The 15N natural abundance of the N lost from an N-saturated subtropical forest in southern China


  • Keisuke Koba,

    1. Faculty of Agriculture, Tokyo University of Agriculture and Technology, Tokyo, Japan
    Search for more papers by this author
  • Yunting Fang,

    Corresponding author
    1. Faculty of Agriculture, Tokyo University of Agriculture and Technology, Tokyo, Japan
    2. Key Laboratory of Vegetation Restoration and Management of Degraded Ecosystems, South China Botanical Garden, Chinese Academy of Sciences, Guangzhou, China
    3. Now at Key Laboratory of Vegetation Restoration and Management of Degraded Ecosystems, South China Botanical Garden, Chinese Academy of Sciences, Guagzhou, China
    Search for more papers by this author
  • Jiangming Mo,

    1. Key Laboratory of Vegetation Restoration and Management of Degraded Ecosystems, South China Botanical Garden, Chinese Academy of Sciences, Guangzhou, China
    Search for more papers by this author
  • Wei Zhang,

    1. Key Laboratory of Vegetation Restoration and Management of Degraded Ecosystems, South China Botanical Garden, Chinese Academy of Sciences, Guangzhou, China
    Search for more papers by this author
  • Xiankai Lu,

    1. Key Laboratory of Vegetation Restoration and Management of Degraded Ecosystems, South China Botanical Garden, Chinese Academy of Sciences, Guangzhou, China
    Search for more papers by this author
  • Lei Liu,

    1. Key Laboratory of Vegetation Restoration and Management of Degraded Ecosystems, South China Botanical Garden, Chinese Academy of Sciences, Guangzhou, China
    Search for more papers by this author
  • Tao Zhang,

    1. Key Laboratory of Vegetation Restoration and Management of Degraded Ecosystems, South China Botanical Garden, Chinese Academy of Sciences, Guangzhou, China
    Search for more papers by this author
  • Yu Takebayashi,

    1. Faculty of Agriculture, Tokyo University of Agriculture and Technology, Tokyo, Japan
    Search for more papers by this author
  • Sakae Toyoda,

    1. Department of Environmental Science and Technology, Tokyo Institute of Technology, Yokohama, Japan
    Search for more papers by this author
  • Naohiro Yoshida,

    1. Department of Environmental Science and Technology, Tokyo Institute of Technology, Yokohama, Japan
    2. Department of Environmental Chemistry and Engineering, Tokyo Institute of Technology, Yokohoma, Japan
    Search for more papers by this author
  • Keisuke Suzuki,

    1. Faculty of Science, Shinshu University, Matsumoto, Japan
    Search for more papers by this author
  • Muneoki Yoh,

    1. Faculty of Agriculture, Tokyo University of Agriculture and Technology, Tokyo, Japan
    Search for more papers by this author
  • Keishi Senoo

    1. Graduate School of Agriculture and Agricultural Life Sciences, University of Tokyo, Tokyo, Japan
    Search for more papers by this author

Corresponding author: Y. Fang, Institute of Applied Ecology, The Chinese Academy of Science, 72 Wenhua Rd., Shenyang 110016, China. (


The 15N-enrichment of plants and soils is believed to indicate characteristics of the open nitrogen (N) cycle in terrestrial ecosystems because N lost from an ecosystem is presumably 15N-depleted through isotopic fractionation. However, because of a lack of an appropriate analytical methodology to confirm that supposition, the δ15N value for total dissolved nitrogen (TDN, the sum of ammonium, nitrate, and dissolved organic N) in stream water from forests has been measured only rarely. This report describes the δ15N values for TDN, ammonium, and nitrate in precipitation and stream water, together with those for soil-emitted nitrous oxide (N2O; measured once) in an N-saturated subtropical forest in southern China. Concentration-weighted δ15N values of TDN were −0.7‰ in precipitation and +1.2‰ in stream water. The difference in δ15N between soil (+3.9‰) and TDN in the stream water was 2.7‰. In contrast, soil-emitted N2O was strongly 15N-depleted (−14.3‰): 18‰ lower than that of the soil. Our results demonstrate that the discharged N loss is 15N-depleted only slightly compared with soil N, and gaseous N losses can be a strong driver for raising the terrestrial ecosystem δ15N. Our findings suggest that the relation between ecosystem δ15N and the open N cycle can be interpreted better by considering the net discrimination against 15N determined by the balance between gaseous and discharge N losses. Steady state 15N budget calculations proposed by Houlton and Bai (2009) can provide important information about the gaseous N fluxes, which are difficult to measure directly. The steady state calculation for the relationships among gaseous N loss, apparent isotopic fractionation during gaseous N loss, and isotopic signature of N inputs suggests that precise measurements of unmeasured components (e.g., dry deposition, NO and N2 emission) are quite important for better estimation of gaseous N losses from the ecosystem.

1. Introduction

The natural abundance of nitrogen (N) isotopes provides unique information about the N cycle in terrestrial ecosystems [Högberg, 1997]. The following observed trends, some of which were determined in laboratory experiments, are helpful to derive information related to the N cycle from natural abundance of 15N (δ15N) of plants (δ15Nplant) and soils (δ15Nsoil) [Nadelhoffer and Fry, 1994; Högberg, 1997]:

(a) Input of N from precipitation and N2 fixation generally supplies N with low δ15N (i.e., <0‰) [Heaton, 1986; Shearer and Kohl, 1986], which leads to the expectation that the ecosystem (soil and plant) δ15N (δ15Necosystem) in early succession is low because N input via N2 fixation and precipitation are considerable in such ecosystems. Vitousek et al. [1989] actually observed this expected trend in Hawaiian ecosystems.

(b) Plants are more 15N-depleted than soils [Fry, 1991], which engenders the assumptions that N available to plants such as ammonium (NH4+) and nitrate (NO3) produced in the soil have low δ15N values and that isotopic fractionations would occur during plant N uptake and assimilation. Direct δ15N measurements of available N in soil revealed that the δ15N of soil available N is generally lower than that of bulk soil, but that it varies [Binkley et al., 1985]. Isotopic fractionation during plant uptake with or without mycorrhizae has been estimated in laboratory experiments as varying from negligible to considerable [Mariotti et al., 1982; Yoneyama et al., 1991; Handley et al., 1993].

(c) Nitrification, producing NO3 which can then be discharged from the terrestrial ecosystem, can produce strongly 15N-depleted NO3 [Mariotti et al., 1981]. One can reach an important expectation for the relation between N loss and δ15Necosystem that terrestrial ecosystems with large N losses via discharge are 15N-enriched because NO3 is regarded as a major N form that is lost from terrestrial ecosystems via discharge.

(d) Large isotope effects can occur during gaseous N losses via ammonia volatilization [Mizutani et al., 1985], nitrification [Yoshida, 1988], and denitrification [Mariotti et al., 1981]. Here, one can generalize the expectation together with (c) for the relation between N loss and δ15Necosystem: terrestrial ecosystems with large N loss, whether gaseous loss or hydrological loss, are expected to be 15N-enriched.

These observed trends and expectations in 15N discrimination engender the basic premise that the fractionation of 15N/14N during biological processes and loss of 15N-depleted N increase δ15Necosystem over time. In other words, ecosystems have long been regarded as becoming 15N-enriched through losing processes of N such as N discharge (as NO3) and gaseous emissions of ammonia, nitrous oxide (N2O), nitric oxide (NO), and N2 because all these N compounds which are lost from the ecosystem are regarded as 15N-depleted as a result of isotope effects [Högberg and Johannisson, 1993]. Unfortunately, because of technical difficulties in monitoring concentration and δ15N of N loss, direct tests to guarantee the validity of this basic premise have been conducted only rarely. Nevertheless, this basic premise spurs us to examine the openness of the N cycle in a given terrestrial ecosystem, which is otherwise quite difficult to evaluate using conventional analytical tools. Austin and Vitousek [1998] evaluated the openness of the N cycle in forest ecosystems in Hawaii by evaluating δ15Nplant and δ15Nsoil. Their results revealed that δ15Necosystem is higher in dry ecosystems with less precipitation than in wet ecosystems with high precipitation. They inferred that dry forest ecosystems are more open in terms of N cycling than wet forest ecosystems are because dry ecosystems are water-limited and are ultimately not N-limited. The same trends in δ15Nplant and δ15Nsoil across precipitation gradients have been confirmed in other studies conducted in Hawaii [Schuur and Matson, 2001], in Australia [Schulze et al., 1998; Murphy and Bowman, 2009], and throughout the world [Amundson et al., 2003; Craine et al., 2009]. Martinelli et al. [1999] also compared δ15Nplant and δ15Nsoil in temperate and tropical forests. They reported that the higher δ15N in tropical forests reflects their N-richness and open N cycles.

Recent anthropogenic N inputs have induced N saturation in many forest ecosystems [Agren and Bosatta, 1988; Aber, 1992]. Changing the ecosystem N status from an N-limited to an N-saturated state is similar to changing the N cycle from one of low N availability to one of high N availability, and from closed to open. Net nitrification, NO3 leaching, and gaseous N loss can increase in N-saturated forests [Aber, 1992]. Consequently, soils and plants in N-saturated forests have been expected to be more 15N-enriched than those in an N-limited forest because of the large losses of the 15N-depleted N compounds. Indeed, δ15N has been used to assess the status of N saturation based on the premise described above. Högbom et al. [2002] observed elevated foliar δ15N with elevated NO3 leaching. In Europe, Emmett et al. [1998] observed that δ15Nplant and δ15Nsoil values increase along an N deposition gradient. The observations that net N mineralization and nitrification rates are positively correlated with soil δ15N [Pardo et al., 2002; Templer et al., 2007] and foliar δ15N [Garten and van Miegroet, 1994; Pardo et al., 2006; Craine et al., 2009] suggest the potential use of δ15N for examination of the N saturation status.

Despite its importance in support of the basic premise, the relation between N loss from the forest ecosystem and δ15Necosystem has not been examined adequately. For gaseous N losses in forms such as N2O and NO, previous reports have described, as expected, that these emitted gases from the soil in the field are more 15N-depleted than bulk soil N is [Kim and Craig, 1993; Pérez et al., 2000, 2001; Yamulki et al., 2001; Menyailo et al., 2003; Tilsner et al., 2003; Li and Wang, 2008; Meijide et al., 2010]. For discharge N loss, few studies provide both information of the δ15Nsoil and the discharged N for mutual comparison. Houlton and Bai [2009], after recently summarizing δ15N data of bulk soil N and discharge N losses (mainly as NO3) from 13 study sites, concluded that the discharge N loss does not discriminate against soil 15N. That important finding by Houlton and Bai [2009] undermines the validity of the basic premise. Further evaluation of the relation in δ15N between soil and discharged NO3 loss is necessary. Moreover, most δ15N values available from the literatures that Houlton and Bai [2009] summarized were δ15N values of NO3. In the relevant literatures, one study [Houlton et al., 2006] out of eight provided the δ15N value for total dissolved nitrogen (TDN) in the stream water, which is the sum of NO3, NH4+, and dissolved organic nitrogen (DON). The concentration of NH4+ in stream water is generally quite low [e.g., Barnes et al., 2008]. For that reason, its contribution to TDN with respect to concentration and δ15N is quite small. However, the importance of DON in N leaching has been increasingly recognized recently [Perakis and Hedin, 2002; Fang et al., 2009]. Consequently, the δ15N value of TDN (≈NO3 + DON) is more appropriate than the δ15N value of NO3, as representative of δ15N of discharged N loss. However, no appropriate method enabling high-throughput analysis of δ15N of NO3 and TDN has been available until recently.

The δ15N value for DON in stream water has been assumed as approximately equal to that of bulk soil N [Amundson et al., 2003] because DON is produced in soil mainly via decomposition and dissolution of soil organic matter, which might not induce a large isotope effect. However, results of several studies have demonstrated the possibility that δ15N of DON in the soil, the origin of DON in stream water, might be higher than that of bulk soil N. Actually, Handley et al. [1999], based on their global survey of δ15Necosystem, proposed that 15N-enriched DON loss might be responsible for low δ15Necosystem in wet and cold climates. Houlton et al. [2007] first reported δ15N values for soil DON (extracted using 2M KCl solution) in Hawaiian tropical forest soils. Those values were higher than those of bulk soil N in five out of six forest sites. Takebayashi et al. [2010] and Koba et al. [2010b] also reported higher δ15N values of soil DON extracted using 2M KCl solution than those of bulk soil N. However, Houlton et al. [2006] found that discharged TDN has lower δ15N than that of bulk soil in four out of six Hawaiian forest sites, although soil DON has higher δ15N than bulk soil in most of the study sites [Houlton et al., 2007]. Consequently, the lack of information related to δ15N of TDN in stream water must be resolved to test the validity of the basic premise based on the assumption that the discharge N loss increases the δ15Necosystem.

This report describes results of a survey of δ15N values for different N species (TDN, NH4+, and NO3) in precipitation and in stream water, and those for soil-emitted N2O in an N-saturated, subtropical forest in southern China. In this N-saturated forest, large losses of discharge N [Fang et al., 2008], soil-emitted N2O [Tang et al., 2006], and NO [Li et al., 2007] are observed. We can expect that such an open N cycle with large N losses in this study site would induce remarkably 15N-depleted N losses as discharged and gaseous N from the forest. The objectives of the study were to determine mechanisms regulating the δ15N of discharged N and the role of TDN and gaseous N losses in altering δ15Necosystem values.

2. Materials and Methods

2.1. Sampling Site

The study was conducted in an old-growth (age > 400 years) evergreen broadleaved forest [Zhou et al., 2006] in the Dinghushan Biosphere Reserve (DHSBR) in the central region of Guangdong province in southern China (112°33′E, 23°10′N). The region, located in a subtropical/tropical moist forest life zone, has a monsoon climate. The mean annual rainfall of 1927 mm shows a distinct seasonal pattern, with 75% of rainfall occurring during March–August and only 6% occurring during December–February [Huang and Fan, 1982]. The annual mean relative humidity is 80%. The mean annual temperature is 21.0°C, with average coldest (January) and warmest (July) temperatures of 12.6 and 28.0°C, respectively. The major species in the forest are Castanopsis chinensis Hance, Schima superba Chardn. and Champ., Cryptocarya chinensis (Hance) Hemsl., Cryptocarya concinna Hance, Machilus chinensis (Champ. Ex Benth.) Hemsl., and Syzygium rehderianum Merr. and Perry in the canopy and sub-canopy layers. The soil is a lateritic red earth formed from sandstone [He et al., 1982]. The soil pH, total C, total N, C/N ratio for the 0–10 cm mineral soil depth were, respectively, 3.8, 4.6%, 0.19%, and 22 [Fang et al., 2006].

This old-growth forest has eliminated any N-limitation through both long-term N accumulation over more than 400 years and high atmospheric N deposition in recent decades. Total dissolved N inputs (including organic N) in throughfall were estimated as 52 kg N ha−1 yr−1 in 2005 [Fang et al., 2008]. Total N leaching via surface runoff and from below the main rooting zone is 67 kg N ha−1 yr−1, which is 15 kg N ha−1 yr−1 greater than the atmospheric N input [Fang et al., 2008]. In this forest, the soil N2O emission rate is as high as 2.6–4.7 kg N ha−1 yr−1 [Tang et al., 2006; W. Zhang et al., 2008b]; soil NO emissions are also reported as high as 6.1–6.9 kg-N ha−1yr−1 [Li et al., 2007] from this forest. These previous studies highlight rapid and open N cycling in this forest [Mo et al., 2006].

2.2. Sampling

During the study period from Sept. 2007 to Dec. 2008, both precipitation and stream water were collected every two weeks, although the sampling frequency was reduced in winter because of low precipitation and discharge. Bulk precipitation in an open area in the reserve was collected using two open glass funnels (15 cm diameter), each connected to a 2.5 L sampling bottle with black polypropylene tubes. Stream water was collected from the weir of the 8 ha catchment. All water samples were filtered within 24–48 h of collection through 0.45 μm filters in the laboratory; then they were stored in plastic bottles at 4°C until chemical analysis.

Headspace air was collected from four static chambers for analysis of N2O concentrations and δ15N of the N2O once in Aug. 2007, before the sampling campaign for the solutions. Four static chambers in a plot of 100 m × 200 m located in the middle of the mountain slope in the study forest [W. Zhang et al., 2008a, 2008b] were used. The static chamber consisted of an anchor ring and a removable chamber top. The plot was far from the bottom of the mountain slope with a stream (ca. 50 m). Soil moisture was moderate; the soil was not waterlogged [W. Zhang et al., 2008b]. Removable chamber tops (35 cm height × 25 cm diameter) were set to the anchor rings when incubation started; after 1.5 h, the headspace atmosphere was collected into evacuated 200 mL glass flasks. Analysis of the δ15N of N2O necessitated a large accumulation of emitted N2O in the static chamber headspace. Therefore, we used a longer duration (1.5 h) than that used in typical N2O emission studies (e.g., 30 min with 10 min sampling intervals given by Tang et al. [2006]). An earlier report of a study at this site described that the increase in N2O concentrations remained linear for up to 2 h following chamber closure [W. Zhang et al., 2008b]. It is noteworthy that a single sampling campaign is insufficient to explore variations in δ15N values of emitted N2O. For better estimation for δ15N of N2O, additional investigation of δ15N of N2O must be conducted with more-intensive regular sampling and sampling following rain events.

2.3. Chemical Analysis

Concentrations of dissolved inorganic N (DIN; NH4+ plus NO3) for water samples were determined using ion chromatography (IC, DX-120 with AS-14 column; Dionex Corp., Osaka, Japan). The TDN concentration was determined using persulfate oxidation to NO3 [Miyajima et al., 2005] with subsequent determination of NO3 concentration using IC. The dissolved organic N concentration was calculated as the difference between TDN and DIN concentrations (DON = TDN − DIN). For 3 out of 42 samples (including rainwater and stream water), the DON concentrations were calculated as negative values; those values were designated as zero. The summer samples (10 samples, May–July of 2008) were accidentally lost after DIN measurements. For these samples, DON concentrations and isotope data were therefore not available.

The definition of δ15N is that δ15N = [Rsample / Rstandard] −1, where R is 15N/14N and atmospheric N2 is defined as the standard. The δ15N values of NO3 were measured using the “denitrifier method” [Sigman et al., 2001; Casciotti et al., 2002] with an isotope ratio mass spectrometer (IRMS; Delta XP; Thermo Fisher Scientific K.K., Yokohama, Japan) coupled with a Precon (Thermo Fisher Scientific K.K.), a gas chromatograph (GC; HP6890; Hewlett-Packard Co., Palo Alto, CA, USA) equipped with a Poraplot column (25 m × 0.32 mm), and a GC interface III (Thermo Fisher Scientific K.K.). The denitrifying bacterium, Pseudomonas aureofaciens (ATCC#13985), was used to convert 20 nmol of NO3 into N2O in 20 mL vials before isotope analysis. Several standards (USGS32, 34, and 35, and IAEA-NO3) were used to obtain the calibration curve to correct for drift and blank. The average standard deviations for replicate analysis of an individual sample were ±0.2‰ for δ15N.

We also measured δ15N of NH4+, TDN (and consequently DON) for all water samples following the methods described by Koba et al. [2010a]. In brief, NH4+ was concentrated on the glass fiber filter using the diffusion method [Holmes et al., 1998], then the concentrated NH4+ was digested to NO3 using persulfate [Tsunogai et al., 2008; Lachouani et al., 2010]. Finally, δ15N of the converted NO3 from the NH4+ was measured using the denitrifier method [Houlton et al., 2006]. Unlike conventional EA-IRMS analyses of NH4+ [Stephan and Kavanagh, 2009; Koba et al., 2010a], the method we used was not able to estimate the recovery % of NH4+ because the NO3 concentration in the persulfate-oxidized solution was difficult to measure as a result of the melted GF/D. Furthermore, the N2O amount introduced into GC-IRMS must be constant (ca. 30 nmol) to avoid nonlinear response of GC-IRMS, which prevented us from using the peak area of sample N2O for the recovery check. The recovery of NH4+ during diffusion procedures that we routinely used was estimated as appropriate (98–102%) with EA-IRMS [Koba et al., 2010a]. Therefore, we checked our analytical precision using several standards with different δ15N. After the blank correction, correlation between δ15N of the standards (USGS 25, 26, and IAEA-N2; δ15Ntrue) and blank-corrected δ15N (δ15Nmeasured) was calculated as δ15Ntrue = δ15Nmeasured × 1.11 − 0.59 (R2 = 0.999). The blank-corrected δ15N data of samples were applied to this equation to obtain the final δ15N of NH4+. The precision (standard deviation) calculated through repeated measurement of in-house (NH4)2SO4 (δ15N = −1.2‰) was ±0.5‰ for δ15N of NH4+.

The δ15N values of TDN were measured using the same method with persulfate-digested samples [Knapp et al., 2005]. The δ15N of DON was calculated using the δ15N values of DIN, TDN, and the concentrations of DIN and TDN [Houlton et al., 2006] according to the following.

equation image

Therein, [ ] denotes the concentration of the enclosed species. Calibrations for δ15N of TDN were conducted using the calibrated in-house standards (alanine, glycine, and histidine [Koba et al., 2010a]). Good correlation between δ15N of the standards and measured δ15N was obtained for TDN (δ15Ntrue = δ15Nmeasured × 1.02 − 0.52, R2 = 0.999) after the blank correction. The precision (standard deviation) of δ15N was ±0.2‰ for TDN. The propagated errors for the calculation of δ15N value of DON were determined using a Monte Carlo method as described by Knapp et al. [2010]. By assuming that the standard deviations of the measurements were 0.2‰ for NO3 and TDN, and 0.5‰ for NH4+, and by assuming 0.3 μM for concentrations of NH4+, NO3, and TDN, and by assuming average concentrations and isotopic values for precipitation and stream water, the analytical errors for TDN in precipitation and stream water were inferred respectively as ±2.3‰ and 1.0‰. These large analytical errors, especially for precipitation, prevented us from presenting a confident discussion of DON in this report, but the calculated concentrations and δ15N values are reported for reference.

Nitrous oxide concentration and its δ15N were measured using GC-IRMS (MAT 252; Thermo Fisher Scientific K.K.) [Toyoda et al., 2004] with calibrated N2O standard gas (349 μL L−1, δ15N = −0.8‰). The emission rates of N2O were calculated as the increment of N2O concentration between the atmosphere (338 μL L−1) and the incubated headspace for 1.5 h using the volumes of the chamber (19.6 L) and basal area (0.05 m2). The δ15N of soil-emitted N2O was calculated as follows.

equation image

Statistical analyses were conducted using R software (R Development Core Team, R: A language and environment for statistical computing, 2008, R Found. for Stat. Comput., Vienna, An α level of 0.05 was inferred as indicating significance. Friedman tests were used for comparison of concentrations and isotopic values among different N compounds across the study. Multiple comparisons of mean values of concentrations and δ15N among N forms were performed using Bonferroni correction. Spearman's correlation coefficient (rs) was used to assess correlation among isotopic data, concentration data, and hydrological data. Pearson's correlation coefficient (r) was also used to assess correlation between δ15N of soil N and δ15N of the leaching N with data presented by Houlton and Bai [2009] and the data obtained in this study.

Data for the daily precipitation and streamflow rate were provided by the Dinghushan Forest Ecosystem Research Station, Southern China Botanical Garden. Annual N input and N output were the sums of the products of concentrations and water amounts for respective sampling periods.

3. Results

3.1. Precipitation Chemistry

Concentrations of all N species in precipitation fluctuated greatly, with lower values during April–June (Figure 1a). The NH4+ concentrations were 3.7–107.3 μM (50.2 ± 28.8 μM as mean ± S.D., n = 23). The NO3 concentrations were 11.6–115.4 μM (average 46.9 ± 24.1 μM, n = 23; Figure 1a). Total dissolved N concentrations averaged 121.8 ± 52.3 μM, ranging from 60.6 to 238.5 μM (n = 18; Figure 1a). The DON concentrations averaged 23.6 ± 11.4 μM with a range of 6.9–40.3 μM (n = 15). The respective contributions of N compounds to TDN were 46 ± 21% for NH4+, 40 ± 8% for NO3, and 17 ± 15% for DON. No significant difference was found in concentrations among these three N forms. The concentration of each N compound showed no significant relation with the amount of precipitation (Figure 1e), except for the significant negative correlation between the amount of precipitation and DON concentration (rs = −0.68, p < 0.05). The N inputs were estimated as 11 and 12 kg N ha−1 yr−1 for NH4+ and NO3, respectively, using concentration data (Figure 1a) and the amount of precipitation (Figure 1e). The input of TDN was also estimated as 31 kg N ha−1 yr−1 when the average TDN concentration (121.8 μM) was assigned to the five missing samples in summer.

Figure 1.

Concentrations and δ15N of different N forms (TDN as solid squares, NO3 as open circles, and NH4+ as solid triangles) in (a, b) precipitation, (c, d) stream water, and (e) daily precipitation (bar), with discharge (line).

Isotopic signatures of N compounds dissolved in the precipitation also varied greatly, showing no clear seasonal pattern, except for NH4+, which showed low values in August and September (Figure 1b). The δ15N values of NH4+ were all negative (−8.8 ± 4.5‰ as mean ± S.D., n = 18), ranging from −16.6 to −1.3‰. In contrast, δ15N values of NO3 were all positive (+6.8 ± 1.8 ‰, range +5.2 to +11.3‰, n = 18; Figure 1b). The δ15N of TDN averaged −0.8 ± 2.6‰, ranging from −4.9 to +3.2‰ (n = 18; Figure 1b). The δ15N calculated range in DON was very wide, ranging from −69.8 to +63.6‰ (+5.0 ± 27.9‰ as mean ± S.D.; n = 15). The δ15N value of NH4+ was significantly lower than NO3 or DON (p < 0.0001). The δ15N values of each N compound showed no significant relation with the concentration or amount of precipitation. The concentration-weighted average δ15N values were −7.7, +5.3, +3.9, and −0.7‰, respectively, for NH4+, NO3, DON, and TDN (Table 1).

Table 1. The δ15N Values and N Fluxes in DHSBR Evergreen Broadleaved Forest
  103δ15NFlux (kg-N ha−1 yr−1)
  • a

    Concentration-weighted average.

  • b

    Averaged value calculated by concentration and δ15N data by Koba et al. [2010b] and bulk density data at 0–50cm soil depth (W. Saito, unpublished data, 2008).

  • c

    Arithmetric average of foliar δ15N for five species (n = 13) by Koba et al. [2010b].

  • d

    Flux-weighted average from Table 2.

  • e

    N2O flux data by Tang et al. [2006].

Mineral soil at 0—50 cm depthBulk N3.9b 
Plantsc −4.6 
Gaseous N lossN2O−14.3d3.2e
Stream wateraTDN1.243

3.2. Stream Water Chemistry

Concentrations of NO3 and TDN in stream water fluctuated greatly, although concentrations of NH4+ were stable (Figure 1c). Concentrations of NH4+ in the stream water were always low (10.3 ± 7.3 μM as mean ± S.D.), ranging 0–32.2 μM (n = 29; Figure 1c). The NO3 concentration in the stream water was 170.8 ± 106.4 μM (31.7– 458.7 μM, n = 29; Figure 1c). The TDN concentration was 206.1 ± 79.8 μM (58.0–354.2 μM, n = 24; Figure 1c), which included the DON concentration of 27.1 ± 27.1 μM (9.4–133.3 μM, n = 24, Figure 2d). The contributions of respective N compounds to TDN were 4 ± 4% for NH4+, 82 ± 10% for NO3, and 13 ± 9% for DON. The contributions of NO3 to TDN were high throughout the study duration (Figure 1c).

Figure 2.

Relations between concentrations and δ15N values for (a) TDN, (b) NO3, (c) NH4+, and (d) DON in stream water during the study period. Significant correlation was found for TDN (rs = −0.49, p < 0.05), for NO3 (rs = −0.71, p < 0.0005), and for NH4+ (rs = −0.49, p < 0.05). No significant correlation was found between DON concentration and δ15N of DON in the stream water.

The δ15N of NH4+ in the stream water was low (−9.2 ± 3.9‰ as mean ± S.D.), ranging from −14.7 to +0.2‰ (n = 23; Figure 1d), although δ15N of NO3 in the stream water was high (2.2 ± 1.4‰ with the range of +0.4 to +6.8‰, n = 24; Figure 1d). The δ15N value of TDN in the stream water was +1.5 ± 1.2‰, ranging from −0.4 to +5.4‰ (n = 24), and the δ15N DON was −2.5 ± 7.7‰ with a wide range (−19.9 to +10.7‰, n = 24, Figure 2d). The δ15N of NH4+ was lower than that of NO3 and DON (p < 0.0001). Significant negative correlation was found between δ15N values and concentrations for NH4+, NO3, and TDN (Figures 2a–2c), although no such correlation was found for DON (Figure 2d). The concentration-weighted average δ15N values were, respectively, −7.2, +1.4, −1.5, and +1.2‰ for NH4+, NO3, DON, and TDN (Table 1).

3.3. Soil-Emitted N2O and 15N Distribution

The soil emission rate of N2O was 39.4 ± 31.0 μg-N2O m−2 hr−1 (0.25 ± 0.10 nmol m−2 s−1), ranging 7.9–81.7 μg-N2O m−2 hr−1 (0.05–0.52 nmol m−2 s−1, n = 4; Table 2). The δ15N for emitted N2O were, respectively −29.1 to −10.4‰ with mean and SD of −18.6‰ and 7.8‰ (n = 4; Table 2). No significant relation was observed between the emission rates and δ15N values. The flux-weighted average of δ15N was −14.3‰ (Table 1).

Table 2. N2O Emitted From DHSBR Evergreen Broadleaved Forest Soil
 N2O Emission (μg N2O m−2 hr−1)103δ15N
Chamber 129.4−18.9
Chamber 27.9−29.1
Chamber 338.7−15.9
Chamber 481.7−10.4

Figure 3 presents variations in δ15N for each component of the evergreen broadleaved forest in DHSBR, including those of plants and soil N [Koba et al., 2010b]. The values of δ15N ranged from −29‰ for soil emitted N2O to +11‰ for soil TDN. Although a large difference in δ15N was found among different N compounds in precipitation, soil, and stream water, the variation in δ15N of TDN in stream water was small (Figure 3).

Figure 3.

Frequency distributions of δ15N of N cycle components in the study forest. Data for plants and soil N were derived from Koba et al. [2010b].

4. Discussion

4.1. Concentrations and Isotopic Signature of N Compounds in Precipitation

In this study, the δ15N of NH4+ was much lower than the mean δ15N of NO3 (Figures 1b and 3) in the precipitation. Such a pattern has also been reported frequently in many areas [Freyer, 1978; Garten, 1993; Russell et al., 1998; Y. Zhang et al., 2008]. Xiao and Liu [2002] measured δ15N of DIN in Guiyang city in southwestern China. They interpreted their low δ15N value of NH4+ as indicative of the extensive use of excretory wastes in agriculture and the release of ammonia from soil. They also reported positive δ15N of NO3 (+2.0 and +4.1‰ for light and heavy rainfall, respectively), which is comparable to our data (Figure 1b). In an earlier study, such positive δ15N values of NO3 were also found in Guangzhou city (90 km east of our study site) (+3.6‰ on average [Fang et al., 2011]). These positive δ15N values of NO3 can be interpreted as evidence of the contribution of NOx produced by coal combustion [Heaton, 1986; Elliott et al., 2007; Fang et al., 2011].

The δ15N of TDN was −4.9 to +3.2‰ with a smaller seasonal fluctuation than that of δ15N of NH4+ (Figures 1b and 3). The δ15N for TDN in precipitation has been reported much less frequently: Houlton et al. [2006] reported a range of −4 to +4 ‰ in Hawaii; Knapp et al. [2010] reported −8 to +8‰ in Bermuda; and Rolff et al. [2008] reported −4.2 to +12.3‰ in the Baltic Sea. Our data are within these ranges. Knapp et al. [2010] found that TDN has higher δ15N (−2.3‰) than NO3 (−4.5‰). In contrast, NO3 had higher δ15N than TDN by 7.7 ± 0.7‰ in our study (n = 18; Figure 1b). In our study, the low δ15N of NH4+ (−16.6 to −1.3‰; Figure 3) and large contribution of NH4+ to TDN (46 ± 21%) are responsible for the lower δ15N of TDN than that of NO3 (Figure 1b). Overall, the results reported herein (Figures 1b and 3 and Table 1) suggest δ15N of input N (δ15Ninputs) because δ15N of TDN in the precipitation was only slightly negative (−0.7‰). However, it is noteworthy that different N species had different δ15N (Figure 3), and that δ15Ninputs can vary according to the relative contributions of the respective N species.

4.2. Concentrations and Isotopic Signature of N Compounds in Stream Water

The predominance of NO3 in TDN in stream water found in this study (Figure 1c) caused the δ15N of NO3 to dominate the δ15N value of TDN (Figure 1d). The lower δ15N of TDN by 1.0 ± 0.1‰ (n = 24) than that of NO3 (Figures 1d and 3) is partly attributed to the low δ15N of NH4+ (Figures 1d and 3). Nadelhoffer et al. [1999] also reported 15N-depleted NH4+ (−7.3‰, ranging from −13.8 to −2.8‰) in stream water. This low δ15N of NH4+ in stream water is explainable partly by the cation-exchange of NH4+ onto the cation-exchange sites in soils before the discharge: during cation exchange, 15N is retained preferentially at the exchange site in the soil [Delwiche and Steyn, 1970]. Low δ15N values of NH4+ in soil solution compared with those of bulk soil N have been reported [Koopmans et al., 1997; Koba et al., 2003]. The similarity of δ15N values of NH4+ between those in precipitation and in stream water (Figure 3) implies some direct discharge of precipitation NH4+ into the stream water. However, such a direct discharge of precipitation NH4+ seems unlikely because of the strong adsorption capacity of NH4+ onto soil. Furthermore, short residence time of NH4+ in this forest soil (several hours) was found using the 15N-dilution method [Isobe et al., 2012] because of the high demand of NH4+ by heterotrophic and autotrophic microbes. Consequently, NH4+ in precipitation would be consumed or adsorbed in the soil and would be less likely to be discharged into stream water.

The low contribution of DON to TDN in stream water (13% on average; Figure 1d) compared with the high contribution of DON extracted by 2M KCl to the soil available N at this study site (81% on average) [Koba et al., 2010b] reflects the consumption and/or sorption of DON before its discharge from the forest [Rastetter et al., 2005]. Such sorption and/or consumption of DON, together with possible different chemical features between DON extracted by 2M KCL and DON dissolved in the stream water, probably accounts for the difference in δ15N between soil and stream water (Table 1). The isotope effects for the DON absorption and consumption in the soil are not known. Houlton et al. [2006, 2007] reported the same general trend that soil DON has higher δ15N and stream water DON has lower δ15N than bulk soil N, as we observed (Figure 3). As in the case of dissolved organic carbon [e.g., Kaiser et al., 2001], DON consists of many organic materials. Therefore, it is quite difficult to evaluate the expected apparent isotope effect between soil and stream water.

Significant negative correlations between the concentration and δ15N were found for TDN, NO3, and NH4+ (Figures 2a–2c). Surface soil tends to have high concentrations of TDN and NO3 with low δ15N values in the DHSBR, although no clear vertical trend is found for NH4+ [Koba et al., 2010b]. Consequently, it is likely that the discharge of soil TDN and NO3 from the surface soils during storms results in high concentrations of TDN and NO3 with low δ15N values in stream water.

The concentration-weighted average δ15N values of the discharged NO3 and TDN were, respectively, +1.4 and +1.2‰ (Table 1). The weighted averages of δ15Nsoil were +3.9‰ for 0–50 cm of mineral soils (Table 1), as calculated from the concentration and δ15Nsoil by Koba et al. [2010b] and soil bulk density data at each depth (W. Saito, unpublished data, 2008). The differences between δ15N of soil bulk N and discharged N (NO3 or TDN) were much less (3‰) than the expected difference based on the reported isotopic fractionation during ammonia-oxidation in the nitrification (14–38‰) [Mariotti et al., 1981; Casciotti et al., 2003]. This large isotopic fractionation can be regarded as a consequence of the isotopic fractionation between NH4+ and ammonia [Hermes et al., 1985], and of fractionation during the subsequent oxidization of hydroxylamine to nitrite (summarized by Casciotti et al. [2011]). Even including our new δ15N data of discharged TDN and bulk soil N in the summarized data set of Houlton and Bai [2009], the discharge loss of N is not 15N-depleted in comparison with soil globally. Significant correlation was found between δ15Nsoil and δ15N of the leaching N (δ15N of leaching N = (0.97 ± 0.16) × δ15Nsoil − (0.96 ± 0.65), R2 = 0.78, p(slope) < 0.0001 and p(intercept) = 0.17) with data from Houlton and Bai [2009] and our data (Table 1).

Such a small difference in δ15N between bulk soil N and discharged NO3 can be explained by the underexpression of the intrinsic, large isotope effect of nitrification [Mariotti et al., 1981; Casciotti et al., 2003] when nitrifiers consume the NH4+ completely. The net nitrification rates in this soil are the same as the net mineralization rates (Y. Fang, unpublished data, 2011), indicating the complete consumption of NH4+ to NO3. Such high demand of NH4+ by nitrifiers is also confirmed using the 15N-dilution method [Isobe et al., 2012]. When NH4+ is consumed completely by nitrification, the δ15N value of NO3 is equal to that of initial NH4+, which is also close to δ15Nsoil [Koba et al., 1998]. This underexpression of the isotope effect by nitrification is expected to be the reason for the high δ15N value of NO3. Occurrence of denitrification presents another possible explanation for the small difference in δ15N between bulk soil N and discharged NO3. However, the lack of a clear increase in δ18O of discharged NO3 corresponding to the increase in δ15N found in this forest (Y. Fang, unpublished data, 2011) suggests that denitrification, if it occurs, consumes NO3 completely without showing any increase in δ15N and δ18O of discharged NO3 [Bai and Houlton, 2009].

4.3. Isotopic Signature of N2O Emitted From the Forest Soil

As expected, the soil-emitted N2O showed much more 15N-depletion than soil N (Table 1 and Figure 3). The difference in weighted-average δ15N values was 20.4‰ between NH4+ in the soil and N2O, and 15.9‰ between NO3 in the soil and N2O (Table 3). Previous reports have described that the nitrification produces strongly 15N-depleted N2O with a difference of ca. 30–70‰ between NH4+ and N2O [Yoshida, 1988; Sutka et al., 2006]. However, N2O produced by denitrification is less 15N-depleted because of the smaller isotopic fractionation (<37‰) [Koba et al., 2009] and because N2O reduction during denitrification can increase δ15N of N2O [Ostrom et al., 2007]. In fact, δ15N of N2O does not provide conclusive information as to which process—nitrification or denitrification—is principally responsible for the emission of N2O, and for N2O reduction to N2 at our study site. It should be described that our measurement was conducted during only a single sampling occasion because of the difficulties of isotopic measurement of N2O. Nevertheless, many earlier studies [Kim and Craig, 1993; Pérez et al., 2000, 2001; Yamulki et al., 2001; Menyailo et al., 2003; Tilsner et al., 2003; Meijide et al., 2010] and our data clarify that soil-emitted N2O is 15N-depleted. We did not attempt to measure δ15N of NO and N2 in this study because of technical difficulties in measurement. However, Li and Wang [2008] recently developed a method for δ15N of NO and reported slightly lower δ15N values of NO emitted in a vegetable field (−18.9 to −48.9‰) than the δ15N values for the soil emitted N2O in this study. Consequently, the gaseous loss of N from the forest ecosystems as N2O and NO is very likely to increase the δ15N of available N, and thereby the δ15Necosystem, as Houlton and Bai [2009] reported.

4.4. Steady State Calculation for the 15N-Enrichment of Soil

We applied a simple model [Houlton and Bai, 2009] to investigate the manner and extent to which gaseous and discharge N losses can affect the N budget in this watershed. As reported based on results of many studies [Handley et al., 1999; Brenner et al., 2001; Amundson et al., 2003], δ15Nsoil can be determined in the steady state as follows, being independent from δ15Nplant [Houlton et al., 2006; Bai and Houlton, 2009; Houlton and Bai, 2009].

equation image
equation image

In those equations, fgaseous_N_loss and fleaching_N_loss respectively denote the fractions of loss by gases and by leaching, and εG and εH respectively represent the effective isotope effects associated with gaseous and leaching losses. Soil emission rates were found to be 2.6 to 4.7 kg N ha−1 yr−1 for N2O [Tang et al., 2006; W. Zhang et al., 2008b] and 6.1 to 6.9 for NO [Li et al., 2007]. It is quite difficult to quantify the gaseous loss of N as N2 [Groffman et al., 2006; Yang and Silver, 2012] and the lack of information on this N2 emission does not allow us to discuss the importance of this process quantitatively. The N loss via N2 by denitrification might be minor in this watershed in terms of the low annual precipitation compared to the wet tropical forests in Hawaii, where N2 loss overwhelms combined N2O and NO losses [Houlton et al., 2006], and in terms of low soil pH value at this study site, which makes the importance of N2O greater than that of N2 [Šimek and Cooper, 2002]. Even so, the gaseous N loss at our study sites is expected to be more than 8.7 kg N ha−1 yr−1, thus fgaseous_N_loss > 0.17 [ = 8.7/(8.7 + 43)].

Combination of (3) and (4) gives us

equation image

We can simulate the 15N-enrichment of the soil according to the N losses using equations (3) and (4) [Fry, 2006]. Using equation (5), we can also estimate fgaseous_N_loss, which is conventionally difficult to measure.

According to Table 1, we assign the required values for the calculation as δ15Nsoil = 3.9‰ and εH = δ15Nsoilδ15N of TDN in stream water = 3.9 − 1.2 = 2.7‰. Figure 4, produced using these input data, illustrates that fgaseous_N_loss decreases dramatically with the increase in εG. The thick line in Figure 4 shows the case of δ15Ninputs = −0.7‰ observed in this study. When we assign εG as 18.2‰ (δ15Nsoilδ15N of N2O = 3.9 − (−14.3) = 18.2, Table 1), fgaseous_N_loss is calculated as 0.12 (dotted arrows in Figure 4).

Figure 4.

Steady state calculation for εG, fgaseous_N_loss, and δ15Ninputs according to equation (5). Shaded areas (fgaseous_N_loss < 0, and >1) show that the estimations are meaningless because the definition of the fgaseous_N_loss is not observed. The values in ‰ on the lines are δ15Ninputs. The thick line represents the relation in the case of δ15Ninputs = −0.7‰ (Table 1), and fgaseous_N_loss is calculated as 0.12 with εG = 18.2 [3.9 − (−14.3)] (dotted arrows). Then starting from the open circle on the x axis, when δ15Ninputs = 1.2 ‰, the fgaseous_N_loss is 0.

The apparent mismatch between calculated fgaseous_N_loss (0.12) and field-observed minimum fgaseous_N_loss (>0.17) is attributable to the unreliability of εG or δ15N of N2O caused by insufficient representativeness of our δ15N of N2O, which was measured only once (Table 2), and to a lack of information of δ15N of NO and N2, and possibly to the inconsistent steady state assumption of this watershed. The low δ15N of NO emitted from soil by Li and Wang [2008] (−19.9 to −48.9‰), which are exclusive data as far as we know, can decrease the δ15N of gaseous N loss (lower than −14.3 ‰ as δ15N of N2O only; Table 2) and thereby increase εG greatly (larger than 18.2 based on δ15N of N2O only; Table 2). Assuming that δ15N of NO is −35‰, the δ15N of gaseous N loss is calculated as −26.6‰ [(4.7 × (−14.3) + 6.9 × (−35))/(4.7 + 6.9) = −26.6]. Thereby, εG is 30.5 [(3.9 − (−26.6) = 30.5], resulting in small fgaseous_N_loss (0.07; Figure 4 with δ15Ninputs = −0.7‰), implying much greater total N loss from this watershed of 165.7 kg N ha−1 yr−1 [11.6/0.07 = 165.7], which is completely unrealistic. These mismatches strongly suggest the importance of the measurement of εG in the field [Bai and Houlton, 2009] or δ15N of the lost N via gases (especially for N2 with no data available) to evaluate the unmeasured gaseous N fluxes using equations (3)(5).

It is noteworthy that the N input and output for the study forest were not balanced (Table 1). N output can be at least 51.7 kg N ha−1 yr−1 (43 + 8.7) via leaching and gaseous N loss, 20 kg N ha−1 yr−1 greater than N input in precipitation estimated in this study (Table 1). This imbalance might indicate that the study forest ecosystem was losing N from the ecosystem and/or the importance of dry deposition, which we did not measure. To evaluate the effect of unmeasured input of N and its δ15N on fgaseous_N_loss, we conducted sensitivity analysis with variable δ15Ninputs. With δ15Nsoil = 3.9‰ and εH = 2.7‰, equation (5) can be modified as shown below.

equation image

This simple relation among δ15Ninputs, εG, and fgaseous_N_loss (Figure 4) enables us to discuss the importance of δ15Ninput. When δ15Ninputs = 1.2‰, fgaseous_N_loss is 0% irrespective of εG. In addition, in most of the cases of δ15Ninputs > 1.2‰, meaningful fgaseous_N_loss whose value falls in the range of 0–1 cannot be obtained even with variable εG (Figure 4). The inclusion of dry deposition in the N input in equation (5) is important because the δ15N of dry deposition, which has been seldom measured in prior studies, tends to be high [Yeatman et al., 2001; Elliott et al., 2009; Kawashima and Kurahashi, 2011] compared with wet deposition measured in this study. Future studies must be conducted with precise measurements of δ15Ninputs, including the monitoring of dry deposition, to apply the equation of Houlton and Bai [2009] (equations (3)(5)) effectively.

5. Conclusion

This comprehensive investigation revealed wide variation in δ15N in each N component within a single forest: from −29‰ (for soil-emitted N2O) to +11‰ (extractable soil TDN) (Figure 3). Our results demonstrate that input TDN via precipitation has a δ15N value close to zero, consistent with previous studies in other sites. However, δ15N varied greatly between NH4+ and NO3. Gaseous output of N2O was strongly 15N-depleted relative to the soil, whereas discharged TDN and NO3 had only slightly lower δ15N than the soil had. These results suggest, as Houlton and Bai [2009] demonstrated clearly, that the balance between gaseous N loss with strong 15N discrimination and discharged N loss with slight or null discrimination determines the net effect of 15N discrimination, which must be considered when δ15Necosystem is used as a parameter indicating the open characteristics of the N cycle in a given ecosystem.

It is noteworthy that we used an incomplete data set: missing data in summer when frequent rain and discharge events occur (Figure 1e), and a single measurement of gaseous N loss (only for N2O) from the soil (Table 2). Especially, gaseous N loss can occur in different N forms (NO and N2 in addition to N2O) and can show large spatiotemporal variations in 15N natural abundance and emission rates. Further studies of gaseous N losses with more frequent samplings can improve our understanding of the relation between δ15Necosystem and N loss. In this study, we devoted particular attention to TDN to evaluate the hydrological input and output of N to explore the importance of DON (Table 1). The δ15N of DON in the stream water was not so important in this study in 15N budget because of its low contribution to TDN (Figure 1d and Table 1). However, we emphasize that TDN in precipitation should be monitored in 15N budget calculation because of the high sensitivity of the steady state equation to δ15Ninputs; when we use δ15N of NO3 as δ15Ninputs as in the work of Houlton and Bai [2009], steady state calculations cannot be functional because of the higher δ15N of NO3 than δ15Nsoil (Table 1 and Figure 4). Furthermore, δ15N of dry deposition should be included for the precise estimation of δ15Ninputs in future studies for better use of the steady state equation to estimate fgaseous_N_loss.


This work was supported by the National Natural Science Foundation of China (40703030), National Basic Research Program of China (2009CB421101), and by grants-in-aid for Scientific Research of the Japan Society for Promotion of Science (21310008), grants from the Ministry of Education, Culture, Sports, Science, and Technology (MEXT), Japan (19658060, 20780113, 21310008, 19310019, and 19201004), a grant from the Global Environment Research Fund (C-052), and Mitsui and Co., Ltd. Environment Fund (R08-C108) and by the NEXT Program (GS008) of the Japan Society for the Promotion of Science (JSPS). Y. F. was supported by the JSPS with a Postdoctoral Fellowship for Foreign Researchers and a grant-in-aid for JSPS Fellows (20–08421). K. K. was also supported by the Program to Create an Independent Research Environment for Young Researchers from MEXT. Meteorological and hydrological data were provided by Dinghushan Forest Ecosystem Research Station. We thank X. M. Fang, Y. Sasaki, and K. Isobe for their assistance in sample collection, chemical analysis, and fruitful discussion. We also appreciate B. Houlton and E. Bai for their summarized data set used to confirm the relation between δ15N of leaching N loss and δ15N of soil. We gratefully acknowledge three anonymous reviewers and the Editor for their constructive comments and advice related to an earlier version of this manuscript.