Impacts of natural factors and farming practices on greenhouse gas emissions in the North China Plain: A meta‐analysis

Abstract Requirements for mitigation of the continued increase in greenhouse gas (GHG) emissions are much needed for the North China Plain (NCP). We conducted a meta‐analysis of 76 published studies of 24 sites in the NCP to examine the effects of natural conditions and farming practices on GHG emissions in that region. We found that N2O was the main component of the area‐scaled total GHG balance, and the CH 4 contribution was <5%. Precipitation, temperature, soil pH, and texture had no significant impacts on annual GHG emissions, because of limited variation of these factors in the NCP. The N2O emissions increased exponentially with mineral fertilizer N application rate, with y = 0.2389e0.0058x for wheat season and y = 0.365e0.0071x for maize season. Emission factors were estimated at 0.37% for wheat and 0.90% for maize at conventional fertilizer N application rates. The agronomic optimal N rates (241 and 185 kg N ha−1 for wheat and maize, respectively) exhibited great potential for reducing N2O emissions, by 0.39 (29%) and 1.71 (56%) kg N2O‐N ha−1 season−1 for the wheat and maize seasons, respectively. Mixed application of organic manure with reduced mineral fertilizer N could reduce annual N2O emissions by 16% relative to mineral N application alone while maintaining a high crop yield. Compared with conventional tillage, no‐tillage significantly reduced N2O emissions by ~30% in the wheat season, whereas it increased those emissions by ~10% in the maize season. This may have resulted from the lower soil temperature in winter and increased soil moisture in summer under no‐tillage practice. Straw incorporation significantly increased annual N2O emissions, by 26% relative to straw removal. Our analysis indicates that these farming practices could be further tested to mitigate GHG emission and maintain high crop yields in the NCP.

the main component of the area-scaled total GHG balance, and the CH 4 contribution was <5%. Precipitation, temperature, soil pH, and texture had no significant impacts on annual GHG emissions, because of limited variation of these factors in the NCP.
The N 2 O emissions increased exponentially with mineral fertilizer N application rate, with y = 0.2389e 0.0058x for wheat season and y = 0.365e 0.0071x for maize season.
Emission factors were estimated at 0.37% for wheat and 0.90% for maize at conventional fertilizer N application rates. The agronomic optimal N rates (241 and 185 kg N ha −1 for wheat and maize, respectively) exhibited great potential for reducing N 2 O emissions, by 0.39 (29%) and 1.71 (56%) kg N 2 O-N ha −1 season −1 for the wheat and maize seasons, respectively. Mixed application of organic manure with reduced mineral fertilizer N could reduce annual N 2 O emissions by 16% relative to mineral N application alone while maintaining a high crop yield. Compared with conventional tillage, no-tillage significantly reduced N 2 O emissions by ~30% in the wheat season, whereas it increased those emissions by ~10% in the maize season. This may have resulted from the lower soil temperature in winter and increased soil moisture in summer under no-tillage practice. Straw incorporation significantly increased annual N 2 O emissions, by 26% relative to straw removal. Our analysis indicates that these farming practices could be further tested to mitigate GHG emission and maintain high crop yields in the NCP.

K E Y W O R D S
farming practice, fertilizer, meta-analysis, methane, natural factor, nitrous oxide

| INTRODUCTION
Global atmospheric concentrations of greenhouse gases (GHGs) such as CO 2 , N 2 O, and CH 4 have continued to increase, which has further heightened public and scientific concerns (IPCC, 2014;Wei, Zhang, Chen, Zhang, & Zhang, 2012). N 2 O and CH 4 , mainly derived from the agricultural sector , have 265 and 28 times greater global warming potentials than CO 2 over a time horizon of 100 years (IPCC, 2014). Although a number of climate change mitigation measures have been adopted in China during recent years, requirements for further mitigation of the continued increase in GHG emission are still much needed .
The North China Plain (NCP) occupies 23% of national cropland area (Ding, Cai, Cai, Yagi, & Zheng, 2007) and accounts for 43% of total winter wheat (Triticum aestivum L.) and summer maize (Zea mays L.) production in China . High land productivity in the NCP has relied on intensive farming practices since the 1980s (Liao, Wu, Meng, Smith, & Lal, 2015), which are characterized by frequent irrigation (Wang, Yu, Wu, & Xia, 2008) and high levels of mineral nitrogen (N) fertilizer application (550-600 kg N ha −1 year −1 ; Ju et al., 2009). However, in the near future, greater crop yields with reduced GHG emissions must be achieved in China to meet the dual goals of ensuring food security and reducing negative environmental impacts The State Council of China, 2016).
Agricultural practices regulate soil N and carbon (C) dynamics and thereby affect the fluxes of N 2 O and CH 4 (Adviento-Borbe, Haddix, Binder, Walters, & Dobermann, 2007;Mutegi, Munkholm, Petersen, Hansen, & Petersen, 2010). Natural factors also affect or interact with farming practices, thereby influence N 2 O and CH 4 emissions (Chatskikh, Olesen, Berntsen, Regina, & Yamulki, 2005;Čuhel et al., 2010;Gu et al., 2013;Jansen, 2009;Smith, 1997;Vidon, Marchese, Welsh, & Mcmillan, 2016). In recent decades, many site-specific studies have been conducted to explore the impacts of fertilization (Tan et al., 2017;Yan, Yao, Zheng, & Liu, 2015), tillage (Tian et al., 2012;Wei et al., 2012), and crop residues Huang, Gao, Christie, & Ju, 2013) on GHG emission and crop yield in the NCP. However, these individual studies were not able to provide a generalized understanding across this large region. Therefore, a comprehensive quantitative analysis of published studies regarding the NCP is necessary to assess the overall relationship between GHG emissions and natural and farming factors. Meta-analysis was selected for this purpose, because it is a powerful method to integrate site-specific results and draw overall conclusions at regional and global scales (Gurevitch, Curtis, & Jones, 2001;Luo, Wang, & Sun, 2010).
Previous meta-analyses for China's agricultural soils have examined the relationship between natural and farming factors and GHG emissions (Lu, Huang, Zou, & Zheng, 2006;Zhao et al., 2016). However, they did not focus on winter wheat-summer maize (WW-SM) rotation, the typical and major farming system in the NCP, and their conclusions did not provide technical support for GHG mitigation in the region. In this study, we calculated both response ratios and average amounts of GHG emission under different natural factors and farming practices. Regression analysis has also been used to obtain relationships between N 2 O emissions, emission factors (EFs, percentage of fertilizer-induced N 2 O emission), crop yields, and N application rates. We aimed at quantifying the comprehensive responses of GHG emissions to major farming practices and natural factors in the NCP, which will facilitate large crop yields and GHG mitigation in the region.

| Data collection
We conducted a literature survey of peer-reviewed papers published prior to April 2016 and collected data on N 2 O/CH 4 emissions, climate and soil factors, farming practices, and crop yields for WW-SM systems in the NCP region. All the papers were obtained from the databases of China National Knowledge Infrastructure (CNKI, the largest Chinese academic journal database) and Web of Science. We conducted a preliminary search using the keywords "N 2 O," "CH 4, " and "NCP." We then selected papers based on the following selection criteria: (1) Studies must have been of the NCP under WW-SM cropping systems; (2) measurements of N 2 O and/or CH 4 fluxes must have been made under field conditions in the entire growth period of the wheat and/or maize cropping season, using static chamber methods; (3) cumulative GHG fluxes during the entire season, measurement frequency, and the number of field replications had to be reported. By applying these selection criteria, 76 papers were selected for study (56 for N fertilization, 19 for tillage, 29 for straw management, 13 for slow-release fertilizer (SRF) application, and 24 for organic fertilizer application; Appendix S1). Some authors published their results on grain yield and GHG emission separately in different papers, so in some cases missing yield data were collated from different publications by the same authors. For each study, the GHG emission or crop yield for each individual treatment combination was separated as distinct single data points in our meta-analysis. Unless available in the original literature, precipitation and temperature during the experimental period of each study were obtained from the China Meteorological Data Service Center (http://data.cma.cn). To avoid bias toward multiyear studies, the mean value of measurements in different years was used as a single observation when experiments were repeated over time, except for analysis of the effects of weather conditions (precipitation and temperature).

| Calculation of total GHG balance
We used the IPCC coefficients to calculate CO 2 -equivalents (CO 2 -eq) of N 2 O and CH 4 emissions over a 100-year time horizon (298 and 25 for N 2 O and CH 4 , respectively; IPCC, 2007). The overall CO 2 -eq of N 2 O and CH 4 emission was expressed as total GHG balance (Cherubini, 2010). Area-scaled and yield-scaled data represented the total GHG balance per unit crop field (ha) and per unit crop yield (Mg), respectively. The equations are as the follows.

| Natural factors
CO 2 -equivalent N 2 O and CH 4 emissions for fertilization levels of ≥200 kg N ha −1 season −1 or ≥400 kg N ha −1 year −1 from each study were selected to evaluate the impacts of soil pH and soil texture on GHG emissions. Soil pH was divided into two levels (6.5-7.5 and >7.5), which represent neutral and alkaline soils, respectively. Soil textures in the meta-analysis were categorized according to the USDA classification system. To avoid limiting the number of samples in each texture class, we classified the textures by clay content into two types, sandy loam and loam to clay loam. We used the methods of Linquist, Van Groenigen, Adviento-Borbe, Pittelkow, and Van Kessel (2012) to conduct the meta-analysis, and the equations used were as follows.
Equation (3) was used to calculate weighted mean values of GHG emissions or area-scaled total GHG balance under different natural conditions, in which Y i is the observation of GHG emission or total GHG balance at the ith site, M is the mean value of CO 2 -eq GHG emission or area-scaled total GHG balance (kg CO 2 -eq ha −1 season −1 or year −1 ), and W i is the weight for observations at the ith site, which was calculated using Equation (4). In that equation, n is the number of replicates in the field experiment, f is the number of GHG measurements per month, and obs is the total number of observations at the ith site. To prevent studies with high sampling frequencies from being assigned extreme weights, a maximum value f = 5 was assigned when GHG fluxes were measured more than once per week. Linear regression was used to examine the relationship of N 2 O emissions with precipitation and temperature during the experimental period.

| Farming practices
Response ratio (R) was used to evaluate the impacts of farming practices on N 2 O emissions, CH 4 emissions, crop yield, and total GHG balance (area-scaled and/or yield-scaled). Only studies that included side-by-side comparisons were selected for this calculation. The rates of applied N were separated into three levels (50-150, 150-250, and 250-350 kg N ha −1 season −1 or 100-300, 300-500, and 500-700 kg N ha −1 year −1 ). N fertilizers in the selected studies were mainly ammonium-based (e.g., urea) in the study region (Ju et al., 2009). In and application of SRF. We divided the tillage measures into no-tillage (NT) and conventional tillage (CT), and straw management into straw incorporation and straw removal. To evaluate the effect of straw incorporation under N fertilization, the effects of straw incorporation on N 2 O emission were further separated into with and without N fertilizer application. CH 4 emissions were all found to be negative in the side-by-side comparisons. We used CH 4 uptake in the calculation of response ratios to avoid confusion when understanding effect sizes.
The natural log of the response ratio (lnR) was calculated as an index of the effect size: where X t and X c are measurements of N 2 O emission, CH 4 uptake, yield, or total GHG balance (area-scaled and/or yield-scaled) for the treatment and control (Table 1), respectively. The mean of the response ratios (R) was calculated from lnR values of individual studies using Equation (6): where W i is the weighting factor, estimated by Equation (4). To facilitate interpretation, results of the R analysis were reported as percentage change under the treatment relative to the control ([R − 1] × 100).
In addition to the calculation of R, we calculated absolute values of mean GHG emission and area-scaled total GHG balance under different levels of N application or farming practice. Mean values were then evaluated using the same approach as described in Section 2.2.2, with M in Equation (3) representing the mean value of N 2 O emissions (kg N 2 O-N ha −1 season −1 or year −1 ), CH 4 emissions (kg CH 4 -C ha −1 season −1 or year −1 ), or area-scaled total GHG balances (Mg CO 2 -eq ha −1 season −1 or year −1 ) under various treatments.

| Statistical and regression analysis
All studies that reported either N 2 O emission or crop yield were included to determine best-fit regression curve models for N 2 O emission or yield as functions of the N application rate. Linear, exponential, quadratic, and linear-plateau models (Cerrato & Blackmer, 1990) were tested with each dataset. We used the Statistical Analysis System (SAS Institute, 1998) package for statistical analyses and evaluation of significance levels. If statistical significance was detected for several models at the critical level of 5%, we then selected the model with the

| Meta-analysis
The meta-analysis was performed using MetaWin 2.1 (Rosenberg, Adams, & Gurevitch, 2000). A random-effect model was used to calculate the mean effect size. We used bootstrapping (4,999 iterations) to generate these mean emissions, total GHG balances, effect sizes, p-values, and 95% bootstrapped confidence intervals (95% CIs). Mean effect sizes were only considered significantly different if their 95% CIs did not overlap. Sensitivity analysis was conducted for absolute values and response ratios to test whether the weighted and unweighted approach give similar results. The results using the weighted approach were very similar to that using unweighted approach, hence we only report the results of the former approach herein.

| Natural factors
When all observations were taken into account, average N 2 O emissions during the wheat season, maize season, and annual period were 320 (232-400, 95% CI), 983 (841-1,153, 95% CI) and 1,492 (1,264-1,742, 95% CI) kg CO 2 -eq kg ha −1 , respectively (Figure 1a-c). This indicates significantly higher N 2 O emissions in the maize season (about three times that of the wheat season; p < .05). Average CH 4 emissions were all found to be negative, suggesting that the agricultural soils of the NCP act as an overall sink for atmospheric CH 4 . When expressed as CO 2 -eq, the CH 4 uptake was much less than N 2 O emission, that is., <5% of the area-scaled total GHG balance, indicating that the overall area-scaled total GHG balance was predominantly determined by N 2 O emission.
Therefore, we mainly address the trends of N 2 O emission in this section.
The N 2 O emission tended to be higher in loam to clay loam textured soils than in sandy loam soils, but a significant difference between these two soil textures was only detected for the wheat season ( Figure 1a; p < .05). No pronounced differences in CH 4 uptake or area-scaled total GHG balance were found between soil texture categories (p > .05).
In the wheat season, N 2 O emissions and area-scaled GHG balances in soils with pH of 6.5-7.5 were significantly greater than those with pH > 7.5 (p < .05; Figure 1a), but pronounced differences were not found for maize season and at annual scale (p > .05; Figure 1b,c). Across all periods, no statistical differences of CH 4 emission were detected between neutral (pH 6.5-7.5) and alkaline (pH > 7.5) soils (p ˃ .05; Figure 1a

| N application rate
N 2 O emissions under the lowest N application rate (50-150 kg N ha −1 season −1 or 100-300 kg N ha −1 year −1 ) were 0.57, 0.51, and 1.37 kg N 2 O-N ha −1 for the wheat season, maize season, and annual period, respectively. The N 2 O emissions increased dramatically to 1.14, 2.24, and 3.86 kg N 2 O-N ha −1 , respectively, under the highest N application rate (250-350 kg N ha −1 season −1 or 500-700 kg N ha −1 year −1 ; p < .05; Table 2). The area-scaled total GHG balance showed trends similar to N 2 O emission, which increased from 0.60 CO 2 -eq ha −1 year −1 under the lowest N application rate to 1.75 CO 2 -eq ha −1 year −1 for the highest rate (Table 2). N application rates also had a significant effect on the absolute amount of CH 4 uptake in the maize season (p < .01; Table 2).
Relative changes in N 2 O emission remained relatively small at low N application rates, but increased sharply at higher rates (Figure 3a,d,g).
This was most evident at annual scale, in which the relative change was as great as 500% under the highest N application rate (500-700 kg N ha −1 year −1 ), nearly twice that under the low N application rate (100-300 kg N ha −1 year −1 ) (p < .05; Figure 3g). However, N application rates had no significant effect on relative changes of CH 4 uptake (Figure 3b,e,h), except for low rates (100-300 kg N ha −1 year −1 ) at annual scale, for which the CH 4 uptake significantly increased, by 10.2% (p < .05).
Exponential models fit a significant relationship between N 2 O emission and N rate (p < .01; Figure 4a-c), especially so for the maize season (R 2 = 0.52). This indicates that the N 2 O emission increased exponentially in response to increasing N application rate. The EF of N 2 O generated from the exponential model also showed a nonlinear relationship with N application rate (Figure 4a-c). The relationship between crop yield and N application rate could be described by quadratic or linearplateau models (p < .01; Figure 4d

| Tillage
The effect of tillage on N 2 O emission showed different trends between the wheat and maize seasons. In the wheat season, N 2 O emission significantly declined by nearly 30% under NT (p < .05; Figure 5a) as compared with CT. In contrast, N 2 O emission was significantly enhanced (by ~10%) for the maize season (p < .05; Figure 5b). At annual scale, there were no significant overall differences in the N 2 O emission (p ˃ .05; Figure 5c) between NT and CT management. In contrast, the effect of NT on CH 4 uptake was consistent between the various growth seasons. Compared with CT, NT significantly (p < .05) reduced CH 4 uptake, that is., 31.6%, 19.9%, and 23.3% for the wheat season, maize season, and annual period, respectively (Figure 5a-c). No-tillage slightly but significantly decreased crop yield relative to CT (p < .05) and was 11.2%, 2.7%, and 3.3% for the wheat season, maize season, and annual period, respectively (Figure 5a-c). The areascaled total GHG balances showed similar trends as N 2 O emissions, which decreased significantly by 33% for wheat season (p < .05; Figure 5a) and increased significantly by 16% for maize season (p < .05; Figure 5b) under NT. However, there was no difference for the annual period ( Figure 5c). NT significantly increased yield-scaled total GHG balance by 18.8% in the maize season (p < .05; Figure 5b) but had no effect during the wheat season or annually (p > .05; Figure 5a,

| Straw incorporation, application of organic manure, and SRF
Regardless of N fertilization, N 2 O emission increased with straw incorporation relative to straw removal, especially in maize season (29.9%, p < .05; Figure 6b) and the annual period (25.8%, p < .05; Figure 6c). The relative increase of N 2 O emission from straw incorporation tended to be greater under no N fertilization as compared with N fertilization. The area-scaled total GHG balance under straw incorporation significantly increased by 28.4% in maize season (p < .05; Figure 6b), but was similar to straw removal in wheat season (p > .05; Figure 6a). The side-by-side comparison showed significant reductions in CH 4 uptake under straw incorporation compared with straw removal, which were 17.5%, 9.5%, and 10.0% for the wheat season, maize season, and annual period, respectively (p < .05; Figure 6a-c).
Crop yield under straw incorporation tended to be higher than that under straw removal, especially in wheat season (15.4%) and annual period (25.8%) (Figure 6a,c). This resulted in a decline of yield-scaled total GHG balance in the wheat season (p < .05; Figure 6a). These results indicate that straw incorporation enhanced N 2 O emission and reduced CH 4 uptake, but achieved a greater crop yield. However, no significant differences in absolute values of N 2 O emission, CH 4 uptake, area-scaled total GHG balance, or yield were found between these two straw practices (Table 4).

| GHG emission from NCP
Average N 2 O emissions over the NCP (Figure 1) were lower than those of a previous global analysis (Linquist et al., 2012) is normally oxidized, making these soils sink for atmospheric CH 4 in dry farmland systems (e.g., Hu et al., 2013;Powlson, Goulding, Willison, Webster, & Hütsch, 1997;Robertson & Grace, 2004). In addition, the radiative forcing potential of N 2 O is ~12 times greater than that of CH 4 (IPCC, 2007), which has an additional (disproportionate) impact on its estimated contribution to the area-scaled total GHG balance (Six, Ogle, Conant, Mosier, & Paustian, 2004). These results highlight that GHG mitigation actions in the NCP should mainly target N management and N 2 O.
F I G U R E 5 Effect of no-tillage on N 2 O emission, CH 4 uptake, yield, and total GHG balance (area-scaled and yield-scaled) for (a) wheat season, (b) maize season, and (c) annual period relative to conventional tillage. Data are expressed as mean percentage changes with 95% confidence intervals (represented by error bars

| Natural factors
Heavy rainfall may stimulate N 2 O emission in the NCP Yan et al., 2013), but this effect was only observed during maize season in our analysis (Figure 2b). The wheat season in the NCP coincided with very weak precipitation (Wang et al., 2008), so irrigation was more frequently used in that season (~300-500 mm) than in maize season (~100-200 mm). Therefore, the impact of Effects of soil pH, soil texture, and temperature on N 2 O emission or area-scaled total GHG balance were also, in most cases, not significant in current study (p > .05; Figures 1 and 2d-f). Only one study site (Taian of Shandong Province; Appendix S1) in our database had soil pH <7.4, so pH values in neutral (pH 6.5-7.5) and alkaline soils (pH > 7.5) of the NCP were too similar to produce significant distinctions of GHG emission. Similarly, the narrow range of mean temperature (mostly 7-9°C in wheat season and 24-26°C in maize season; Figure 2d,e) and soil texture (sandy loam to clay loam; Figure 1) across the experimental sites might not have been sufficiently variable to generate significant differences in GHG emission.

| N fertilization
The availability of soil N determines N 2 O emissions from soils Liu & Zhang, 2011;Van Groenigen, Velthof, Oenema, Van Groenigen, & Van Kessel, 2010). The relative changes of N 2 O emission at low-to-moderate N application rates remained relatively constant compared with no N fertilization, but increased sharply at higher N application rates (Figure 3a,d,g). When N is added beyond plant or microorganism demand (Kim, Hernandez-Ramirez, & Giltrap, 2013;Li et al., 2001), more N remains in the soil, which can then be lost through N 2 O emission (Gerber et al., 2016;Hoben, Gehl, Millar, Grace, & Robertson, 2011;Kim et al., 2013;McSwiney & Robertson, 2005). In our case, the exponential model gave the best fit for the relationship between N 2 O emission and N rate (p < .01; Figure 4a-c).
Overuse of N fertilizer may even lead to a decline in crop yield (Ju, Liu, Zhang, & Roelcke, 2004;Liu, Ju, Zhang, Pan, & Christie, 2003;Zhu & Chen, 2002). Our simulation showed that calculated AONR The IPCC uses 1% as the default value for EF for upland crops (IPCC, 1997). However, EFs usually are not constant and increase nonlinearly with increasing N rates (Kim et al., 2013;Shcherbak, Millar, & Robertson, 2014). The EFs obtained in our study were 0.37% and 0.90% for the wheat and maize seasons, respectively (Figure 4a,

| Straw incorporation
In our study, N 2 O emissions following the incorporation of wheat and maize straw were higher than that under straw removal, particularly in maize season ( Figure 6). This was because of increasing anaerobic conditions and enhanced denitrification when straw was returned to soils (Chen, Li, Hu, & Shi, 2013;Mutegi et al., 2010;Shan & Yan, 2013). However, under no N fertilization, the relative increase in N 2 O emission from straw incorporation tended to be greater than under N fertilization ( Figure 6). This may be explained by the higher background N 2 O emission in N fertilized soils and the decrease in soil dissolved organic carbon under the combined application of mineral N and crop straw Shan & Yan, 2013;Yao et al., 2009).
Similarly, straw incorporation can supply substrate and create anaerobic microsites for methanogenesis, which inhibits CH 4 oxidation . This is corroborated by our observation that CH 4 uptake under straw incorporation was significantly reduced by 17.5%, 9.5%, and 10.0% relative to straw removal in the wheat season, maize season, and annual period, respectively (p < .05; Figure 6a-c).
Although straw incorporation may induce greater soil-derived N 2 O emissions, it also promotes soil organic C sequestration (Liu, Lu, Cui, Li, & Fang, 2014;Meng et al., 2016) and avoids substantial, uncontrolled GHG emission from straw burning in the NCP (Lu et al., 2010;Smith et al., 2008). Moreover, we found that annual crop yield under straw incorporation increased significantly by ~9% relative to straw removal (p < .05; Figure 6c), similar to a study in Europe (6%; Lehtinen et al., 2014). The impact of straw incorporation on GHG emission should be further comprehensively assessed.

| Slow-release N fertilizer
There have been divergent results of SRF impacts on N 2 O emission, either positive (Akiyama et al., 2013;Li et al., 2015) or negative (Bordoloi & Baruah, 2016;Ji et al., 2013). In the present analysis, SRF reduced annual N 2 O emissions by 13.1%, but this was not statistically significant (p > .05; Figure 7a). The effect of SRF on N 2 O emission is modulated by environmental conditions , the observation period (Hou, Akiyama, Nakajima, Sudo, & Tsuruta, 2000), and crop demand for N (Akiyama, Yan, & Yagi, 2010). Even with no significant reduction in the N 2 O emission, the potential benefits of SRF for reduced NH 3 volatilization and N leaching should not be neglected (Shaviv & Mikkelsen, 1993).

| Mixed application of organic and mineral fertilizer
Compared with M, annual N 2 O emissions significantly increased under M+O (17.0%; p < .05; Figure 7a), probably because of the increased supply of C and anaerobic conditions favoring denitrification (Anderson & Levine, 1986;Kamewada, 2007;Velthof, Kuikman, & Oenema, 2003). O appeared to reduce N 2 O emissions but also significantly decreased crop yield (p < .05), because of the lack of synchronicity of N supply with crop demand under O treatment (Skinner et al., 2014;Tuomisto, Hodge, Riordan, & Macdonald, 2012). In contrast to M+O, the significant reduction in annual N 2 O emission under RM+O (16.9%; p < .05; Figure 7a) was because of lesser N supply from reduced mineral N fertilizer . RM+O also slightly increased crop yield (Figure 7c). Hence, application of reduced mineral N with organic manure is a promising alternative farming practice to meet the demands of reducing GHG emissions while maintaining crop yield in the NCP.

| Limitations of our analysis
It should be pointed out that literatures reporting N 2 O and CH 4 emissions and crop production are relatively limited for the NCP, so this may weaken the efficacy of the meta-analysis. For instance, we cannot reach robust conclusions on tillage, natural factors, and their interaction effects. For the analysis approach, we did a sensitivity analysis that indicated that the weighted and unweighted approaches gave very similar results, for both absolute values and response ratios for GHG emission and crop production as influenced by natural and farming factors. To the best of our knowledge, this study is the first on GHG emissions affected by major farming practices and natural factors in the NCP, which may provide technical support for GHG mitigation in the region.

ACKNOWLEDGMENTS
This work received support from the National Science and Technology