Plant and Soil

, Volume 311, Issue 1, pp 211–234

Quantification of N2O fluxes from soil–plant systems may be biased by the applied gas chromatograph methodology

Authors

    • State Key Laboratory of Atmospheric Boundary Layer Physics and Atmospheric Chemistry, Institute of Atmospheric PhysicsChinese Academy of Sciences
  • Baoling Mei
    • State Key Laboratory of Atmospheric Boundary Layer Physics and Atmospheric Chemistry, Institute of Atmospheric PhysicsChinese Academy of Sciences
  • Yinghong Wang
    • State Key Laboratory of Atmospheric Boundary Layer Physics and Atmospheric Chemistry, Institute of Atmospheric PhysicsChinese Academy of Sciences
  • Baohua Xie
    • State Key Laboratory of Atmospheric Boundary Layer Physics and Atmospheric Chemistry, Institute of Atmospheric PhysicsChinese Academy of Sciences
  • Yuesi Wang
    • State Key Laboratory of Atmospheric Boundary Layer Physics and Atmospheric Chemistry, Institute of Atmospheric PhysicsChinese Academy of Sciences
  • Haibo Dong
    • State Key Laboratory of Atmospheric Boundary Layer Physics and Atmospheric Chemistry, Institute of Atmospheric PhysicsChinese Academy of Sciences
  • Hui Xu
    • Institute of Applied EcologyChinese Academy of Sciences
  • Guanxiong Chen
    • Institute of Applied EcologyChinese Academy of Sciences
  • Zucong Cai
    • Institute of Soil ScienceChinese Academy of Sciences
  • Jin Yue
    • State Key Laboratory of Atmospheric Boundary Layer Physics and Atmospheric Chemistry, Institute of Atmospheric PhysicsChinese Academy of Sciences
  • Jiangxin Gu
    • State Key Laboratory of Atmospheric Boundary Layer Physics and Atmospheric Chemistry, Institute of Atmospheric PhysicsChinese Academy of Sciences
  • Fang Su
    • China Agricultural University
  • Jianwen Zou
    • Nanjing Agricultural University
  • Jianguo Zhu
    • Institute of Soil ScienceChinese Academy of Sciences
Regular Article

DOI: 10.1007/s11104-008-9673-6

Cite this article as:
Zheng, X., Mei, B., Wang, Y. et al. Plant Soil (2008) 311: 211. doi:10.1007/s11104-008-9673-6

Abstract

With regard to measuring nitrous oxide (N2O) emissions from biological sources, there are three most widely adopted methods that use gas chromatograph with an electron capture detector (GC–ECD). They use: (a) nitrogen (N2) as the carrier gas (DN); (b) ascarite as a carbon dioxide (CO2) trap with DN (DN-Ascarite); and (c) a mixture gas of argon and methane as the carrier (AM). Additional methods that use either a mixture of argon and methane (or of CO2 and N2) as a make-up gas with the carrier nitrogen or soda lime (or ascarite) as a CO2 trap with the carrier helium have also been adopted in a few studies. To test the hypothesis that the use of DN sometimes considerably biases measurements of N2O emissions from plants, soils or soil–plant systems, experiments were conducted involving DN, AM and DN-Ascarite. When using DN, a significant relationship appeared between CO2 concentrations and the apparent N2O concentrations in air samples. The use of DN led to significantly overestimated N2O emissions from detached fresh plants in static chamber enclosures. Meanwhile, comparably lower emissions were found when using either the DN-Ascarite or AM methods. When an N2O flux (from a soil or a soil–plant system), measured by DN in combination with sampling from the enclosure of a static opaque chamber, was greater than 200 μg N m−2 h−1, no significant difference was found between DN and DN-Ascarite. When the DN-measured fluxes were within the ranges of <−30, −30–0, 0–30, 30–100 and 100–200 μg N m−2 h−1, significant differences that amounted to −72, −22, 5, 38 and 64 μg N m−2 h−1, respectively, appeared in comparison to DN-Ascarite. As a result, the DN measurements in rice–wheat and vegetable fields overestimated both annual total N2O emissions (by 7–62%, p < 0.05) and direct emission factors for applied nitrogen (by 6–65%). These results suggest the necessity of reassessing the available data determined from DN measurements before they are applied to inventory estimation. Further studies are required to explore appropriate approaches for the necessary reassessment. Our results also imply that the DN method should not be adopted for measuring N2O emissions from weak sources (e.g., with intensities less than 200 μg N m−2 h−1). In addition, we especially do not recommend the use of DN to simultaneously measure N2O and CO2 with the same ECD.

Keywords

Nitrous oxideEmissionsGas chromatographyCarrier gasSoilPlantInventory

Abbreviations

GC–ECD

Gas chromatograph equipped with an electron capture detector.

DN

A GC–ECD method that uses pure N2 alone as the carrier gas but does not introduce a high-concentration make-up gas with an electronegative group/atom or with an ionization potential lower than that of N2 (such as carbon dioxide or methane) into the detector while allowing carbon dioxide (CO2) in the air samples to enter the detector.

DN-Ascarite

Different from the DN method only in that an ascarite (a type of sodium-hydroxide-coated silica) filter column is connected to the beginning of the separation column. As a result of this ascarite filter, CO2 and water vapor in the air samples cannot enter the detector.

AM

A GC–ECD method that uses an argon–methane mixture instead of pure N2 as the carrier gas while having the same GC configuration as the DN method.

DN-AM

A GC–ECD method that uses N2 as the carrier gas and directly introduces a mixture of methane and argon (5–10% CH4 in Ar) into the detector as a make-up gas.

DN–CO2

A GC–ECD method that uses N2 as the carrier gas and directly introduces a mixture of CO2 and N2 into the detector as a make-up gas.

He-Ascarite

A GC–ECD method that uses He as the carrier gas, along with an ascarite (or soda lime) filter column connected to the beginning of the separation column.

Introduction

Nitrous oxide (N2O) is the fourth largest single contributor to positive radiative forcing, which contributes to ongoing global warming (IPCC 2007). Natural and managed soils have been identified as the biggest sources of atmospheric N2O. In addition, living plants also release N2O (e.g., Chen et al. 1990; Smart and Bloom 2001). To quantify the intensities from these sources, N2O emission measurements from natural and managed soils, soil–plant systems (e.g., Stehfest and Bouwman 2006) and plants (e.g., Hakata et al. 2003; Pihlatie et al. 2005; Zou et al. 2005) have been carried out worldwide. Global estimates of 1.7–4.8 and 3.3–9.0 × 1012 g N2O–N yr−1 for agricultural terrain and soils under natural vegetation, respectively, have been given (IPCC 2007). However, additional efforts relying on methods that are capable of accurately quantifying N2O emissions from soils and plants alone or soil–plant systems are still required to reduce the present estimation uncertainties.

The majority of the available measurements have been conducted by collecting air samples in static chambers and then analyzing the N2O concentration in the samples with a gas chromatograph equipped with an electronic capture detector (GC–ECD). Other methods, such as micrometeorological techniques using fast response detectors like tunable diode laser absorption spectroscopes (Lavillea et al. 1999), have also been used, albeit less frequently. There are a number of different GC–ECD methods involved in N2O emission measurement. DN, DN-Ascarite and AM have been the most widely used. DN refers to the method using pure N2 as the carrier gas (e.g., Arah et al. 1994; Loftfield et al. 1997; Clough et al. 2006; Purbopuspito et al. 2006). With respect to this method, no make-up gas with a lower ionization potential (like carbon dioxide or methane) than that of N2 is directly introduced into the detector, while carbon dioxide (CO2) from the air sample is allowed to fully enter the detector. For the DN-Ascarite method, an ascarite (a type of sodium-hydroxide-coated silica) filter column is connected to the beginning of the separation column (e.g., Butterbach-Bahl et al. 1997; Breuer et al. 2000; Kiese et al. 2003; Zheng et al. 2000; Holst et al. 2007). The difference between the DN and DN-Ascarite methods is that the ascarite filter traps CO2 from air samples before it enters the separation column; thus, CO2 does not enter the detector. AM refers to a method using a mixture of argon and methane as the carrier (e.g., Mosier and Mack 1980; Loftfield et al. 1992; Henrich and Haselwandter 1997; Hadi et al. 2000; Maljanen et al. 2003; Nishimura et al. 2004; Li et al. 2004; Ding et al. 2007). A CO2 trap is usually not adopted for AM. In addition, a few researchers use a mixture of argon and methane (5% CH4 in Ar) as a make-up gas with N2 as a carrier (hereinafter referred to as DN-AM) to increase the sensitivity of the ECD (Nykänen et al. 1995; Regina et al. 2004). A few other researchers use soda lime (a mixture of Ca(OH)2 and NaOH as a CO2 absorbent) with helium as a carrier gas for the GC–ECD (hereinafter referred to as He-Ascarite) to analyze N2O (Kester et al. 1997). When DN is used in a GC–ECD without a back-flushing design coupled with a switch between the separation column and the detector, ascarite must be used to remove CO2 from the air samples in order to improve N2O separation (e.g., Butterbach-Bahl et al. 1997; Holst et al. 2007). This is because the CO2 peak appears between, and partially overlaps the oxygen (O2) and N2O peaks, obscuring the results. However, the DN method can very effectively separate N2O from CO2 on a chromatogram of an air sample (Fig. 1a and d) when the sample is analyzed with a GC–ECD equipped with a back-flushing design using a switch to connect and disconnect the column and the ECD (e.g., Flessa et al. 1995; Wang and Wang 2003). For such GC–ECD configurations, ascarite was not adopted to remove CO2 (e.g., Wang and Wang 2003; Teepe et al. 2004; Purbopuspito et al. 2006). In a number of studies using DN in such instruments, both the N2O and CO2 contents of air samples were simultaneously detected with an ECD (e.g., Loftfield et al. 1997). DN is much more economical and convenient for routine maintenance in remote regions when compared with the AM method, whose use may be limited because of high costs and other inconveniences. The DN method has therefore been widely adopted for measuring N2O emissions from various sources (e.g., Arah et al. 1994; Loftfield et al. 1997; Clough et al. 2006; Purbopuspito et al. 2006; Zou et al. 2005; Du et al. 2006) after being first reported in the early 1970s (Wentworth and Freeman 1973). So far, however, few studies have been conducted to compare the N2O emissions measured with the DN, DN-Ascarite and AM methods.
https://static-content.springer.com/image/art%3A10.1007%2Fs11104-008-9673-6/MediaObjects/11104_2008_9673_Fig1_HTML.gif
Fig. 1

Chromatographic signals of nitrous oxide (N2O) and carbon dioxide (CO2) in dry ambient air as a response to different gas chromatography methods. a, b, c Shimadzu GC-14B2 (300°C detector temperature) using DN, DN-Ascarite and AM, respectively. d, e, f Agilent 6820 (330°C detector temperature) using DN, DN-Ascarite and AM, respectively. The concentrations of N2O and CO2 in dry ambient air were 330 × 10−9 mol mol−1 and 403 × 10−6 mol mol−1, respectively. Both N2O and CO2 peaks were detected by the electron capture detectors. Detailed configurations of the gas chromatographs are described in Table 1. Definitions of DN, DN-Ascarite and AM are given in the “Abbreviations” section of the text

When observing N2O emissions from a tropical forest floor with translucent static chambers (Werner et al. 2006), we noticed a significant difference in the N2O effluxes between DN and DN-Ascarite (19 ± 10 vs. 5 ± 3 μg N m−2 h−1 for 26 observations, p < 0.001, one-way ANOVA and paired t-test). In this case, the DN-Ascarite method was used in an SRI 8610C GC (SRI Instruments, CA, USA). A back-flushing design coupled with a column-detector switch was not adopted in the GC instrument (e.g., Butterbach-Bahl et al. 1997; Werner et al. 2006; Holst et al. 2007). The DN method was used in an Agilent 4890D GC (Agilent Technologies, Palo Alto, California, USA), in which a back-flushing design coupled with a column-detector switch was adopted (Wang and Wang 2003). The N2O peak was separated very well from the O2 and H2O vapor peaks in the SRI 8610C GC, which did not allow CO2 to enter the detector, and from the CO2 and H2O peaks in the Agilent 4890D GC, which did not allow O2 to enter the detector. Most recently, we observed a similar difference in measured N2O fluxes from a typical semi-arid steppe between the two GC–ECD methods when using DN-Ascarite in the same SRI 8610C, and using DN in either an Agilent 6890 (Agilent Technologies, Palo Alto, CA) or an Agilent 6820 (Agilent Technologies, Shanghai, China) instrument (Holst et al. 2007). The configurations for all Agilent GC–ECDs were the same as described by Wang and Wang (2003). However, when using DN-Ascarite in the Agilent and SRI GC–ECDs, we observed statistically comparable N2O fluxes (Holst et al. 2007). This suggests that the significant difference in measured N2O fluxes between the DN and DN-Ascarite methods might not have been caused by the different GC–ECD types. We speculate that the differences were caused by variable CO2 concentrations in the air samples because DN-Ascarite removed this component from the air samples, but DN did not.

If the CO2-induced influence is valid, we hypothesize that N2O fluxes from soils, plants or soil–plant systems could often be biased when they are measured using DN to analyze air samples from static chamber enclosures because the CO2 abundance is not constant during sampling (e.g., Xu et al. 2006; Zou et al. 2005). In this study, we explore this issue through comparative experiments in both the laboratory and the field.

Materials and methods

Instrumental methods

Our experiments involved a number of different GC–ECD types: the Agilent 6820, the Agilent 4890D, the HP5890II (Hewlett Packard, Palo Alto, California, USA) and the Shimadzu GC-14B (Shimadzu Seisakusho, Kyoto, Japan). The details of the instrument configurations are listed in Table 1. With regard to N2O detection using each of these GC–ECDs (with an exception for a Shimadzu instrument), we compared the DN, DN-Ascarite and AM methods. For the Shimadzu GC-14B1 (see Table 1), AM was exclusively used. For each method, the carrier flow rate was set at about 25 cm3 min−1, which yielded a retention time of about 3.5 min for N2O peak appearance. Further details and criteria of the gas-flow circuits for all instruments are as described in Wang and Wang (2003). For DN-Ascarite, a filter column (stainless steel, ID 1/4 in., 10 cm long) filled with ascarite (20–30 mesh) (Chemicalreagents, Beijing, China) was connected to the beginning of the separation column (3 m long). For AM, the N2 carrier was directly substituted with an argon–methane mixture and the ascarite filter column was removed.
Table 1

Configuration of the gas chromatograph instruments involved in this study

Instrument name

Back-flushinga

Column-ECD switch

Column for special gas species

Working temperature (°C)

Nitrous oxide (N2O)

Carbon dioxide (CO2)

Methane (CH4)

Column

ECD

FID

NCC

Shimadzu GC-14B1

Yes

Yesb

Porapak Q, 80–100 mesh

NA

NA

65

300

NA

NA

Shimadzu GC-14B2

Yes

Yesc

Porapak Q, 80–100 mesh

NA

13XMS, 60–80 mesh

55

300

200

NA

HP5890II 2

Yes

Yesc

Porapak Q, 80–100 mesh

Porapak Q, 60–80 mesh

13XMS, 60–80 mesh

55

330

200

375

Agilent 4890D2

Yes

Yesc

Porapak Q, 80–100 mesh

Porapak Q, 60–80 mesh

13XMS, 60–80 mesh

55

330

200

375

Agilent 68202

Yes

Yesc

Porapak Q, 80–100 mesh

Porapak Q, 60–80 mesh

13XMS, 60–80 mesh

55

330

200

375

ECD electron capture detector; FID flame ionization detector; NCC nickel catalytic converter for reducing CO2 into CH4 and thus detecting it with the FID (e.g., Wang and Wang 2003); NA not available. Definitions of DN, DN-Ascarite and AM are given in the “Abbreviations” section of the text

1Only AM (5% methane in argon) was adopted

2AM (10% methane in argon), DN and DN-Ascarite were examined for each instrument

aTo prevent water vapor (H2O) and other components from entering into the second separation column and thus the detector (e.g., Loftfield et al. 1997; Wang and Wang 2003)

bTo prevent oxygen (O2) and CO2 from entering the detector

cTo prevent O2 from entering the detector (e.g., Wang and Wang 2003)

Laboratory experiments

Laboratory experiments involving all three methods were conducted to compare (a) the detection precision for N2O concentrations in dry ambient air, (b) the ECD signals of a fixed N2O concentration as a response to differing CO2 concentrations, and (c) the N2O emission rates of detached fresh plants in static and gas-tight enclosures.

Dry ambient air with N2O and CO2 concentrations of 330 × 10−9 and 403 × 10−6 mol mol−1, respectively, was employed for laboratory precision comparisons (n ≥ 10) and N2O signal responses to variations in CO2 concentrations. In this dry ambient air, CO2 concentrations were modified by the addition of pure CO2. As a small amount of pure CO2 gas (<1 ml) was added to a large air volume (approx. 1,000 ml), the resulting change in N2O concentration was negligible.

We investigated 12 different species to compare N2O emission rates from detached fresh plants measured with different GC–ECD methods. The species were paddy rice (Oryza sativa L.), radish (Raphanus sativus L.), oil rape (Brassica chinensis var. oleifera), wheat (Triticum aestivum L.), cabbage (Brassica chinensis L.), mulberry (Morus alba L.), cypress (Juniperus formosana Hayata), lablab (Dolichos lablab L.), poplar (Populus simonii Carr.), eggplant (Solanum melongena L.), amaranth (Amaranthus mangostanus L.) and Chinese cabbage (Brassica pekinensis (Loureiro) Rupr.). The measurements were conducted by enclosing washed plants (seedlings of cabbage, radish, Chinese cabbage and amaranth) or detached fresh leaves or branches (aboveground parts of paddy rice, wheat and amaranth; leaves of the remaining species) in gas-tight plastic, translucent chambers (i.e., common mineral water bottles) exposed to bright daytime indoor light. Fresh plant seedlings, leaves or branches were collected from the field or vegetable markets near the laboratories. Each chamber, which contained a volume of 600–650 ml, was closed with a rubber stopper. About 30–50 g (fresh weight) of plant matter was incubated within the enclosure. Various enclosure durations (ranging from 10 min to 6 h) over the span of previous studies (e.g., Zhang 2001; Zhang et al. 2002a, b; Huang et al. 1992; Chen et al. 1992, 1995, 2002, 2003; Li and Chen 1993) were adopted. Incubation and sampling were conducted at daytime under room temperature. Before a chamber was filled with plants, it was flushed with ambient air. Immediately after filling the chambers, an air sample was taken with a plastic syringe (60 ml in volume), and then the chamber was closed. After incubation for a given duration, another air sample was taken before the chamber was opened. The incubated plants were then immediately dried in an oven in order to measure the dry weight. At least five replicates were conducted for each treatment of a given plant species. Sampling time and air temperature were recorded for each sample. After the air samples were collected, they were stored within the syringes and analyzed within a few hours, using different GC–ECD methods. For most plant species, the comparisons between DN and DN-Ascarite were conducted at the field site located in the Yangtze delta, where two Agilent 4890D instruments were installed. For these two GC–ECDs, DN or DN-Ascarite had been exclusively used prior to this study. In the comparison trials, each air sample was repeatedly analyzed using DN and DN-Ascarite using either of the two instruments. For the Agilent 4890D or other GC–ECD types installed in the laboratories other than at this site, DN, DN-Ascarite and AM were compared by measuring the N2O emission rates of detached fresh plants in static enclosures (the DN and DN-Ascarite trials were not conducted with the Shimadzu GC-14B1). When shifting from one method to another, the carrier flow rates were slightly adjusted to allow for the best N2O separation. For each air sample, the concentrations of both N2O and CO2 were measured. The CO2 concentration was analyzed with a flame ionization detector (FID) instead of an ECD. The GC configurations for CO2 analysis are given in Table 1. Pure N2, at a flow rate of about 25 cm3 min−1, was used as the carrier for CO2 detection with a GC–FID coupled with a nickel catalytic converter (van Bochove et al. 2000; Wang and Wang 2003). The CO2 retention time was about 1.4 min. Further details regarding the GC configurations for CO2 analysis with a GC–FID can be found in Wang and Wang (2003). Nitrous oxide emission rates from plants (F, in μg N h−1 kg−1 dry weight) were determined using the following equation.
$$F = k \cdot {P \mathord{\left/{\vphantom {P {1013}}} \right.\kern-\nulldelimiterspace} {1013}} \cdot {{273} \mathord{\left/{\vphantom {{273} {\left( {273 + T} \right) \cdot {\text{ $ \rho $ }} \cdot {{\Delta C} \mathord{\left/{\vphantom {{\Delta C} {\Delta t}}} \right.\kern-\nulldelimiterspace} {\Delta t}}}}} \right.\kern-\nulldelimiterspace} {\left( {273 + T} \right) \cdot {\text{ $ \rho $ }} \cdot {{\Delta C} \mathord{\left/{\vphantom {{\Delta C} {\Delta t}}} \right.\kern-\nulldelimiterspace} {\Delta t}}}} \cdot {V \mathord{\left/{\vphantom {V M}} \right.\kern-\nulldelimiterspace} M}$$
(1)

In Eq. 1, ΔC denotes the difference between the initial and final N2O concentrations (10−9 mol mol−1) during an enclosure duration (Δt, in h); V is the headspace volume (l); W is the dry plant weight (kg); ρ denotes the density of N2O at 273 K and 1,013 hPa, which is 1.25 g N l−1; T is the mean air temperature during incubation (°C); P is the air pressure during incubation (hPa), which is taken to be approximately 1,013 hPa because the altitudes of our laboratories (3–34 m) are very close to average sea level; and k is a coefficient (0.001) for dimensional conversion. Not knowing the variations in air pressure due to altitude differences among the laboratory locations and the seasonal changes of the atmosphere may bias the quantified N2O emission rates by 2%, at most.

Field measurements

The DN and DN-Ascarite methods were compared by measuring N2O fluxes from two soil–plant systems from September 20th, 2005 to December 17th, 2006. One selected soil–plant system was rotationally cultivated with paddy rice and winter wheat. The other was alternatively cultivated with vegetables. Both systems are typical in the Yangtze delta. The rice–wheat cropping system accounts for approximately 15–25% of the cropland area, ∼6.8 × 106 ha, in the Yangtze delta. The remaining paddies in this region are cultivated with rice–rapeseed or rice–winter fallow rotations. Vegetables are cultivated on only a small portion of the croplands in this region.

The field site (32°35′N, 119°42′E) was located at a very flat cultivated area (with an altitude of 3–4 m above the average sea level). The site is exposed to a humid northern subtropical monsoon climate with an annual mean air temperature of 15°C, an annual total precipitation of 979 mm, and a frost-free period of 220 days. The mean air temperature of the hottest month (July) is 28.0°C, and the coldest month (January) averages 2.7°C. There is a sandy loam soil in both cropping systems, with a clay (<0.002 mm) fraction of 14%, a sand (2–0.02 mm) fraction of 58%, a bulk density of 1.16 g cm−3, a pH (1:5 H2O) value of 8.0–8.2, an organic carbon content of 15.7–18.4 g C kg−1, and a total nitrogen content of 1.46–1.58 g N kg−1. Our field experiments were performed on uniform vegetable fields and uniform rice–wheat rotation fields. The fields for vegetables had been cultivated with rice–wheat/rapeseed rotations until they were changed to their present usage 25 years ago. The vegetable fields are about 200 m away from the rice–wheat fields selected for this study. Each cropping system was treated with nitrogen fertilizers following the local conventional regimes while a control was included with no nitrogen fertilizer. For the nitrogen-fertilized treatments, fertilizers were added to the rice–wheat (basal: compound fertilizer (N : P2O5 : K2O = 15% : 15% : 15%); top-dressing: urea) and vegetable fields (basal: farmyard organic manure plus urea; top-dressing: urea) at rates of 300 and 1,400 kg N ha−1 yr−1, respectively. Additional field management of both cropping systems followed local conventional practices. Four randomly replicated field plots were selected for the control treatment in the rice–wheat cropping system, while three were chosen for each remaining treatment in both cropping systems.

Mini-plots with a dimension of 0.5 × 0.5 m were made by permanently installing a stainless steel base frame in the center of each replicate plot. Measurements of N2O fluxes from each mini-plot were carried out once every 2–4 days. On an observation day, a single measurement for each plot was conducted during 9:00–11:00 am. To measure an N2O flux, an opaque static chamber adapted to plant height (with a dimension of length × width × height = 0.5 × 0.5 × 0.5 m or 1.0 m) was mounted onto the base frame collar. A gas-tight enclosure was assured by sealing it with water. Using 60-ml plastic syringes, five air samples were taken from the chamber headspace during a 30-min enclosure period at intervals of approximately 6 min. After sample collection, the chamber was immediately removed to minimize physical/physiological influences on the plants within the mini-plot. Within a few hours, each sample was repeatedly analyzed with the Agilent 4890D near the experimental fields using the DN and DN-Ascarite methods. The N2O concentrations in the air samples were determined with a gas cross-calibrated with the standards from Air Liquide (Munich, Germany) and the National Center for Standard Matters (Beijing, China). Using the equation described by Zheng et al. (2000), N2O fluxes were determined according to the measured concentration increase rates within the enclosures, the chamber headspace height, the air pressure (obtained from the automatic climate station near the experimental fields) and the air temperature within the enclosures (recorded when sampling). Further details of the chamber structure, operation procedures and description of the GC instruments are found in relevant publications (Wang and Wang 2003; Zheng et al. 2006; Xu et al. 2004).

In general, N2O release from regional croplands can be broken down into direct and background emissions (Gu et al. 2007). Direct release can be estimated by multiplying the amount of added nitrogen fertilizer with a direct emission factor (EFd), which in turn is defined as the ratio of nitrogen fertilizer losses through N2O emission to the applied nitrogen amount (Zheng et al. 2004; IPCC 2006). The background release can be estimated by multiplying the cropland area with an annual rate of background N2O emission (BNar), which in turn is defined as the total amount of N2O emission from a unit area of cropland receiving no nitrogen fertilizer in the current year (Gu et al. 2007). The direct emission factor is a key parameter for compiling an N2O emission inventory following the IPCC methodologies (IPCC 2006) or quantitatively evaluating the mitigation effects of a management measure or a practical technique on N2O emissions from a cropland. To quantify EFd and BNar, total annual N2O emissions (on an area basis) under fertilized and unfertilized conditions need to be estimated with field measurements. To estimate the annual total N2O emissions under either fertilization condition, each single observation was regarded as the measure of the daily flux and was thus directly scaled up to a 24-h period, ignoring diurnal variations. Meanwhile, the observation gaps for those days without a measurement were filled with the mean value of the observations in the two closest days, ignoring day-to-day variations. Ignoring diurnal variation might have yielded an uncertainty of 17–21% in the quantification of annual emissions (Zheng et al. 2004). The gap-filling by interpolation, ignoring of day-to-day variation, might have led to an uncertainty of 8–22% (Zheng et al. 2004). However, there is no way to avoid these uncertainties due to low-frequency intermittent measurements. These uncertainties would likely mask the differences in annual total N2O emission amounts between the DN and DN-Ascarite measurements or between the nitrogen-fertilized and unfertilized treatments, unless the differences were significantly larger than the uncertainty magnitudes. In addition to the above uncertainties due to unknown temporal variation, the uncertainties for the annual emission estimates of individual fertilization treatments due to spatial variation were given as the standard errors of field plot replicates. The standard errors for both fertilization treatments were directly propagated to determine the uncertainty of the EFd estimate.

Statistical analysis

Significance levels among different field treatments were determined using one-way ANOVA tests, while those for regression curves were determined using the F-test. The significance levels between any two GC–ECD methods were determined using paired t-tests. The software package SYSTAT 5.05 for Windows from SPSS Inc. (Chicago, USA) was applied to perform the statistical analysis.

Results

ECD signals of N2O and CO2 when using DN, DN-Ascarite and AM

Figure 1 shows the chromatographic responses of the N2O and CO2 signals for a dry ambient air to DN, DN-Ascarite and AM. The Shimadzu GC-14B2 and Agilent 6820 (Table 1) were used to produce these results. Similar responses were also found (data not shown) when analyzing the same ambient air with the Agilent 4890D and HP5890II instruments (Table 1). As compared to DN, the other two methods produced much stronger ECD signals for N2O. Meanwhile, at the normal oven and detector temperatures for N2O analysis, DN yielded a CO2 signal with a positive peak appearing in the Shimadzu GC (Fig. 1a) but a negative peak in the Agilent instrument (Fig. 1d). No CO2 peak emerged when DN-Ascarite was adopted (Fig. 1b and e) because the CO2 in the air sample did not enter the separation column or the detector. When the AM method was used, no CO2 signal appeared in the Agilent 6820, whereas a reduced but obvious CO2 peak emerged in the Shimadzu GC-14B2.

When analyzing ambient air, different N2O peak sizes appeared across the three methods (Fig. 1). DN usually yielded the smallest peak areas (Fig. 2a,c,e and d). With respect to the three methods used in the same GC instrument, Fig. 2 illustrates that there was a trend toward better detection precision when the N2O peak area was larger. Using DN to analyze dry ambient air, the GC instruments listed in Table 1 (excluding the Shimadzu GC-14B1) yielded precisions of 2.15% ± 0.51% (n ≥ 10, standard deviation), while both DN-Ascarite and AM had much better (p < 0.01) precision, namely 0.63% ± 0.34% and 0.57% ± 0.15%, respectively. No significant difference in precision was detected between DN-Ascarite and AM. This indicates that DN-Ascarite may be used as a substitute for AM to analyze N2O in dry air samples.
https://static-content.springer.com/image/art%3A10.1007%2Fs11104-008-9673-6/MediaObjects/11104_2008_9673_Fig2_HTML.gif
Fig. 2

Chromatographic peak areas of nitrous oxide (N2O) and the precision of concentration detection in response to different GC–ECD methods. a, b Shimadzu GC-14B2. c, d Agilent 6820. e, f HP5890II. g, h Agilent 4890D. Detailed configurations of the instruments are described in Table 1. The N2O peak areas and precisions were determined with dry ambient air as described in the footnotes of Fig. 1. Definitions of DN, DN-Ascarite and AM are given in the “Abbreviations” section of the text

ECD signals of N2O corresponding to varying CO2 concentrations

Using the three methods with each of the instruments listed in Table 1 (excluding the Shimadzu 14-B1), we analyzed a series of dry air samples containing constant N2O but variable CO2. Figure 3 shows the results from the Shimadzu 14-B2 and the Agilent 4890D. Similar results (data not shown) were obtained from all the GC–ECDs (excluding the Shimadzu 14-B1). Obviously, the detected N2O concentrations of the air series remained almost constant when they were analyzed by DN-Ascarite or AM, as expected. However, the results in Fig. 3 indicate that changes in the CO2 concentrations of the air samples significantly increased the N2O concentrations detected by DN. If the CO2 concentration span was wide enough (like that shown in Fig. 4d), the correlation of the N2O concentrations detected by DN with the CO2 concentrations was well fit to a Logistic function. This means that a positive and linear correlation appeared when CO2 concentrations were relatively low, e.g., less than 10,000 mol mol−1 (Fig. 3), but when CO2 concentrations increased further, a positive nonlinear correlation occurred until the increase rate in detected N2O concentrations against CO2 concentrations finally was leveled out (Fig. 4d). The magnitude of the CO2 effect was comparable among the different GC–ECD types. This is shown by the comparable slopes for the regressions of N2O concentrations against CO2 concentrations, which were 0.034 and 0.030 for the Shimadzu GC-14B and the Agilent 4890D, respectively (Fig. 3a and b). However, we found that the CO2 effect on N2O signals yielded by DN occurred intensively only for an ECD that had been exclusively used with DN or DN-Ascarite. For any GC–ECD type, once the methods of DN-CO2 (using 10% CO2 in pure N2 as a make-up gas for DN) or AM (using 5–10% CH4 in pure Ar as the carrier) had been continuously used for a period longer than two weeks, the N2O signal amplification caused by increases in CO2 concentrations was significantly reduced, in some cases by as much as an order of magnitude. This is illustrated by the different slopes of the regressions shown in Fig. 3b and d (0.030 vs. 0.0037 for the same Agilent 4890D instrument). Although the CO2 effect was significantly reduced, the precision of N2O detection was still worse (i.e., >1%) when the carrier was changed from AM or DN-CO2 back to DN. The mechanisms behind this interesting phenomenon remain unknown.
https://static-content.springer.com/image/art%3A10.1007%2Fs11104-008-9673-6/MediaObjects/11104_2008_9673_Fig3_HTML.gif
Fig. 3

Signals for constant N2O concentrations measured by different GC–ECD methods in response to varying CO2 concentrations. a Shimadzu GC-14B2, using DN: [N2O] = 0.034 [CO2] + 333.5 (R2 = 0.94, p < 0.001). b Agilent 4890D, for DN: [N2O] = 0.030 [CO2] + 312.1 (R2 = 0.92, p < 0.001). c Agilent 4890D, for DN: [N2O] = 0.0037 [CO2] + 328.5 (R2 = 0.79, p < 0.01). The various CO2 concentrations were created by adding small amounts of pure CO2 to the dry ambient air, as described in the footnotes of Fig. 1, and analyzed with the flame ionization detector of the Agilent 4890D. Detailed descriptions of the instruments are found in Table 1. The GC–ECDs for (a) and (b) had exclusively used DN until the trails in this study. Before the trails in this study, the instrument for (c) had been using AM or DN-CO2 for more than 2 weeks. Definitions of DN, DN-Ascarite, AM and DN-CO2 are given in the “Abbreviations” section of the text

https://static-content.springer.com/image/art%3A10.1007%2Fs11104-008-9673-6/MediaObjects/11104_2008_9673_Fig4_HTML.gif
Fig. 4

N2O and CO2 concentrations of air samples from detached fresh plants within static enclosures and plant N2O emission rates quantified with different gas chromatography methods. a, e Shimadzu GC-14B2, using DN in (a): [N2O] = 0.0344 [CO2] + 315.2 (R2 = 0.98, p < 0.001). b, f Agilent 6820, for DN in (b): [N2O] = 0.0144 [CO2] + 339.2 (R2 = 0.46, p < 0.01). c, d, g Agilent 4890D, for DN in (c): [N2O] = 0.040 [CO2] + 327.7 (R2 = 0.79, p < 0.01), and for DN in (d): [N2O] = 1,800 / [1 + 3.548] exp [−6.86 × 10−5 [CO2]) (R2 = 0.996, p < 0.001). Detailed descriptions of the instruments are found in Table 1. Definitions of DN, DN-Ascarite and AM are given in the “Abbreviations” section of the text. The same capital letters in individual panels of (e), (f) and (g) indicate no significant difference at p < 0.05, whereas different letters in the same panel indicate a significant difference at p < 0.01. The data in (a), (c), (e) and (g) were obtained from measuring Amaranthus mangostanus L. seedlings, and those in (b) and (f) were from measuring Brassica chinensis L. seedlings. The data in (d) integrate the results from multiple plant species: Morus alba L., Oryza sativa L., Populus simonii Carr., Juniperus formosana Hayata and Raphanus sativus L

N2O emission rates of detached fresh plants measured by DN, DN-Ascarite and AM

Using DN, DN-Ascarite and AM for each GC–ECD instrument (Table 1), we measured the N2O concentrations of air samples collected from enclosures containing detached fresh plants. The CO2 concentrations of the same samples were simultaneously measured with the FIDs of the Agilent GC instruments (Table 1). As Fig. 4a–d illustrate, a positive correlation between N2O and CO2 concentrations was significant when using DN (p < 0.01). However, when either DN-Ascarite or AM was used for N2O analysis, no significant correlation between N2O and CO2 concentrations was detected. These results are similar to those for the dry air samples with a constant N2O concentration but varying CO2 concentrations. Consequently, significant differences in the N2O emission rates of detached fresh plants occurred between DN and DN-Ascarite, as well as between DN and AM (p < 0.01), while no obvious differences were observed between DN-Ascarite and AM for all GC–ECDs (Fig. 4e–g). This suggests that the DN method may positively bias N2O emissions from detached fresh plants as compared to the DN-Ascarite or AM methods if CO2 accumulates within the chamber enclosures. For example, in Table 2, the N2O emission rates of various plant species determined using DN were approximately two orders of magnitude higher than those resulting from DN-Ascarite (24.43 vs. 0.07 μg N h−1 kg−1 dry weight of plants). The exaggerating effects of CO2 accumulation on the N2O emission rates measured using DN generally occurred for all the plant species listed in Table 2. This is exemplified in Fig. 5, wherein the differences in the N2O emission rates from Brassica chinensis L. seedlings between either DN and DN-Ascarite or DN and AM were shown to significantly increase with CO2 accumulation rates. Thus, the N2O emission rate measured using DN is also referred to as the apparent N2O emission rate. A correction to the N2O concentrations that determined the apparent emission rates, based on reference to the simultaneously measured CO2 concentrations, could yield results comparable with those measured using DN-Ascarite (Table 2). The impact of CO2 accumulation on N2O emission rates measured using DN might be attributable to the differences in CO2 concentrations between the calibration gas and the air samples.
https://static-content.springer.com/image/art%3A10.1007%2Fs11104-008-9673-6/MediaObjects/11104_2008_9673_Fig5_HTML.gif
Fig. 5

Effects of CO2 accumulation within the chamber enclosures on differences in measured plant N2O emission rates between DN and DN-Ascarite or DN and AM. An example is shown with 20 replicated observations for Brassica chinensis L. seedlings. The measurements were conducted with an Agilent 6820 instrument (see Table 1). Definitions of DN, DN-Ascarite and AM are given in the “Abbreviations” section of the text. Filled diamonds and the coarse line, respectively, show the differences between DN and AM and the regression curve of the differences against CO2 accumulation rates, Y = 68.06X − 17.5 (R2 = 0.60, p < 0.001). Open diamonds and the fine line, respectively, show the differences between DN and DN-Ascarite and the regression curve, Y = 36.36X − 2.9 (R2 = 0.34, p < 0.01). In these regression functions, Y denotes differences in N2O emission rates, and X denotes CO2 accumulation rates

Table 2

Nitrous oxide (N2O) emission rates from detached fresh plants measured with DN and DN-Ascarite by collecting air samples from static chamber enclosures

Plant species

N

N2O emission rates (μg N h−1 kg−1 dry weight)a

Significance (p value)*

DN (apparent)

DN (corrected)

DN-Ascarite

Paddy rice (Oryza sativa L.) shoots

15

13.70 ± 2.04

0.45 ± 0.13

0.30 ± 0.11

<0.001

Radish (Raphanus sativus L.) plants

40

31.58 ± 3.50

0.03 ± 0.14

−0.11 ± 0.10

<0.001

Oil rape (Brassica chinensis var. oleifera) seedlings

20

20.05 ± 2.85

−0.62 ± 0.19

−0.46 ± 0.12

<0.001

Wheat (Triticum aestivum L.) seedlings

10

33.12 ± 3.17

−0.52 ± 0.73

−0.67 ± 0.31

<0.001

Cabbage (Brassica chinensis L.) seedlings

5

74.12 ± 2.69

1.81 ± 0.59

0.77 ± 0.15

<0.001

Mulberry (Morus alba L.) leaves

5

24.57 ± 0.39

0.13 ± 0.07

0.20 ± 0.04

<0.001

Cypress (Juniperus formosana Hayata) branches

5

7.2 ± 0.2

NA

−0.1 ± 0.1

<0.001

Lablab (Dolichos lablab L.) leaves

5

16.82 ± 0.33

−0.12 ± 0.22

−0.19 ± 0.15

<0.001

Poplar (Populus simonii Carr.) leaves

5

26.43 ± 0.52

0.06 ± 0.06

0.11 ± 0.06

<0.001

Eggplant (Solanum melongena L.) leaves

5

9.82 ± 0.26

0.26 ± 0.22

0.11 ± 0.10

<0.001

Amaranth (Amaranthus mangostanus L.) seedlings

12

6.47 ± 1.44

−0.11 ± 0.04

−0.15 ± 0.30

<0.01

Chinese cabbage (Brassica pekinensis (Loureiro) Rupr.) seedlings

20

29.31 ± 1.83

1.33 ± 0.99

1.04 ± 0.97

<0.001

Mean

 

24.43

0.25

0.07

<0.001

Standard deviation

 

18.21

0.73

0.48

 

N number of replicates or observations; NA not available. The columns “DN (apparent)”, “DN (corrected)” and “DN-Ascarite” present the N2O emission rates directly measured with DN, corrected from DN measurements by reference to simultaneously measured CO2 concentrations and with DN-Ascarite measurements, respectively. Definitions of DN and DN-Ascarite are given in the “Abbreviations” section of the text

*Significance level for the difference between DN (apparent) and DN-Ascarite, which was determined with one-way ANOVA and paired t-tests.

aMean ± standard error; for N > 5, the results of enclosure durations ranging from 0.5 to 3 h were averaged; otherwise, the emission rates resulted from an enclosure duration of ca. 3.4 h (with a range of 2.7 to 4.1 h)

Differences in N2O fluxes from soil–plant systems between DN-Ascarite and DN

In the field campaign conducted with vegetable and rice–wheat cropping systems during the period from September 20th 2005 to December 17th 2006, 2,363 pairs of N2O fluxes (Figs. 6 and 7) were measured by DN and DN-Ascarite. The data are summarized in Tables 3 and 4. Those fluxes/emissions measured using DN were also termed in this study as apparent N2O fluxes/emissions.
https://static-content.springer.com/image/art%3A10.1007%2Fs11104-008-9673-6/MediaObjects/11104_2008_9673_Fig6_HTML.gif
Fig. 6

Nitrous oxide (N2O) fluxes from vegetable fields measured using DN and DN-Ascarite in a GC–ECD (Agilent 4890D) by collecting air samples with opaque static chambers. Detailed descriptions of the instruments are found in Table 1. Definitions of DN and DN-Ascarite are given in the “Abbreviations” section of the text

https://static-content.springer.com/image/art%3A10.1007%2Fs11104-008-9673-6/MediaObjects/11104_2008_9673_Fig7_HTML.gif
Fig. 7

Nitrous oxide (N2O) emission fluxes from rice–wheat rotation fields measured using DN and DN-Ascarite in a GC–ECD (Agilent 4890D) by collecting air samples with opaque static chambers. Detailed descriptions of the instruments are found in Table 1. Definitions of DN and DN-Ascarite are given in the “Abbreviations” section of the text

Table 3

Nitrous oxide (N2O) flux, total annual emissions and annual direct emission factors for rice–wheat and vegetable fields: comparisons between DN and DN-Ascarite

 

Unfertilized

Fertilized

DN-Ascarite

DN (apparent)

DN (corrected)

DN-Ascarite

DN (apparent)

DN (corrected)

Rice–wheat fields (fertilized at 300 kg N ha−1 yr−1)

Emission (μg N m−2 h−1)

11 (−74 ∼ 114)a

19 (−149 ∼ 232)b

15 (−77 ∼ 232)c

21 (−84 ∼ 915)a

34 (−172 ∼ 802)b

23 (−100 ∼ 802)a

[837]

[837]

[837]

[637]

[637]

[637]

Annual (kg N ha−1)

1.0 ± 0.2a

1.6 ± 0.2a

1.3 ± 0.1a

1.8 ± 0.4a

2.9 ± 0.3b

2.0 ± 0.3a

[4]

[4]

[4]

[3]

[3]

[3]

Annual direct emission factor (%)

   

0.28 ± 0.22

0.46 ± 0.16

0.26 ± 0.16

Vegetable fields (fertilized at 1,400 kg N ha−1 yr−1)

Emission (μg N m−2 h−1)

15 (−49 ∼ 190)a [419]

27 (−133 ∼ 255)b [419]

20 (−61 ∼ 255)c [419]

976 (−43 ∼ 45,438)a [468]

1,039 (−181 ∼ 52,599)b [468]

1,016 (−109 ∼ 52,599)a [468]

Annual (kg N ha−1)

0.8 ± 0.5a

1.4 ± 0.7a

1.1 ± 0.3a

57.1 ± 1.9a

60.9 ± 1.9b

59.7 ± 1.9a

[3]

[3]

[3]

[3]

[3]

[3]

Annual direct emission factor (%)

   

4.02 ± 0.17

4.25 ± 0.19

4.18 ± 0.16

Definitions of DN and DN-Ascarite are given in the “Abbreviations” section of the text. The mean and range are given for all flux observations. Data for annual total emissions or direct emission factors are means (±standard errors) of plot replicates. The datum in each pair of brackets “[]” indicates the number of flux observations or field plot replicates. The same superscript letters indicate no significant difference, while the different superscript letters indicate significant differences among DN-Ascarite, DN (apparent) and DN (corrected) at the level of p < 0.05 (paired t-tests and one-way ANOVA). Data in the “DN (apparent)” column resulted from the DN-measured fluxes. Data given in the “DN (corrected)” column were obtained from corrected DN measurements. In the correction, each apparent flux (measured by DN) was classified into one of the flux class ranges, and then the corresponding value from the “Y–X” column in Table 4 was added to the apparent flux

Table 4

Comparison of N2O emissions measured with DN and DN-Ascarite for rice–wheat and vegetable fields

Ranges of N2O fluxes measured by DN (μg N m−2 h−1)

N2O fluxes (μg N m−2 h−1)

Pair number

Significance level

DN (apparent)

DN-Ascarite

Y − X

Y / X

Mean (X)

SD

Mean (Y)

SD

A

<−30

−63

34

8

22

72

301

***

B

−30 ∼ 0

−14

7

8

15

22

254

***

C

0 ∼ 30

17

25

12

25

−5

0.72

551

***

D

30 ∼ 100

57

19

19

27

−38

0.34

882

***

E

100 ∼ 200

136

27

72

65

−64

0.53

212

***

F

>200

2,872

6,999

2,713

6,324

−159

0.94

163

ns

Data in the “DN (apparent)” and “DN-Ascarite” columns summarize the fluxes measured using DN and DN-Ascarite, respectively. The symbol “***” indicates a significance level of p < 0.001, while “ns” indicates not significant at p < 0.05 (paired t-tests and one-way ANOVA). SD denotes standard deviation within the range class. Definitions of DN and DN-Ascarite are given in the “Abbreviations” section of the text

The N2O fluxes measured by DN-Ascarite during the entire observational period ranged from −74 to 114 (mean: 11) and −84 to 915 (mean: 21) μg N m−2 h−1 in the unfertilized and fertilized rice–wheat fields, respectively, and from −49 to 190 (mean: 15) and −43 to 45438 (mean: 976) μg N m−2 h−1 in the unfertilized and fertilized vegetable fields, respectively (Table 3). The N2O emissions under fertilized conditions were significantly more intensive (p < 0.01) than those under unfertilized conditions. The differences were by a factor of ∼2 for the rice–wheat cropping system and ∼65 for the vegetable fields. Meanwhile, the N2O fluxes from the vegetable cropping system were significantly higher than those from the rice–wheat system by a factor of ∼1.4 under unfertilized conditions (p < 0.05) and ∼47 under fertilized conditions (p < 0.001). The difference under unfertilized conditions was most likely due to the prior cropping system change from a rice-based regime to long-term vegetable cultivation. The much higher difference between the vegetable and rice–wheat fields under fertilized conditions is most likely attributable to (a) the more intensive nitrogen fertilization for the vegetable fields (1,400 vs. 300 kg N ha−1 yr−1) and (b) the effect of the cropping system change. The latter explanation may be supported by the much higher differences in N2O emissions than those seen in the nitrogen application rates (by a factor of ∼47 vs. ∼2.7). These results may indicate that the long-term change of a rice-based cropping system to vegetable cultivation significantly stimulates N2O emission.

The N2O fluxes measured by DN during the entire observational period ranged from −149 to 232 (mean: 19) and −172 to 802 (mean: 34) μg N m−2 h−1 in the unfertilized and fertilized rice–wheat fields, respectively, and from −133 to 255 (mean: 27) and −181 to 52,599 (mean: 1,039) μg N m−2 h−1 in the unfertilized and fertilized vegetable fields, respectively (Table 3). On average, these values were ∼50% higher (p < 0.01) than those measured by DN-Ascarite. The largest difference (70%, p < 0.05) appeared in the unfertilized rice–wheat field (mean N2O fluxes: 19 vs. 11 μg N m−2 h−1), while the smallest difference (10%, p < 0.05) appeared for the fertilized vegetable field (mean fluxes: 1039 vs. 976 μg N m−2 h−1). These results indicate an overestimation of N2O fluxes due to the use of DN in combination with sample collection using static opaque chambers. Thus, an N2O flux measured by this approach is referred to as an apparent flux. For intensive N2O emissions (for instance the fertilized vegetable case), however, apparent fluxes only slightly biased the actual fluxes.

Correction terms for the DN-determined fluxes

Among the 2,363 N2O fluxes measured by DN, only 163 fluxes (indicated by “F” in Fig. 8) showed values greater than 200 μg N m−2 h−1. Within this range, the fluxes measured by DN-Ascarite and those by DN revealed a positive linear regression (Fig. 8b) with a slope of 0.91 when the intercept was zero (p < 0.0001). This slope indicates that the fluxes within this range measured by DN were higher by ∼9% on average when compared to DN-Ascarite, but this difference was not significant. Accordingly, DN may be only directly applicable for very intensive sources from which N2O fluxes are generally greater than 200 μg N m−2 h−1 (this criterion may be specific to the chamber designs, sampling procedures and ecosystem types of this study). The majority fluxes measured using DN (93%) were lower than 200 μg N m−2 h−1, and were significantly higher than those measured using DN-Ascarite (Table 4). However, the magnitude of the differences between the two methods varied with the levels of DN-measured N2O fluxes. Accordingly, we defined six flux classes: A, B, C, D, E and F. The DN fluxes in each class were within a specific range, given in Table 4 (first two columns), and are illustrated in Fig. 8. When a DN-measured flux was negative, the actual flux might have been severely underestimated (namely more negative than the true flux), while it might have been well-represented by the flux determined using DN-Ascarite. In contrast, when a DN-measured flux was positive but less than 200 μg N m−2 h−1, it might have significantly overestimated the actual emission. However, when a DN-measured flux was closer to zero, the under- or over-estimation was smaller. This is illustrated in Fig. 8a by the deviations of the DN-measured fluxes away from the diagonal. This is also shown by the values of the differences given in the “Y − X” column of Table 4 for the specific ranges of DN-measured fluxes (i.e., 72, 22, −5, −38 and −64 μg N m−2 h−1 for <−30, −30–0, 0–30, 30–100 and 100–200 μg N m−2 h−1, respectively). The values of the differences could be regarded as additive items for the correction of DN-measured fluxes by referring to the N2O fluxes observed using DN-Ascarite. These additive correction items are specific to the chamber dimensions, sampling procedures and ecosystem types of this study.
https://static-content.springer.com/image/art%3A10.1007%2Fs11104-008-9673-6/MediaObjects/11104_2008_9673_Fig8_HTML.gif
Fig. 8

Correlation between nitrous oxide (N2O) fluxes measured by DN and DN-Ascarite. An Agilent 4890D (see Table 1) was used for all measurements. a The DN-measured N2O fluxes ranged from −0.30 to 0.30 mg N m−2 h−1. b The correlation between the fluxes measured by DN-Ascarite and those by DN. The zero-intercept regression in (b) was obtained from 163 pairs of data for which each DN-measured N2O flux was greater than 0.2 mg N m−2 h−1 (including the 20 fluxes with each greater than 4 mg N m−2 h−1 but not shown). The capital letters indicate the ranges of the DN-measured N2O fluxes referred to in Table 4. Definitions of DN and DN-Ascarite are given in the “Abbreviations” section of the text

Differences in annual total N2O emissions as determined using DN-Ascarite and DN

Table 3 presented the annual estimates of N2O emissions on a hectare basis for individual field treatments and the standard error of each estimate due to spatial variations. The uncertainties due to unknown temporal (diurnal and day-to-day) variations were not included in the given errors. The data based on the observations using DN-Ascarite show that application of nitrogen fertilizer to the vegetable fields at a rate of ∼1,400 kg N ha−1 yr−1 significantly increased the annual total amount of N2O emissions by about 73 times (p < 0.001) compared to unfertilized conditions. Nitrogen fertilization at a rate of 300 kg N ha−1 yr−1 also tended to increase annual total amount of N2O emission from the rice–wheat fields by about 80%, but this difference was not statistically significant. The failure to detect a significant fertilization effect in the rice–wheat cropping system was most likely due to two factors. One was the small sample size (four replicates for the unfertilized treatment and three for the fertilized treatment) relative to the large spatial variation (with a mean coefficient of variation (CV) of 41–45%) among the replicate plots located within an area of ca. 3 ha. Another factor was that the actual difference might have been masked to a considerable extent by the uncertainties due to ignorance of the extent of diurnal and day-to-day variations. DN measurements yielded higher total annual N2O emissions compared to DN-Ascarite measurements. However, a significant difference (7% for the vegetable plots and 62% for the rice–wheat fields, p < 0.05) appeared only under fertilized conditions. The difference for background emissions (under unfertilized conditions) tended to be higher (59–83%), but was not statistically significant. The failure to detect a significant difference under unfertilized conditions was most likely due to the large spatial variation (with a CV of 28–66%). The large variation might have been attributed at least partially to the small field plot replicates (poorly covering spatial variability) and the low frequency of intermittent measurements (which poorly sampled diurnal and day-to-day temporal variability).

Differences in annual direct N2O emission factors between DN and DN-Ascarite

Based on the N2O fluxes measured by DN-Ascarite, the EFds of the rice–wheat and vegetable fields were determined to be 0.28% and 4.02%, respectively (Table 3). In comparison, the EFds yielded by DN measurements were higher (p < 0.05) by ∼65% for the rice-wheat fields and ∼6% for the vegetable fields (Table 3). This suggests that DN measurements would significantly overestimate the EFds, but overestimation would become smaller as the EFd magnitude increased.

Discussion

Influences of CO2 concentration on the ECD signals of N2O

As the above results demonstrate, the N2O concentration of an air sample may be biased by using DN in a GC–ECD when the CO2 component is allowed to enter the detector. N2O concentration modifications are thought to be attributable to differences in the CO2 concentrations between the air sample and the calibration gas. In practice, however, it is not possible to ensure equal CO2 concentrations between the calibration gas and air samples. As a result, biases in quantification of N2O concentrations of air samples using DN in a GC–ECD may be inevitable. These factors may further result in an inevitable bias in the quantification of an N2O flux if CO2 concentrations in the air samples for determining this flux are variable. The significant CO2 effects upon the N2O signal yielded by DN may be principally explained by comparing the reactions in the detector under conditions using DN, DN-Ascarite, AM, DN-AM and DN-CO2 in a GC–ECD that is equipped with a back-flushing and a separation column-detector switch.

When using DN-Ascarite, CO2 is not allowed to enter either the second separation column or the ECD, and the following reactions (Rxns 14) may occur within the detector.
$${\text{ $ \beta $ }} + {\text{N}}_2 \to {\text{ $ \beta $ }}\prime + {\rm N}_2^ + + {\text{e}}$$
(Rxn 1)
$${\text{ $ \beta $ }} + {\text{N}}_2 \to {\text{ $ \beta $ }}\prime + {\rm N}_2^ * $$
(Rxn 2)
$${\text{N}}_2 {\text{O}} + {\text{e}} \to {\text{N}}_2 + {\text{O}}^ - $$
(Rxn 3)
$$2{\text{O}}^ - + 2{\text{N}}_2^ + \to {\text{O}}_2 + 2{\text{N}}_2 $$
(Rxn 4)

While the carrier N2 flows through the detector, secondary electrons (e), ionic \({\text{N}}_2 \left( {{\text{N}}_2^ + } \right)\) and slow primary electrons \(\left( {{\text{ $ \beta $ }}\prime } \right)\) are produced by the collision between primary electrons (β) from 63Ni and N2 (Ševčík 1977). Then, the voltage/current baseline is formed by the movement of the secondary electrons and \({\text{N}}_2^ + \). When N2O enters the detector, it captures some of the secondary electrons through Rxns 34 and thus reduces the baseline current/voltage. As a result, a negative peak is formed. The area of this peak depends upon N2O abundance in the detector, which in turn is determined by the N2O concentration in the air injected into the separation column. When the instrument records and outputs the N2O signal, the peak direction is reversed. Thus, we see a positive N2O peak appearing on the chromatography. When the operation of a back-flushing and a ECD-separation column switch and the use of an ascarite trap have successfully prevented all other components (such as O2, H2O and CO2) from entering the detector, the reactions to form a N2O signal are relatively clear and simple (like Rxns 14). Thus, a high sensitivity of N2O detection can be obtained by DN-Ascarite (e.g., Fig. 2).

When DN is used in a GC–ECD with a back-flushing and a column-detector switch, CO2 is allowed to enter both separation columns connected in series. It flows out of the second separation column immediately after O2 but before N2O. The O2 component is vented out by the switch before it enters the detector, but CO2 is usually not vented because it is immediately followed by N2O. When CO2 flows through the detector, Rxns 59 may occur in addition to Rxns 14.
$${\text{ $ \beta $ }} + {\text{CO}}_2 \to {\text{ $ \beta $ }}\prime + {\text{C}}{\rm O}_2^ + + {\text{e}}$$
(Rxn 5)
$${\text{N}}_2^ * + {\text{CO}}_2 \to {\text{N}}_2^\prime + {\text{CO}}_2^ + + {\text{e}}$$
(Rxn 6)
$$2{\text{O}}^ - + 2{\text{CO}}_2^ + \to {\text{O}}_2 + 2{\text{CO}}_2 $$
(Rxn 7)
$${\text{CO}}_2 + {\text{e}} \to {\text{CO}}_2^ - $$
(Rxn 8)
$${\text{CO}}_2^ - + {\text{N}}_2^ + \to {\text{CO}}_2 + {\text{N}}_2 $$
(Rxn 9)
The much lower ionization potential of CO2 compared with N2 by 2.0 eV (Table 5) most likely stimulates these reactions. There may be at least two types of reactions involving CO2. One may be that CO2 collides with primary electrons or super-excited long-lived states of dinitrogen \(\left( {{\text{N}}_2^ * } \right)\) and thus produces \({\text{CO}}_2^ + \) and secondary electrons (Rxns 57). These reactions do not form a separate CO2 peak, but they may affect the N2O signal either by increasing the secondary electron abundance on which the N2O signal relies (Rxns 56) or by accelerating electron capture by N2O (Rxns 3 and 7). These effects depend upon CO2 abundance in the detector and, therefore, upon the concentration in the air sample (see Fig. 3a and b). Another reaction type may be that CO2 captures secondary electrons (Rxns 89) and forms a CO2 peak on the chromatography (see Fig. 1a and d). The latter reaction type may not affect the N2O signal if the CO2 and N2O peaks are well-separated on the chromatography (Fig. 1a and d). The CO2 signals, as well as their responses to detector temperature, are different among GC–ECD types (see Fig. 1). The mechanisms behind these CO2 signal responses to ECD temperature still require further clarification.
Table 5

Ionization potentials (IP) of gas species

Species

IP (eV)

Gas

IP (eV)

N2

15.6a

H2S

10.4f

Ar

15.8a

O2

12.1g

He

24.6b

CH4

12.5a, g

CO2

13.6c

NO

9.3h

N2O

12.9c

CS2

10.1i

NH3

10.2d

COS

11.2j

H2O

12.6e

  

aŠevčík 1977

bSanper et al. 1995

cLee et al. 1988

dTanner and Baranova 1999

eStearns and Wentworth 1996

fHelfritch et al. 1993

gPhuoc 2000

hLudowise et al. 1996

iStapelfeldt et al. 1997

jMatsunaga and Watanabe 1967

When AM is used in a GC–ECD with a back-flushing and a column-detector switch, CO2 is also allowed to enter both separation columns and then the detector. The current/voltage baseline is formed by reactions between primary electrons and Ar atoms or CH4 molecules (Rxns 1011). Due to the much lower ionization potential of CH4 compared with Ar by 3.3 eV (Table 5), a considerable fraction of the secondary electrons may be produced by the CH4 ionization (Rxn 11). The N2O signal is formed by Rxns 3 and 1213 for secondary electron capture.
$${\text{ $ \beta $ }} + {\text{Ar}} \to {\text{e}} + {\text{ $ \beta $ }}\prime + {\text{Ar}}^ + $$
(Rxn 10)
$${\text{ $ \beta $ }} + {\text{CH}}_4 \to {\text{e}} + {\text{ $ \beta $ }}\prime + {\text{CH}}_4^ + $$
(Rxn 11)
$$2{\text{O}}^ - + 2{\text{CH}}_4^ + \to {\text{O}}_2 + 2\,{\text{CH}}_4 $$
(Rxn 12)
$$2{\text{O}}^ - + 2{\text{Ar}}^ + \to {\text{O}}_2 + 2{\text{Ar}}$$
(Rxn 13)
$${\text{CO}}_2^ - + {\text{CH}}_4^ + \to {\text{CO}}_2 + {\text{CH}}_4 $$
(Rxn 14)
$${\text{CO}}_2^ - + {\text{Ar}}^ + \to {\text{CO}}_2 + {\text{Ar}}$$
(Rxn 15)

When the abundance of CH4 in the detector is high enough to prevent other substances with ionization potentials lower than that of Ar from reacting with primary electrons, the N2O signal results from Rxns 3 and 1213 can be detected with high precision (Fig. 2b,d,f and h). However, CO2 from the air sample still has the opportunity to produce secondary electrons by colliding with primary electrons if the abundance of CH4 in the detector is not high enough, due to the lower ionization potential of CO2 compared with Ar by 2.2 eV (Table 5). In this case, the N2O signal detected using AM may be slightly biased by the variation of CO2 concentrations. However, an appropriate CH4/Ar ratio is simultaneously determined by several factors such as the primary electron source intensity of 63Ni, the highest CO2 concentrations of samples, detector temperature, and so on. It needs to be determined experimentally under the conditions for sample analysis.

Under an AM condition, CO2 may also capture the secondary electron and then yield a CO2 peak (Rxns 8 and 1415), but this CO2 signal may vary among GC–ECD types (see Fig. 1c and f). The reason for this variation still requires clarification.

When an Ar–CH4 mixture is used as a make-up gas with N2 as a carrier (i.e., DN-AM) for a GC–ECD with a back-flushing and a column-detector switch, the ionization of N2, Ar and CH4 by primary electrons (Rxns 1 and 1011) jointly forms the current/voltage baseline that is present in the detector. Because the ionization potential of CH4 is much lower compared with either Ar or N2 (Table 5), a considerable portion of the secondary electrons may be produced by the CH4 ionization (Rxn 11). The N2O signal is yielded by the electron capture reactions (Rxns 34 and 1213). When the abundance ratio of CH4 to N2 plus Ar is high enough to effectively prevent the ionization process of other substances with ionization potentials that are lower than N2 and Ar, a high sensitivity of N2O analysis can be realized (e.g., Nykänen et al. 1995; Regina et al. 2004). If the CH4 abundance is not high enough, variable CO2 concentrations of air samples may exert influences upon the N2O signal, similar to the situation using the AM method.

Using soda lime (or ascarite) as a CO2 absorbent with the carrier helium (He-Ascarite) in a GC–ECD with a back-flushing and a detector-column switch to detect N2O (Kester et al. 1997), the reactions that form the baseline current/voltage and N2O peak are quite similar with those of DN-Ascarite. The differences between the He- and DN-Ascarite methods only appear in Rxns 1617, wherein the secondary electrons are produced by ionization of He, which has the highest ionization potential (Table 5) among all the rare gases (Sanper et al. 1995). He+ is neutralized in the reactions of secondary electron capture by N2O.
$${\text{ $ \beta $ }} + {\text{He}} \to {\text{ $ \beta $ }}\prime + {\text{He}}^ + + {\text{e}}$$
(Rxn 16)
$$2{\text{O}}^ - + 2{\text{He}}^ + \to {\text{O}}_2 + 2{\text{He}}$$
(Rxn 17)

When using a mixture of 10% CO2 in N2 as a make-up gas for N2 as a carrier (DN-CO2), the current/voltage baseline is formed by Rxns 12 and 56 and the N2O peak by Rxns 34 and 79. Due to the high CO2 abundance yielded by directly introducing the make-up gas into the detector, the influences of variations in the sample concentrations of CO2 or non-CO2 substances (NH3, H2O, H2S, O2, CH4, NO, CS2, COS, etc.) with lower ionization potentials compared with CO2 (Table 5) are effectively inhibited. As a result, a high precision (better than 0.5%) can usually be obtained by automatic sample injection for detection of N2O in ambient air (the authors’ unpublished data). In practice, of course, the CO2 to N2 ratio needs to increase if there are high sample concentrations of CO2 or non-CO2 substances with lower ionization potentials. However, these cases seldom occur because almost all unexpected substances are prevented from entering the detector by the operation of the back-flushing and column-detector switch.

When a back-flushing operation is not adopted in a GC–ECD, substances (like H2O vapor) that are flowing slower than N2O may accumulate in the separation column and may enter the detector simultaneously with N2O. In this case, the analysis operation has to be frequently stopped for separation column recovery.

When a separation column-ECD switch operation is not adopted in a GC–ECD, the substances (like O2 and CH4) that flow faster than N2O can enter the detector. Then they may influence the N2O signal by providing additional secondary electrons due to their lower ionization potentials as compared to the carrier substances (N2, Ar, He) (Table 5). In this case, O2 may also influence N2O detection due to a partial overlap of their peaks if the carrier flow rate is not slow enough.

If the ascarite column is not well-maintained and becomes wet or has an uneven density, the previously trapped/produced H2O vapor may be released or the H2O vapor from the current air sample may pass through the filter. After the H2O vapor flows out of the separation column, it fully enters the ECD immediately after N2O when the separation column-ECD switch is not adopted. Although there is a switch operation, a slight flow rate modification of the carrier gas (which cannot be avoided if the environmental air temperature is not stable enough) may allow H2O vapor to partially enter the detector. In these cases, the H2O vapor may produce additional secondary electrons due to its much lower ionization potential than Ar, N2 or He (Table 5) and influence the N2O signal of the sample(s). Meanwhile, it produces a very wide H2O peak that usually overlaps with the N2O peak of the following sample.

In the above regards, we suggest against the use of DN to simultaneously detect N2O and CO2 with an ECD. We recommend a substitute of the DN-Ascarite/He-Ascarite and DN with DN-CO2, AM or DN-AM that are much more convenient for routine maintenance. In addition, a GC–ECD without settings for a back-flushing and a separation column-detector switch is not recommended for analyzing N2O in air samples.

Nevertheless, the reaction mechanisms in an ECD may be much more complicated than previously thought. Few studies have satisfactorily described the reaction mechanisms or mechanically explained the CO2 effects on the N2O signal. Although the above reactions may logically explain the effects of CO2 on N2O analysis, experimental evidence is still required to confirm some of them. In addition, the mechanisms remain unclear for the phenomenon illustrated in Fig. 3b and c. These figures show a significant reduction of the CO2 effects upon the N2O signal that is also greatly reduced. These reductions occurred when using DN in a GC–ECD that had been continuously used with AM or DN-CO2 for at least two weeks. Further study is still required to elucidate the mechanisms for this reduction.

N2O emission from plants

Since the early report that detached fresh soybean organs may produce N2O (Chen et al., 1990), a number of studies have reported N2O emissions from plants (e.g., Chen et al. 1992, 2002; Hakata et al. 2003; Huang et al. 1992; Li and Chen 1993; Zhang 2001; Zhang et al. 2002a, b) and have indicated a considerable contribution of plant emissions to the N2O release from soil–plant systems (e.g., Chen et al. 2003; Pihlatie et al. 2005; Smart and Bloom 2001; Xu et al. 2001; Zou et al. 2005). In a considerable portion of these studies, the DN method was used in GC–ECDs to analyze air samples collected from chamber enclosures. However, our results (Table 2) suggest that the N2O emissions from plants may be biased if air samples are collected with chamber enclosures wherein CO2 concentrations vary and are analyzed with DN in GC–ECDs. The bias may be caused by the variation in the CO2 concentrations among air samples, due to a positive correlation between CO2 concentrations and DN-detected N2O concentrations (Figs. 3 and 4). The CO2 within static chamber enclosures is generated by the balance between its release by respiration and uptake by photosynthesis. When sampling with opaque static chambers, CO2 accumulation within the enclosure is dramatic due to an absence of photosynthesis. In this case, a positive bias may be evident for N2O emission from plants. This may at least partially explain the positive correlation between in situ observed N2O emission from wheat shoots and CO2 emissions due to dark respiration (Zou et al. 2005). When plants enclosed in translucent chambers are exposed to a less bright light, photosynthesis may uptake less CO2 than is emitted from respiration, and CO2 may thus accumulate. Therefore, the DN method resulted in overestimated N2O emissions from the detached fresh plants that were exposed to a laboratory light within an translucent enclosure (CO2 accumulated at a rate of 4.3 ± 1.8 mg CO2 h−1 g−1); however, the DN-measured (apparent) N2O emission could be satisfactorily corrected (Table 2) with the N2O–CO2 relationship (e.g., Fig. 4d). On the contrary, if plants in an enclosure are exposed to bright light or sunshine (e.g., Chen et al. 1995, 2002), the DN method may underestimate N2O emission due to CO2 depletion within the enclosure. However, further experimental evidence is still required to prove this hypothesized relationship of DN-measured N2O emission versus CO2 at sub-ambient CO2 concentrations. Nevertheless, the evidence from our study indicates that reassessment may be necessary for the published data on plant N2O emissions (e.g., Chen et al. 1995, 2002, 2003; Xu et al. 2001; Zhang 2001; Zhang et al. 2002a, b; Zou et al. 2005) that have resulted from the measurements using DN in combination with sampling in chamber enclosures.

Biases towards N2O fluxes from soil–plant systems yielded by DN measurements

When air samples are collected from a static chamber enclosure and are analyzed using DN in a GC–ECD, biases may occur in the quantification of N2O emissions from soils or soil–plant systems. This was illustrated by our data (Tables 3 and 4, Fig. 8a), which were measured with opaque static chambers. For instance, significant positive biases occurred when the apparent N2O fluxes (i.e., those measured by DN) from soil–plant systems were positive but less than 200 μg N m−2 h−1. The significant, positive correlation (Fig. 9c) between the apparent N2O effluxes and the CO2 emissions at 1 to 858 (mean 146) mg C m−2 h−1 (p < 0.001) may partially explain the overestimation (Tables 3 and 4). When apparent N2O fluxes were greater than 200 μg N m−2 h−1, they were also significantly and positively correlated to the simultaneously measured CO2 emissions (Fig. 9a–b), while they were not obviously different from the fluxes measured by DN-Ascarite (Table 4). This may imply that, when N2O emission occurs intensively, using DN may not bias the real emissions even though CO2 accumulation within the chamber enclosures occurs at high rates. One may argue that the absence of a significant difference might have been due to a reduction in the CO2-trap efficiency of ascarite under high CO2 concentrations. However, the CO2 concentrations in association with the most intensive CO2 emissions ranged from 400 to 1,300 μmol mol−1. At these CO2 concentration levels, no significant decrease of trap efficiency was detected (Fig. 3).
https://static-content.springer.com/image/art%3A10.1007%2Fs11104-008-9673-6/MediaObjects/11104_2008_9673_Fig9_HTML.gif
Fig. 9

Correlation between DN-measured N2O fluxes from soil–plant systems and simultaneously observed CO2 fluxes due to ecosystem (soil plus plant) respiration. The given data were measured with static opaque chambers in the typical rice–wheat and vegetable cropping systems of the Yangtze delta. Panel (b) re-plots the data within the shade of panel (a), while panel (c) re-plots the data outside the shade of panel (a). In the regressions shown in (b) and (c), Y denotes DN-measured fluxes, and X denotes the natural logarithm of CO2 fluxes. The definition of DN is given in the “Abbreviations” section of the text

One may think of correcting apparent N2O fluxes by referring to the simultaneously measured CO2 concentrations. We were able to do so for the apparent N2O emissions from detached fresh plants within enclosures, where CO2 concentration changes were large (Table 2), but we were not able to make the correction for the apparent N2O fluxes from soil–plants systems. The significant correlation between apparent N2O signals and CO2 concentrations (Figs. 3 and 4) can be detected only when CO2 changes are large (e.g., usually over 1,000 μmol mol−1). For more than 95% of the observations that were conducted in the rice–wheat and vegetable fields, the CO2 changes within the static, opaque chamber enclosures were less than 700 μmol mol−1. For such a small change, it is usually difficult to obtain a significant correlation between CO2 and apparent N2O concentrations (as shown in Figs. 3 and 4). Thus, large errors would be resulted if a regression function obtained for larger CO2 changes (see Fig. 3) was applied to small CO2 changes to correct the apparent N2O concentrations. With respect to DN measurements in soil–plant systems or bare soils, CO2 may be a cause of a bias, but it is not the exclusive cause. Non-CO2 substances with lower ionization potentials compared with N2 (such as O2, H2O, CH4, as well as unknown components released from soils or soil–plant systems) may exert influences if they have similar retention times to that of N2O or if they enter the detector due to a modified carrier flow rate. The influences of non-CO2 substances, which have not yet been identified, may be considerably important when the CO2 change within a chamber enclosure is relatively small. Therefore, significant regression of corrected N2O concentrations against sampling times could not be obtained for a large portion of observations when the correction was conducted according to CO2 concentrations (data not shown). Accordingly, the N2O fluxes resulting from insignificant regressions of corrected N2O concentration against sampling time would be regarded as null data. As a result, the correction based on CO2 changes would greatly reduce the detection precision of N2O fluxes. In order to avoid this problem, we proposed an approach to directly correct the apparent N2O fluxes. The correction factors of this approach were given by referring to the N2O fluxes measured with DN-Ascarite. Using this approach, each apparent N2O flux was classified into one of the ranges that were encoded as A, B, C, D, E or F (Fig. 8), and then the corresponding value from the “Y − X” column of Table 4 was added to this apparent flux. Table 3 summaries the corrected results in the “DN (corrected)” column. As the data in this table show, the corrected N2O fluxes were significantly reduced as compared to the uncorrected ones (p < 0.05), while they were still slightly higher than the fluxes measured by DN-Ascarite (p < 0.05 for the unfertilized treatments). Subsequently, reduced annual N2O emissions resulted from the correction (only statistically significant under fertilized conditions). The outcome was statistically comparable with that resulting from the DN-Ascarite measurements. As a result, the correction yielded an EFd value of 0.26% for the rice–wheat fields, which was 43% lower than the estimate from the uncorrected fluxes and was comparable with that (0.28%) resulting from the DN-Ascarite measurements (Table 3). For the vegetable fields, the correction yielded an EFd value of 4.18%, which was only 2% lower than the estimate from the uncorrected fluxes but also 4% higher than that resulting from the DN-Ascarite measurements (Table 3).

The above calibration results suggest that the proposed approach for correcting DN-measured N2O fluxes can result in acceptable annual emissions and direct emission factors for applied nitrogen, although it is very rough. Of course, the applicability of the additive correction factors presented in the “Y − X” column of Table 4 is limited to chamber designs, sampling procedures and ecosystem types similar to those of this study. For situations different from those of this study, new specific correction factors need to be explored. Additionally, the proposed correction approach cannot completely solve the DN-measurement problems that lead to considerable biases, although it can significantly improve N2O fluxes and, thus, annual estimates as well as direct emission factors. Therefore, DN should be avoided when taking measurements of N2O emissions, especially from those sources with low fluxes. Moreover, further studies are required to determine appropriate approaches for reassessment of the available N2O emission data quantified using DN.

Among the N2O fluxes from the soil–plant systems measured by DN (Figs. 6 and 7), a considerable portion (21.6%) was more negative than those measured by DN-Ascarite. Reasons for this difference are not yet known. Thus, further study is needed to elucidate the mechanisms behind this difference.

Conclusions

With regard to the detection of N2O concentrations in ambient air samples, using DN-Ascarite and AM yielded relatively high precision (better than 1%), but using DN yielded low precision (worse than 1%). For air samples with varying CO2 concentrations and a constant N2O concentration, a positive correlation appeared between the CO2 concentration and the apparent N2O concentration, as measured by DN. The relationship was significant and conformed to a Logistic function over a large CO2 concentration span. Accordingly, measuring N2O emission from plants by DN led to an obvious overestimation as compared to DN-Ascarite or AM. With regard to DN-measured N2O emissions from a soil–plant system or a bare soil, a flux greater than 200 μg N m−2 h−1 was comparable with that simultaneously measured by DN-Ascarite. Otherwise, significant differences appeared between DN and DN-Ascarite, with the magnitudes of the differences varying with the intensity of the actual N2O emission. As a result, direct N2O emission factors for nitrogen fertilizers applied into cultivated fields, as well as background emission from croplands, were overestimated by DN measurements. These results suggest the necessity of reassessing available N2O emission data that were determined using DN measurements. These results also imply that the DN method should be avoided when measuring N2O emissions from weak sources with low fluxes. Nevertheless, the DN method coupled with sample collection by static chambers is still applicable for intensive sources with high N2O fluxes.

Acknowledgements

This study was jointly supported by the National Natural Science Foundation of China, the Chinese Academy of Sciences, and the European Union (grant numbers: 40425010, 40331014, KZCX2-yw-204, KZCX3-SW-440, NitroEurope IP 017841). We sincerely thank Mr. Guangren Liu, Prof. Wen Zhang, Dr. Zaixing Zhou, Dr. Shenghui Han, Mr. Gang Liu, Prof. Yong Han, Mr. Huajun Tong, Ms. Rui Wang, Mr. Jia Deng and Mr. Zhisheng Yao for their substantial assistance.

Copyright information

© Springer Science+Business Media B.V. 2008