Methanogenic Pathway and Fraction of CH4 Oxidized in Paddy Fields: Seasonal Variation and Effect of Water Management in Winter Fallow Season

A 2-year field and incubation experiment was conducted to investigate δ13C during the processes of CH4 emission from the fields subjected to two water managements (flooding and drainage) in the winter fallow season, and further to estimate relative contribution of acetate to total methanogenesis (Fac) and fraction of CH4 oxidized (Fox) based on the isotopic data. Compared with flooding, drainage generally caused CH4, either anaerobically or aerobically produced, depleted in 13C. There was no obvious difference between the two in transport fractionation factor (εtransport) and δ13C-value of emitted CH4. CH4 emission was negatively related to its δ13C-value in seasonal variation (P<0.01). Acetate-dependent methanogenesis in soil was dominant (60–70%) in the late season, while drainage decreased Fac-value by 5–10%. On roots however, CH4 was mostly produced through H2/CO2 reduction (60–100%) over the season. CH4 oxidation mainly occurred in the first half of the season and roughly 10–90% of the CH4 was oxidized in the rhizosphere. Drainage increased Fox-value by 5–15%, which is possibly attributed to a significant decrease in production while no simultaneous decrease in oxidation. Around 30–70% of the CH4 was oxidized at the soil-water interface when CH4 in pore water was released into floodwater, although the amount of CH4 oxidized therein might be negligible relative to that in the rhizosphere. CH4 oxidation was also more important in the first half of the season in lab conditions and about 5–50% of the CH4 was oxidized in soil while almost 100% on roots. Drainage decreased Fox-value on roots by 15% as their CH4 oxidation potential was highly reduced. The findings suggest that water management in the winter fallow season substantially affects Fac in the soil and Fox in the rhizosphere and roots rather than Fac on roots and Fox at the soil-water interface.


Introduction
Paddy fields are an important source of the greenhouse gas, methane (CH 4 ), contributing to 5-19% of the total global CH 4 emission [1]. Proper water management is considered to be one of the most important options for regulating CH 4 emission from paddy fields [2,3]. Generally, the fields are either intermittently irrigated or continuously flooded during the rice-growing season, and either drained without any irrigation except for rain water or kept flooded in the winter fallow season. Compared with continuous flooding, intermittent irrigation significantly decreases CH 4 emission from rice fields during the rice-growing season by 40-70% [4][5][6]. Similarly, drainage, relative to flooding, in the winter fallow season not only prevents CH 4 emission from the fields directly in the current season, but also sharply reduces CH 4 emission indirectly during the following rice-growing season [7][8][9]. Although effects of water management in the winter fallow season on CH 4 flux from the fields are considerably reported, its effect on the processes of CH 4 emission, including CH 4 production, oxidation and transportation, remains unclear. The stable carbon isotope technique, an important method for identifying processes of CH 4 emission from rice fields, has been widely used through measuring carbon isotopic ratios [10][11][12]. In addition, it can be used to quantify contributions of various CH 4 sources and provide information about carbon isotopes for global CH 4 budget [13,14]. To our knowledge so far, very little study has been done on the measurement of stable carbon isotopes in the fields during the ricegrowing season as affected by water management in the winter fallow season.
Methanogenesis is the precondition of CH 4 emission from paddy fields and mainly occurs through two pathways. One is H 2 / CO 2 reduction with the participation of specific hydrogenotrophic methanogens that use H 2 or organic molecules as H donor (CO 2 +4H 2 R CH 4 +2H 2 O). The other is acetate fermentation with the participation of acetotrophic methanogens (CH 3 COOH R CH 4 + CO 2 ). In general, the latter plays a more important role than the former in CH 4 formation [15,16]. If d 13 C-values of the CH 4 , CO 2 and acetate involved in methanogenesis are measured, contributions of the two pathways can be estimated by using the stable carbon isotope technique [17,18]. Theoretically, acetate fermentation and H 2 /CO 2 reduction accounts for 67% and 33%, respectively, of the total methanogenesis. Practically, relative contributions of the two pathways vary with rice cultivar, rice growth, water management, and environmental conditions, etc. [4,10,11,19]. During the rice-growing season, drainage can significantly enhance soil Eh, causing increase in oxidizing substances like Fe 3+ , sulphate and nitrate, and their inhibition of acetotrophic methanogens, thus reducing acetate-dependent methanogenesis [4,20]. In the winter fallow season, water management also significantly affects soil Eh, CH 4 production and then CH 4 emission from the fields during the following ricegrowing season [8], but its impact on relative contributions of the two main pathways of methanogenesis remains poorly known. CH 4 oxidation, which occurs at the root-soil interface and soilwater interface, is very important to regulating paddy CH 4 emission. By comparing CH 4 emission from the field or CH 4 production from aerobic incubation with methanogenesis in the strict anaerobic environment at the early stage, it was found that as much as 50-90% of the CH 4 was oxidized before escaping into the atmosphere [21][22][23]. By using the stable carbon isotope method to quantify the fraction of CH 4 oxidized in the paddy fields, recent studies in America and Italy indicated that it was less than 50% [10,12,24,25]. In China however, the fraction of CH 4 that was oxidized in a paddy field under intermittent irrigation during the rice-growing season was measured by this means to be up to 80% [4]. It was significantly higher than those in the fields under continuous flooding as above mentioned. Moreover, CH 4 oxidation potential was relatively higher in intermittently irrigated paddy soil than in continuously flooded soil [4], which suggests that CH 4 oxidation is highly impacted by water management during the rice-growing season. It is further indicated that oxidization of endogenous CH 4 in the paddy fields seems to be more obvious in China, particularly in the fields that are intermittently irrigated during the rice-growing season. Although CH 4 oxidation potential in paddy soil in a whole year has been reported [9], the percentage of CH 4 oxidized in the field as affected by water management in the winter fallow season is still unknown.
Therefore, a 2-year field and incubation experiment was carried out in the paddy fields subjected to two types of water management (flooding and drainage) in the winter fallow season. Seasonal CH 4 emission fluxes, CH 4 in soil pore water and floodwater, CH 4 in the aerenchyma of the plants, CH 4 production and oxidation in fresh paddy soil and rice roots, and their respective d 13 CH 4 -values during the rice-growing season were measured. The objectives of this study were: (1) to investigate impact of water management in the winter fallow season on CH 4 production, oxidation and emission and their d 13 CH 4 ; and (2) further to evaluate its effect on pathways of CH 4 production and fraction of CH 4 oxidized in the fields by using the isotopic measurements.

Field Description and Experimental Design
With the authorization of the Institute of Agricultural Science, Zhenjiang City, the experiment was carried out at Baitu Town, Jurong City, Jiangsu Province, China (31u589N, 119u189E). Main features of the experiment field have already been described in detail before [8]. The experiment was designed to have two treatments, three replicates each, i.e. winter fallow under continuous flooding (Flooding) and winter fallow without irrigation except for rain water (Drainage). Measurements of this study were performed during the 2008 and 2009 rice-growing seasons. Rice stubble and wild weeds were all removed from the experimental plots after rice harvest in the winter fallow season. For rice transplanting in the next rice-growing season, all the plots were ploughed the way the local farmers do. During the rice-growing season, they were continuously flooded and only drained for rice harvest. Rice seedlings (Cultivar ''Oryza sativa L. Huajing 3'') were transplanted at their 3-4-leaf stage on June 22 and 26 in 2008 and 2009, respectively. Urea was applied at a rate of 300 kg N ha 21 , of which 50% was done as basal fertilizer, 25% as tillering fertilizer, and 25% as panicle fertilizer. Both Calcium superphosphate and Potassium chloride were applied as basal fertilizer at a rate of 450 and 225 kg ha 21 , respectively. For further details of the farming practices during the two years, please see Zhang et al. [8].
Field Sampling and Measuring CH 4 flux was observed using the static closed chamber method [8]. To measure the flux, gas samples were collected at 4-7-day intervals using 18 mL vacuum vials. For determining isotopic signature (d 13 C) of the emitted CH 4 , gas samples were taken at 10,20-day intervals using 500 mL bags (Aluminium foil compound membrane, Delin gas packing Co., Ltd, Dalian, China) with a small battery-driven pump [26]. The first sample was collected after the chamber was closed for 3-5 min, and the second at the end of the 2 h closure. Isotopic signature (S) of the emitted CH 4 was calculated using the equation below: where A and B stands for CH 4 concentration (ml L 21 ) in the samples at the beginning and at the end, respectively, while a and b for the corresponding d 13 CH 4 -values (%) of the gas samples, separately. Simultaneously, soil redox potential (Eh) at the depth of 10 cm was measured, using Pt-tipped electrodes (Hirose Rika Co. Ltd., Japan) and an oxidation-reduction potential meter with a reference electrode (Toa . Soil temperature at the depth of 10 cm was measured with a hand-carried digital thermometer (Yokogawa Meter and Instruments Corporation, Japan). Soil pore water samples were collected using a Rhizon soil moisture sampler (10 RHIZON SMSMOM, Eijkelkamp Agrisearch Equipment, Giesbeek, Netherlands) [26]. The samplers were installed (in triplicate) in the plots prior to rice transplanting and then left in the soil over the whole observational periods. Samples (,5 mL) were firstly extracted using 18 mL vacuum vials to flush and purge the sampler before sampling. Then ,10 mL of soil solution was drawn into another vial for further analysis. Simultaneously, 10 mL of floodwater was collected using a plastic syringe and then transferred in to an 18 mL vacuum vial. Finally, the pressure of all sampling vials was equilibrated by filling in pure N 2 gas. After heavy shaking by hand, the airs in the headspace of the vials were directly analyzed for CH 4 on the GC-FID, and their corresponding d 13 CH 4 -values were determined using the isotope ratio mass spectrometer. CH 4 concentrations (C CH4 ) in pore water and floodwater were calculated using the following equation: where m stands for mixing ratio of CH 4 in the headspace of a vial (mL L 21 ), M V for volume of an ideal gas (24.78 L mol 21 at 25uC), G V for volume of the gas headspace of the vial (L), and G L for volume of the liquid in the vial (L). Samples of the gas in the aerenchyma of and emitted from the plants were taken using specially designed PVC bottomless pots [27]. The pot, 30 cm in height and 17 cm in diameter, was designed to have a water-filled trough around its top, avoiding any possible gas exchange during the sampling times. A PVC plate (18 cm in diameter) with a hole (the diameter could be adjusted to the growing plant) in the center was placed on top of each pot, allowing the plant to grow through the hole and keep divided into two parts. Then, one plant inside the pot was cut right above the plate while the other remained intact as the control. Finally, chambers (306306100 cm) were laid on the pots, and gas samples in the headspace of the chambers were collected simultaneously with a small battery-driven pump.
Triplicate soil cores were collected from each experimental plot using a stainless steel corer (7 cm diameter625 cm length) and then prepared into mixture [9]. Samples (in triplicate) from the mixture, about 50 g (dry weight) each, were taken and promptly transferred into 250-mL Erlenmeyer flasks separately, and turned into slurries with N 2 -flushed de-ionized sterile water at the ratio of 1:1 (soil/water). During the whole process, the samples were constantly flushed with N 2 to remove O 2 and CH 4 , and the flasks containing these samples were sealed for anaerobic incubation. Some other flasks with air headspace were sealed directly for aerobic incubation. They were all stored in N 2 at 4uC for further analysis within 8 h. A small portion of the soil sample was dried for 72 h at 60uC for determination of isotopic composition of the organic carbon.
Complete rice plants together with roots were carefully collected from the experimental plots at each main rice growth stage, i.e. tillering (TS), booting (BS), grain-filling (FS) and ripening (RS) stages, in 2009 [8]. The roots were washed clean with N 2 -flushed demineralized water and cut off from the green shoots at 1-2 cm from the root with a razor blade. The fresh roots, 20 g each portion, were then put into flasks, separately, for further preparation and processing in the same way as for the soil. The shoots were dried at 60uC for 72 h for dry weight measurement and then stored at room temperature for determination of isotopic composition of the organic carbon.

Fresh Soil and Roots Incubations
CH 4 production potentials were measured for the paddy soil and rice roots under anaerobic incubation. The flasks were flushed with N 2 consecutively for six times through double-ended needles connecting a vacuum pump to purge the air in the flasks of residual CH 4 and O 2 . Simultaneously, methanogenesis was determined aerobically using flasks with air headspace directly. They were subsequently incubated at a temperature the same as measured in the field for 50 h in darkness. Gas samples were collected twice with a pressure lock syringe, and analyzed 1 h and 50 h later after the flasks were heavily shaken by hand, for CH 4 on the GC-FID. CH 4 production was calculated using the linear regression of CH 4 increasing with the incubation time.
CH 4 oxidation potentials were determined for the paddy soil and rice roots under aerobic incubation with high CH 4 concentration supplemented, using equipment the same as described above. Firstly, pure CH 4 was injected into each flask to make a high concentration inside (,10000 mL L 21 ). Then, the flasks were incubated in dark under the same temperature as measured in the field and shaken at 120 r.p.m. CH 4 depletion was measured by sampling the headspace gas in the flask after vigorous shaking for subsequent GC-FID analysis. The first sample was collected generally 30 min after pure CH 4 was injected. Samples were then taken in 2-3 h intervals during the first 8 h of the experiment. They were left in the flasks over night and measured the next day in 2 h intervals again. CH 4 oxidation was calculated by linear regression of CH 4 depletion with incubation time.

Analytical Methods
CH 4 was quantified using the gas chromatograph (GC) equipped with a flame ionization detector (FID) [28]. The isotopic composition (d 13 C) of CH 4 and CO 2 was determined with a Finnigan MAT-253 Isotope Ratio Mass Spectrometer (IRMS, Thermo-Finnigan, Bremen, Germany) using the continuous flow technique [26,29]. The IRMS had a fully automatic interface for pre-GC concentration (Pre-Con) of trace gases, and the precision of repeated analyses was 60.196% (n = 9) with 2.02 mL L 21 CH 4 injected. Gas samples were first blown into the chemical trap with Mg(ClO 4 ) 2 and ascarite by He flow (20 mL min 21 ). Over 99.99% of the CO 2 and H 2 O in the samples was absorbed and removed. CH 4 in the samples was then converted into CO 2 in a combustion reactor at about 1000uC. Subsequently, it was flowed into the freezing traps with liquid nitrogen (-196uC) and the GC for further separation. The separated gases were finally transferred into the mass spectrometer for d 13 C determination. The dried soil and plant samples were analyzed for carbon isotope composition with a Finnigan MAT-251 Isotope Ratio Mass Spectrometer (Thermo Finnigan, Bremen, Germany). Soil organic carbon contents were measured by wet oxidation using dichromate in acid medium followed by the FeSO 4 titration method.

Calculations
Isotope ratios are expressed in the standard delta notation: where R stands for 13 C/ 12 C of the sample and the standard, respectively, using PDB carbonate for the standard. Carbon isotope fractionation factor during acetate fermentation (e acetate/CH4 ) or H 2 /CO 2 reduction (a CO2/CH4 ) for methanogenesis was defined by Hayes [30]: where d 13 C acetate , d 13 CH 4 (acetate) and d 13 CH 4 (H2/CO2) is the d 13 C values of acetate, CH 4 produced from acetate and from H 2 /CO 2 , respectively. Relative contribution of acetate to total CH 4 (F ac ) was calculated using the following mass balance, assuming that acetate fermentation and H 2 /CO 2 reduction were the only sources of methanogenesis in the rice fields [10][11][12]: where d 13 CH 4 stands for d 13 C value of total CH 4 . In addition, the fraction of CH 4 that was oxidized (F ox ) in the fields was estimated using the equation given by Stevens and Engelkemeir [13] and Tyler et al. [12]: where d13CH 4(original) stands for carbon isotopic signature of the primarily produced CH 4 , d 13 CH 4(oxidized) for carbon isotopic signature of the residual CH4 after oxidization, of which the calculation was done using a semi-empirical equation [12]: and a ox stands for isotope fractionation factor due to CH 4 oxidation by the methanotrophs, and e transport for isotope fractionation factor due to CH 4 transport by the plants.

Statistics
Statistical analysis was done using the SPSS 18.0 software for Windows (SPSS Inc., Chicago). Differences between the two treatments in mean (n = 3) CH 4 concentration, mean CH 4 production and oxidation potentials, and mean soil Eh were determined through one-way analysis of variance (ANOVA) and least significant difference (LSD) test. Relationships between CH 4 fluxes and emitted d 13 CH 4 (n = 18), between mean CH 4 production potential and soil Eh (n = 11), and between CH 4 oxidation potential and soil temperature (n = 11) were assessed using correlation analysis. Statistical significant differences and correlations were set at P,0.05.

CH 4 Emission and d 13 C
CH 4 emissions (Fig. 1a, e) were significantly higher from flooded fields than from drained fields as had been reported before [8]. Different variation patterns were observed in the d 13 C of the emitted CH 4 in 2008 and 2009 seasons (Fig. 1b, f). Generally, the emitted CH 4 tended to be 13 C-enriched in 2008 with its d 13 Cvalue increased from -69 to -51% in Treatment Flooding, and from -65 to -47% in Treatment Drainage (Fig. 1b). However, more complicated changes were observed in 2009, showing that the emitted CH 4 was relatively enriched in 13 C at the beginning and at the end of the season, and relatively 13 C-depleted in the middle of the season (Fig. 1f). However, little difference was found between Treatments Flooding and Drainage, with d 13 C-values being in the range of -68 , -48% and -71 , -53%, respectively ( Significant difference in soil Eh was observed between the two treatments ( Fig. 1c, g, Table 1). Soil Eh was very close to 0 mV at the beginning of the season and remained much lower in Treatment Flooding than in Treatment Drainage throughout the two seasons, with the averaged value of -165 and -88 mV in 2008, and -153 and -26 mV in 2009, respectively. Soil temperature at the depth of 10 cm generally peaked around D25 (25 days after rice transplanting) and then gradually declined till the end of the season (Fig. 1d, h) (Fig. 2a, c). For Treatment Drainage however, CH 4 concentration decreased gradually from 300 to 200 mmol L 21 during the 2008 season (Fig. 2a), whereas it was the highest (,200 mmol L 21 ) in the middle and the lowest (,50 mmol L 21 ) at the beginning and the end of the 2009 season (Fig. 2c). The averaged CH 4 concentration during the two seasons was generally higher in Treatment Flooding than in Treatment Drainage (Table 1). d 13 C-value of the CH 4 was relatively stable during the 2008 season though it increased and then slightly decreased (Fig. 2b). As a whole, CH 4 was much more 13 C-enriched in Treatment Flooding (-65%) than in Treatment Drainage (-67%) over the 2008 season ( Fig. 2b, P,0.05). In the 2009 season however, d 13 C-value fluctuated sharply within the range from -65 to -55% to -70% or so (Fig. 2d). No obvious difference in mean d 13 C-value (, -60%) was observed between the two treatments in 2009 ( Fig. 2d, P.0.05).
CH 4 concentration in floodwater of the field in 2009 was measured simultaneously. No more than 7 mmol L 21 of CH 4 was detected though little data were obtained (Fig. 2c). On the other hand, CH 4 in floodwater became more and more 13 C-enriched towards the end of the season, with the d 13 C-value increased from -50 to -40% (Fig. 2d). Little difference in d 13 C-value was observed as well between Treatments Flooding and Drainage (Fig. 2d, P.0.05). Compared with porewater CH 4 , floodwater CH 4 was much more enriched in 13 C (Fig. 2d, P,0.05).

Plants Emitted and Aerenchymatic CH 4 and d 13 C
To quantify stable carbon isotope fractionation during the CH 4 emitted through the aerenchyma of the plants, d 13 C-values of the emitted CH 4 and aerenchymatic CH 4 were measured simultaneously. On the three sampling days during the 2009 season, the emitted CH 4 was relatively stable with its d 13 C-value stable around -60% ( Table 2). The aerenchymatic CH 4 as expected, was significantly 13 C-enriched compared to the emitted CH 4 , with the d 13 C-values varying in the range of -51 , -42%, and being about -47% on average for the two treatments ( Table 2). As a result, the isotope fractionation factor due to CH 4 transport (e transport ) was determined to be in the range from -16 to -11% in Treatment Flooding and from -14 to -12% in Treatment Drainage. As a whole, no obvious difference in mean value of e transport (, -13%) was observed between the two treatments ( Table 2, P.0.05).

CH 4 Production Under Anaerobic Incubation and d 13 C
CH 4 production potentials of the slurries of paddy soil were measured during the 2008 and 2009 rice seasons (Fig. 3a, d). Methanogenesis started more quickly and became more intense in Treatment Flooding than in Treatment Drainage over the two seasons (Fig. 3a, d), peaked around D40-60 for the former and around D80 for the latter. Mean production potential was significantly higher in Treatment Flooding than in Treatment Drainage during the two seasons (Table 1, P,0.05). The produced CH 4 was relatively stable in d 13 C (, -60%) in Treatment Flooding while fluctuated sharply (from -72 to -55%) in Treatment Drainage in 2008 (Fig. 3b). In 2009 however, it was gradually enriched in 13 C for the two treatments, with d 13 C-value ranging from -70 to -60% (Fig. 3e). In addition, the mean d 13 CH 4 -value in Treatment Flooding appeared to be slightly more positive than that in Treatment Drainage over the two seasons, varying in the range of -63 , -58% and -66 , -63%,  Table 1. Mean CH 4 concentration (mmol L 21 ) in soil pore water, mean CH 4 production and oxidation potentials (mgCH 4 g d 21 ), and mean soil Eh (mV) during the 2008 and 2009 rice seasons (mean 6 SD, n = 3). respectively. The produced CO 2 became isotopically heavier step by step, causing d 13 C-value to decrease from -20 , -15% at the beginning of the season to around -10% at the end of the season, and it was relatively more 13 C-enriched in Treatment Flooding than in Treatment Drainage over the two seasons (Fig. 3c, f, P.0.05).
Abundant methanogenesis was measured on the fresh rice roots under anaerobic incubation in 2009 (Fig. 4a). The production of CH 4 increased sharply and peaked around D60, just like the soil (Fig. 3d). Then it decreased gradually till the end of the season. As a whole, it was significantly higher in Treatment Flooding than in Treatment Drainage (Fig. 4a, Table 1, P,0.05). Similar to CH 4 produced in the soil, CH 4 produced on the roots was depleted in 13 C at the beginning of the season (Fig. 4b). Subsequently, it became more 13 C-enriched, with its d 13 C-values ranging from -90 to -75%. No significant difference was observed in mean d 13 Cvalue between Treatment Flooding (-83%) and Treatment Drainage (-81%). However, it was much more negative compared to the CH 4 produced in the soil in d 13 C-value (Figs. 3e and 4b, P,0.01). The d 13 C-value of produced CO 2 ranged from -22 to -17% over the two seasons and no obvious difference was observed between the two treatments ( Fig. 4c, P.0.05).

CH 4 Production Under Aerobic Incubation and d 13 C
Less than 0.3 mgCH 4 gsoil 21 d 21 was produced in the soil under aerobic condition over the 2009 season (Fig. 5a), and 1.0-1.5 mgCH 4 groots 21 d 21 was on the roots at the beginning of the season and below 0 mgCH 4 groots 21 d 21 at the end of the season (Fig. 5c). The produced CH 4 was very stable over the season both in the soil and on the roots, with d 13 C-values around -58 , -55% and -44 , -41%, respectively. Opposite to the CH 4 produced under anaerobic condition, it was significantly more enriched in 13 C on the roots than in the soil (Fig. 5b, d, P,0.01). Generally, the d 13 C-value was more positive in Treatment Flooding than in Treatment Drainage (Fig. 5b, d). In addition, the CH 4 produced under aerobic condition was significantly 13 C-enriched relative to that produced under anaerobic condition, especially those from the roots (Figs. 3e, 4b and 5b, d, P,0.01).

CH 4 Oxidation Under Aerobic Incubation Amended with High CH 4 Concentration
Similar variation patterns of the CH 4 oxidation potentials in the paddy soil during the 2008 and 2009 seasons were observed, showing a peak at the beginning of the season and a steep slope till the end of the season (Fig. 6a, b). Although the potential was relatively lower in Treatment Flooding than in Treatment  Drainage in the middle of the season but slightly higher both at the beginning and at the end of the season (Fig. 6a, b), no significant difference was observed between the two treatments in mean of the potential (Table 1, P.0.05). Notably, a significant positive relationship was observed between CH 4 oxidation potential and soil temperature in temporal variation (r = 0.703-0.859, P,0.05).
Intensive oxidation signal on the fresh roots was observed, which was also the strongest (400-600 mgCH 4 groots 21 d 21 ) at the beginning of the season and declined to the lowest (150-400 mgCH 4 groots 21 d 21 ) at the end of the season (Fig. 6c). Throughout the 2009 season, CH 4 oxidation potential on the roots was significantly higher in Treatment Flooding than in Treatment Drainage (Fig. 6c, P,0.05).

Organic Carbon in Soil and Plant Samples
During the 2008 season, the content of organic carbon in the soil was 1.0260.08% in Treatment Flooding and 1.1160.05% in Treatment Drainage, and it seemed to increase during the 2009 season, reaching 1.6560.01% and 1.8360.10%, respectively. Soil organic carbon in Treatment Drainage was very stable in d 13 C (-27.9%) during the two rice seasons, whereas it was slightly 13 Cenriched in Treatment Flooding, with d 13 C-value increasing from -28.1% in 2008 to -27.0% in 2009. Organic carbon in the plant samples showed little change throughout the 2009 season, with d 13 C-value of -28.9%, -29.2% and -28.7% on D27, D66 and D108, respectively, although it was slightly lighter in contrast to the organic carbon in the soil.

Effects on Stable Carbon Isotopes
The processes of CH 4 emission involved in CH 4 production, oxidation and transportation in the fields were well identified by measuring stable carbon isotopes (Fig. 7). In paddy fields, besides the applied organic fertilizers, plant photosynthesis and degradation of soil organic carbon are the two most important substrate sources for methanogenesis [31]. As substrates for CH 4 production, d 13 C-value of organic carbon in the plant samples (-29%) seemed to be slightly negative than that in the soil samples (-27%) (Fig. 7). Previous observations also showed that organic carbon was slightly lighter in the plant than in the soil [10,27,32]. Intensive carbon isotope fractionation generally happens when methanogenic precursors form CH 4 . Around 10-20% occurs during CH 4 production through acetate fermentation while 50-70% during CH 4 production through H 2 /CO 2 reduction [16,33]. As a consequence, CH 4 from the former (-60 , -50%) is usually more positive than that from the latter (as negative as -110%) [34]. The CH 4 produced in the soil was more 13 C-enriched in Treatment Flooding than in Treatment Drainage (Fig. 3b, e) and also more (-65%) than that on the roots (-80%) in both treatments (Fig. 7). This shows that flooding, compared with drainage in the winter fallow season, increased the relative contribution of acetate to CH 4 production and that aceticlastic methanogenesis in the soil was more important than that on the roots (Fig. 8). Early anaerobic measurements indicated that CH 4 from the roots was more depleted in 13 C than that from the soil [17,27].    In rice-based ecosystems, the produced CH 4 , except for the portions oxidized and emitted into the atmosphere, is temporarily retained in the soil as entrapped CH 4 and dissolved CH 4 in soil pore water [35]. As mainly in the form of bubbles, CH 4 in soil pore water probably remains unoxidized and is usually considered to be the original CH 4 produced in the field in many reports [11,12,36]. However, Krüger et al. [10] found that CH 4 in soil pore water poorly represented the produced CH 4 . In the present study, it might be partially oxidized as well in the rhizosphere during the periods of D30-60 in the 2009 season (Fig. 6d), though its mean d 13 C-value lingered around -60% and close to that of the CH 4 produced in anaerobic soil over the season (Fig. 7). When porewater CH 4 is released through the soil-water interface in paddy fields, it will be considerably oxidized, leaving the remainder temporarily in the floodwater. Since 12 CH 4 is consumed faster than 13 CH 4 by soil microbes, the residual CH 4 is then 13 C-enriched [34]. As a consequence, floodwater CH 4 (-45%) was more 13 C-enriched than porewater CH 4 (-60%). The observation of d 13 C-value of the CH 4 produced in aerobic soil being more positive than that in anaerobic soil (Fig. 7) further demonstrates that CH 4 oxidation is intensive at the soil-water interface. In addition, rice roots can excrete O 2 thus forming an important CH 4 -oxidizing zone in the rhizosphere. What is more, fresh rice roots per se have a strong CH 4 oxidation capacity [10,27,37]. CH 4 aerobically produced on the roots appeared to be more enriched in 13 C (-45%) than that in the soil (-55%). It suggests that rice roots in the rhizosphere may be more important than the soil per se as driving force for CH 4 oxidation, thus causing more 12 CH 4 consumed and leaving more 13 CH 4 remained (Fig. 7).
Aerenchymatic CH 4 (, -47%) was similar to oxidized CH 4 in d 13 C-value (Fig. 7), which demonstrates that it has been strongly oxidized in the rhizosphere. In Italian paddy fields, Krüger et al. [10] also found that it stayed around -50% throughout the ricegrowing season. After being emitted through transportation of the plants, aerenchymatic CH 4 was much heavier than emitted CH 4 (Fig. 7), due to the fact that 12 CH 4 was transported from the plants at a faster rate than 13 CH 4 [38]. By subtracting d 13 C-value of aerenchymatic CH 4 from d 13 C-value of emitted CH 4 the transport fractionation by the plants is quantified [10][11][12]. In theory, the transport fractionation is relatively small due to small CH 4 transport capacity of the plants at the beginning of the season. It gets strengthened together with the growth of the plants during the middle of the season but weakens again till the end of the season because of aging of the roots and plants. Consequently, value of fractionation (e transport ) was the lowest during the middle of the season, and relatively high at the beginning and the end of the season because it was shown as negative (Table 2). Moreover, it was averaged around -13%, suggesting that water management in the winter fallow season has little effect on CH 4 transport fractionation during the following rice-growing season. Similar e transport was also observed in other field experiments [10][11][12].
The d 13 C-value of emitted CH 4 fluctuated largely during the 2008 and 2009 rice seasons (Fig. 1b, f), and they were negatively related to CH 4 emission in seasonal variation (r = -0.695 , -0.546, P,0.01). An analogous relationship between them was also observed in other experiments [4,26], which was considered to be the combined effect of CH 4 production, oxidation and transport in the fields [26,39,40]. Although water management in the winter fallow season played a key role in CH 4 emission from the rice fields (Fig. 1a, e), it had little impact on d 13 C-value of emitted CH 4 (Fig. 1b, f). For the two seasons, mean value was , -60%, being in the range of the measurements in previous report [26]. Compared with flooding, drainage had CH 4 relatively more depleted in 13 C (Fig. 3b, e), but the CH 4 would become enriched in 13 C again after it was oxidized because F ox -value in the latter was 5-15% higher (for detailed description, please see Section Effects on CH 4 oxidation below). In addition, there was no obvious difference in e transport between the two treatments (Table 2). Therefore, the 13 C-depleted CH 4 in Treatment Drainage was supposed to be offset by the higher fraction of CH 4 oxidation, thus making the d 13 C-value of emitted CH 4 from the two treatments similar.

Effects on CH 4 Production
Previous studies demonstrated that water management in the winter fallow season significantly affected CH 4 production during the following rice-growing season [8,9]. In the present study, it showed an important effect on CH 4 production of the fields by significantly affecting soil Eh. Methanogens are a kind of extreme anaerobic bacteria, which produce CH 4 under strict reductive conditions. Compared to the fields flooded in the winter fallow season, the fields drained were probably much lower in population and activity of methanogens [41][42][43] and it generally took a longer time for methanogens to revive during the following rice-growing season [44]. Therefore, drainage delayed and decreased CH 4 production in soil by significantly increasing soil Eh (Figs. 1c, g and 3a, d, Table 1). On roots as well, the lower the soil Eh, the higher the CH 4 production (Table 1) because roots get to age and decay faster if they are constantly under a stronger reductive condition [45,46]. A significant negative correlation between mean CH 4 production and soil Eh (Table 1, r = -0.805, P,0.01 for soil and r = -0.994, P,0.01 for roots, respectively) better demonstrated that soil Eh significantly affected by water management in the winter fallow season was a key factor that influenced CH 4 production. CH 4 concentration in soil pore water being generally lower in Treatment Drainage than in Treatment Flooding ( Table 1) further showed that drainage decreased CH 4 production in the fields.
In paddy fields, CH 4 mainly comes from acetotrophic and hydrogenotrophic methanogenesis. Methanol-dependent methanogenesis may possibly be another contributor to the total CH 4 production, though, insignificant [47]. Relative contribution of acetotrophic methanogenesis (F ac ) to the total CH 4 production can be calculated by following Eq. (7) if a fractionation factor of a CO2/CH4 = 1.079 is used for CO 2 -dependent methanogenesis and d 13 CH 4 (acetate) = -43% is for acetate-dependent methanogenesis based on the following reports [17,19]. In an Italian paddy soil, Fey et al. [19] found that a CO2/CH4 , decreasing with increasing temperature, was 1.083 at 10uC, 1.079 at 25uC, and 1.073 at 50uC, which was in good agreement with the relationships in marine sediment [48] and methanogenic cultures [49]. Therefore, a CO2/CH4 = 1.079 was applied because the temperature during the two seasons varied in the range of 20-30uC with an average of 24uC. On the other hand, Fey et al. [19] demonstrated that d 13 CH 4 (acetate) increased with increasing temperature, e.g., from -50 , -46% at 10uC to -45 , -36% at 25uC, and to -43 , -31% at 37uC. Moreover, the d 13 CH 4 (acetate) -values of -43 , -36% have even been applied considerably to experiments in the fields during the rice-growing season [10][11][12]36]. Due to lack of measurements, a constant value of -43% was hence used in the present study for better comparison with these reports. What is more important, it was more reasonable and suitable because d 13 C-value of the soil organic carbon-substrate for methanogenesis, in this study was similar to those observed before [10,17,19]. Although they might be different in microbe habitats and varied with temperature and rice growth [10,17,19], the same values of a CO2/CH4 and d 13 CH 4 (acetate) above mentioned have also been used [4,26,27].
The findings show that variation of F ac -value in paddy soil during the 2008 rice season was similar to that during the 2009 rice season in pattern. That is, acetate-dependent methanogenesis dominated in the late season, while it was not so much important in the mid season, with F ac -value being over 60-70% and less than 40%, respectively (Fig. 8a, b). In Italian paddy fields, measurements also show that it was dominant at the end of the season [10]. Water management in the winter fallow season had an important impact on methanogenic pathways of paddy soil during the following rice-growing season. Generally, CH 4 from acetate cleavage dominated in Treatment Flooding during the two rice seasons, having a mean F ac -value of 53-65%, which was 5-10% higher than in Treatment Drainage (Fig. 8a, b). Drainage increased production of oxidants, such as Fe 3+ or sulphate, along with the increase in soil Eh [20,50]. As a result, the growth and activity of methanogens was probably out-competed by iron-or sulphate-reducing bacteria [51,52]. More importantly, acetotrophic methanogens seemed to be inhibited to a larger extent than hydrogenotrophic methanogens [20,53]. This suggests that soil Eh is an important indicator of pathways of methanogenesis in paddy fields-the higher the soil Eh, the more inhibited the acetotrophic methanogens, and the less the contribution of acetate to the total methanogenesis. In the present study therefore, drainage in the winter fallow season significantly increased soil Eh (Fig. 1c, g) and obviously decreased methanogenesis in paddy soil compared to flooding (Fig. 3a, d), and hence the contribution of acetate-dependent methanogenesis, probably ascribed to the fact that acetate-utilizing methanogens was more inhibited by any increase in soil Eh (Fig. 1c, g) [20]. Intermittent irrigation during the rice-growing season significantly reduced the contribution of acetate to CH 4 production, of which the finding could further verify this point of view [4].
The relative contribution of acetate-dependent methanogenesis on rice roots was similar to that in paddy soil in 2009, which was the lowest (almost 0%) in the mid season but rose up to the highest (,40%) at the end of the season (Fig. 8c). In total, F ac -value was 1-32% in Treatment Flooding and 0-38% in Treatment Drainage, being much lower than that in the soil (Fig. 8). It indicates that methanogenesis on fresh rice roots is mostly from H 2 /CO 2 , and it is little affected by water management in the winter fallow season. Previous reports also show that F ac -value of rice roots was less than 40% in most of the season [10]. In an incubation experiment with rice roots D75-80 old, Conrad et al. [17] found that CH 4 mainly came from H 2 /CO 2 -dependent methanogenesis as well throughout the entire observation, with an average F ac -value of 47%. Compared with that of soil, the relative contribution of acetate to the total methanogenesis on the roots was lower by approximately 30% (Fig. 8), which is likely attributed to the difference in population of their dominant methanogens [54][55][56]. More exact measurements using stable isotope probing techniques have further demonstrated that CH 4 production on roots depends mainly on H 2 /CO 2 reduction triggered by RC-I methanogens (Rice Cluster I Archaea) [57,58]. On the other hand, organic carbon slightly lighter in plant samples than in soil samples might be a possible reason for F ac -value being much lower in paddy soil than on rice roots.

Effects on CH 4 Oxidation
CH 4 oxidation in soil seemed to be highly influenced by soil temperature rather than water management in the winter fallow season. Firstly, no significant difference was observed between flooding and drainage in mean oxidation potential during the 2008 and 2009 seasons (Table 1). Secondly, it varied with soil temperature (Figs. 1d, h and 6a, b), and a positive relationship was observed over the two seasons (r = 0.703-0.859, P,0.05), which is in good agreement with the previous report [9]. An appropriate soil temperature favors growth of methane-oxidizing bacteria, thus enhancing their capacity of CH 4 oxidation [59]. The higher the soil temperature within the range of 12.5-34.8uC, the higher the CH 4 oxidation rate [60], which is consistent with our observations. Considerable measurements on fresh roots have shown that the roots per se have a high CH 4 oxidation capacity [10,37,61]. In the present study, CH 4 oxidation on the roots was the strongest at the beginning of the season and weakened later on (Fig. 6c), being in agreement with the previous reports [10,27]. Drainage compared to flooding in the winter fallow season significantly decreased CH 4 oxidation potential on the roots (Fig. 6c), probably attributed to the effect of flooding highly increasing CH 4 production (Fig. 4a). Higher concentration of CH 4 stimulated growth and activity of the methanotrophs on the surface of the roots, thus raising their CH 4 oxidation capacity [59].
The fraction of CH 4 oxidation (F ox ) can be quantified by measuring d 13 C-value of CH 4 from various compartments of the paddy fields with a special model in case some parameters (a ox and e transport ) are already available [10][11][12][13]. The potential shift in the carbon isotopes during the CH 4 oxidation (fractionation factor a ox = 1.025-1.038) was firstly determined in methanotrophs enriched cultures [62] and then considerably in landfill cover soils at a temperature of about 25uC [63][64][65]. Interestingly, the value of 1.025-1.038 has been widely applied to field conditions [10][11][12]24,25] though the knowledge of a ox in paddy soil is still incomplete. Very recently, we have found a ox = 1.025-1.033 at 28.3uC in a Chinese paddy soil [27]. Consequently, the value of 1.038 was used in the present study due to the similar temperature during the seasons, and more reasonable results would be obtained (Fig. 9). On the other hand, the transport fractionation factor e transport was equivalent to the difference in 13 C between emitted and aerenchymatic CH 4 ( Table 2), ranging from -16 to -11% in the 2009 season. In 2008 however, no corresponding measurements were performed. Nevertheless, the averaged value of -13% in 2009 was applied to the 2008 field data (Fig. 9), because it was also very close to previous observations [10][11][12]. Since fractionation factors (a ox and e transport ) are influenced by temperature, microbes, soil property, and rice growth [10,64], more attention thereby need further be paid to getting reliable and exact values of CH 4 oxidation in paddy fields.
Similar to the potentials of CH 4 oxidation, the fraction of CH 4 oxidized in the rhizosphere was relatively high (as high as 60-90%) in the first half of the rice growth period during the 2008 and 2009 seasons and relatively low (,10-30%) in the remainder periods (Fig. 9a, b). In Italian paddy fields, measurements also show that CH 4 oxidation was very important at the beginning of the season but became slight later, with F ox -value decreasing rapidly from approximately 40 to 0% [10,20,24]. Under unfertilized microcosms, Conrad and Klose [25] obtained that F ox -value decreased from about 15% in the beginning to about 5% at the end, which was probably attributed to nitrogen-limitation of the methanotrophs [10,20,24]. On the whole, mean value of F ox was 35-55% in Treatment Drainage, being 5-15% higher than that in Treatment Flooding during the 2008 and 2009 seasons (Fig. 9a,  b). It suggests that compared to flooding in the winter fallow season drainage can increase the proportion of CH 4 oxidized during the following rice-growing season. Probable reason was that drainage significantly decreased CH 4 production (Fig. 3a, d) while it did not simultaneously affect CH 4 oxidation in the field (Fig. 6a, b).
When CH 4 in soil pore water passed through the soil-water interface into the floodwater, intensive signals of CH 4 oxidation were observed by following changes in isotopic signature between them (Fig. 7). An obvious oxidation signal was also observed of the dissolved CH 4 approaching to soil surface [11,12]. Therefore, F oxvalue was reasonably calculated based on d 13 C-value of CH 4 in pore water for d 13 CH 4 (initial) and on d 13 C-value of CH 4 in floodwater for d 13 CH 4 (final) . It was high on D16 and D88, but relatively low on D47 and D50, especially in Treatment Flooding ( Table 3). As the emission of CH 4 in the fields goes absolutely through aerenchyma of the plants in the middle of the season, a very high percent of the CH 4 is therefore consumed in the rhizosphere and a low percent oxidized at the soil-water interface (Table 3). On the contrary, CH 4 emits into the atmosphere mainly through bubble ebullition and molecular diffusion in the early and the late rice-growing season, the CH 4 is probably oxidized at the soil-water interface, showing a relatively high F ox -value as a  (Table 3). On the whole, there was no difference in mean F ox -value between the two treatments (47%) throughout the entire observational period, indicating that water management in the winter fallow season has little impact on CH 4 oxidation at the soil-water interface. Notably, Krüger et al. [10] had even pointed out that porewater CH 4 was a poor indicator of produced CH 4 . Therefore, CH 4 in soil pore water on D47 and D50 in the present study might be oxidized partially as well. As a result, its d 13 C-value was possibly not fit to stand for d 13 CH 4 (initial) , which would bias the estimation of CH 4 oxidation therein. Actually, the CH 4 produced in paddy fields would be mostly oxidized in the rhizosphere because over 90% of the CH 4 is considered to emit into the atmosphere through the aerenchyma of the plants while less than 0.1% released via ebullition and diffusion [22,23,66]. Moreover, the absolute rates of CH 4 oxidation at the soil-water interface were significantly lower than those in the rhizosphere [66][67][68]. Therefore, although the fraction of CH 4 oxidized at the soil-water interface appears to be very high (Table 3), the amount of the CH 4 must be significantly lower than that oxidized in the rhizosphere, and may be negligible.
In lab conditions, the difference between anaerobic and aerobic CH 4 productions in the soil was apparently attributed to CH 4 oxidation at the soil-water interface [69]. In the present study, CH 4 from aerobic incubation has undergone intensive oxidization, relative to CH 4 produced anaerobically (Figs. 3d, e and 4a, b). This was reflected both in significantly low production rate (Fig. 5a, c) and more positive d 13 C-value of the produced CH 4 (Fig. 7). Therefore, the fraction of CH 4 that was oxidized in aerobic condition could be estimated directly by using Eq. (8) based on d 13 C-values of anaerobically produced CH 4 for d 13 CH 4 (initial) (Figs. 3e and 4b) and on d 13 C-values of aerobically produced CH 4 for d 13 CH 4 (final) (Fig. 5b, d). Although little is known about the isotope fractionation when CH 4 oxidation occurs on rice roots, a ox = 1.038 was tentatively used to estimate F ox -value thereupon as well as in paddy soil. Results show that a variation pattern of F oxvalue was similar to that in the rhizosphere, which was the highest in the first half of the season but tended to get lower in the second (Fig. 9). For soil, F ox -value ranged from ,5 to 50% and was slightly affected by water management in the winter fallow season (Fig. 9c). For roots however, it was over 100% in the most of the season, suggesting that fresh rice roots consume almost all the produced CH 4 by themselves in lab conditions. Moreover, it was 15% lower in Treatment Drainage than in Treatment Flooding (Fig. 9d). It is a matter of fact that little CH 4 was produced in aerobic incubation and it even became negative in growth at the end of the season (Fig. 3c). In addition, drainage significantly decreased the CH 4 oxidation capacity of the field relative to flooding (Fig. 6c). Methanotrophs are found to attach closely to, or even live inside, rice roots [37,70,71]. The roots per se have a strong CH 4 oxidation capacity indeed (Fig. 6c). On the other hand, it further indicates that value of a ox = 1.038 may be unreasonable for roots in estimation of F ox , because in field conditions, CH 4 production directly or indirectly from the roots must not be completely oxidized and F ox -value should be lower than 100%. Therefore, more investigation of fractionation factor a ox in paddy soil, in particular on rice roots, should be performed to better quantify CH 4 oxidation in the fields.

Conclusions
Through the field and laboratory experiments, we investigated d 13 C in every process of CH 4 emission from rice fields as affected by water management in the winter fallow season and further estimated pathways of CH 4 production and fraction of CH 4 oxidation using the stable carbon isotope technique. Compared Figure 9. Temporal variation of the fraction of CH 4 oxidized (F ox ) in the rhizosphere and at the surfaces of paddy soil and rice roots. F ox in (a) and (b) was calculated with Eq. (8) using 1.038 for a ox , d 13 C-values of CH 4 anaerobically produced in soil (Fig. 3b, e) for d 13 CH 4 (initial) , and d 13 C-values of emitted CH 4 (Fig. 1b) minus -13.0% for both treatments in 2008 but (Fig. 1f) minus -13.3% for flooding and -12.8% for drainage in 2009 for d 13 CH 4 (final) . F ox in (c) and (d) was calculated in 2009 with Eq. (8) using 1.038 for a ox , d 13 C-values of CH 4 anaerobically produced in soil (Fig.  3e) and on roots (Fig. 4b) for d 13 CH 4 (initial) , and d 13 C-values of CH 4 aerobically produced in soil (Fig. 5b) and on roots (Fig. 5d) for d 13 CH 4 (final) . TS, BS, FS and RS represent tillering, booting, grain-filling and ripening stages, respectively. Bars represent standard errors (n = 3). doi:10.1371/journal.pone.0073982.g009 with flooding, drainage generally caused the produced CH 4 depleted in 13 C. Although drainage significantly decreased CH 4 emission, it had little effect on d 13 C-value of emitted CH 4 , as well as the transport fractionation factor e transport . Acetate-dependent methanogenesis dominated in the soil in the late season, but H 2 / CO 2 -dependent methanogenesis occurred mostly on the rice roots over the season. Drainage decreased the contribution of acetate to CH 4 production by 5-10%. In field conditions, ,10-90% of the CH 4 was oxidized in the rhizosphere, while ,30-70% at the soilwater interface. In lab conditions, less a half of the CH 4 was oxidized in the soil, while almost all on the roots. Moreover, CH 4 oxidation was more important in the first half of the season as well as in the rhizosphere. Drainage increased the fraction of CH 4 oxidized in the rhizosphere by 5-15%, which is possibly attributed to the fact that CH 4 production decreased significantly while CH 4 oxidation did not simultaneously. Measuring d 13 C-values of the CH 4 from different pools in the rice fields is useful for quantifying the methanogenic pathway and the fraction of CH 4 oxidized in these fields. More importantly, it is useful for better understanding the processes of CH 4 emission, which may provide useful information for setting up an isotope model. Such a model may be of a great help to national or global CH 4 budget. Therefore, more attentions should be paid to the paddy fields with more different patterns of agricultural management at a larger scale. Table 3. Fraction of CH 4 oxidized (F ox ) at the soil-water interface during the 2009 rice season (mean 6 SD, n = 3).  (Fig. 2d); b d 13 C-values of CH 4 in floodwater (Fig. 2d)