Nitrous Oxide Emission and Grain Yield in Chinese Winter Wheat–Summer Maize Rotation: A Meta-Analysis

Abstract

A systematic understanding of nitrous oxide (N₂O) emission and grain yield in winter wheat–summer maize rotation, one of the most important cereal cropping systems in China, is still lacking. The primary aim of this study was to quantify the N₂O emissions and grain yield, as well as responses to mitigation strategies, in this intensively managed agroecosystem. We conducted a pairwise meta-analysis by compiling a comprehensive dataset of annual N₂O emissions (n = 530) and grain yields (n = 352) from peer−reviewed publications. The N₂O emissions increased with nitrogen (N) fertilizer input rates following a linear model (r² = 0.295, p < 0.001), giving a specific emission coefficient and background emission of 0.71% and 0.5 kg N ha⁻¹ yr⁻¹, respectively. The grain yields responded to the N input rates following a linear-plateau model (r² = 0.478, p < 0.001), giving an optimal N input rate and maximum grain yield of 405 kg N ha⁻¹ yr⁻¹ and 15.5 t ha⁻¹ yr⁻¹, respectively. The meta-analyses revealed that reducing N fertilizers (approximately 50% of the full N input), water-saving irrigation, reduced or no tillage, and applying enhanced efficiency fertilizers significantly decreased N₂O emissions (range: −45% to −9%) and increased or did not impact grain yields (range: −1% to 3%). We recommend that reducing agricultural inputs (i.e., N fertilizers, irrigation, and tillage) is a feasible N₂O mitigation strategy in the intensively managed winter wheat–summer maize rotation that can be employed without additional environmental risks.

1. Introduction

Nitrous oxide (N₂O) is a powerful greenhouse gas with a global warming potential of 273 over a 100-year timescale [1]. The concentration of atmospheric N₂O has increased by 23% since the industrial era, mostly owing to the use of nitrogen (N) fertilizers for food production [1]. Globally, agricultural soils are the largest anthropogenic source, releasing 2.1−3.8 Tg N₂O-N per year [1]. There has been an urgent need to develop N₂O mitigation strategies without crop yield penalties [2,3].

Soil N₂O is primarily produced via microbial processes such as nitrification and denitrification [4]. It is well documented that soil aeration, the availability of N substrates, and soil organic carbon (SOC) content are among the most robust controllers of N₂O fluxes from various agroecosystems [5,6,7,8]. These factors are closely related to climate (e.g., temperature and precipitation), soil properties (e.g., soil pH, texture, and organic matter content), and field management practices (e.g., N input rate, organic amendments, and tillage), thus leading to the large spatial and temporal variations in N₂O emissions [9,10,11,12]. Numerous field experiments have been conducted to investigate the responses of N₂O emissions to various mitigation strategies, such as reducing N input rates [13,14,15], application of enhanced efficiency fertilizers (EEFs) or biochar [16,17,18], and reduced or no tillage [19,20]. The grand mean effects of those mitigation strategies were generally quantified with meta-analyses by pooling the observations of all agroecosystems at global or regional scales [11,21,22]. Of note, those mitigation strategies may also impact crop yields [11,21,22]. Meanwhile, the results of several meta-analyses highlighted the importance of crop types by revealing the differences in N₂O emissions, crop yields, and responses to mitigation strategies among agroecosystems [21,23,24,25]. Thus, a critical analysis for a specific agroecosystem is necessary to better understand the variations in N₂O emissions and crop yields.

Winter wheat–summer maize rotation is one of the most important cereal cropping systems in China, normally undertaken with intensive management in order to pursue high grain yields [13,26]. For instance, conventional N input rates are up to 700−1000 kg N ha⁻¹ yr⁻¹, approximately 2−3 times the optimal input rates recommended by experts [14,20,27]. Such a high overuse of N fertilizers has raised serious concerns about N₂O emissions and other N losses [3,13,26,28]. N₂O emissions and grain yields have been measured by numerous field measurements during the past few decades [10,29,30,31]. However, a systematic understanding of N₂O emission and grain yield is still lacking.

We conducted a pairwise meta-analysis of the winter wheat–summer maize rotation by compiling a comprehensive dataset of annual N₂O emissions and grain yields from peer-reviewed publications. The primary aim of this study was to quantify N₂O emissions and grain yields, as well as their responses to mitigation strategies in this intensively managed agroecosystem.

2. Materials and Methods

2.1. Data Collection and Processing

Peer-reviewed publications were collected by searching the Web of Science (http://apps.webofknowledge.com, accessed on 22 September 2022) and China National Knowledge Infrastructure database (https://www.cnki.net/, accessed on 22 September 2022) with various combinations of the keywords “winter wheat”, “summer maize”, “cereal crops”, “nitrous oxide”, “N₂O”, and “greenhouse gas”. The references were carefully refined to winter wheat–summer maize rotation. Only field measurement data were included, without considering laboratory or pot experimental data and modeling results. The N₂O fluxes were always measured by using manual static chambers, with three or more spatial replicates. The consecutive measurement period covered at least an experimental year. The standard deviations (SD) of annual N₂O emissions and grain yields were readily available or can be calculated.

The dataset totally comprised 530 and 352 measurements of N₂O emissions and grain yields, respectively, from 61 peer-reviewed publications up to May, 2022 (Tables S1–S4). The records included site location, climate (annual temperature and precipitation), experimental years, soil properties (soil texture, pH, and SOC content), field management practices (N sources and input rates, tillage, and irrigation), annual N₂O emissions, and grain yields.

The experimental sites were mostly across East, North, Northwest and Southwest China (Table S1). Conventional N fertilizers refers to synthetic fertilizers, crop residues, and organic fertilizers (such as animal manure, oil cake, and composts), while EFFs include control- or slow-released fertilizers, nitrification inhibitors, and urease inhibitors.

The observations in unfertilized treatments were not used in further data processing, statistical analysis, or meta-analysis except when specifically mentioned. The measurements in treatments amended with Table S2 or Table S3 were only used for meta-analysis and were excluded when defining the relationships of N₂O emissions or grain yields against N input rates. The SOC content was grouped into two classes, as SOC <10 and >10 g C kg⁻¹. Soil texture was categorized into three classes of soils: fine—(clay loam, sandy clay loam, silty clay loam, sandy clay, silty clay, and clay), medium—(loam, silty loam, and silt), and coarse-textured—(sand, loamy sand, and sandy loam). Climate types were temperate and subtropical monsoon climates.

2.2. Statistical Analysis

Statistical analysis was performed using R software (version 4.2.1, the R Foundation, Vienna, Austria; https://www.r-project.org/, accessed on 22 September 2022). A non-parametric Kolmogorov–Smirnov test was applied to test if the data series followed normal or log-normal distributions (p > 0.05). The responses of N₂O emissions and grain yields to N input rates (including those measurements in unfertilized treatments) were tested by fitting a linear and linear-plateau model, respectively, following ordinary least squares regression (p < 0.05). The slope and intercept of the linear regression between N₂O emissions and N input rates were defined as emission coefficient and background N₂O emission, respectively [32]. Additionally, quantile regressions were applied to show regression lines of N₂O emissions against N input rates (including those measurements in unfertilized treatments) at the 5% and 95% quantile levels by using the quantreg package [33].

2.3. Meta-Analysis

The pairwise management practices are: reducing N fertilizers versus full N input rate; water-saving irrigation versus full water irrigation; reduced or no tillage versus conventional tillage; mixtures of synthetic fertilizers and organic fertilizers or full organic fertilizers versus full synthetic fertilizers; crop residue retention (as a substitution of synthetic fertilizers or an additional N source) versus removal; and application of EEFs or biochar versus conventional fertilizers. The proportion of reduced N fertilizers to full input rates ranged from 10% to 75% across the studies (Table S1). We chose three categories (<25%, 25–50%, and >50%) to involve adequate measurements and at least three publications per category for meta-analysis. Except for reducing N fertilizers and additional crop residue retention, the effects of other mitigation strategies were always tested at the same N input rates.

The impacts of management practices were calculated by the natural log of response ratio (RR) as follows:

$$\ln (RR)=\ln (\frac{X_t}{X_c})=\ln (X_t)-\ln (X_c)$$

(1)

where X_t and X_c are the N₂O emissions or crop yields of the experimental and control treatments, respectively. The variance (var) was calculated as follows:

$$var=\frac{S{D}_t{}^2}{n_t{X}_t{}^2}+\frac{S{D}_c{}^2}{n_c{X}_c{}^2}$$

(2)

where SD_t and SD_c are the standard deviations of the experimental and control treatments, respectively. n_t and n_c are the sample sizes of the experimental and control treatments, respectively.

The ln(RR) series always followed normal distributions (p > 0.153). The grand mean and 95% confidence intervals (CIs) were generated by a bootstrapping procedure with 999 iterations using a random effects model featured in the meta package in R [34]. The effects were back-transformed and reported as percentage changes. The effects were considered significant if the 95% CIs did not overlap with zero.

3. Results

3.1. Grain Yields

Annual grain yields in the fertilized treatments varied from 6.5 to 22.1 t ha⁻¹ yr⁻¹ (n = 238; Table S1). The data series followed a normal distribution (p = 0.536), with a median of 15.2 t ha⁻¹ yr⁻¹. The grain yields responded to the N input rates following a linear-plateau model (r² = 0.478, p < 0.001), giving an optimal N input rate and maximum grain yield of 405 kg N ha⁻¹ yr⁻¹ and 15.5 t ha⁻¹ yr⁻¹, respectively (Figure 1).

3.2. N₂O Emissions

Annual N₂O emissions from the fertilized treatments varied from 0.3 to 30.7 kg N ha⁻¹ yr⁻¹ (n = 371; Table S1). The data series followed a log-normal distribution (p = 0.756), with a median of 2.9 kg N ha⁻¹ yr⁻¹. The N₂O emissions increased with N input rates following a linear model (r² = 0.295, p < 0.001), giving a specific emission coefficient and background emission of 0.71% and 0.5 kg N ha⁻¹ yr⁻¹, respectively (Figure 2). The emission coefficients estimated at 5% and 95% quantile levels were 0.23% and 1.56%, respectively (Figure 2).

The difference between N₂O emissions from soils with SOC content < and >10 g C kg⁻¹ was significant (p < 0.001), with median values of 2.41 and 3.17 kg N ha⁻¹ yr⁻¹, respectively (Figure 3). Further, medium-textured soils released larger N₂O emissions than fine- and coarse-textured soils did (p < 0.001), with median values of 2.95, 2.69, and 2.39 kg N ha⁻¹ yr⁻¹, respectively, while the differences between the fine- and coarse-textured soils were non-significant (Figure 3). Conversely, N₂O emissions did not differ significantly (p = 0.188) between the subtropical and temperate monsoon climates, with median values of 2.87 and 2.90 kg N ha⁻¹ yr⁻¹, respectively (Figure 3).

3.3. Effects of Mitigation Strategies

Reducing N fertilizers, water-saving irrigation, and reduced or no tillage were effective in mitigating N₂O emissions (Figure 4), with grand mean effects of −29% (95% CIs: −30% to −28%), −9% (95% CIs: −10% to −8%), and −45% (95% CIs: −46% to −44%), respectively. Both reducing N fertilizers and reduced or no tillage increased grain yields by 2% (95% CIs: 1% to 3%), while the effect of water-saving irrigation was non-significant (95% CIs: −4% to 1%). Of note, grain yields were increased by 3% (95% CIs: 2% to 4%) and decreased by 6% (95% CIs: −10% to −2%) when the N-reducing proportions were 25−50% and >50%, respectively, while the effect was non-significant (95% CIs: −6% to 1%) when the reducing proportions were <25% (Figure 4).

Mixtures of synthetic and organic fertilizers significantly increased N₂O emissions by 10% (95% CIs: 7% to 13%), while the impacts on grain yields were non-significant (95% CIs: −2% to 1%) (Figure 5). Full organic fertilizers decreased N₂O emissions and grain yields by 10% (95% CIs: −13% to −7%) and −3% (95% CIs: −7% to −1%), respectively (Figure 5). Crop residue incorporation, either as a substitution of synthetic fertilizers or an additional N source, increased N₂O emissions by 7% (95% CIs: 3% to 11%) and 16% (95% CIs: 14% to 18%), respectively, and increased grain yields by 13% (95% CIs: 9% to 18%) and 3% (95% CIs: 1% to 4%), respectively. Application of EEFs mitigated N₂O emissions by 41% (95% CIs: −42% to −39%) and increased grain yields by 3% (95% CIs: 2% to 4%). Biochar amendments mitigated N₂O emissions by 9% (95% CIs: −10% to −8%), while the impacts on grain yields were not available (Figure 5).

4. Discussion

4.1. Optimal N Input Rate and Maximum Grain Yield

Generally, grain yields respond to the N input rates following a linear-plateau or quadratic model, which provides an optimal N input rate and maximum grain yield for wheat and/or maize production at various regional scales (

Table 1. Comparison of the estimated optimal N input rate (kg N ha⁻¹ yr⁻¹/season⁻¹) and maximum grain yields (t ha⁻¹ yr⁻¹/season⁻¹) for wheat and maize production in different regions.

Region.	Crop	Optimal N Input	Maximum Grain Yield	Model	n	Reference
		kg N ha−1 yr−1/season−1	t ha−1 yr−1/season−1
China	winter wheat–summer maize	405	15.5	linear-plateau	288	This study
North China Plain	winter wheat	129	6.6	linear-plateau	59	[35]
North China Plain	summer maize	289	8.5	quadratic	101	[36]
North China Plain	winter wheat–summer maize	314	14.4	linear-plateau	37	[10]
Asia	wheat	315	6.3	quadratic	n.a.	[21]
Asia	maize	313	8.3	quadratic	n.a.	[21]

n.a., not available.

). In this study, we estimated an optimal N input rate of 405 kg N ha⁻¹ yr⁻¹, above which grain yields did not appear to increase with N input rates (Figure 1). Approximately 60% of the observations in the dataset were based on crop fields receiving N fertilizer above the optimal N input rate (Table S1), denoting an intensive overuse of N fertilizers in winter wheat–summer maize rotation. The estimated maximum grain yield was slightly higher than the median of the observations (15.5 versus 15.2 t ha⁻¹ yr⁻¹). Our estimates were similar to the sum (optimal N input rate of 418 kg N ha⁻¹ yr⁻¹ and maximum grain yield of 15.1 t ha⁻¹ yr⁻¹) of the estimates made for winter wheat (optimal N input rate of 129 kg N ha⁻¹ yr⁻¹ and maximum grain yield of 6.6 t ha⁻¹ yr⁻¹) and summer maize (optimal N input rate of 289 kg N ha⁻¹ yr⁻¹ and maximum grain yield of 8.5 t ha⁻¹ yr⁻¹) by Cui et al. [35] and Wang et al. [36], respectively, in the North China Plain. However, higher [21] or lower [10] optimal N input rates were also reported for wheat and maize production (Table 1). The differences probably owing to the variations in the involved data.

4.2. Variations in N₂O Emissions

The N₂O emissions from the winter wheat–summer maize rotation (range: 0.3 to 30.7 kg N ha⁻¹ yr⁻¹) were largely within the wide ranges that have been observed in various agroecosystems at global or regional scales (Table 1). In this study, the N input rates significantly contributed to the large variations in N₂O emissions (Figure 1). The specific emission coefficient (0.71%) was lower than the IPCC [37] default (1%) for upland agroecosystems and those (range: 0.8% to 1.8%) reported by previous studies (

Table 2. Comparison of the observed N₂O emissions (kg N ha⁻¹ yr⁻¹/season⁻¹) from the fertilized treatments and the estimated emission coefficient (%) and background emissions (kg N ha⁻¹ yr⁻¹/season⁻¹) for various crop types at global or regional scales.

Region	Crop	N2O Emissions	Emission Coefficient	Background Emission	Reference
		kg N ha−1 yr−1/season−1	%	kg N ha−1 yr−1/season−1
China	winter wheat–summer maize	0.3–30.7	0.71	0.50	This study
world	upland crops	0.89–8.0	1.25	1.0	[32]
world	maize	0.1–11.5	1.06	1.15	[39]
world	wheat	0.1–9.3	1.21	0.59	[39]
world	sugarcane	0.03–9.56	1.21	0.93	[23]
world	fruit orchard	−0.12–26.0	1.36	0.73	[25]
China	vegetable	0.03–48.4	0.8	16.9	[40]
China	greenhouse vegetable	0.2–41.8	0.95	0.78	[24]
China	tea plantation	1.19–32.7	1.8	1.70	[41]

). The estimated background emissions (0.50 kg N ha⁻¹ yr⁻¹) were also lower than the mean of Chinese cropland (0.96 kg N ha⁻¹ yr⁻¹) [38] and those (range: 0.73 to 16.9 kg N ha⁻¹ yr⁻¹) listed in Table 2. It is necessary to consider the substantial differences in emission coefficients and background emissions among those agroecosystems (Table 2) when estimating global or regional N₂O budgets.

The quantile regression gave wide uncertainty regarding the emission coefficient (0.23% and 1.56% at 5% and 95% quantile levels, respectively; Figure 1), suggesting that the relative importance of other regulating factors needs to be evaluated. Previous studies revealed that soil properties and climate further contributed to the large variations in N₂O emissions from agroecosystems such as fruit orchards and tea plantations [25,41], and we suspect that these happened to the winter wheat–summer maize rotation as well. Firstly, our results showed that N₂O emissions from soils with SOC content <10 g C kg⁻¹ were significantly lower (p < 0.001) than those from soils with SOC content >10 g C kg⁻¹ (Figure 3). High SOC content is generally accompanied by high organic matter content that favors microbial activities and N₂O production by providing N and carbon (C) substrates via decomposition [12,42]. Secondly, N₂O emissions from medium-textured soils were significantly larger (p < 0.001) than those from fine- and coarse-textured soils (Figure 3). Several studies reported that fine-textured soils favored N₂O emissions by developing anaerobic conditions in soil profiles [25,43] and/or retaining N substrates from leaching [44]. Conversely, other studies revealed that fine-textured soils limited N₂O emissions by promoting N₂O reduction via denitrification [9,45]. Thus, the impacts of soil texture on N₂O emission are variable, depending on the prevailing processes through which N₂O is produced or consumed. Several field measurements revealed that nitrification and denitrification contributed equally to N₂O emissions from the winter wheat–summer maize rotation [20,31], while medium-textured soils favored both processes. We also used a backward stepwise regression to determine the responses of N₂O emissions to multiple factors, such as the N input rate, soil texture, and SOC content. However, soil texture and SOC content were excluded during the analysis, suggesting that their effects were rather complex and could not be explained by a linear model. Finally, climate types did not appear to influence N₂O emissions (Figure 3). The results of several meta-analyses revealed that precipitation was a major regulating factor, generally with large N₂O emissions in wet relative to dry areas [21,44]. We propose that irrigation may partially fill the water input gap between the temperate and subtropical zones and subsequently decrease the differences in N₂O emissions.

4.3. N₂O Mitigation Strategies

Reducing N input significantly decreased N₂O emissions by 29% and increased grain yields by 2% (Figure 4), confirming that N fertilizers were overused in the winter wheat–summer maize rotation. In general, the response ratio of N₂O emissions related negatively to the proportion of reduced N fertilizers (r² = 0.336, p < 0.001; Figure 6). This result indirectly reflected the positive dependence of N₂O emissions on N input rates (Figure 2). Of note, the results appeared contradictory in that grain yields increased when use of N fertilizers was reduced (Figure 4) and a linear-plateau model was fitted to these two variables (Figure 1). Indeed, several studies reported a quadratic model between grain yields and N input rates (Table 1), denoting a potential for yield losses when N fertilizers were overused. Nonetheless, reducing more than 50% of the full amount of N fertilizers was not recommended because of the significant loss of grain yields (Figure 4). Additionally, water-saving irrigation and reduced or no tillage decreased N₂O emissions and increased or did not impact grain yields (Figure 4). We suggest that reducing agricultural inputs (i.e., N fertilizers, irrigation, and tillage) is a feasible N₂O mitigation strategy in the intensively managed winter wheat–summer maize rotation.

Mixtures of synthetic and organic fertilizers and full organic fertilizers increased and decreased N₂O emissions, respectively (Figure 5). This was probably because the mixtures simultaneously provided N and C substrates and subsequently promoted microbial activities and N₂O production, whereas full organic fertilizers might limit the readily available N substrates [46,47,48]. This explanation was partially supported by the stimulating effects of crop residues on N₂O emissions (Figure 5). Although the N₂O emissions were enhanced by 7–16% (Figure 5), the combination of synthetic fertilizers with organic materials (including organic fertilizers and crop residues) still has the possibility of mitigating net greenhouse gas emissions by sequestering SOC [12,31]. It is well documented that SOC sequestration rates increase with the extra C input rates (i.e., the substitution ratio) [49]. However, full organic fertilizers were not recommended due to the significant loss of grain yields (Figure 5). Further, we suggest that crop residues be directly returned to soil when pursuing high grain yields, rather than being converted to organic fertilizers (Figure 5).

Application of EEFs or biochar significantly decreased N₂O emissions by 9–41% (Figure 5). This finding is in agreement with the results of previous meta-analyses, which showed that applying EEFs or biochar was effective in mitigating N₂O emissions (range: −63% to −30%) across various agroecosystems [25,50,51]. In this study, the observations of EEFs and biochar amendments were always conducted in alkaline soils (pH > 7.5; Tables S2 and S3). Thus, caution must be exercised because application of EEFs may stimulate ammonium evaporation from alkaline soils [51], while biochar may further increase soil pH [52].

5. Conclusions

We observed that N₂O emissions and grain yields responded to N input rates following a linear and linear-plateau model, respectively. The specific emission coefficient (0.71%) and background emissions (0.50 kg N ha⁻¹ yr⁻¹) were small relative to other upland agroecosystems. The estimated optimal N input rate (405 kg N ha⁻¹ yr⁻¹) and maximum grain yield (15.5 t ha⁻¹ yr⁻¹) were comparable to previous reports for winter wheat and summer maize production.

The meta-analyses revealed that reducing N fertilizers (approximately 50% of the total N input), water-saving irrigation, reduced or no tillage, and applying EEFs decreased N₂O emissions and increased or did not impact grain yields. We suggest that reducing agricultural inputs (i.e., N fertilizers, irrigation, and tillage) is a feasible N₂O mitigation strategy in the intensively managed winter wheat–summer maize rotation that can be employed without additional environmental risks.