Upward diffusion of nitrous oxide produced by denitrification near shallow groundwater table in the summer: a lysimeter experiment

Abstract

Movement of nitrous oxide (N₂O) produced in subsoil and shallow groundwater is important in determining the direct and indirect N₂O emissions from agricultural soils. From the results of our previous study in a lysimeter-contained Gray lowland soil in the summer, we hypothesized that if a large amount of N₂O is produced near shallow groundwater table in the summer, it will diffuse upward to the atmosphere. To examine this hypothesis, we conducted a one-year experiment in the same lysimeters for the cultivation of soybean–wheat double cropping (SW) or upland rice (UR). Dissolved N₂O concentration in the drainage water in the UR plots exceeded 0.4 mg N L⁻¹ in the summer, whereas that in the SW plots remained <0.1 mg N L⁻¹. Analyses of the concentrations of nitrate and dissolved N₂O in the drainage water and their nitrogen and oxygen isotopic compositions (δ¹⁵N and δ¹⁸O) during the summer revealed that denitrification was the main process for the N₂O production near the groundwater table. There was a significant positive correlation between the dissolved N₂O concentration and soil-surface N₂O flux in the summer. Calculated upward diffusive N₂O fluxes at three soil depths by Fick's law also supported our hypothesis. The δ¹⁵N values of N₂O in the soil-surface flux were similar to those in the shallow groundwater in the UR plots during the summer. Such similarity was not found in the SW plots. We conclude that our hypothesis was confirmed by the above results. Comparison of the monitored data with other seasons indicates that low soil water content was a driving force for the upward N₂O diffusion as well as the high dissolved N₂O concentration.

Keywords:

• denitrification
• gas diffusion
• groundwater
• nitrous oxide
• indirect emission.

Introduction

Nitrous oxide (N₂O) is predicted to be 298 times as potent a greenhouse gas as carbon dioxide (CO₂) on a 100-year time scale (Forster et al. 2007), and is indirectly involved in the catalytic destruction of stratospheric ozone. Agriculture is the largest anthropogenic N₂O source, and accounted for 58% of the global anthropogenic N₂O emissions in 2005 (Smith et al. 2007). Nitrogen (N) fertilization inevitably enhances the production and emission of N₂O through its effects on microbial nitrification and denitrification processes in agricultural soils. However, it is well known that soil-surface N₂O emissions have high spatial variability (Clemens et al. 1999; Yanai et al. 2003). Akiyama et al. (2006) compiled long-term field monitoring data on soil-surface N₂O emissions from Japanese agricultural fields (246 measurements from 36 sites), and reported that the mean emission from poorly drained upland fields was more than four times that of fields with well-drained soils. In such wet soils, N₂O is produced mainly via denitrification (Davidson 1991).

The N₂O production by denitrification is not just confined to topsoil; subsoil and shallow groundwater can also be hot spots. Understanding the mechanism and fate of N₂O produced in subsoil is important for accurately assessing the direct and indirect N₂O emissions (Clough et al. 2005). The vertical transfer of N₂O from the groundwater table to the atmosphere depends on biophysical soil properties, such as the rates of N₂O production and consumption and the rates of convective and diffusive transport (Deurer et al. 2008; Well et al. 2001). According to Mosier et al. (1998), N₂O emitted from groundwater to the atmosphere by upward diffusion is also considered as an indirect emission associated with N leaching.

Clough et al. (1999) conducted a short-term soil column experiment, and reported that 0.4% of the ¹⁵N-labeled nitrate (), injected at a depth of 0.8 m in the presence of a carbon (C) substrate, diffused upward to the atmosphere as N₂O. In a German aquifer, Weymann et al. (2009) used a ¹⁵N tracer method and found that groundwater-derived N₂O was diffused upward to the atmosphere though its amount was hardly significant. On the basis of the results of a six-year field experiment in Gray lowland soil, Kusa et al. (2010) reported that 14% of soil-surface N₂O emission was derived from N₂O produced in the subsoil by a concentration gradient method. These studies have demonstrated that N₂O produced in subsoil and shallow groundwater can contribute more or less to soil-surface N₂O emission.

In our previous study conducted in lysimeters with a shallow groundwater table, soil-surface N₂O flux and dissolved N₂O concentration in drainage water increased synchronously during the summer (Minamikawa et al. 2010). As far as we know, such a relationship has not been observed in any previous studies. Given the accumulated knowledge of N₂O dynamics in a soil profile, we proposed the following hypothesis to explain the observed relationship: if a large amount of N₂O is produced near the groundwater table in the summer, it will diffuse upward and significantly contribute to soil-surface N₂O emission.

The objective of the present study was to examine this hypothesis. We measured soil-surface N₂O emissions, N₂O concentrations in the soil profile, and dissolved N₂O emissions in the lysimeter drainage water for one year. Diffusive N₂O fluxes in the soil profile were calculated with Fick's law. We also measured nitrogen and oxygen isotopic compositions (δ¹⁵N and δ¹⁸O) of and N₂O in the drainage water during the summer to discuss the production process of N₂O dissolved in the drainage water. Then we compared the δ¹⁵N and δ¹⁸O values of N₂O in the soil-surface flux with those in the shallow groundwater and the drainage water to discuss a link between dissolved N₂O concentration and soil-surface N₂O flux. In addition, we discussed whether our hypothesis could apply to the other seasons.

Materials and Methods

Study site and crop cultivation

We carried out a one-year experiment at the lysimeter facility of the National Institute for Agro-Environmental Sciences (NIAES), Tsukuba, Ibaraki, Japan (36°01′N, 140°07′E) from March 2007 to April 2008. Each lysimeter contained Gray lowland soil (Gleyic Fluvisols) and had a cross-sectional area of 9 m² (3 m × 3 m) and a 1 -m soil depth (Fig. 1). The bottom of each lysimeter was filled with a layer of sand and gravel. Total C and N contents in the topsoil (0 to 0.2 m depth) were 18.8 g C kg⁻¹ and 1.6 g N kg⁻¹, respectively, and those in the subsoil (0.2 to 1 m depth) were 13.7 g C kg⁻¹ and 1.4 g N kg⁻¹, respectively. Soil textures in the topsoil and subsoil were classified into Light Clay (36% clay, 34% silt, and 30% sand) and Silty Clay Loam (19% clay, 46% silt, and 35% sand), respectively. The topsoil's pH (H₂O) was 5.7 and its bulk density was 1.01 g cm⁻³. The groundwater table had been set to a depth of 0.9 m from the soil surface since the winter of 2005 (Fig. 1) to simulate soil water condition in actual lowland fields, but evapotranspiration could lower the groundwater level. Subsurface drainage water was discharged through the bottom drain outlet whenever enough precipitation occurred. We collected samples of drainage water and shallow groundwater from the deep and shallow (0.9-m depth) valves, respectively (Fig. 1).

Figure 1. Schematic diagram of a lysimeter. The dotted line represents a pre-installed soil gas sampler.

We examined two kinds of upland cropping systems, each with two replicates: double-cropping of soybean (Glycine max L. cv. Enrei) in the summer with wheat (Triticum aestivum L. cv. Norin 61) in the winter (SW), and single cropping of upland rice (Oryza sativa L. cv. Toyohatamochi) (UR). Single cropping of paddy rice (Oryza sativa L. cv. Nipponbare) had been conducted in all four plots for more than ten consecutive years before 2002. Thereafter, crop rotation between paddy rice and upland crops was conducted at intervals of two years: cultivation of upland crops in 2002 and 2003, followed by cultivation of paddy rice in 2004 and 2005, and cultivation of upland crops in 2006 and 2007.

Field management followed conventional practices for the region. Total urea application rates in the SW and UR plots were 12 and 6 g N m⁻², respectively. Fused magnesium phosphate and potassium chloride were applied as basal dressing for soybean, wheat, and upland rice at 6, 10, and 10 g P₂O₅ or K₂O m⁻², respectively. At crop maturity, all aboveground biomass was harvested and oven-dried for three days at 80°C to determine dry mass. The wheat and rice crop residues were incorporated into the topsoil of each plot after harvesting, whereas soybean residues were removed. Other details of lysimeter facility and crop cultivation are described in Minamikawa et al. (2010).

Gas and water measurements

All the measurements were conducted at one place or for one sample in each plot. We continuously monitored the soil-surface N₂O flux six times a day in each lysimeter by the automated chamber system (Nishimura et al. 2005b). The chamber covered 0.81 m² (0.9 m × 0.9 m), and its height was adjustable to 0.6 or 1.2 m according to crop height. We analyzed N₂O concentration inside the chamber using a gas chromatograph (GC) equipped with an electron-capture detector (GC-14B, Shimadzu, Kyoto, Japan), and determined soil-surface N₂O flux from a linear increase in the N₂O concentration for 30 min. Total soil-surface N₂O emission was calculated by integrating the fluxes over time. The volumetric soil water content at a depth of 0.1 m was monitored using a time-domain reflectometry (TDR) moisture sensor (CS616S, Campbell Scientific Instruments, Logan, UT, USA), and the water-filled pore space (WFPS) was calculated from the volumetric water content and soil porosity (0.61 ± 0.03 m³ m⁻³).

We collected soil gas samples at a depth of 0.15, 0.45, and 0.75 m in the soil profile once a week using pre-installed stainless-steel tube samplers (outer dimension 1.26 mm, inner dimension 0.90 mm), at a distance of 0.5 m from the lysimeter's side wall (Fig. 1). We analyzed the concentrations of N₂O and CO₂ using the automated gas analysis system which consisted of two GCs (Sudo 2006).

We collected the drainage water samples whenever drainage occurred after precipitation and also once a week without regard to precipitation. The concentrations of dissolved N₂O and CO₂ in the drainage water were measured by a headspace technique (Minamikawa et al. 2010) using the automated analysis system. We monitored the volume of drainage water with a tipping bucket pluviometer (UIZ-TB200, Uizin, Tokyo, Japan), and calculated the total emissions of dissolved N₂O and CO₂ by integrating daily emissions (i.e., dissolved gas concentration × daily drainage volume). Shallow groundwater was sampled for the stable isotope analysis during the summer (discussed later, in the section “Stable isotope analysis”).

Aliquots of the drainage water sample were passed through a 0.45-µm membrane filter. We analyzed the concentrations of and ammonium () using an ion chromatograph (DX-500, Dionex, Sunnyvale, CA, USA) and the concentration of dissolved organic carbon (DOC) with using a total organic carbon analyzer (TOC-V, Shimadzu, Kyoto, Japan). Total amounts of leached and DOC were calculated in the same manner as were the amounts of dissolved gases.

Estimation of diffusive N₂O flux in the soil profile

Vertical diffusive N₂O fluxes at a depth of 0.075, 0.3, and 0.6 m were calculated from the measured concentration gradients between the two neighboring depths on the basis of Fick's law:

where D_s is the gas diffusion coefficient in soil, C is the gas concentration in soil, and Z is soil depth. We adopted the model of Buckingham (1904) to estimate the relative gas diffusion coefficient (D_s/D_a), in which D_s/D_a is derived from the air-filled porosity squared, because the dataset on this relationship for Gray lowland soil (Kawamoto et al. 2005) apparently fitted to this model. Then D_s was obtained from multiplying D_s/D_a by the gas diffusion coefficient in air (D_a) at 20°C (0.1632 cm² s⁻¹) (Massman 1998). The air-filled porosity was calculated by subtracting the volumetric water content from the fixed value of porosity (0.61 m³ m⁻³). The volumetric water content at each depth in the soil profile was estimated supposing that it linearly increases from the measured value at a depth of 0.1 m to 1 m³ m⁻³ at the groundwater table (0.9 m depth). According to our preliminary checking, there was no obvious difference between the approximation to a straight line and a sigmoid curved line. Soil temperatures of the entire profile were fixed to be 20°C. Atmospheric N₂O concentration (0 m depth) was fixed to be 0.31 µL L⁻¹. We did not estimate diffusive N₂O flux near the boundary between soil and groundwater because of a lack of required information.

Stable isotope analysis

Water and gas samplings for the stable isotope analysis were conducted more than once a week from late June to September (from July to August for gas). The δ¹⁵N and δ¹⁸O of and N₂O in the water and gas samples were measured using an isotope ratio mass spectrometer (Mat 252, Finnigan Mat, Bremen, Germany) equipped with a GC (5890 Series II, Hewlett-Packard, Waldbronn, Germany) (GC-IRMS). The measurement precision for the GC-IRMS was ±0.22‰ for δ¹⁵N and ±0.28‰ for δ¹⁸O. The standard for ¹⁵N was atmospheric dinitrogen (N₂), and that for ¹⁸O was Vienna Standard Mean Ocean Water (VSMOW). The required amount of N₂O for reliable measurement by the GC-IRMS was 15 nmol, and the required volume of N₂O gas was dependent on the N₂O concentration of each gas sample.

The δ¹⁵N and δ¹⁸O of in the drainage water ( _drain) were analyzed by the denitrifier method (Sigman et al. 2001; Casciotti et al. 2002), in which was converted quantitatively into N₂O by denitrifying bacteria that lack N₂O reductase activity (Pseudomonas chlororaphis f. sp. aureofaciens, ATCC 13985). We extracted and purified the N₂O produced by the bacteria using a preparation line (PreCon, Thermoquest, Bremen, Germany), and analyzed it using the GC-IRMS. Details of the apparatus are given in Sigman et al. (2001) and Casciotti et al. (2002). Gas samples of dissolved N₂O in water were prepared using the headspace method.

We also measured the δ¹⁵N and δ¹⁸O of N₂O in the air inside a closed chamber (N₂O_chamber) and of N₂O in the soil air at a depth of 0.15 m (N₂O_soil). An incomplete adjustment of the GC-IRMS caused some loss of data. We installed a static closed chamber (0.75 m ×0.2 m × 0.1 m height) between crop rows without including crops for two hours to collect N₂O derived from the soil (N₂O_flux-c). We also collected atmospheric ambient N₂O (N₂O_air) at a height of 2 m above the ground on site. To simplify calculations of gas movement within the soil profile, we assumed that only gas diffusion was responsible for gas movement, and that the diffusion was driven by the N₂O concentration gradient. Therefore, although there could be both upward and downward N₂O fluxes in the soil profile, we assumed that only the upward flux occurred. This means that the gross downward N₂O flux was assumed to be zero and that the gross upward N₂O flux equaled the net upward N₂O flux. Hence, the amount of N₂O_chamber consisted of the amount of N₂O_air inside the chamber just after placement plus the amount of N₂O_flux-c. Given these assumptions, the amount (molar basis) of N₂O can be replaced by its concentration (µL L⁻¹ basis), and thus, the concentration of N₂O_chamber (=C_chamber) consists of the concentration of N₂O_air (C_air) and the calculated concentration of N₂O_flux-c (=C_chamber − C_air). Then, δ¹⁵N and δ¹⁸O of N₂O_flux-c were calculated from the following isotope mass-balance equation:

As for N₂O_soil, we used the same assumptions as those for N₂O_chamber, except for the direction of gas fluxes. Instead, we assumed that the soil at the 0.15 -m depth functioned as a semi-closed system, in which N₂O could enter and leave based on the concentration gradient. Hence, N₂O_air was assumed to enter the system, but downward movement to depths >0.15 m was assumed to be zero due to the existence of a plow pan below 0.15 m. The N₂O concentration in the system (C_soil) could increase as a result of the supply of N₂O derived from the soil (N₂O_flux-s), though some escaped outside the system. On this basis, we used the following equation to calculate δ¹⁵N and δ¹⁸O of N₂O_flux-s:

We preliminarily checked the validity of the assumptions used for Eq. 3 by comparing the calculated δ¹⁵N and δ¹⁸O of N₂O_flux-s with or without the entry of atmospheric N₂O against those of N₂O_flux-c. We did not report δ¹⁵N and δ¹⁸O values of N₂O_flux-c and of N₂O_flux-s when C_chamber and C_soil were below 0.33 µL L⁻¹, as we considered this value to be the lower threshold value for reliable analysis. For example, given δ¹⁵N-N₂O_soil = −10.0‰, δ¹⁸O-N₂O_soil = 40.0‰, C_soil = 0.50 µL L⁻¹, δ¹⁵N-N₂O_air = 6.0‰, δ¹⁸O-N₂O_air = 42.0‰, and C_air = 0.31 µL L⁻¹, calculated δ¹⁵N-N₂O_flux-s (−41.4‰) and δ¹⁸O-N₂O_flux-s (36.7‰) have an uncertainty of ±0.72‰ and ±1.55‰, respectively, due to error propagation by instrument precision.

We compared the δ¹⁵N and δ¹⁸O values of N₂O_flux-c and N₂O_flux-s with those of N₂O in the drainage water (N₂O_drain) and N₂O in the shallow groundwater (N₂O_sgw).

Statistical analysis

A paired t-test (p < 0.05) was performed to examine the effect of cropping system on the harvested aboveground biomass, drainage volume, leached and DOC, and N₂O and CO₂ emissions using statistical software (JMP 8.0, SAS Institute, Cary, NC, USA). Residual normality between dissolved N₂O concentration and soil-surface N₂O flux was preliminary analyzed with the Shapiro-Wilk's test for the spring (May 2007 and March–April 2008), the summer (June–August 2007), the autumn (September–November 2007), and the winter (December 2007–February 2008). Because the normality was rejected for all the seasons, the relationship between them was examined using the Spearman rank correlation coefficient. For the same reason, the relationship between the concentrations of DOC and dissolved N₂O through the year was examined by plot using the Spearman rank correlation coefficient.

Results

Crop biomass and leaching of and DOC

Aboveground biomasses of soybean and wheat were similar to those of local crops, whereas that of upland rice was 75% lower than expected (Table 1) probably because of its weak drought tolerance. The normal biomass in the SW plots and the low precipitation in August (23.5 mm versus the normal value of 121.8 mm) decreased the topsoil WFPS to < 30% (Fig. 2a and b). Accordingly, the groundwater level in the SW plots lowered by >0.2 m from the initial depth during this period (data not shown), which made it impossible to collect samples of the shallow groundwater. Such water movement was also observed in the previous studies (Thorburn et al. 1995; Zhang et al. 1999). The total drainage volume in the SW plots was significantly lower than that in the UR plots, and each accounted for an average of 32% and 54% of the total precipitation (1280 mm yr⁻¹), respectively (Table 1).

Figure 2. Seasonal courses of (a) daily precipitation and mean air temperature; (b) topsoil water-filled pore space (WFPS) and cumulative drainage volume; soil-surface nitrous oxide (N₂O) flux in (c) the soybean–wheat double cropped (SW) plots and (d) the upland rice (UR) plots; (e–g) soil N₂O profile, and the concentrations of (h) dissolved N₂O; (i) nitrate (); (j) dissolved carbon dioxide (CO₂), and (k) dissolved organic carbon (DOC) in the drainage water. Horizontal bars in (d) indicate the period of crop cultivation. Arrows indicates urea application and the rate (g N m⁻²). Basal dressing of urea in the UR plots (3 g N m⁻²) was conducted on 26 April 2007.

Table 1. Total crop biomass, drainage water volume, leached nitrate () and dissolved organic carbon (DOC), and the emissions of nitrous oxide (N₂O) and carbon dioxide (CO₂)

		SW1	SW2	UR1	UR2	t-test
Harvested aboveground biomass	(g DW m^−2yr^−1)	501/1100^†	517/1146^†	116	120	***
Drainage volume	(mm yr^−1)	421	400	677	699	**
Leached	(g N m^−2yr^−1)	2.77	2.88	14.4	13.6	**
Leached DOC	(mg C m^−2yr^−1)	638	564	789	984	NS
Soil-surface N2O emission	(mg N m^−2yr^−1)	139	122	307	150	NS
Dissolved N2O emission	(mg N m^−2yr^−1)	51.8	52.1	108	39.1	NS
Dissolved CO2 emission	(g C m^−2yr^−1)	6.64	7.08	7.88	6.92	NS
^†Data for soybean and wheat, respectively.
** and *** represent significant difference by t-test at p = 0.01 and 0.001, respectively; NS represents no significant difference by t-test at p = 0.05. SW, soybeanwheat double cropping; UR, upland rice.

The concentration ranged from 7.4 to 34.0 mg N L⁻¹ in the UR plots, whereas it ranged from 0.5 to 19.0 mg N L⁻¹ in the SW plots (Fig. 2i). Heavy precipitation often supplied to the resident groundwater. The total leached -N in the UR plots was significantly greater than that in the SW plots (Table 1). The concentration was generally <0.1 mg N L⁻¹ through the year (data not shown). The DOC concentration in the SW plots exceeded 2 mg C L⁻¹ from July to October (Fig. 2k); however, the total leached DOC did not differ between the SW and UR plots (Table 1).

N₂O dynamics

Soil-surface N₂O flux in the SW plots had sharp increases just after fertilizer N applications, and a slight increase in the flux occurred from March to April (Fig. 2c). Soil-surface N₂O flux in the UR plots had a broad and great increase from August to October that was interrupted by a large precipitation event in early September (Fig. 2a and d).

The N₂O concentration in the soil profile increased with depth through the year (Fig. 2e–g). From July to November, N₂O concentrations at the three depths stayed high. In addition, N₂O concentrations in the SW plots had slight increases from March to April. The N₂O concentrations in the SW plots decreased temporarily from early August to early September, the period of lowered groundwater level. The CO₂ concentration in the soil profile also increased with depth through the year, and the highest concentration at a depth of 0.75 m exceeded 20 mL L⁻¹ in all the plots (data not shown).

Dissolved N₂O concentrations in the drainage water of the UR plots reached 0.5 mg N L⁻¹ from July to October, which was interrupted by heavy precipitation (Fig. 2a and h) that subsequently diluted and displaced dissolved N₂O in the resident groundwater. Dissolved N₂O concentrations in the SW plots remained <0.1 mg N L⁻¹ from July until heavy precipitation in late October. The highest concentrations of dissolved N₂O (0.28 to 0.51 mg N L⁻¹) were comparable to those observed in our previous study (0.47 to 0.89 mg N L⁻¹; Minamikawa et al. 2010), and within the range of those observed in actual agricultural drainage water (0.03 to 9.98 mg N L⁻¹; Heincke and Kaupenjohann 1999, and references therein). Dissolved CO₂ concentration in the SW plots exceeded 20 mg C L⁻¹ from May to August and in April (Fig. 2j). In the SW plots, there were significant negative correlations between the concentrations of DOC and dissolved N₂O (Fig. 3).

Figure 3. Relationships between dissolved organic carbon (DOC) concentration and dissolved nitrous oxide (N₂O) concentration in the drainage water. Statistical significance was tested by the Spearman rank correlation coefficient. SW, soybean–wheat double cropping; UR, upland rice.

As shown in Fig. 4, there were significant positive correlations between dissolved N₂O concentration and soil-surface N₂O flux for all the seasons. However, the magnitude of soil-surface N₂O flux against dissolved N₂O concentration in the spring and winter was generally lower than that in the summer and autumn. Calculated diffusive N₂O fluxes at the three depths in the UR plots (Fig. 5) had similar seasonal course and magnitude to the soil-surface N₂O flux (Fig. 2d), whereas those in the SW plots were generally lower than the soil-surface flux (Fig. 2c).

Figure 4. Relationships between dissolved nitrous oxide (N₂O) concentration in the drainage water and soil-surface N₂O flux for (a) the spring, (b) the summer, (c) the autumn, and (d) the winter. Statistical significance for all four plots was tested by the Spearman rank correlation coefficient. SW, soybean–wheat double cropping; UR, upland rice.

Figure 5. Seasonal courses of diffusive nitrous oxide (N₂O) flux at a depth of (a) 0.075 m, (b) 0.3 m, and (c) 0.6 m in the soil profile calculated by Fick's law. SW, soybean–wheat double cropping; UR, upland rice.

δ¹⁵N and δ¹⁸O values of _drain and N₂O_drain

The concentrations in the drainage water steadily decreased from June to September, except for the heavy precipitation events (Fig. 6a). The δ¹⁵N and δ¹⁸O of _drain followed courses opposite to those for the concentration (Fig. 6b and c). The slope for the relationship between δ¹⁵N and δ¹⁸O (Δδ¹⁸O/Δδ¹⁵N) of _drain between the two heavy precipitation events (i.e., from 19 July to 31 August) ranged between 0.61 and 0.66 in the SW plots and between 0.82 and 0.90 in the UR plots (Fig. 7a).

Figure 6. Detailed time courses of the concentration, nitrogen (δ¹⁵N), and oxygen (δ¹⁸O) of (a–c) nitrate concentrations in the drainage water ( _drain) and (d–f) nitrous oxide concentrations in the drainage water (N₂O_drain) from June to September 2007. Arrows indicates the timing of urea application and the rate (g N m⁻²). SW, soybean–wheat double cropping; UR, upland rice.

Figure 7. Relationships between nitrogen (δ¹⁵N) and oxygen (δ¹⁸O) for (a) nitrate concentrations in the drainage water ( _drain) and (b) nitrous oxide concentrations in the drainage water (N₂O_drain) from 19 June to 31 August 2007. Dashed and solid lines indicate a slope of 0.5 and 1.0, respectively. SW, soybean–wheat double cropping; UR, upland rice.

Dissolved N₂O concentrations in the UR plots increased during this period, except for the heavy precipitation events, whereas those in the SW plots remained at their lowest level of the year (Fig. 6d). There were significant (P < 0.01) and strong (r = −0.79 to −0.81) negative linear correlations between the concentrations of and dissolved N₂O in the UR plots between the two heavy precipitation events (data not shown). The δ¹⁵N and δ¹⁸O of N₂O_drain were unstable, but generally followed similar courses to those of _drain (Fig. 6e and f). The Δδ¹⁸O/Δδ¹⁵N of N₂O_drain between the two heavy precipitation events ranged between 0.35 and 0.68, except for the SW1 plot with some outliers (Fig. 7b). The δ¹⁵N values of N₂O_drain were 10.1‰ to 39.2‰ lower than those of _drain, whereas δ¹⁸O values of N₂O_drain were 44.1‰ to 70.0‰ higher.

Vertical comparison of the δ¹⁵N and δ¹⁸O of N₂O

Soil-surface N₂O flux and N₂O concentration at a depth of 0.15 m in the UR plots reached their highest values in August (Fig. 8a and b). The δ¹⁵N values of N₂O_flux-c and N₂O_flux-s ranged from −17.7‰ to 1.8‰ in the SW plots and from −40.6‰ to −6.7‰ in the UR plots (Fig. 8c and e). Differences in δ¹⁵N between N₂O_flux-c and N₂O_flux-s were generally less than ±5‰. The δ¹⁸O values of N₂O_flux-c and N₂O_flux-s ranged between 35.5‰ and 55.8‰ (Fig. 8d and f).

Figure 8. Detailed time courses of (a) soil-surface nitrous oxide (N₂O) flux; (b) N₂O concentration at a 0.15 m depth; (c) nitrogen (δ¹⁵N) and (d) oxygen (δ¹⁸O) of N₂O_flux-c; (e) δ¹⁵N and (f) δ¹⁸O of N₂O_flux-s; (g) δ¹⁵N and (h) δ¹⁸O of N₂O_sgw; and (i) δ¹⁵N and (j) δ¹⁸O of N₂O_drain from July to August 2007. Arrow in graph (a) indicates urea topdressing in the upland rice (UR) plots (3 g N m⁻²). Graphs (g) and (h) lack data of the soybean–wheat double cropping (SW) plots due to the lowered groundwater level.

The δ¹⁵N of N₂O_sgw in the UR plots ranged from −44.7‰ to −16.8‰, and were 1.9‰ to 28.9‰ lower than the δ¹⁵N of N₂O_drain (Fig. 8g and i). The δ¹⁵N of N₂O_drain in the SW plots ranged from −31.5‰ to 8.4‰ (Fig. 8i). The δ¹⁸O of N₂O_sgw in the UR plots ranged from 39.1‰ to 49.4‰, whereas those of N₂O_drain in the UR and SW plots ranged from 44.6‰ to 70.2‰ (Fig. 8h and j).

The δ¹⁵N values of N₂O_flux-c and N₂O_flux-s in the SW plots were generally between 5‰ lower and 15‰ higher than δ¹⁵N of N₂O_drain (Fig. 8c, e and i). The δ¹⁸O values of N₂O_flux-c and N₂O_flux-s in the SW plots were generally 10‰ to 20‰ lower than those of N₂O_drain (Fig. 8d, f and j). The time courses of the δ¹⁵N values of N₂O_flux-c and N₂O_flux-s in the UR plots were similar to those of δ¹⁵N of N₂O_sgw, but the δ¹⁵N values of N₂O_flux-c and N₂O_flux-s were generally 5‰ to 10‰ higher than the δ¹⁵N of N₂O_sgw (Fig. 8c, e and g). Differences in δ¹⁸O values among N₂O_flux-c, N₂O_flux-s, and N₂O_sgw were generally ±5‰ in the UR plots (Fig. 8d, f and h).

Discussion

Denitrification near the groundwater table

The concentration, δ¹⁵N, and δ¹⁸O of _drain in the SW and UR plots between the two heavy precipitation events (Fig. 6a–c) exhibited typical time courses for the progress of denitrification. The high values of δ¹⁸O for N₂O_drain were probably due to the oxygen in N₂O which originated from both and water (Casciotti et al. 2002; Well and Flessa 2009). The Δδ¹⁸O/Δδ¹⁵N for _drain (0.61 to 0.90) fell within the range observed in groundwater, lake, and pure laboratory culture of denitrifying bacteria (0.47 to 1.02; Lehmann et al. 2003; Singleton et al. 2007; Granger et al. 2008). In addition, dissolved O₂ concentration (DO) in the drainage water was generally <1 mg L⁻¹, except for heavy precipitation events (data not shown). These results clearly confirm that denitrification occurred near the groundwater table in the SW and UR plots during the summer.

Apparent differences in δ¹⁵N between N₂O_drain and _drain (10.1‰ to 39.2‰) fell within the range for N isotope enrichment factor for denitrification in culture experiments and natural environments (0‰ to 40‰; Lehmann et al. 2003, and references therein) though further denitrification process that reduces N₂O to N₂ may have shrunk the difference in δ¹⁵N. The Δδ¹⁸O/Δδ¹⁵N for N₂O_drain (0.35 to 0.68) was comparable to those observed for N₂O produced by denitrification (Meijide et al. 2010; Ostrom et al. 2010). In addition, the significant negative linear correlations between the concentrations of and dissolved N₂O in the drainage water of the UR plots can be explained by N₂O production by denitrification, even if nitrification had a certain contribution to N₂O production. Furthermore, if nitrification is the main process of N₂O production, there will be a positive relationship between the concentrations of and dissolved N₂O because N₂O and are an intermediate and the end products of nitrification, respectively. Such a relationship was observed in previous studies (e.g., Ueda et al. 1993; Mühlherr and Hiscock 1998), but we did not find it in the SW and UR plots (see Fig. 2h and i). Mass balance of N between the detected (<0.1 mg N L⁻¹, data not shown) and the existing N₂O in the drainage water that temporally exceeded 0.4 mg N L⁻¹ (Fig. 2h) also indicates that nitrification was not the main process for N₂O production. Therefore, we conclude that denitrification was the main process for N₂O production near the groundwater table in our lysimeters during the summer. Although we did not measure the δ¹⁵N of in the shallow groundwater, the observed δ¹⁵N of N₂O_sgw in the UR plots was relatively low in August (<−40‰, Fig. 8g), which may suggest that nitrification had some contribution to N₂O production. As for the other seasons, denitrification may also have been the main process for N₂O production because of the constant low and DO through the year, as well as for the reason described below.

Cropping system had strong influences on leaching of and DOC in the soil profile (Table 1, Fig. 2b, j and k). The high concentration in the UR plots was due to the low N uptake by upland rice (Fig. 2i). As reported in Kusa et al. (2010), macropores and cracks peculiar to Gray lowland soil would be a cause of rapid leaching of . It is well known that denitrification in subsoil and groundwater is limited by a lack of readily available C (e.g., Hedin et al. 1998; Hill et al. 2000; von der Heide et al. 2008). From May to July and in the following April, CO₂ production and consumption in the SW plots (see Fig. 2i and j) were relatively well balanced stoichiometrically for denitrification. During this period, dissolved N₂O concentration in the SW plots also decreased correspondingly (Fig. 2h). Weymann et al. (2010) observed a significant negative relationship between the highest dissolved N₂O concentration and DOC concentration in incubation experiments on the groundwater in a German aquifer. Such relationships were found only in the SW plots (Fig. 3a and b). As reviewed in Kuzyakov and Domanski (2000) and Chantigny (2003), it is obvious that the greater plant biomass is, the greater root exudates and plant-derived C leach. Therefore, the low concentrations of and dissolved N₂O in the SW plots would be explained by appropriate N uptake by crops and enhanced denitrification due to the great wheat biomass.

However, as was the case for the UR plots (Fig. 3c and d), there have been several cases in which DOC was not a significant predictive variable for the magnitude of N₂O production by denitrification partly because of its poor bioavailability (e.g., Deurer et al. 2008). Observed concentration range of DOC in the UR plots (Fig. 3c and d) was narrower and lower than those in the previous studies that showed the significant relationship (e.g., Hill et al. 2000; von der Heide et al. 2008), and thus the range in the UR plots may not have developed strict reductive conditions preferable to N₂O reduction to N₂. More detailed studies of the bioavailability of DOC will be necessary to clarify the difference in N₂O production by denitrification between the two cropping systems.

Upward diffusion of N₂O produced near the groundwater table

Significant positive correlation between dissolved N₂O concentration and soil-surface N₂O flux in the summer (Fig. 4b) strongly supports our hypothesis. The correlation was also observed in the other seasons (Fig. 4a, c and d), suggesting that our hypothesis would apply to the other seasons. As for the seasonal difference in the magnitude of the upward N₂O diffusion, Van Groenigen et al. (2005) also observed a similar discrepancy between the summer and winter in a sandy soil field with a shallow groundwater table (0.5 to 1.2 m depth). They argued that high WFPS in the winter would increase the diffusion time for N₂O from the subsoil to the atmosphere, thereby increasing the likelihood for further N₂O reduction to N₂. We expect that this explanation would also be valid in the present study. Therefore, the comparison of the measured data among the seasons indicates that soil water content (WFPS; Fig. 2b) was an important factor determining the magnitude of upward N₂O diffusion in our lysimeters.

Calculated diffusive N₂O fluxes in the soil profile of the SW plots (Fig. 5) indicate that soil-surface N₂O flux was mostly derived from the topsoil. In contrast, those in the UR plots indicate that soil-surface N₂O fluxes were mainly derived from >0.6 m depth during the summer and autumn. To examine the N₂O movement at further depths in the UR plots, we used dissolved N₂O concentration in the drainage water, which can also be expressed as the equilibrated N₂O concentration in the vapor phase by Henry's law. For example, the highest dissolved N₂O concentration in the UR1 plot (0.510 mg N L⁻¹) was supposed to equilibrate with 635 µL L⁻¹ of N₂O at 19.2°C of the measured water temperature. Although all the converted N₂O concentrations in the UR plots were not shown here, they (2–727 µL L⁻¹) were always higher than the measured N₂O concentrations at a depth of 0.75 m through the year (0.3–94.2 µL L⁻¹; Fig 2g). This indicates that the N₂O source for the upward diffusion was at >0.75 m depth.

Inconsistency of δ¹⁵N and δ¹⁸O values of N₂O_drain with those of N₂O_flux-c or N₂O_flux-s in the SW plots (Fig. 8c–f, i and j) indicates that the soil-surface N₂O flux was mainly derived from other source(s) than the upward diffusion. The δ¹⁵N values of N₂O_flux-c and N₂O_flux-s were similar to but not exactly consistent with those of N₂O_sgw in the UR plots (Fig. 8c, e and g). This discrepancy may suggest that there were other possible N₂O sources and processes that changed the δ¹⁵N values of N₂O_flux-c and N₂O_flux-s. Urea application often induced temporal N₂O fluxes (Fig. 2c and d), probably because of nitrification in the topsoil. The δ¹⁵N of the N₂O_flux-c just after the topdressing was −27.5‰ and −19.0‰ in the UR plots (Fig. 8c). In addition, we measured the δ¹⁵N of the N₂O_flux-c just after urea incorporation in May 2008 (−22.6‰). These values were comparable to those observed in the UR plots in late August when the soil-surface N₂O flux remained high. Therefore, from the viewpoint of isotopic signatures of N₂O, we could not exclude the topsoil nitrification from possible N₂O sources for soil-surface emission during the summer. However, in our lysimeters, Nishimura et al. (2005a) reported that, considering topsoil and contents, the contribution of topsoil nitrification to soil-surface N₂O emission was relatively low. Therefore, we assumed that topsoil nitrification alone could hardly explain the broad and high N₂O fluxes in the UR plots from the summer to the autumn. Denitrification in the topsoil was also unlikely during this period because the topsoil WFPS was around 30% to 40% (Fig. 2b), which is low enough to significantly impede N₂O production by denitrification (Davidson 1991). In contrast, N₂O reduction to N₂ due to entrapment (dissolution) of N₂O in soil water (Clough et al. 1999, 2005) and in anaerobic microsites (Meijide et al. 2010) would be plausible mechanisms that made heavier the δ¹⁵N of N₂O that was moving upward. Therefore, our natural abundance method was limited in examining the upward N₂O diffusion by itself; however, the obtained results were not inconsistent with our hypothesis, considering other possible sources and processes.

From the combination of the above-mentioned three indirect observations, we conclude that our hypothesis was verified in the present study. It will be difficult to simultaneously explain the three results by other sources and processes. Two kinds of upland cropping systems provided distinctive experimental conditions for the dissolved N₂O concentration during the summer. The upward N₂O diffusion would have occurred in the UR plots that had high dissolved N₂O concentration in the drainage water during the summer. In contrast, the upward N₂O diffusion would have been negligible in the SW plots that had the low dissolved N₂O concentration during the summer. The δ¹⁸O of N₂O_sgw in the UR plots could not be used as an isotopic signature for the upward movement because these values (approx. 40‰ to 50‰; Fig. 8h) were close to that of the atmospheric N₂O. Of course, the usage of ¹⁵N-labeled and N₂O (e.g., Clough et al. 2006) would provide direct evidence for the upward diffusion. However, we did not adopt it to avoid a long-term residence of ¹⁵N-labeled compounds in the soil and groundwater for the following experiments, and also to discuss simultaneously the production process of N₂O dissolved in the drainage water by a natural abundance method.

Soil-surface N₂O emission from August to October in the UR plots accounted for 63% to 66% of the annual total, versus 25% to 30% in the SW plots. Given that most soil-surface N₂O emission in the UR plots during this period was derived from that produced near the groundwater table, its quantitative significance is potentially important, as compared to any previous studies. Weymann et al. (2009) reported that only 0.04–0.28% of the soil-surface N₂O flux originated from groundwater-derived N₂O. The highest dissolved N₂O concentrations in their study were between 0.014 and 0.016 mg N L⁻¹, which were one order of magnitude lower than our study. The N₂O concentrations in the soil profile of their study were between 0.297 and 0.427 µL L⁻¹ at depths of 0.3–0.9 m, which were the same order as the atmospheric ambient concentration. The reason for this negligible contribution would be due to high permeability of the soil (Podzols and Gleysols, Deurer et al. 2008). On the other hand, we assume that there were two plausible causes that explain the high contribution in our lysimeters. First, the structure of the bottom of the lysimeter was a closed system that did not allow the resident groundwater to flow out without additional precipitation supply. As indicated by the constant low DO values, such a structure would have developed the redox conditions favorable to denitrification. Another cause is the enhanced upward N₂O diffusion during the summer due to the soil drying, as mentioned above. The macropores and cracks could also enhance the upward N₂O diffusion as reported in Kusa et al. (2010). It is necessary to investigate the case in actual agricultural fields on lowland soils and poorly drained soils.

Conclusion

The present study examined the following hypothesis: if a large amount of N₂O is produced near the shallow groundwater table in the summer, it will diffuse upward to the atmosphere. Dissolved N₂O concentration in the drainage water in the UR plots exceeded 0.4 mg N L⁻¹ during the summer, whereas that in the SW plots remained <0.1 mg N L⁻¹. Analyses of the concentrations of and dissolved N₂O and their δ¹⁵N and δ¹⁸O during the summer revealed that denitrification was the main process for N₂O production near the groundwater table. There was a significant positive correlation between the dissolved N₂O concentration and soil-surface N₂O flux in the summer. The course and magnitude of diffusive N₂O fluxes at the three depths in the soil profile in the UR plots were almost consistent with those of soil-surface N₂O flux during this period. The δ¹⁵N values of N₂O in the soil-surface flux were similar to those in the shallow groundwater in the UR plots during the summer, whereas such similarity was not found in the SW plots. We conclude that our hypothesis was confirmed by the above results. Comparison of the monitored data with the other seasons indicates that soil water content is a key factor determining the magnitude of upward N₂O diffusion as well as the dissolved N₂O concentration in our lysimeters.

Acknowlegments

We thank Drs. Hiroko Akiyama and Shigeto Sudo (NIAES) for their help using the GCs. We also thank Drs. Karamat Sistani (USDA Agricultural Research Service) and Sadao Eguchi (NIAES) for valuable comments on the manuscript before submission. This study was funded by the Global Environmental Research Fund (S2-3a) of Japan's Ministry of the Environment.