Corresponding author: X. Zheng, State Key Laboratory of Atmospheric Boundary Layer Physics and Atmospheric Chemistry, Institute of Atmospheric Physics, Chinese Academy of Sciences, Beijing 100029, China. (email@example.com)
 Although synthetic nitrogen fertilizers play an important role in increasing cereal grain yields, there have been increased concerns about their intensive utilization and environmental consequences. The overall goal of this study is to gain an insight into the integrated evaluation of greenhouse gas emission and nitric oxide (NO) release and grain yield as affected by nitrogen fertilization in a subtropical rice-wheat rotation system. The assessment was based on four consecutive yearly measurements of the fluxes of methane (CH4), nitrous oxide (N2O) and ecosystem respiration (CO2), and the simultaneous observation of NO emissions in nonrice seasons under three fertilization practices (i.e., the conventional farmers’ practice with common nitrogen application rate, an alternative practice with reduced nitrogen input, and no nitrogen application as a control). Clearly, these trace gas fluxes showed largely intra-annual and interannual variations, highlighting the importance of entire year measurement for multiple years to achieve representative annual estimates. The annual mean CH4 fluxes varied from 95 kg C ha–1 (7.8 kg C t–1 grain) for the farmers’ practice to 205 kg C ha–1 (25.7 kg C t–1 grain) for the control, indicating that nitrogen fertilization inhibited CH4 emissions. Across all the years, the annual N2O emissions increased exponentially with an increasing nitrogen rate and harvested aboveground biomass. The annual N2O emission averaged 1.3–5.3 kg N ha–1(159–444 g N t–1 grain) for all treatments. The annual direct emission factors of N2O-N tended to increase with increasing nitrogen rate and averaged 0.61% and 0.85% for the alternative and farmers’ practices, respectively. Over all nonrice seasons, the seasonal mean NO emissions ranged from 0.15 to 1.4 kg N ha–1(58–253 g N t–1 grain), and were equivalent to 0.43% to 0.54% of the applied nitrogen. Averaging across the 4 years, the annual aggregate emissions of CH4 and N2O were 7.4 t CO2-eq ha–1(928 kg CO2-eq t–1grain) for the control and 5.6–5.7 t CO2-eq ha–1(468–494 kg CO2-eq t–1grain) for the fertilized treatments. Despite the comparable greenhouse effect between the alternative and farmers’ practices, reducing the common N rate by 37% resulted in decreased NO emission and increased nitrogen use efficiency, and negligible effects on economic return from grain yields.
 The exchange of carbon dioxide (CO2), methane (CH4), nitrous oxide (N2O), and nitric oxide (NO) between agricultural ecosystems and the atmosphere considerably influences global warming and atmospheric chemistry [Intergovernmental Panel on Climate Change (IPCC), 2007]. Agricultural practices have been estimated to release 5.1–6.1 Pg CO2-equivalent yr–1 and 1.6 Tg NO-N yr–1, accounting for 10–12% and 5% of the global anthropogenic greenhouse gas (GHG) and NO emissions, respectively [IPCC, 2007]. The estimates for GHG and NO emissions from agricultural fields on the regional and global scales are highly uncertain because the measurements conducted to date are insufficient for temporal and spatial representations of the climate, soil, and management conditions [Stehfest and Bouwman, 2006]. The majority of the field measurements of GHG and NO emissions were conducted in temperate regions, and there have been few flux measurements in tropical and subtropical cropping systems [Stehfest and Bouwman, 2006; Wang et al., 2011]. Stehfest and Bouwman  summarized the information from 1008 measurements of N2O fluxes in agricultural fields and indicated that 13% and 14% of the measurements were performed in tropical and subtropical climates, respectively.
 It is well known that in agricultural soils C- and N- trace gases are produced or consumed by microbial processes, but the magnitude of the exchanges between the soil and the atmosphere is largely affected by the soil temperature, inorganic N (NH4+ and NO3–) content, and moisture content [Conrad, 1996, 2002]. Agricultural practices such as nitrogen fertilization that drive soil N availability may significantly change the global atmospheric budgets of these gases [IPCC, 2007]. It has been generally recognized that the application of synthetic nitrogen fertilizers is one of the main sources of anthropogenic N2O and NO emissions into the atmosphere [Matson et al., 1998], and researchers have shown the depressive effects of this practice on CH4 consumption in upland agricultural soils [Aronson and Helliker, 2010]. Previous reports regarding the effect of N fertilization on CH4 emission from paddy fields are controversial. Some researchers observed that the addition of synthetic N fertilizers increased CH4 emission from rice fields [Lindau et al., 1991; Corton et al., 2000], although other studies reported no change or reduced CH4 emission after N fertilization [Cai et al., 2007; Xie et al., 2010; Yao et al., 2012]. In addition, several studies indicated that the correlation between nitrogen input and NO and N2O emissions may be more complicated [Ludwig et al., 2001; Mcswiney and Robertson, 2005]. For instance, N-trace gas emissions from agricultural fields exhibit a threshold response to nitrogen fertilization. At fertilization rates less than or equal to those required for crop uptake, the utilization of synthetic nitrogen fertilizers tended to create a positively linear response in NO and N2O emissions [Liu et al., 2005; Halvorson et al., 2008]. When fertilizer rates were in excess of the crop requirement, N-trace gas emissions often exponentially increased with increasing N inputs [Van Groenigen et al., 2010; Ma et al., 2010; Hoben et al., 2011]. Overall, nitrogen fertilization not only potentially regulates N2O and NO emissions and CH4 fluxes, but it also strongly affects crop productivity and CO2 emissions [Treseder, 2008]. The addition of nitrogen fertilizer can raise crop yield and biomass, which may have a positive effect on soil C sequestration as a result of the biomass input into the soil from crop residues and roots [Van Groenigen et al., 2006; Liu and Greaver, 2009]. Accordingly, a comprehensive assessment of the nitrogen fertilization effects on GHG fluxes requires considering the net exchanges of the C- and N-trace gases simultaneously because of the trade-offs among them [Liu and Greaver, 2009; Shang et al., 2011]. Meanwhile, it has been suggested that GHG emissions (or global warming potential (GWP)) should be assessed as a function of crop yield, i.e., yield-scaled GWP [Shang et al., 2011; Linquist et al., 2012].
 As reported by Heffer , more than 50% of the global synthetic nitrogen fertilizers are applied in the production of the major cereals (i.e., rice, wheat, and maize). China is a major agricultural producer, and recent studies have shown that Chinese farmers have demonstrated an excessively high utilization or overuse of synthetic nitrogen fertilizers [Ju et al., 2009]. For example, the average utilization of synthetic nitrogen fertilizers in China was in excess of 200 kg N ha–1 in 2000, which was much higher than the normal use in India (under 100 kg N ha–1) and in developed countries (approximately 120 kg N ha–1) [Huang et al., 2012]. The widespread overuse of synthetic nitrogen fertilizers has resulted in large nitrogen losses in the form of N-trace gases (e.g., N2O and NO) and low nitrogen use efficiency (NUE) [Matson et al., 1998; Ju et al., 2009]. Therefore, improving nitrogen fertilizer management in high-intensity agricultural systems is of great importance in the mitigation of C- and N-trace gas fluxes and in crop production sustainability. In the Chinese subtropics, an annual rice-winter wheat rotation cropping system is used in approximately 13 million hectares, which provides approximately 63% of the total cereal grain yields (http://www.stats.gov.cn/english/statisticaldata/) [China Statistical Yearbook, 2010]. However, the majority of previous studies on GHG fluxes from paddy fields under N fertilization management were only performed during the rice growing period [e.g., Cai et al., 1997; Zou et al., 2009; Xie et al., 2010], and few year-long measurements considered the entire rice-winter wheat rotation cycle. In a one-year study conducted in Nanjing, Zou et al.  found pronounced linear relationships between the N2O emissions and the synthetic N application rate during the rice and winter wheat seasons and in the annual rotation cycle. The emission factors of N2O-N (i.e., the percentage of nitrogen fertilizer lost as N2O-N emission in the current season or year) were 1.08%, 1.49%, and 1.26% for the rice season, winter wheat season, and annual rotation cycle, respectively. The linear correlations between N2O emissions and the synthetic N application rate were also observed in a single-year field measurement on a rice-wheat rotation cropping system under intermittent irrigation management, with an annual emission factor of 0.85% [Liu et al., 2010]. Clearly, integrated evaluations of N fertilization management alternatives in terms of their performance in decreasing C- and N-trace gas fluxes while remaining feasible for sustaining grain yields are considerably lacking for the rice-winter wheat rotation cropping system.
 In this study, we present the results of a 4 year field measurement in which CH4, N2O, CO2, and NO fluxes and crop productivity were observed simultaneously in a subtropical rice-winter wheat rotation cropping system under three synthetic N fertilization rates. The main objectives of this study were to (a) characterize the responses of C- and N-trace gas fluxes to fertilization rates, with fluxes expressed on both an area- and yield-scaled bases; (b) determine the emission factors of N2O and NO for different fertilization rates and cropping years; and (c) assess the efficacy of using reduced N fertilization to minimize the net GHG emission while sustaining grain yield or improving NUE.
2 Materials and Methods
2.1 Study Site and Experimental Design
 The field measurements were conducted in paddy soils (32°35′5″N, 119°42′0″E) that were located in the north of Jiangsu province, China. This region is a typical rice-winter wheat cropping area, which has a northern subtropical monsoon climate. The mean annual temperature and mean annual rainfall are approximately 15.9°C and 924 mm, respectively. The soil is classified as a fluvisol (http://www.fao.org/ag/Agl/agll/wrb/doc/wrb2006final.pdf) [IUSS Working Group WRB, 2006] with sandy loam texture (58% sand and 14% clay) and a pH of 8.0 in the upper 15 cm. The soil organic carbon and nitrogen content in the cultivated layer averaged 18.4 and 1.45 g kg–1, respectively.
 The experiment was performed over the course of four consecutive rice-wheat rotation cycles from June 2004 to June 2008. The study consisted of three N fertilizer application rates, arranged in a randomized complete block design with three replications in each year (see Table S1 in the auxiliary material). These application rates included the following: the conventional farmers’ practice (FP) for the region, as determined via farm survey; the alternative practice (AP), which was in accordance with agronomists’ recommendations and applied less N fertilizer in the rotation cycle; and a control (CK) that received no N fertilizer. In the local farmers’ practice, 250 and 225 kg N ha–1 of compound fertilizer and urea were applied in the rice- and wheat-growing seasons, respectively. Compound fertilizer combined with urea was applied with three splits for the rice and wheat growing season as follows: 36% as basal fertilizer, 24% as tillering fertilizer, and 40% as panicle fertilizer in the rice season, and 50% as basal fertilizer, 10% as jointing fertilizer, and 40% as panicle fertilizer in the wheat season (Table S1). In the alternative practice, 150 kg N ha–1 of compound fertilizer and urea was applied in the rice- and wheat-growing seasons, consistent with the recommended rates for this region [Zhu and Chen, 2002]. Similar to the farmers’ practice, split applications of N fertilizer were adopted in the alternative practice, with two splits for the rice season (60% as basal fertilizer and 40% as panicle fertilizer) and three splits for the wheat season (50% as basal fertilizer, 10% as jointing fertilizer, and 40% as panicle fertilizer). In addition, based on the local convention, phosphorous and potassium fertilizers were applied at equal rates of 70 and 75 kg ha–1 as basal fertilization for all treatments during the rice and wheat growing periods, respectively.
 From 2004 to 2008, rice seedlings were usually transplanted in mid-June and harvested in middle or late October. In the following season, winter wheat is usually sown in early November and harvested in early June of the following year (Table S1). The harvested crop residues were fully removed from the field for other uses, and the soil was plowed prior to rice transplanting and wheat sowing. In agreement with the common water practices in this region and in other areas of China, intermittent irrigation with midseason aeration (i.e., flooding-drainage-reflooding-moist) was adopted in all field plots during the rice period, while no artificial irrigation occurred in the following wheat season [Yao et al., 2010].
2.2 Carbon- and Nitrogen-Trace Gas Flux Measurements
 The N2O and CH4 fluxes and the ecosystem respiration (CO2) were measured simultaneously over the entire rotation cycles using static opaque manual gas sampling chambers, as described by Yao et al. . Based on our previous results, which showed that up to 97% of the total NO emissions from rice-wheat rotation systems were released in the upland wheat season [Zhou et al., 2007], the NO fluxes in this study were monitored only during the nonrice period (i.e., the fallow and wheat growing season) using the same method as in the GHG measurements [Mei et al., 2009]. In the center of each experimental plot, a rectangular stainless steel frame (50, 50, and 15 cm in length, width and height, respectively) was inserted into the soil and maintained in place throughout the entire rotation cycle, except when it was removed for necessary farming practices (e.g., tillage). At the time of the gas sampling, the frame was covered by a vented insulated stainless steel chamber with a bottom area of 0.25 m2 and a height of 50 or 100 cm depending on crop growth, and gas samples were taken with a 60 mL polypropylene syringe at regular intervals (0, 8, 16, 24, and 32 min after covering). The gas concentrations in the samples were analyzed within 4 h using a gas chromatography instrument (Agilent 4890D, Agilent Technologies, Palo Alto, CA, USA) equipped with an electron capture detector for N2O detection and a flame ionization detector for CH4 and CO2 detection (a nickel catalyst applied for converting CO2 to CH4). The DN-Ascarite and DN-CO2 methods described by Zheng et al.  were used for the N2O analysis. Additional details for the instrumental configurations for analyzing the gases were as described by Zheng et al. . The fluxes were calculated from the rate of change in the gas concentrations in the enclosed chamber headspace with time [Hutchinson and Livingston, 1993].
 The NO fluxes were determined by sampling twice from the chamber enclosure that was used for the GHG measurements [Mei et al., 2009]. One sample was taken using a KNF/N86KNDC pump (Neuberger Inc., Germany) at the beginning of the enclosed-chamber period, and the other was taken at the end of this period, and the samples were injected into an evacuated bag made of inert aluminum-coated plastic. Within 1 h of collection, the NO in the sampling bags was measured using a TE-42C chemiluminescent NO-NO2-NOx analyzer (Thermo Environmental Instruments Inc., USA). The flux was calculated from the concentration difference between the two samples collected at the beginning and end of the chamber enclosure period. NO flux was corrected for the headspace temperature and ambient air pressure effects [Zheng et al., 2003]. The NOx analyzer was calibrated prior to and at the end of each growing season using a TE-146i dilution-titration instrument (dynamic gas calibrator).
 In addition, soil CO2 emission was measured using 15 cm × 20 cm × 20 cm (length × width × height) polyvinyl chloride chambers from November 2004 to August 2005. The chamber frames were installed between the rows of the plants. The gas samples for CO2 detection were drawn using 20 mL syringes at 0, 5, 10, 15, and 20 min, respectively. The detection and calculation methods were the same as described above for the ecosystem respiration measurements.
 Generally, the C- and N-trace gas fluxes were observed between 09:00 A.M. and 11:00 A.M. because the soil temperature during this time was close to the daily mean soil temperature [Yao et al., 2009]. The gas samples were usually taken approximately two to three times per week at intervals of 2 to 3 days, while measurements were conducted every day for approximately one week during the expected high emission periods, e.g., in the midseason aeration stage during the rice season and following fertilization and substantial precipitation during the nonrice season.
2.3 Auxiliary Measurements
 The soil temperature at a depth of 5 cm was measured once a day between 09:00 A.M. and 11:00 A.M. using a manual thermocouple thermometer (JM624, Tianjin Jinming Instrument Co. Ltd., China). The air temperature and daily precipitation were recorded using an on-site automatic meteorological station. The field floodwater depth in the rice season was monitored daily using a stainless steel ruler, and the soil (0–8 cm) moisture in the nonrice season was measured daily adjacent to the frames using a portable frequency domain reflector probe (RDS Technology Co., Ltd Jiangsu, Nanjing, China). To determine the inorganic N (NH4+ and NO3–) content in the soil, soil samples were taken from three points per experimental plot at depths of 0 to10 cm at weekly intervals using a 3 cm diameter gauge auger. Following the sampling, the samples were bulked for each treatment, and the inorganic N content of the field moist soil samples was measured using 0.01 mol L–1 CaCl2 extraction (soil:water = 1:10) and colorimetric techniques [Yao et al., 2010]. The soil inorganic N measurements were conducted from 10 June 2005 to 25 October 2006 due to a shortage of labor. In addition, after the physiological maturity of the rice and wheat, the aboveground biomass, including straw and grain, was harvested manually from five subplots with 0.36 m2 in each experimental plot. The harvested aboveground biomass was oven-dried for 3 days at 80°C and weighed to obtain the dry straw and grain yields. The straw and grain samples were ground and analyzed using the potassium dichromate-volumetric method and the semi-micro Kjeldahl method for total C and N content, respectively [Liu et al., 2012]. The yields of straw-C/N, grain-C/N, and the total aboveground C/N (straw plus grain) were calculated by multiplying the straw and grain C/N concentrations by the dry straw and grain yields, respectively.
2.4 Data Processing and Statistical Analysis
 The fluxes of N2O, NO, and CH4 and the ecosystem respiration (CO2) for each treatment and sampling date were determined as the mean of the three fluxes from the three spatial replications. The daily C- and N-trace gas fluxes for the dates between samplings were estimated using linear interpolation, and the total cumulative emissions were calculated as the sum of all daily fluxes for the rice season, nonrice season, and entire rotation cycle for each year. To more effectively evaluate the effects of the fertilizer practices on the C- and N-trace gas fluxes, the seasonal and annual cumulative emissions were expressed on both an area- and grain yield-scaled bases. The CO2 equivalents (CO2-eq) of seasonal and annual CH4 and N2O fluxes were calculated using GWP indices of 25 and 298, respectively, over a 100 year time horizon [IPCC, 2007]. The fertilizer-induced direct emission factors (EFds) for N2O-N and NO-N emissions were calculated by subtracting the total cumulative emissions of N2O and NO in the control treatment from the corresponding cumulative emissions in each fertilized treatment and dividing the result by the fertilizer application rate. The fertilizer NUE was calculated by subtracting the aboveground N yield in the control treatment from the aboveground N yield in each fertilized treatment and dividing the result by the applied total amount of N fertilizer.
 The statistical analyses were conducted using SPSS 12.0 (SPSS Inc., Chicago, IL, USA) and Origin 7.0 (Origin Lab Corporation, USA). The differences in the total amounts of N2O, CH4, CO2, and NO fluxes from the rice-winter wheat systems from 2004 to 2008, as affected by the N fertilizer treatment, year, and their interaction, were examined using a two-way analysis of variance. The differences in the grain yields and total amounts of the C- and N-trace gas fluxes among the three treatments during the experimental period were examined by a one-way ANOVA with Tukey's multiple range test. The relationships between the NO, N2O, CH4, and CO2 fluxes and the environmental variables (e.g., the straw and grain yields, inorganic N content, and fertilizer rate) were evaluated using a linear or nonlinear regression process.
3.1 General Climatic Variables
 The annual mean air temperatures were 15.6, 15.7, 16.8, and 15.5°C, respectively, for the 2004/2005, 2005/2006, 2006/2007, and 2007/2008 cropping years (Figure 1). The amount of rainfall varied significantly intra-annually and interannually. The annual precipitation was lower in the 2004/2005 (659 mm) and 2007/2008 (748 mm) cropping years, higher in 2005/2006 (1139 mm) and comparable in 2006/2007 (980 mm), compared to the multiyear average (924 mm) for this site. During the rice-growing season (i.e., June to October), the precipitation is usually sufficient and 49–70% of the annual rainfall occurred during this period in the studied years.
 The soil (5 cm) temperature showed a temporal pattern similar to that of the air temperature in the four consecutive rice-wheat rotation cycles (Figure 1). The annual mean soil temperatures were 14.8, 15.0, 15.1, and 14.5 for the 2004/2005, 2005/2006, 2006/2007, and 2007/2008 cropping years, respectively, which was lower than the corresponding average of the air temperature. The soil moisture content, expressed as WFPS (water-filled pore space) ranged from 35% to 100% WFPS during the nonrice seasons. As a result of the substantially varied rainfall distribution in different years, the soil WFPS showed significantly different seasonal patterns during the nonrice seasons because it was regulated by precipitation. During the measurement period, there were no significant differences in the soil temperature and moisture among the different fertilizer practices.
3.2 Soil Inorganic N Content and Agronomic Variables
 The seasonal variations in the soil inorganic N (NH4+ and NO3–) content over the rice-wheat rotation cycle were primarily regulated by the N application (Figures 2a and 2b). The relatively high content of NH4+ and NO3– was primarily observed within 10 days after fertilization. The mean NH4+ contents for the control, alternative and farmers’ practices were 6.1, 8.4, and 11.7 mg N kg–1 soil dry weight (SDW), respectively, during the two rice seasons and 5.3, 5.8, and 8.2 mg N kg–1SDW during the nonrice season, respectively. A reverse pattern was observed for the mean NO3– content, which was relatively lower in the rice season (2.2, 2.4, and 3.9 mg N kg–1 SDW for the control, alternative and farmers’ practices, respectively) than in the nonrice season (4.6, 10.6, and 11.8 mg N kg–1 SDW for the control, alternative and farmers’ practices, respectively). During the entire rotation cycles, the annual mean inorganic N contents were 9.3, 14.2, and 18.3 mg N kg–1SDW for the control, alternative and farmers’ practices, respectively, indicating that the inorganic N contents increased with an increasing N rate.
 Straw and grain yields in the fertilized plots (the alternative and farmers’ practices) were significantly higher compared to the control, but there were no significant difference between the alternative and farmers’ practices (Table 1). Two-way ANOVA analyses showed that the grain yields of rice varied significantly with the fertilizer and the year but were independent of the interaction of these factors. The grain yields of wheat were significantly affected by the fertilizer and year and their interaction (Table S2). The grain yields for the control, alternative and farmers’ practices ranged from 4.8 to 7.3 t ha–1 for rice and from 2.0 to 6.1 t ha–1 for wheat across the 4 years. For the N concentrations of straw and grain, the farmers’ practice treatment did not differ from the alternative practice, and the control had lower values than the two fertilized treatments. For the C concentrations of straw and grain, there were no significant differences among the treatments (Table 1). The fertilizer NUE for rice ranged from 34% to 56% (mean: 41%) in the alternative practice and from 27% to 42% (mean: 36%) in the farmers’ practice, indicating that the reduced N input improved the NUE of rice. For the wheat seasons, the average NUE was 60% for the alternative practice, which was higher than that observed for the farmers’ practice (48%). The annual values for the NUE in the alternative and farmers’ practices were 51% and 42%, respectively, when averaged across the 4 year rotation cycles.
Table 1. The Characteristics in Straw and Grain Yields (Mean ± Standard Error, in t ha–1) and Their Carbon and Nitrogen Concentrations (in %) at Physiological Maturity, and the Estimated Nitrogen Use Efficiency (in %) Under Different Fertilizer Practices During the Rice and Wheat Seasons of 2004–2008
CK, no nitrogen fertilizer; AP, the alternative practice that applied 150 kg N ha–1 for both the rice and wheat seasons; FP, the farmers’ practice that applied 250 and 225 kg N ha–1 for the rice and wheat season, respectively.
SY, straw yield; TC, total carbon concentration; TN, total nitrogen concentration; GY, grain yield; NUE, nitrogen use efficiency. Different letters within the same column indicate statistically significant differences among treatments in each season.
 In the rice seasons, the N2O fluxes were stimulated by each of the fertilizer applications and then gradually declined. In addition, another pronounced flux appeared at the end of each rice season, with the exception of the 2004 rice season, due to drainage (Figures 3a–3d). Statistical analyses indicated that the total cumulative N2O emissions over the rice seasons were greatly affected by fertilizer treatment but were independent of the year and the interaction of these factors (Tables 2 and S2). Generally, seasonal N2O emissions increased with increasing nitrogen rates for each rice season. Over the four rice seasons, area-scaled seasonal N2O emissions were an average of 211% higher in the alternative practice than the control (P < 0.05), and they were approximately 95% higher in the farmers’ practice than in the alternative practice (P < 0.05) (Table 2). When expressed relative to rice grain yield, the yield-scaled N2O emissions, on average, showed the same trend as the area-scaled emissions except the farmers’ practice treatment did not significantly differ from the alternative practice (Table 3). For all rice seasons, the calculated direct emission factors of N2O-N (EFds) ranged from 0.21% to 1.00% and from 0.30% to 1.22%, with mean values of 0.53% and 0.76% for the alternative and farmers’ practices, respectively (Table 2).
Table 2. Seasonal Cumulative Fluxes of Methane (CH4, in kg C ha–1), Ecosystem Respiration (CO2, in t C ha–1), Nitrous Oxide (N2O, in kg N ha–1) and Nitric Oxide (NO, in kg N ha–1), and Direct Emission Factors of N2O (, in %) and NO (EFd − NO, in %) Under Different Fertilizer Practices During the Rice Growing and Nonrice Seasons of 2004–2008
Table 3. Yield-Scaled Seasonal Fluxes of Methane (CH4, in kg C t–1), Ecosystem Respiration (CO2, in t C t–1), Nitrous Oxide (N2O, in g N t–1) and Nitric Oxide (NO, in g N t–1) Under Different Fertilizer Practices During the Rice Growing and Nonrice Seasons of 2004–2008
 Similar to the rice season, the nitrogen fertilizer events applied in the nonrice season triggered the emission of large amounts of N2O but no large increase in emission occurred following the first topdressing (i.e., the jointing fertilizer application). For the NO fluxes, their seasonal variations showed a similar pattern to those of N2O (Figure 3). The cumulative N2O emissions over the nonrice season varied significantly with fertilizer treatment and year, but the totals were not significantly affected by the interaction of these factors. In contrast, the seasonal NO emissions were significantly affected by the interactions between the fertilizer and the year, apart from by the fertilizer and year separately (Tables 2 and S2). Compared to the control, the application of nitrogen fertilizer in the alternative and farmers’ practices significantly increased N2O and NO emissions. In the alternative and farmers’ practices, there was a linear relationship between the NO emissions and soil inorganic N (NH4++NO3–) content during the nonrice season (R2 = 0.36–0.67, P < 0.01, Figure 2d). However, across all the treatments, NO emissions increased exponentially with increasing soil inorganic N contents (R2 = 0.71, P < 0.01, Figure 2d). Meanwhile, across all nonrice seasons, seasonal NO emissions showed an exponential response to both the nitrogen rate and the aboveground biomass (straw plus grain) (Figures 4c and S1). Over the four nonrice seasons, the area-scaled seasonal N2O emissions were increased an average of 1.2 and 2.4 times higher in the alternative and farmers’ practices, respectively, compared to the control (P < 0.05). However, there was no significant difference in the yield-scaled N2O emissions among the treatments (Table 4). With respect to the averaged NO emissions over all of the nonrice seasons, as expressed either on an area basis or relative to wheat yield, the alternative and farmers’ practices showed dramatically higher emissions than did the control. However, the alternative and farmers’ practices did not significantly differ from each other (Tables 2 and 3). Similar to the rice season, the EFds of N2O-N and NO-N in the nonrice seasons showed temporal variations for different years. The mean EFds for all nonrice seasons were 0.72% and 0.97% for N2O-N in the alternative and farmers’ practices, respectively, and 0.43% and 0.54% for NO-N in the alternative and farmers’ practices, respectively (Table 2).
Table 4. Annual Cumulative Emissions of Methane (CH4), Ecosystem Respiration (CO2) and Nitrous Oxide (N2O) Expressed on Both an Area- and Yield-Scaled Basis, and Annual Direct Emission Factors of N2O (), Grain Yields (GY) and Nitrogen Use Efficiency (NUE) Under Different Fertilizer Practices During the Rice-Wheat Rotation Cycles of 2004–2008
 Over the four consecutive annual rotation cycles, the cumulative N2O emissions ranged from 1.12 to 1.63 kg N ha–1 for the control, from 2.21 to 4.98 kg N ha–1 for the alternative practice, and from 3.65 to 7.63 kg N ha–1 for the farmers’ practice. The annual N2O emissions were significantly affected by fertilizer and year and by the interaction of these factors (Tables 4 and S2). Compared to the control, the 4 year mean annual N2O emissions were significantly increased by 141% and 308% for the alternative and farmers’ practices, respectively (P < 0.05). When expressed relative to grain yield, the yield-scaled N2O emissions were consistently lower for the control than for the farmers’ practice (P < 0.05), but there was no significant difference between the control and alternative practice (Table 3). Across the sampling dates from June 2005 to October 2006, during which the N2O fluxes and soil inorganic N content were measured simultaneously, a positive linear correlation was found between the N2O emissions and the total inorganic N (NH4++NO3–) content in the alternative and farmers’ practices (R2 = 0.18–0.55, P < 0.01, Figure 2c). However, across all the treatments, a nonlinear response curve best described the increases in the N2O emissions with increasing soil inorganic N content (R2 = 0.62, P < 0.01, Figure 2c). In addition, for the data measured in different time scales, the exponential model best described the correlation between the cumulative N2O emissions and the nitrogen rate or the aboveground biomass (straw plus grain) for the rice season, nonrice season and annual rotation cycle (Figures 4 and S1). For the annual rotation cycles, the EFds of N2O-N were estimated to be an average of 0.61% and 0.85% as a result of the fertilizer management of the alternative and farmers’ practices, respectively.
3.4 Methane Fluxes and Ecosystem Respiration
 During the rice growing periods, the CH4 fluxes increased gradually in the early stage and decreased sharply in the midseason aeration stage. The amplitudes of the CH4 fluxes were then primarily regulated by dry and wet soil conditions, i.e., water regimes (Figures 1 and 5e–5h). The three fertilization treatments (CK, AP, and FP) had similar seasonal patterns but varied in the magnitudes of the CH4 fluxes for each of the rice seasons. The seasonal cumulative CH4 emissions depended greatly on the fertilizer addition, year, and the interaction of these factors (Tables 2 and S2). Compared to the control, the alternative practice tended to decrease the seasonal CH4 emissions, although this influence was not statistically significant. However, the farmers’ practice significantly decreased the seasonal CH4 emissions by an average of 54% compared to the control (Table 2). Across all rice seasons, the trend and magnitude of the fertilization effects on the yield-scaled CH4 emissions relative to the control were similar to their effects on the area-scaled CH4 (Tables 2 and 3).
 In the nonrice seasons, the CH4 fluxes out of or into the soil were often sporadic. There were no regular and consistent seasonal patterns for the field treatments (Figure S2). For the nonrice seasons, all of the treatments acted as small net sinks of atmospheric CH4. The seasonal CH4 uptake ranged from 0.63 kg C ha–1 for the control in the 2004/2005 nonrice season to 1.99 kg C ha–1 for the farmers’ practice in the 2007/2008 nonrice season. The 4 year mean uptake was between 1.16 and 1.25 kg C ha–1, with no significant difference observed among the treatments (Table 2). Obviously, substantial CH4 emission was observed during the rice season, which was significantly greater than the uptake in the nonrice season. Therefore, all of the treatments acted as a net source of atmospheric CH4 over the time period of one year. The statistical analyses showed that annual CH4 emissions varied significantly with the fertilizer and year (Tables 4 and S2). Compared to the control, there was no significant influence of N fertilizer on the mean annual CH4 emissions in either the alternative or the farmers’ practices, although the mean annual CH4 emissions tended to be lower by 38–54%. However, when expressed relative to grain yield, the average annual CH4 emission across the 4 year rotation cycles was significantly inhibited by 70% in the farmers’ practice compared to the control (Table 4).
 In this study, the CO2 emission was measured at the ecosystem level (including soil and plants) using a dark static chamber. There were similar seasonal patterns of ecosystem respiration in all treatments during the rice-growing and nonrice seasons (Figures 5a–5d). During the rice growing periods, the ecosystem respiration gradually increased following rice transplanting and reached a peak in mid-August with the maximum biomass of plants. In contrast, the CO2 emissions decreased gradually from the beginning of the nonrice season to mid-February because of lower temperatures. As the wheat turned green and the temperature increased, the ecosystem respiration slowly reached its peak in mid-May (Figures 1 and 5). The amounts of CO2 emitted over the rice or nonrice seasons were significantly influenced by the application of nitrogen fertilizer (Table S2). The increased fertilizer application rate (i.e., in the alternative and farmers’ practices) stimulated the ecosystem respiration both in the rice and in the nonrice seasons (Table 2). When the amounts of ecosystem respiration from the rice and nonrice seasons were combined, the annual CO2 emissions ranged from an average of 11.25 to 15.11 t C ha–1 over the four annual rotation cycles. Across the different fertilizer rates and crop years, a linear response curve best described the correlations between the total ecosystem respiration and aboveground biomass (straw plus yield) at the seasonal and annual scales (R2 = 0.49–0.65, P < 0.01). Relative to the control, the mean annual ecosystem respiration was increased by 28% and 34% for the alternative and farmers’ practices, respectively (Table 4). When expressed based on grain yields, the total CO2 emissions did not differ significantly between the treatments for the rice and nonrice seasons or for the annual cycles (Tables 3 and 4).
3.5 Seasonal and Annual Aggregate Emissions of CH4 and N2O
 As shown in Figure 6, the aggregate emissions of CH4 and N2O during the rice seasons ranged from 3.50 t CO2-eq ha–1 in the farmers’ practice of the 2005 rice season to 12.08 t CO2-eq ha–1 in the control of the 2004 rice season, with a mean value of 5.38 t CO2-eq ha–1. Over the four rice seasons, the averaged aggregate emissions of CH4 and N2O ranged from 4.26 to 7.05 t CO2-eq ha–1, and nitrogen fertilization tended to decrease the net GHG emissions. Similarly, the application of N fertilizer showed inhibitory effects on the mean yield-scaled aggregate emissions of CH4 and N2O, which ranged from 649 to 1290 kg CO2-eq t–1.
 During the nonrice periods, the aggregate emissions of CH4 and N2O were between 0.33 and 1.94 t CO2-eq ha–1 and averaged to 0.89 t CO2-eq ha–1. Among the treatments, the alternative and farmers’ practices significantly increased the GHG emissions by 131% and 264% on average, respectively, compared to the control (P < 0.05). However, there were no significant differences in the yield-scaled aggregate emissions of CH4 and N2O (146–271 kg CO2-eq t–1) for any treatments (Figure 6).
 The aggregate emissions of CH4 and N2O for the rice-growing seasons, expressed on area and grain yield bases, were approximately 5.0 and 3.6 times higher than those of the nonrice seasons. During the rice seasons, the GHG emissions were largely regulated by the CH4 emissions, which, on average, accounted for 97%, 88%, and 71% of the aggregate emissions for the control, alternative and farmers’ practices, respectively. In contrast, the N2O emissions substantially influenced the combined climatic impact for the nonrice seasons, which contributed approximately 103–111% of the aggregate emissions for all treatments. Averaged over the 4 years, the ranges of the annual aggregate emissions of CH4 and N2O for all treatments were 5.66–7.43 t CO2-eq ha–1 when expressed on an area basis and 468–928 kg CO2-eq t–1 when expressed on a grain yield basis (Figure 6). Although the application of N fertilizer increased the aggregate emissions of CH4 and N2O during the nonrice periods, the annual aggregate emissions tended to be higher for the control than for the fertilized treatments (i.e., the alternative and farmers’ practices) due to the substantial decreases in the CH4 emissions from the fertilized treatments in the rice seasons.
 It is noteworthy that in this study the exchanges of N2O, NO, CH4, and CO2 between the rice-wheat rotation agroecosystem and the atmosphere were measured using static manual chambers with the appropriate sampling schedule (i.e., preferably sampling daily for several days after fertilization, during a midseason aeration stage in the rice season, and after substantial precipitation in the nonrice season and otherwise sampling at progressively wider intervals). This sampling method provides reliable results in terms of coping with the temporal variability of these gases, but it introduces an extent of uncertainty into the flux estimate, e.g., an approximately 18% overestimation for the N2O measurements [Yao et al., 2009].
4.1 Nitrogen Fertilization Affecting GHG and NO Fluxes
 Our findings corroborate the argument that the application of synthetic N fertilizers strongly affects the emissions of environmentally important trace gases, such as CH4, CO2, N2O, and NO from agricultural fields [Chu et al., 2007; IPCC, 2007]. Over the entire annual rotation cycle, the rice-wheat systems are sources for atmospheric CH4 because the total CH4 emissions in the rice seasons were only marginally offset by CH4 uptake in the nonflooded wheat seasons. Nonetheless, our results indicate that the application of synthetic N fertilizers caused a general suppression of annual CH4 emissions across multiyear measurements, irrespective of their expression on either an area- or yield-scaled bases (Table 4). Numerous studies have demonstrated that CH4 production, oxidation and transport are affected by synthetic nitrogen fertilizers. However, the magnitude and even the direction of these responses varied among field studies, which observed positive, negative and no influences of nitrogen inputs on CH4 emissions [Bodelier and Laanbroek, 2004; Cai et al., 2007]. In the majority of ecosystems, CH4 production and consumption occur simultaneously [Liu and Greaver, 2009], and so the decrease in annual CH4 emissions under the fertilized treatments was presumably caused by the activities of methanogenic archaea and methanotrophic bacteria. In rice paddy fields, the utilization of synthetic N fertilizers, especially ammonium-based fertilizers, has been shown to improve the growth and activities of methane-oxidizing bacteria, especially in the rhizosphere of rice [Bodelier et al., 2000], thus resulting in the oxidation of more CH4. For our studied alluvial soil with sandy loam texture, probably CH4 consumption under urea-based fertilization was further stimulated by partially aerobic soil conditions due to the porous and highly percolating nature of the soil. Meanwhile, the toxicity of nitrite (NO2–), NO and N2O produced in nitrification or denitrification processes for methanogenic archaea may participate in the suppression of CH4 production by nitrogen fertilization [Klüber and Conrad, 1998], thus reducing the amount of CH4 emitted into the atmosphere.
 With respect to CO2 emission, the application of synthetic N fertilizers significantly increased the annual ecosystem respiration (Table 4). This result is in agreement with previous studies performed in freshwater marshes [Zhang et al., 2007] and planted microcosms [Inselsbacher et al., 2011]. Ecosystem respiration was strongly dependent on plant respiration and microbial decomposition, and N fertilizer plays an important role in plant growth and soil biological reactions, including heterotrophic microorganisms and plant roots [Snyder et al., 2009]. Accordingly, as was also observed in our study (Figure S3), nitrogen fertilization generally enhances soil respiration by stimulating plant root respiration, including microbial respiration in the rhizosphere, as a result of root growth and the consequent effect on root exudates [Dick, 1992]. Meanwhile, we hypothesized that the CO2 emissions from the aboveground living biomass in the fertilized treatments should increase due to a positively significant effect of nitrogen fertilization on crop biomass. As reported in previous studies [e.g., Ding et al., 2007], we observed significant correlations between ecosystem respiration and harvested aboveground biomass. This finding indicates that the aboveground biomass was an important factor affecting CO2 emissions, which supports our hypothesis. However, there were no differences in yield-scaled CO2 emissions between the control and fertilized treatments (Tables 3 and 4), indicating that at these nitrogen application rates, increases in grain yields were at least proportional to increases in CO2 emissions. This result suggests that for a cropping system, probably ecosystem respiration could be used as an index for crop productivity. This observation could be helpful for model testing and validation.
 For the N-containing trace gases, the area- and yield-scaled emissions of NO and N2O from the control were significantly lower than those from the fertilized treatments, and the cumulative emissions increased with increasing amounts of applied N fertilizers (Tables 2 and 4). Previous studies have consistently revealed that nitrogen fertilization stimulates NO and N2O emissions from soils [Matson et al., 1998; Bouwman et al., 2002; Liu et al., 2005; Chu et al., 2007]. However, our study clearly demonstrates that there was a pronounced interannual variability in the cumulative emission of NO and N2O. This result suggests that precipitation, particularly rainfall distribution throughout the experimental period and the amount of each individual rainfall event associated with fertilizer application, plays a key role in influencing the NO and N2O emissions because the soil field management and soil temperature were comparable across the different cropping years. For example, the cumulative rainfall over a period of one week following the third fertilizer application (i.e., panicle fertilization) in the nonrice seasons was between 15.3 and 39.6 mm for the 2004/2005, 2005/2006, and 2006/2007 cropping years, but it was only 2.2 mm for 2007/2008. Hence, the elevated NO and N2O emissions observed in mid-April of 2004/2005, 2005/2006, and 2006/2007 due to nitrogen topdressing were coupled with rainfall events, while no obvious peak in emissions occurred in mid-April of 2007/08 (Figures 1 and 3). Although the impact of the variable weather conditions among the cropping years, our results show significantly positive correlations between emissions and soil inorganic N (NH4++NO3–) content, which is heavily affected by the application rate of synthetic N fertilizers (Figure 2). Meanwhile, across all the fertilizer rates and cropping years, it was clear that the cumulative emission of NO and N2O increased exponentially with an increasing application rate of synthetic N fertilizers (Figure S1). Similarly, Harrison et al.  reported that NO and N2O emissions increased nonlinearly with increasing nitrogen rates in agricultural fields of England. In contrast, Liu et al.  observed that NO and N2O emissions increased linearly with increasing nitrogen application rates in irrigated cropping systems of Colorado. As suggested in numerous earlier studies [e.g., Grant et al., 2006; Van Groenigen et al., 2010], a nonlinear response of N2O emissions to N rates generally occurs once N inputs exceed soil and crop ecosystem uptake capacities. Nonetheless, each agroecosystem and specific growing season usually differs in the amount of nitrogen that is appropriate for ecosystem N uptake capacity, due to the influences of site-specific factors [Snyder et al., 2009]. Given the diversities of the agroecosystems, soil types, management, and weather conditions, therefore, we assumed that nitrogen fertilization interacts with other site-specific factors to regulate the processes of NO and N2O emissions, thereby producing different results for the responses of N-trace gases to nitrogen rates. In addition, in this study we observed significant nonlinear relationships between N-trace gas emissions and harvested aboveground straw and yield (Figure 4), which is comparable to previous studies on winter wheat [Chen et al., 2008; Liu et al., 2012], continuous maize [Mcswiney and Robertson, 2005], and spring barley [Abdalla et al., 2010] cropping systems. This result indicates that N2O emissions from a cropping system might be predicted by crop productivity.
4.2 Direct Emission Factors and Background Emissions of N2O and NO
 In this study, the direct emission factors of N2O and NO tended to increase with increasing synthetic N rates (Tables 2 and 4). This finding is in accordance with reports from Mcswiney and Robertson  and Hoben et al. , but is inconsistent with other studies [Liu et al., 2005; IPCC, 2006] in which the EFds of N2O and NO were constant for different N application rates. Presumably, the differences in responses of the N2O and NO emission factors to nitrogen fertilization among these studies result from differences in climates (e.g., precipitation), soil, and management conditions and in the period and frequency of field measurements [Bouwman et al., 2002; Laville et al., 2011]. In addition, our multiyear measurements of the N2O and NO emission factors showed that there are large interannual variations for the same fertilizer practices, indicating that multiyear studies are required to achieve temporally representative estimates of N-trace gas emissions. Likewise, IPCC  and Wang et al.  suggested that to reduce the uncertainties of N2O emission inventories, the emission factors of N2O should be based on long-term measurements that could fully reflect the variability of the underlying biogeochemical processes.
 The mean seasonal emission factors of N2O were calculated to be 0.53–0.76% and 0.72–0.97% for the rice and nonrice seasons, respectively. These results are comparable to previous estimates for the rice paddies [Zou et al., 2007] and upland croplands [Zheng et al., 2004] in China. For the entire rice-wheat rotation cycles, the annual N2O emission factors in the farmers’ practice averaged to be 0.85 ± 0.18%, which is comparable to the IPCC default value of 1% for upland croplands [IPCC, 2006]. Also, Yan et al.  estimated that the average N2O emission factors were approximately 0.93% for croplands in East, Southeast, and South Asia. However, the N2O emissions from the farmers’ practice were substantially decreased in the recommended alternative practice, giving a mean annual emission factor of 0.61%. Across all of the fertilized treatments, the emission factor of N2O was estimated to be an average of 0.73% for the annual rotation cycles, which is significantly higher than the annual emission factor of 0.38% for the subtropical rice-wheat systems of the Indo-Gangetic plains of India [Pathak et al., 2002]. In contrast, the annual emission factors obtained at this study site were generally lower than those observed in other rice-wheat systems in the same delta region [Zou et al., 2005; Liu et al., 2010]. The primary factors that contributed to the lower N2O emissions from this study site were most likely the lower soil clay content (14% vs. 51–54%) and the higher soil pH (8.0 vs. 6.5–6.7) compared to the sites investigated by Zou et al.  and Liu et al. . It was reported that coarsely textured soils or soils with high pH levels usually inhibit soil N2O emissions [Bouwman et al., 2002; Stehfest and Bouwman, 2006].
 In the nonrice seasons, the seasonal NO emission factors were quantified as 0.43% and 0.54% for the alternative and farmers’ practices, respectively (Table 2). The emission factors of NO in this study fell within the range of the available worldwide observations as reported by Mei et al. . Those authors summarized NO emissions from 39 measurements conducted in nonvegetable croplands and estimated that the emission factors ranged from 0.01% to 3.21%, with a mean value of 0.48%. The NO emission factors observed in our study were consistent with the estimates for croplands in East, Southeast, and South Asia [0.48%, Yan et al., 2003] and upland agricultural fields on calcareous soils in the North China Plain [0.45%, Cui et al., 2012]. However, our seasonal NO emission factors differed from the estimates of previous studies conducted in other rice-wheat systems in this area. Zheng et al.  estimated that the emission factors of NO in the nonrice season ranged from 1.75% to 2.50%, and Zhou et al.  observed emission factors of 0.10% for NO during the nonrice period. By summarizing the results of our study and other studies, we observed that the emission factors of N2O and NO not only showed the large interannual variability, but also exhibited high site-to-site spatial variation. As a result, for the regional or national estimates, site- or region-specific emission factors should be used when possible, and estimates should not be simple extrapolations from the IPCC default values.
 Although some studies quantifying estimates of N2O and NO emissions by either considering or ignoring the background emissions have been widely adopted [Dobbie and smith, 2003; Yan et al., 2003], it is clear that the background emissions of N-trace gases are a major component for developing N2O and NO emission inventories in the agricultural sector [Zheng et al., 2004]. For example, Yan et al.  estimated that the amounts of background N2O and NO emissions accounted for 43% and 40%, respectively, of the total emissions from croplands in East, Southeast, and South Asia. In our study, the background emission of NO in the nonrice seasons ranged from 0.11 to 0.18 kg N ha–1, with a mean value of 0.15 kg N ha–1 (Table 2). Compared with the reported background emissions of 0.15–0.87 kg N ha–1 for NO during the nonrice periods in Chinese rice-winter upland crop rotation systems [Zheng et al., 2003; Yan et al., 2003; Zhou et al., 2010; Deng et al., 2012], our observed emission factors were relatively low. On an annual scale, the mean background emission of N2O was 1.31 kg N ha–1, which is consistent with background emission estimates for Chinese croplands (1.22–1.66 kg N ha–1 yr–1) that were obtained using field observations and process-based models [Li et al., 2001; Yan et al., 2003]. Similarly, Gu et al.  summarized the available data from field measurements and estimated that annual background N2O emissions ranged from 1.14 to 2.53 kg N ha–1 for Chinese rice-based cropping rotations. However, the annual background N2O emissions obtained in this study were substantially lower than the values of 3.62–4.87 kg N ha–1 yr–1 reported by Zou et al. . As mentioned above, the differences in the site-to-site N2O emissions were mainly attributed to different soil properties. As also indicated by Gu et al. , annual background emissions might be highly dependent on total soil N content, organic C content, bulk density, and clay content, individually or collectively.
4.3 Grain Yield, NUE, and net GHG Emissions and Implications for N Fertilizer Management Strategy
 In this study, grain yields for different fertilizer practices averaged 6.2 and 4.5 t ha–1 for rice and wheat, respectively (Table 1). Linquist et al.  estimated the grain yields for 62 study sites worldwide to be an average of 6.1 and 4.8 t ha–1 for rice and wheat, respectively. These results are very close to the results of our study. In addition, the mean annual yields (yields of rice plus wheat, ranging from 8.09 to 12.10 t ha–1) from our study were also comparable to the total yields of 5.97–12.1 t ha–1 for the rice-wheat systems in the Indo-Gangetic plains of India [Pathak et al., 2002; Malla et al., 2005; Bhatia et al., 2005]. Compared to the control, the N inputs significantly increased crop yields at both the seasonal and annual scales, but there were no yield differences between the alternative and farmers’ practices, indicating that the common synthetic N rate used by the local farmers in our study might exceed plant uptake capacity, thus resulting in marginal yield improvement [Ju et al., 2009; Van Groenigen et al., 2010]. Similarly, Mcswiney and Robertson  and Ma et al.  indicated that grain yields generally increased in response to nitrogen inputs before exceeding crop requirements and then leveled off at higher inputs of nitrogen fertilizer. On the other hand, the NUE in the alternative and farmers’ practices was assessed in the present study and ranged from 27% to 56% for rice and from 37% to 70% for wheat (Table 1). Our estimated NUE values were within the range of 19–75% for arable crops across the world, as reported by Van Groenigen et al. . Over the annual rotation cycles, reducing the common application rate by 37% improved the NUE from an average of 42% in the farmers’ practice to 51% in the alternative practice (Table 4). Also, Matson et al.  observed that the nitrogen fertilizer recovery efficiency (i.e., NUE) substantially increased from 46% to 57% by reducing nitrogen inputs in the intensive wheat systems of Mexico.
 Previous studies have observed a trade-off effect between the CH4 and N2O emissions during the rice season with N fertilizer management [e.g., Cai et al., 1997]. Our present results corroborate these findings, and it is important to assess the integrative effects of nitrogen fertilization on net GHG emissions. In this study, the estimated aggregate emissions of CH4 and N2O ranged from 3.50 to 12.08 t CO2-eq ha–1 and from 0.33 to 1.94 t CO2-eq ha–1 for the rice and nonrice seasons, respectively (Figure 6). These estimates fell within the range in GWP across the 328 worldwide observations for the rice and wheat seasons (75–22,237 kg CO2-eq ha–1 for rice and 32–4349 kg CO2-eq ha–1 for wheat) as reported by Linquist et al. . At the same time, our yield-scaled aggregate emissions of CH4 and N2O for rice (mean: 900 kg CO2-eq t–1) were an average of 3.6 times higher than those for wheat (mean: 197 kg CO2-eq t–1), which is also comparable to the values estimated by Linquist et al. . Furthermore, our study extended the earlier findings of Linquist et al.  by demonstrating the entire year estimate for aggregate emissions of CH4 and N2O. Across different fertilization rates and cropping years, the average annual aggregate emissions of CH4 and N2O were estimated to be 6.27 t CO2-eq ha–1 or 630 kg CO2-eq t–1. To our knowledge, research on net GHG emissions over rice-wheat cropping systems is very limited. Malla et al.  estimated the annual aggregate emissions of CH4 and N2O to be 1035–1288 kg CO2-eq ha–1 (or 93–115 kg CO2-eq t–1) in the irrigated rice-wheat system in the Indo-Gangetic plains of India under various fertilizer management practices. The higher annual aggregate emissions of CH4 and N2O at this study site than that of Malla et al.  were likely due to the significantly higher soil organic carbon (18.4 vs. 4.2 g kg–1) and synthetic N application rates (475 vs. 240 kg N ha–1 yr–1), as these two factors are the crucial substrates for soil CH4 and N2O production. In comparison to the control, the application of synthetic N fertilizers tended to mitigate the annual GHG emissions on either an area- or yield-scaled basis (Figure 6). Although the annual aggregate emissions of CH4 and N2O in the farmers’ practice were comparable to those in the alternative practice, the farmers’ practice is not a preferred N management option for the rice-wheat rotation cropping system because it could release large amounts of NO during the nonrice season and lead to low NUE. It should be noted that our analysis of net GHG emissions did not include the changes in the soil organic carbon (which could reflect the net balance between carbon respiration and fixation in a cropping system, as indicated by Shang et al.  and Wang et al. ), because numerous studies have shown that within a typical study period of several years, changes in soil organic C are difficult to detect because the magnitude of change is small and there is a large degree of spatial variation [e.g., Post et al., 2001; Conant et al., 2011]. Therefore, reducing the annual N fertilizer from the farmers’ application rate (~475 kg N ha–1 yr–1) to the recommended rate (~300 kg N ha–1 yr–1) is an acceptable strategy for rice-wheat cropping systems in that there were no significant differences in the annual net GHG emissions or grain yields, but the losses of nitrogen were substantially reduced.
 Overall, this study provided an insight into the integrated evaluation of greenhouse gas emissions and NO release and provided information about the grain yield, as influenced by nitrogen fertilization in typical rice-winter wheat cropping systems. The cumulative emissions of CH4 and N2O and the emission factors of N2O-N varied greatly in different seasons within a year and between different cropping years, highlighting the necessity of year-long investigations across multiple years to achieve representative annual estimates. The seasonality of N2O and NO emissions was primarily regulated by soil mineral nitrogen contents, while the cumulative emissions were exponentially correlated to the nitrogen application rate and the harvested aboveground biomass. Averaged over 4 years, nitrogen fertilizer application tended to decrease the annual CH4 emissions and significantly increase the annual N2O emissions, expressed on either an area- or yield-scaled basis. Meanwhile, nitrogen fertilization substantially increased annual ecosystem respiration (CO2) emissions, but scaling the CO2 emissions by grain yield resulted in similar yield-scaled CO2 emissions from all treatments. Compared to the control, the application of nitrogen fertilizer tended to mitigate the aggregate emissions of CH4 and N2O, but the annual aggregate emissions were comparable between the alternative and farmers’ fertilizer practices. In addition, our results demonstrated that the alternative practice with reduced nitrogen inputs could decrease NO emissions and increase nitrogen use efficiency and can maintain cereal grain yields. Given the comprehensive assessment of environmental, agronomic, and economic aspects of nitrogen fertilizer management, reducing the common application rate by 37% would be a favorable strategy for Chinese subtropical rice-winter wheat rotation cropping systems.
 This work was jointly funded by the Ministry of Science and Technology of China (2012CB417103), the National Natural Science Foundation of China (51139006, 41075109 and 41021004) and the Helmholtz-CAS Joint Laboratory project ENTRANCE. We thank Huajun Tong and Feng Cui for their substantial assistance in the field and laboratory measurements.