natural abundance isotope approach offered by SP avoids alteration of soil conditions inherent in incubation tracer-level studies. Nonetheless, it is clear that SP can be influenced by reduction of N2O in soils [Jinuntuya-Nortman et al., 2008; Ostrom et al., 2007] and our results indicate that reduction must be considered in flux chamber studies. On the basis of the approach outlined by Ostrom et al. [2007] we estimate that although reduction may not occur in some cases, as much as 50% of the N2O produced in some KBS soils is reduced. Because reduction increases SP, reduction does not compromise our conclusion regarding the predominance of bacterial denitrification in N2O production in KBS soils. While the field grown in soybean exhibited a strong signal from either nitrification or fungal denitrification in early summer, our data overwhelmingly indicates that denitrification and nitrifier denitrification is the predominant source of N2O (87%) in agricultural soils at the KBS LTER under management for corn and soybean and in successional fields. In finding that denitrification and nitrifier denitrification are collectively, the primary source of soil-derived N2O our study addresses a long-standing question as to the microbial origins of N2O evolving from soils.


[37] We would like to acknowledge the combined efforts of personnel at the Kellogg Biological Station for assisting in experimental design, particularly Kurt Smemo for his contribution in collecting samples. Hasand Gandhi assisted with instrumentation and analysis and James Humpula provided field assistance. Financial support was provided by the National Science Foundation (DEB 0316908 and LTER program) and the Michigan Agricultural Experiment Station.