The biogeochemistry of nitrous oxide (N2O) was investigated in Lake Kizaki, Japan, where accumulation of N2O in the water column has been observed. The N2O concentration profile showed weak accumulation in the oxic zone, although much higher and much lower N2O concentrations were observed in the deeper oxygen-deficient zone. Intramolecular partitioning of 15N (site preference) of N2O within the oxic zone increased concomitantly with increased N2O concentration. The site preference of the newly produced N2O in the oxic zone was estimated as 33.6‰. This high site preference strongly suggests that this N2O was produced by hydroxylamine oxidation. In regions of the oxygen-deficient zone, the nitrate (NO3−) concentration decreased rapidly, concomitantly with increased nitrogen and oxygen isotope ratios, indicating denitrification. The high site preference and nitrogen isotope ratio of N2O (δ15Nbulk), combined with isotopic data of NO3−, strongly suggest denitrification as the main N2O source. Moreover, site preference and δ15Nbulk data suggest that the existing N2O in the oxygen-deficient zone was already strongly reduced (more than 75%) to N2. Results of this study demonstrate the feasibility of using isotope and isotopomer analyses of N compounds to elucidate the complex biogeochemistry of N2O in an intact ecosystem.
 Nitrous oxide (N2O), a greenhouse gas that is produced in the nitrogen (N) cycle, is produced as a by-product of hydroxylamine (NH2OH) oxidation in nitrification, a combined process of the oxidations of ammonium (NH4+) to nitrite (NO2−) via NH2OH and NO2− to nitrate (NO3−). Autotrophic ammonia oxidizers perform ammonia oxidation from NH4+ to NO2−. Regarding denitrification, which is a microbial NO3− respiration in anaerobic environments, reduction of NO2− following NO3− reduction produces N2O via nitric oxide as an intermediate product. The resultant N2O can be reduced further to di-nitrogen (N2) gas (N2O reduction). Denitrification is performed by heterotrophic denitrifiers mainly under low-oxygen conditions, although autotrophic ammonia oxidizers can also carry out NO2− reduction (“nitrifier-denitrification” [Wrage et al., 2001]). Hereinafter, “nitrification” refers to the whole ammonia oxidation process to NO3−, whereas “denitrification” denotes NO3− reduction to N2. Furthermore, we specifically use the term “N2O reduction” to emphasize this last enzymatic process in denitrification. When referring to “NO2− reduction,” we include the reduction of NO2− to N2O driven by heterotrophic denitrifiers (denitrification) and autotrophic ammonia oxidizers (nitrifier-denitrification).
 Nitrification is regarded as dominating N2O production in oxic ocean water [Yoshinari, 1976; Ostrom et al., 2000; Popp et al., 2002] and in oxic lake water [Yoh et al., 1988; Yoh, 1990; Priscu et al., 2008], although it has been suggested that denitrification can occur even in oxic ocean water [Yoshida et al., 1989; Yamagishi et al., 2005], possibly in association with anaerobic organic particles [Alldredge and Cohen, 1987]. In addition, nitrifier-denitrification can occur even in oxic ocean water because this process might not be a strictly anaerobic process [Shaw et al., 2006]. In low dissolved-oxygen (DO) environments, denitrification is thought to dominate N2O production in oceans [Gruber and Sarmiento, 1997] and in lakes [Mengis et al., 1997]. However, nitrification in oxygen-deficient environments has been reported frequently in oceans (summarized by Molina and Farías ). Aerobic ammonia oxidizers can be active and can even increase the ratio of N2O/NO2− production in such oxygen-deficient environments [Goreau et al., 1980]. Because of this complexity, evaluating the relative importance of NH2OH oxidation, NO2− reduction, and N2O reduction remains difficult [Baggs, 2008; Codispoti, 2010] despite the remarkable differences in oxygen and carbon demands among these processes. Consequently, an understanding of production and consumption processes of the N2O in an ecosystem is important not only to develop a means to control the emissions of this greenhouse gas but also to elucidate the complex N cycle that is regulated partly by the balance of supply and demand for carbon and oxygen.
 Although incubation studies using tracers (e.g., 15N) and inhibitors (e.g., acetylene to stop the ammonia oxidation and N2O reduction) provide important insight into these N transformation processes, noninvasive approaches are important to elucidate the processes occurring in situ. Natural abundances of N and O isotopes (δ15N and δ18O) of N2O have been used to study N2O biogeochemistry [e.g., Kim and Craig, 1993]. The net isotope effect (NIE; the difference in the isotope values between the product and substrate) has been applied to apportion N2O production to nitrification and denitrification [e.g., Kim and Craig, 1990]. However, this approach cannot always work because NIE and the actually expressed isotope fractionation factors can vary temporally, spatially, and according to substrate availability [Ostrom et al., 2002, 2007, 2010; Jinuntuya-Nortman et al., 2008]. In addition, the isotopic signature of the substrate for N2O (inorganic N such as NH4+ and NO3−) in the ecosystem tends to be nonconservative and highly variable [Ostrom et al., 2002]. Along with δ15N and δ18O of N2O, observation of the intramolecular distribution of 15N (site preference) [Yoshida and Toyoda, 2000] presents a new promising tool to determine the N2O produced by NH2OH oxidation from NO2− reduction [Sutka et al., 2006]. Hydroxylamine oxidation produces N2O with high site preference, whereas NO2− reduction produces N2O with low site preference [Sutka et al., 2006]. Furthermore, site preference presents the great advantage that these distinct site preference are independent of the isotopic signature of substrates such as NH4+ and NO3− [Toyoda et al., 2005] against δ15N and δ18O of N2O, which can be influenced strongly by the isotopic signatures of substrates. Recently, Frame and Casciotti  used extensive incubation experiments to estimate site preference for NH2OH oxidation (36.3 ± 2.4‰) and NO2− reduction (−10.7 ± 2.9‰). The use of this clear difference in site preference can elucidate the sources and sinks of N2O in aquatic systems [Popp et al., 2002; Toyoda et al., 2002, 2009; Well et al., 2005; Yamagishi et al., 2005, 2007; Westley et al., 2006; Charpentier et al., 2007; Farías et al., 2009]. Furthermore, the site preference of N2O can be a more powerful parameter for the elucidation of N2O dynamics in aquatic ecosystems when combined with isotope ratios of other N compounds such as NO3− and NH4+ [Priscu et al., 2008].
 Nutrient dynamics in Lake Kizaki, a temperate, dimictic lake in central Japan, were investigated intensively during the 1970s and 1980s. Remarkably different N dynamics characteristics were reported at different depths (summarized by Saijo and Hayashi ). Water circulation is completed in April. Soon thereafter, stratification starts. During April–July, NH4+ accumulates gradually in the oxic zone [Yoh et al., 1988], partly because light inhibits nitrification [Yoshioka and Saijo, 1985]. Although photoinhibition of nitrification in oceans is now generally regarded as unimportant [Yool et al., 2007], photoinhibition of nitrification in Lake Kizaki has been identified clearly in changes in NH4+ and NO3− concentrations, 15N experiments, and in incubation experiments with isolated bacteria from the water in Lake Kizaki [Takahashi et al., 1982; Yoshioka and Saijo, 1984, 1985; Yoshioka, 2001; Y. Sasaki, unpublished data, 2008]. Nitrification, which is photoinhibited until early summer, increases suddenly in July, when stratification is completed and when the ammonia oxidizers recover from photoinhibition [Takahashi et al., 1982; Yoshioka, 2001]. Accumulation of NO3− in oxic hypolimnion is observed along with parallel accumulation of N2O [Yoh et al., 1988]. Significant correlations among NH4+ consumption, the increase in NO3−, the apparent oxygen utilization (AOU), and the N2O accumulation (ΔN2O) suggest that nitrification is related with this increase in the N2O concentration [Yoh et al., 1988]. However, the determination of N2O production processes based on the relation between AOU and ΔN2O is somewhat difficult [Nevison et al., 2003; Frame and Casciotti, 2010] because no universal AOU – ΔN2O relation has been identified to confirm N2O production by nitrification or denitrification. Consequently, other conclusive information is necessary to confirm that nitrification causes the accumulation of N2O in the oxic zone in Lake Kizaki. Moreover, each of NH2OH oxidation and NO2− reduction is related to nitrification, and it remains completely unknown which process is dominant for N2O production as inferred from conventional approaches.
 In deeper hypolimnia, microbial respiration reduces the DO concentration. High concentrations of N2O have been observed there [Yoh et al., 1983]. Incubation experiments with inhibitors suggest the occurrence of both nitrification and denitrification in oxygen-deficient zones [Yoh et al., 1990]. Yoh et al.  confirmed the occurrence of nitrification in a zone with low DO concentration (<3 μmol L−1) by investigating NH4+ accumulation in an incubation study with nitrification inhibitors (acetylene and nitrapyrin). Although their field-incubation study provides important information related to nitrification and denitrification, the relative importance of these processes on N2O accumulation was not clarified by their research. Moreover, the extent of N2O reduction in this zone remains unclear.
 This study was undertaken to examine the degree to which the isotope and isotopomer signatures of N2O correspond to the well-established N-cycling processes of Lake Kizaki. On the basis of results of previous studies, we expected to find (1) that accumulation of N2O in oxic hypolimnia is produced by the activity of aerobic autotrophic nitrifiers [Yoh et al., 1988] and (2) that high N2O concentration prevails in oxygen-deficient hypolimnia because of nitrification and denitrification, with incomplete N2O reduction to N2 [Yoh et al., 1990]. We conducted isotopic and isotopomer analyses for NH4+, NO3−, and N2O in Lake Kizaki in late July, when oxic zone nitrification had just been completed, to test whether these data can reflect in situ N-cycling processes.
2. Materials and Methods
 Samples were collected from the center of Lake Kizaki (36°33′N, 137°50′E, 1.4 km2, maximum depth 28.5 m) on 31 July 2007. Dissolved oxygen (DO) concentrations, water temperatures, and electrical conductivities (EC) were measured using a DO meter (Model85 SCOOT; YSI Inc.). Water samples were collected using Niskin samplers (GO1010X, 5 L and 1.7 L). Samples for N2O concentration and isotopomer ratios were collected in 250 mL glass serum vials, preserved with 1 ml of saturated HgCl2, and then sealed with butyl rubber stoppers. For nutrient concentrations and NO3− and NH4+ isotopic analyses, the samples were filtered through a GF/F glass fiber filter, which had been combusted previously at 450°C. The filtered samples were frozen until further analyses.
 Ammonium concentrations were determined using a phenol–hypochlorite reaction method [Sagi, 1966]. Concentrations of PO43− were determined using the molybdenum-blue method described by Murphy and Riley . Nitrite concentration was measured by complexation with sulfanilamide and subsequent coupling with N-(1-naphtyl)-ethylenediamine to form a red azo dye measured photometrically at 540 nm. Nitrate concentrations were measured using ion chromatography (DX-120 with an AS-14 column; Dionex Corp.).
 The concentration and isotopomer analyses of N2O were performed using an online analytical system comprising a 200 mL gas extraction chamber (Koshin Rikagaku Seisakusho, Tokyo, Japan), a stainless-steel gas transfer line, preconcentration traps, chemical traps for removal of H2O and CO2, and a gas chromatograph/isotope-ratio mass spectrometer (MAT 252; Thermo Fisher Scientific Inc.) [Toyoda et al., 2009]. The notation of the isotopomer ratios is the following:
Therein, 15Rα and 15Rβ represent the 15N/14N ratios of the central (α) and terminal (β) N atoms of N2O, respectively; 15Rbulk and 18R denote average isotope ratios for 15N/14N and 18O/16O, respectively. Subscripts “sample” and “std” signify isotope ratios for the sample and the standard, respectively; also used are atmospheric N2 for N and Vienna Standard Mean Ocean Water (V-SMOW) for O. We define the 15N site preference as an illustrative parameter of the intramolecular distribution of 15N [Toyoda and Yoshida, 1999]:
The measurement precision was usually better than 0.2‰ for δ15Nbulk, better than 0.5‰ for δ18O of N2O (δ18ON2O), and better than 1.0‰ for site preference.
 The δ15N and δ18O of NO3− and were measured using the “denitrifier method,” as described by Sigman et al.  and Casciotti et al. . Isotopic data of NO3− were calibrated using standards: USGS32, USGS34, USGS35, and IAEA-NO3 [Böhlke et al., 2003]. The data are reported relative to atmospheric N2, and the data relative to V-SMOW. The precision achieved through repeated analyses of an in-house standard was typically better than 0.2‰ for and 0.5‰ for A sample collected at 24 m depth contained high NO2−. We measured the δ15N and δ18O of NO2− using the azide method [McIlvin and Altabet, 2005] with NaNO2 and KNO2 in-house standards that had been calibrated by Drs. Karen Casciotti and Matthew McIlvin of Woods Hole Oceanographic Institute. The δ15N and δ18O of NO2− at the depth of 24 m were −8.9‰ and +15.1‰, respectively. We used these values to back calculate the and of the 24 m depth sample using mass balance estimation with the expected differences in isotopic fractionation during NO3− and NO2− reduction by denitrifiers reported by Casciotti and McIlvin ; the true isotopic values were higher by 2.3‰ (for and 1.2‰ (for than the isotopic values obtained from the denitrifier method. For δ15N of NH4+ the NH4+ of 100 mL sample was concentrated onto a GF/D glass fiber filter using the diffusion method [Holmes et al., 1998]. The NH4+ on the GF/D was converted to NO3− using persulfate oxidation and subsequent reduction to N2O with a denitrifier [Koba et al., 2010]. Several reference materials (USGS25, USGS26, IAEA-N2) were used for calibration. The precision, as verified using another in-house standard, (NH4)2SO4, was better than 0.5‰ for
 Statistical analyses were conducted using a statistical software package (R, R Development Core Team, http://cran.r-project.org, 2008) with Pearson's product moment correlation; P values <0.05 were inferred as statistically significant.
 The isotope and isotopomer ratio (δmeasured) and the concentration (C) of the target N compound were fitted to the Rayleigh isotope fractionation model to determine the isotopic discrimination expressed by the isotope enrichment factor ɛ [Mariotti et al., 1981]:
where ɛ = (1 − α) × 1000 and α is the fractionation factor expressed as the ratio of reaction rates for the isotopically light and heavy molecule (KL/KH). When ɛ was assigned, this model was also used to back calculate the original, not-reduced N2O isotope/isotopomer ratio (δoriginal) and concentration (C0) according to the following approximate Rayleigh linearization [Mariotti et al., 1988; Ostrom et al., 2007; Jinuntuya-Nortman et al., 2008; Koba et al., 2009]:
In addition, the isotopic signature of newly produced compound (δproduct) is approximately a function of isotopic signature of substrate (δsubstrate) and ɛ of the production process [Mariotti et al., 1981] as
when the pool size of substrate is sufficiently large and δsubstrate can be assumed to be constant. Equation (5) is also valid with substitutions of δsubstrate to δresidual substrate when the system is an open system [Fry, 2006].
 The water temperature of the surface water was higher than 20°C (Figure 1a). A thermocline was detected at 5–12 m depth. The water temperature decreased gradually below the thermocline to 6.8°C at the bottom. The concentration of DO, which was almost equilibrated with that of the atmosphere at the surface (290 μmol L−1), decreased gradually to 12 m depth (Figure 1a). The DO concentration remained stable at 12–19 m depth and decreased rapidly below 19 m depth (Figure 1a). Low DO was detected at 22 m (30 μmol L−1) and 23 m (0.3 μmol L−1) depths. The DO concentrations were lower than the detection limit below 24 m depth (Figure 1a). We defined the oxic zone as extending from the surface to 22 m depth and the oxygen-deficient zone as extending from 22 m to the bottom.
 Concentrations of PO43− (Figure 1b), NH4+, and NO2− (Figure 1c) were low (<1 μmol L−1) at most depths. They all increased near the bottom, which caused high electrical conductivity (Figure 1b). A peak of NO2− was found at 24 m depth (Figure 1c). Nitrate concentrations increased steadily between the surface water and 22 m depth and decreased rapidly between 23 m and 25 m depth (Figure 1d). Concentrations of N2O, which increased concomitantly with the increase in NO3− between the surface and 22 m depth, peaked at the depth of 24 m as NO2− (Figure 1d). Below 25 m depth, NO3− and N2O quantities became undetectable (Figure 1d).
 The values were stable (3.9‰–4.6‰) from the surface to 10 m depth (Figure 2). The minimum was observed at the depth of 15 m. The values were also stable (7.0‰–8.1‰) from the surface to 10 m depth. They decreased rapidly between 10 m and 22 m depth. Both and increased at depths greater than 22 m. The values were 10.2‰–19.3‰. They were higher than values at all depths (Figure 2).
 The isotopomer ratios of N2O (Figure 3) at the surface were all close to those of atmospheric N2O (6.5‰, 43.7‰, and 18.8‰ for δ15Nbulk, δ18O, and site preference [Yoshida and Toyoda, 2000]). In addition, δ15Nbulk decreased gradually from the surface to the bottom. The δ18ON2O values, which were stable between the surface and 22 m depth (46‰–48‰), increased between 22 and 24 m depth. Below 24 m, δ18ON2O decreased. The site preference values of N2O increased gradually from the surface to 22 m depth, and then decreased below that depth (Figure 3).
4.1. N2O in the Oxic Zone
 We found a parallel increase in NO3− and N2O concentrations in the oxic zone above 22 m depth (Figure 1d), which indicates N2O production by nitrification. Previously, higher nitrification rates and a larger accumulation of NO3− had been observed at deeper depths in the oxic zone of this lake, according to the higher NH4+ availability to ammonia oxidizer [Takahashi et al., 1982] and the longer period following the recovery of ammonia oxidizer from photoinhibition [Yoshioka and Saijo, 1985]. The amount of newly produced N2O is expected to be related to the newly produced NO3− if nitrification is related to the N2O accumulation in the oxic zone. We calculate the net amount of newly produced N2O and NO3− in the water column (netN2O and netNO3−, respectively) by subtraction of the concentration at the surface from that at the target depth. Below 10 m depth, the calculated netN2O and netNO3− increased in direct relation to depth (Figure 4a) and showed a significant relation (netN2O (nM) = 3.9 × netNO3− (μmol L−1) – 23.1, r2 = 0.99, P < 0.005, n = 3). Although this slope (3.9) is higher than the previously observed values in this lake (1.4–1.5 [Yoh, 1990]) and higher than the reported value of 0.5–1.5 in the aerobic culture of Nitrosomonas europaea [Hynes and Knowles, 1984], the netNO3−–netN2O relation indicates that N2O production is related with nitrification. However, some uncertainty remains in the interpretation of this relation. The uptake of NO3− by phytoplankton and microbes is expected to affect the NO3− profile, and consequently, the relation between netNO3− and netN2O. The parallel increase in N2O and NO3− concentrations started from depths greater than 10 m (Figure 1d), where phytoplankton activity is expected to be quite low because of the low relative light intensity (less than 1% of that at the surface) during June–July [Kishino and Hayashi, 2001]. Consequently, uptake of NO3− by phytoplankton was not expected to affect the netNO3−–netN2O relation to any marked degree. However, the microbial uptake of NO3−, which can affect the netNO3−–netN2O relation, cannot be ruled out. Moreover, the netNO3−–netN2O relation can provide no information indicating which is the actual N2O production process: NH2OH oxidation or NO2− reduction by an ammonia oxidizer. Both processes are expected to be related with the nitrification.
 New evidence for the importance of NH2OH oxidation as the N2O production in this oxic zone was derived from site preference of N2O (Figures 3 and 4b). Sutka et al.  reported that N2O produced by pure cultures of the ammonia oxidizing and methane oxidizing bacteria via NH2OH has a high site preference of 33‰; the site preference value for N2O produced by NH2OH (end-member for NH2OH-oxidation N2O) is now updated as 36.3 ± 2.4‰ using results obtained through more detailed analysis [Frame and Casciotti, 2010]. The site preference of the newly produced N2O (netSP) in the oxic zone would be close to this value if NH2OH oxidation in the nitrification process is responsible for the N2O accumulation, as indicated by concentration data. The following mass balance estimation was applied to calculate the netSP:
The netSP of N2O at 14–22 m depth, where netN2O was sufficiently high (>5 nM) for the reliable mass balance calculation (Figure 4b), was calculated using equation (6) as 33.6 ± 5.8‰ (mean ± s.d.; Figure 4b), which is consistent with the site preference for nitrification (36.3 ± 2.4‰ as mean ± s.d. [Frame and Casciotti, 2010]). Consequently, we conclude that NH2OH oxidation was the main source for N2O in the oxic zone in Lake Kizaki.
 Unlike site preference, the δ15Nbulk of newly produced N2O (netδ15Nbulk; Figure 4b) was confounded by variations in NIE and expressed isotopic fractionation. It provided no conclusive information related to the N2O production process, even with help from δ15N of newly produced NO3−Figure 4c). The netδ15Nbulk was estimated as 0.7 ± 0.9‰ (mean ± s.d.; Figure 4b) and the (0.5 ± 4.8‰ as mean ± s.d. for 15–22 m depth; Figure 4c) calculated similarly to netSP with equation (6). Assuming that which was not measured because of its low concentration (Figure 1c), is close to δ15N of sinking particles (−1‰ to 4‰ [Yoshioka et al., 1989]), then (0.5 ± 4.8‰) was close to the δ15N of sinking particles, suggesting that NIE during nitrification was negligible in spite of the reported large isotopic discriminations that occur during ammonia oxidization (−14‰ to −38‰ [Mariotti et al., 1981; Casciotti et al., 2003]). The complete consumption of NH4+ in the nitrification, resulting in the quite low NH4+ concentration and high NO3− concentration (Figures 1c and 1d), was responsible for the observed negligible NIE during nitrification, which further indicates that the netδ15Nbulk could not be 15N-depleted despite the large isotopic fractionation during N2O production via nitrification [Yoshida, 1988; Sutka et al., 2006]. Consequently, netδ15Nbulk observed here was consistent with the interpretation that NH2OH oxidation is the dominant N2O production process.
 The δ18O of newly produced N2O (netδ18ON2O; Figure 4b) was ineffective, even with help from δ18O of newly produced NO3−Figure 4c). The netδ18ON2O (47.6 ± 1.5‰, mean ± s.d.; Figure 4b) was similar to the δ18ON2O (Figure 3), indicating that δ18O of produced N2O was similar to that of atmospheric N2O (45‰–50‰). The δ18ON2O of N2O produced by NH2OH is not constrained well [Koba et al., 2009]. Furthermore, isotopic fractionation or NIE for 18O during NH2OH oxidation is not well understood because no measurement of δ18O of NH2OH has been reported. However, it is noteworthy that netδ18ON2O was higher at deeper depths with lower DO (Figure 1a) than at shallower depths (Figure 4b). Kim and Craig  reported a parallel increase in δ18O for DO and N2O, suggesting that N2O was produced with NH2OH, which should have high δ18O affected by the high δ18O for DO. In this study, δ18O of DO is expected to be high at deeper depths with low DO concentrations because of isotopic fractionation during DO consumption [Bender, 1990]. The high netδ18ON2O observed here in the environment with presumably high δ18O of DO is also consistent with the interpretation that NH2OH oxidation is the dominant process for N2O production in this zone.
 Consequently, site preference indicates that NH2OH oxidation is the dominant N2O production process, which is consistent with netδ15Nbulk netδ18ON2O. Because site preference is independent of the isotopic signatures of substrates, site preference provided conclusive information related to the N2O production process, which was not obtainable using other conventional approaches including the use of netδ15Nbulk netδ18ON2O.
4.2. N2O in the Oxygen-Deficient Zone
 The occurrence of denitrification in the oxygen-deficient zone (22–28 m) was indicated by the decrease in NO3− concentrations and concomitant increases in and (Figures 1d and 2). If we can assume a pseudo closed system for this oxygen-deficient zone as Lehmann et al.  assumed for the suboxic and anoxic hypolimnia, then application of equation (4a) with NO3− concentrations and isotopic data in this zone yields estimates of the isotope effects and of −24.4 ± 7.2‰ and −15.7 ± 3.7‰ (mean ± s.d.), respectively. The estimated and values fall into the range of isotopic effects during denitrification measured using pure-culture denitrifier strains (−5.4‰ to −26.6‰ for and −4.8‰ to −22.8‰ for [Granger et al., 2008]). In our data set, the ratio was 0.64. Granger et al.  reported this ratio as 0.86–1.02 (average 0.96) for NO3− reduction by a membrane-bound nitrate reductase, and 0.62 for reduction by auxiliary periplasmic NO3− reductase with pure-cultured denitrifier strains. Recently, Knöller et al.  reported the range of as 0.33–0.79 with pure-culture denitrifier strains. Additionally, many reports of field reports have described the range of as 0.5–1 [Böttcher et al., 1990; Mengis et al., 1999; Lehmann et al., 2003; Panno et al., 2006]. Consequently, concentration and isotopic signatures of NO3− indicate that NO2− reduction would be the N2O production process in the oxygen-deficient zone. However, it should be noted that only a single profile was measured and that our interpretation for NO3− largely depends on the assumption of a pseudo closed system of this oxygen-deficient zone with respect to NO3−. Such a point-in-time observation from a single vertical profile unfortunately prevents us from fully quantitative discussion on the denitrification. In addition, further studies should definitely include monitoring of the changes in NO3− and N2O profiles by frequent samplings throughout the year in scaling our results over space and time. Because of such limitations in this study, more evidence to confirm whether NO2− reduction is the dominant N2O production process is necessary to support the interpretation with NO3− data. Moreover, N2O reduction should be considered if NO2− reduction is dominant in the oxygen-deficient zone. The low N2O concentrations (nM level; Figure 1d) compared with the changes in NO3− concentration (μmol L−1 level; Figure 1d) suggest that most of the N2O produced by NO2− reduction should be reduced. More evidence of the strong N2O reduction in this oxygen-deficient zone can therefore strongly support the validity of this interpretation that NO2− reduction is the dominant N2O production process.
 The relation between site preference and δ15Nbulk of N2O strongly supports the idea that denitrification is responsible for the N2O detected at 22–24 m depth (Figure 5). The simultaneous production and consumption of N2O can be investigated using an isotopomer diagram [Yamagishi et al., 2007; Koba et al., 2009] in which the δ15N of substrates (NH4+ and NO3−) are used to interpret the relation between site preference and δ15Nbulk (Figure 5). The expected N2O production/consumption in this oxygen-deficient zone leads us to infer that the preexisting N2O at each depth is exhausted because of N2O reduction. The N2O produced by NO2− reduction is assumed to have site preference of −10.7‰ to 0‰ [Sutka et al., 2003, 2004, 2006; Toyoda et al., 2005; Frame and Casciotti, 2010] and the δ15Nbulk of −36‰ to +11‰ according to the measured (3.7‰–10.9‰; Figure 2) and the expected isotopic fractionation during NO3− reduction to N2O (0‰–39‰ [Brandes and Devol, 1997; Toyoda et al., 2005]). Covariations for δ15N and site preference during N2O reduction from 0.7 [Yamagishi et al., 2007] to 1.9 [Jinuntuya-Nortman et al., 2008] as 15ɛN2Obulk/SPɛ, indicated by arrows in Figure 5, are useful to detect the occurrence of N2O reduction. The site preference of measured N2O at 22–24 m depth (23‰–32‰) was much higher than the site preference end-member of N2O produced by NO2− reduction (−10.7‰ to 0‰). Consequently, the measured N2O cannot be explained solely by the NO2− reduction, and the occurrence of N2O reduction is clearly indicated. Actually, all N2O data fell into the shaded areas, where the occurrence of both NO2− reduction and N2O reduction is indicated (Figure 5). The difference between measured site preference and the end-member range of site preference for NO2− reduction (−10.7‰ to 0‰) is useful to estimate the N2O reduction using equation (4b). Assuming that SPɛ is −16.4‰, as calculated for the N2O reduction in the water column [Yamagishi et al., 2007], calculation with equation (4b) shows that more than 75% of preexisting N2O should have been reduced to N2 (92, 87, and 89% for 22, 23, and 24 m depth, respectively, when site preference of NO2− reduction is −10.7‰, and 85%, 75%, and 80% for 22, 23, and 24 m depth, respectively, when site preference of NO2− reduction is 0‰). When we applied other SPɛ values (−2.9‰ to −6.8‰ [Jinuntuya-Nortman et al., 2008; Ostrom et al., 2007]) to equation (4b), the percentages of N2O reduction were calculated as almost 100% (data not shown). Although these calculations are semiquantitative, depending heavily on SPɛ, the δ15N–site preference diagram confirmed that denitrification shaped the high N2O concentrations in this oxygen-deficient zone in Lake Kizaki.
 The δ15N – site preference diagram also shows that NH2OH oxidation was unimportant for 22–24 m depth. We measured in the oxygen-deficient zones to calculate δ15Nbulk of N2O produced by nitrification in the oxygen-deficient zone. The in hypolimnia was high (10.2‰–19.3‰; Figure 2). Yoshioka et al.  also reported high (10.3‰) in Lake Kizaki at 26 m depth. These high values agree with high in Lake Kinneret [Hadas et al., 2009] that was attributed to the remineralization of high δ15N sediment organic matter. The value of δ15N of sediment in Lake Kizaki was reported as ca. 5‰ [Yoshioka et al., 1988]. Remineralization of sediment organic matter is therefore a plausible explanation for our high The N2O produced from NH2OH is assumed to have site preference of 36.3‰ [Frame and Casciotti, 2010] and the δ15Nbulk of −58‰ to −27‰ according to the measured of 22–24 m and the expected isotopic effect during nitrification (−68‰ to −46‰ [Yoshida, 1988; Sutka et al., 2006]). The measured δ15Nbulk and site preference fell outside the area where mixed N2O produced from NH2OH oxidation and NO2− reduction should have been (the hatched area in Figure 5). The measured N2O was situated in the upper right part, outside of the hatched area, which indicates that the occurrence of N2O reduction should be included even in cases where NH2OH oxidation occurs. It is unlikely, however, that all NH2OH oxidation, NO2− reduction, and N2O reduction occur simultaneously at 22–24 m depths because of the different sensitivities and requirements of DO of the respective processes. The reduction of N2O is the most sensitive process to DO [Zumft, 1997]. Therefore, the coexistence of NH2OH oxidation and N2O reduction is unlikely. Above all, the contribution of NH2OH oxidation in oxygen-deficient zones is regarded as less important, as inferred from the site preference-δ15N diagram.
 In the deeper oxygen-deficient zone (from 24 m depth to the bottom) where DO and NO3− were below the detection limit, N2O concentrations decreased (Figure 1d) concomitantly with decreases in δ15Nbulk, site preference, and δ18ON2O (Figure 3). Actually, DO remains as zero during June–October in this deepest zone [Li et al., 1996]. The exhaustion of NO3− and DO implied that denitrification was completed. A decrease in the N2O concentration in the deeper oxygen-deficient zone has been observed in Lake Kizaki [Yoh et al., 1983]. This zone was regarded as a sink of N2O [Yoh et al., 1990], but the decrease in N2O concentration cannot be explained simply and solely by N2O reduction for the reason that the N2O reduction is expected to induce an increase in δ15Nbulk, site preference, and δ18ON2O [Ostrom et al., 2007; Jinuntuya-Nortman et al., 2008]. Diffusion of 15N-depleted and 18O-depleted N2O from the upper zone with high N2O concentration to the deepest zone might explain the decrease in δ15Nbulk and δ18ON2O along with depth. Well and Flessa  reported such diffusive isotopic fractionations for N2O in the air for δ15Nbulk and δ18ON2O. However, Well and Flessa  also reported that site preference would be slightly increased while N2O is diffused in the air, which cannot explain the high site preference at the deepest depth. Moreover, a partial reduction of diffused N2O would further increase δ15Nbulk, δ18ON2O, and site preference, thereby counterbalancing the diffusion effect. Consequently, although the detected N2O in this anoxic zone was regarded as resulting from the combination of N2O diffusion from the upper zone and N2O reduction, we were unable to ascertain the production/consumption processes shaping the low δ15Nbulk, δ18ON2O, and site preference in this zone.
4.3. Implications and Uncertainty of Interpretation in This Study
 Our interpretation of stable isotope data suffered from the large variation of NIE and isotopic fractionation, but the importance of NH2OH oxidation in the oxic zone and N2O reduction in the oxygen-deficient zone in the lake ecosystem cannot be ascertained without using the approach with isotopes. In our study, isotope data can generally be regarded as adding information to biogeochemical investigations through (1) unequivocal identification of processes that cannot be detected otherwise, (2) clear confirmation of the results from concentration data, or (3) indication that interpretations of concentration data are plausible. Our interpretations based on the conventional δ15N and δ18O were limited to qualitative (case 3 shown above) support of the occurrence of NH2OH oxidation in the oxic zone and denitrification in the oxygen-deficient zone. However, our isotope/isotopomer data provided new, more detailed information related to the in situ processes implicating not NO2− reduction but NH2OH oxidation as the dominant N2O production process in the oxic zone, and implying that strong N2O reduction occurred in the oxygen-deficient zone (case 1 shown above).
 Although the isotopomer approach is useful, different approaches such as the use of tracers and inhibitors in incubation experiments and molecular techniques to identify the key player in the targeted N2O production process [e.g., Priscu et al., 2008] are necessary for better quantitative estimation of complex N2O production and consumption processes. Furthermore, our interpretation based on the single profile of concentration, and isotope and isotopomer data should be improved using more frequent observations to scale our results over time and space. Because of the regular pattern of N biogeochemical processes occurring in Lake Kizaki, we interpreted the N2O biogeochemistry using net changes in concentrations and isotopic/isotopomer data for N2O and NO3− (Figure 4). However, it is more appropriate to interpret the relative changes in concentrations, and isotopic and isotopomer data among profiles obtained at different time periods by frequent monitoring of the vertical profiles in Lake Kizaki. Such studies can reduce the uncertainties in our interpretations in this study and allow us to scale our results over space and time. In addition, more investigations of the isotope and isotopomer fractionations during nitrification and denitrification can constrain the uncertainties that hinder the interpretation of N2O biogeochemistry.
 Isotopomer data indicate that NH2OH oxidation is responsible for accumulated N2O in the oxic zone. The simultaneous occurrence of NO2− reduction and N2O reduction determined the N2O profile in the oxygen-deficient zone of the hypolimnion. Results of this study revealed the usefulness of the site preference of N2O together with the isotopic signature of substrates (NH4+ and NO3−) for elucidation of complex N2O biogeochemistry. Our interpretations using stable isotopes rely heavily on isotope effects and fractionation factors for each N transformation process, yielding semiquantitative interpretations. Further investigations of isotope effects such as the relation between substrate concentration and isotopic fractionation factors, and of ratios between fractionation factors for 18O, 15N, and site preference are crucial to support better quantitative estimation of the processes of N2O production and consumption. The thorough isotopic survey including such investigations can have a huge potential to contribute to better understanding of N2O dynamics in oxic/anoxic boundary zone, which further advances the understanding of the importance of coastal zone as N2O source and of the N2O dynamics in open water bodies [Codispoti, 2010].
 We thank K. Yamada, C. Kashiwabara, T. Makita, and other members of both Yoshida Laboratory at Tokyo Tech and Yoh Laboratory at TUAT for fruitful discussions. We also thank Scott Wankel, Nathaniel Ostrom, and an anonymous reviewer for valuable comments about this manuscript. This work was partly supported by Grants-in-Aid for Creative Scientific Research (2518780112 and 20780113) and by the Program to Create an Independent Research Environment for Young Researchers of the Ministry of Education, Culture, Sports, Science and Technology of Japan, the Global Environmental Research Fund (A-0904) of the Ministry of the Environment, and the NEXT Program (GS008).