Trade-offs between high yields and greenhouse gas emissions in irrigation wheat cropland in China

Although the concept of producing higher yields with reduced greenhouse gas (GHG) emissions is a goal that attracts increasing public and scientific attention, the tradeoff between high yields and GHG emissions in intensive agricultural production is not well understood. Here, we hypothesize that there exists a mechanistic relationship between wheat grain yield and GHG emission, and that could be transformed into better agronomic management. A total 33 sites of on-farm experiments were investigated to evaluate the relationship between grain yield and GHG emissions using two systems (conventional practice, CP; high-yielding systems, HY) of intensive winter wheat ( Triticum aestivumL.) in China. Furthermore, we discussed the potential to produce higher yields with lower GHG emissions based on a survey of 2938 farmers. Compared to the CP system, grain yield was 39 % (2352 kg ha −1) higher in the HY system, while GHG emissions increased by only 10 %, and GHG emission intensity was reduced by 21 %. The current intensive winter wheat system with farmers’ practice had a median yield and maximum GHG emission rate of 6050 kg ha −1 nd 4783 kg CO2 eq ha−1, respectively; however, this system can be transformed to maintain yields while reducing GHG emissions by 26 % (6077 kg ha −1, and 3555 kg CO2 eq ha−1). Further, the HY system was found to increase grain yield by 39 % with a simultaneous reduction in GHG emissions by 18 % (8429 kg ha −1, and 3905 kg CO2 eq ha−1, respectively). In the future, we suggest moving the trade-off relationships and calculations from grain yield and GHG emissions to new measures of productivity and environmental protection using innovative management technologies.


Introduction
Increasing population and consumption are placing unprecedented pressure on agricultural and natural resources (Tilman et al., 2002;Burney et al., 2010;Foley et al., 2011).It has been projected that chemical nitrogen (N) fertilizer consumption will increase by 142-169 % to support a 100-110 % increase in global food crop yields from 2005 to 2050 (Tilman et al., 2011;IFA, 2012).Agricultural intensification of the "green revolution" improved crop productivity while simultaneously increasing environmental costs such as greenhouse gas (GHG) emissions (Tilman et al., 2002;Burney et al., 2010).Agriculture, including fertilizer production, directly contributes 10-12 % of global GHG emissions, and this figure rises to 30 % or more when land conversion and emissions beyond the farm gate are included (Smith et al., 2007).The Intergovernmental Panel on Climate Change (IPCC;2007) reported that global GHG emissions would need to peak before 2015 and be reduced on the order of 50-85 % (from 2000 levels) by 2050 if dangerous climate change (i.e., a temperature rise > 2.4 • C) is to be avoided.These intertwined challenges necessitate a new imperative for global agriculture, where higher grain yields are produced with more efficient use of N fertilizer and a reduction in both reactive N losses and GHG emissions.
Several conceptual frameworks have been proposed to guide efforts that could produce higher yields with reduced input or environmental costs.These frameworks include ecological intensification (Cassman, 1999), an evergreen revolution (Swaminathan, 2000), and eco-efficient agriculture (Keating et al., 2010), and they share a view of cropping systems as ecosystems that should be designed to maximize the use of fixed resources (land, light, favorable growing conditions) and optimize the use of agricultural inputs (particularly N and P fertilization) to produce high grain yields.Such systems can draw upon features of traditional agricultural knowledge and add new ecological information to the intensification process (Matson et al., 1997;Chen et al., 2011).While there is agreement regarding the need for such improvements, there are only a few examples of how they can be developed and adapted on a large scale and across hundreds of millions of farmers' fields (Carberry et al., 2013).
Wheat production in the North China Plain (NCP) involves some of the most intensive N applications in the world, and the enrichment of N in soil, water, and air has created serious environmental problems (Cui et al., 2010;Zhang et al., 2012).For example, the N applied by farmers of winter wheat in the NCP is often greater than 300 kg N ha −1 (Cui et al., 2010), even though results from region-wide experiments have demonstrated that the optimal N rate is 128 kg N ha −1 (Cui et al., 2008).This overuse of N fertilizer over the past 10 years has not increased wheat yield, with stagnation at ∼ 4573 kg ha −1 with national average (mean grain yield from 2003 to 2012, FAO, 2013).In contrast, in previous high-yield studies in this region, high wheat yield (≥ 9000 kg ha −1 ) was achieved by optimizing the wheat canopy and using favorable management practices to maximize both the quantity and quality of the wheat canopy (Meng et al., 2013).
Recent studies have shown great promise for increasing N-use efficiency and grain yield in maize production by integrating crop and N management (Chen et al., 2011;Grassini and Cassman, 2012).Here, we hypothesize that there exists a mechanistic relationship between wheat grain yield and GHG emission, and that could be transformed into better combining improved crop management technologies with optimal N management.Two groups of experiments with different on-farm N level management systems were conducted in the key winter wheat growing region of northern China.A conventional practice (CP) plot was managed based on farmers' current practices with a yield of approximately 6000 kg ha −1 ; on a high-yield (HY) plot, an integrated soil-crop system management approach was applied to close the yield gap and maintain the grain yield at approximately 8500 kg ha −1 .We evaluated the trade-off relationships between crop productively and GHG emission for the CP and HY systems.We discuss the potential for shifting the focus of the current farming system to new productivity and environmental protection values to produce higher yields with reduced GHG emissions.

Methods and materials
All experiments were conducted on farm fields at 33 sites in 31 counties from 2007 to 2008, including 15 sites in Henan province (S1 to S15), 4 sites in Hebei province (S16 to S19), 12 sites in Shandong province (S20 to S31), and 2 sites in Shaanxi province (S32 to S33, Supplement Fig. S1 and Table S1).The climate in the experimental region is a warm, temperate, sub-humid, continental monsoon climate with cold winters and hot summers.The annual cumulative mean temperature for days with mean temperatures above 10 • C is 4000-5000 • C, and the annual frost-free period is 175 to 220 days.Annual precipitation is 500 to 700 mm, with approximately 30-40 % of the rainfall occurring during the winter wheat growing season (from the beginning of October to middle of June).The amount and distribution of rainfall vary widely from year to year, and are affected by the continental monsoon climate.The soil types were mainly calcareous fluvo-aquic, yellow brown, cinnamon, yellow cinnamon, meadow sanne, and yellow soils.Details of site coordinates, average annual precipitation, soil texture, soil types and some soil properties are shown in Supplement Table S1.

On-farm field experiments: design, crop management, and sampling procedures
Both systems (CP and HY) were tested at each of the 33 sites under four or five N treatments.Five N treatments in 15 sites in Henan province included no N as a control (CK), and low (50 % of median), median, high (150 % of median), and very high (200 % of median) treatments.Four N treatments at the other 18 sites included no N as a control (CK), and low (50 % of median), median, and high (150 % of median).The amount of N fertilizer for the median N treatment was recommended by local agricultural extension employees based on experience.Detailed information of N application rates for the 33 sites is shown in Table 1.
For both conventional practice (CP) and high-yield (HY) systems, one-third of granular urea (CO(NH 2 ) 2 ) is applied by broadcasting at the time of sowing, and the remainder is applied at the stem elongation stage prior to irrigation.Depending on the weather, winter wheat typically receives three irrigations (about 90 mm per time): one before winter, a second at the stem elongation stage, and another around the anthesis stage.Although the volume of irrigation was not precisely measured for every plot and site, the values were similar for each system at every site.For the CP system, experiments were managed using each individual farmer's current crop management practices, except for N fertilizer application rate.In the HY system, local agronomists recommended new varieties with resistance to disease, environmental stress, and lodging that also had the potential for high yields.These new varieties varied across experimental sites.In addition, the better combinations of planting date and plant populations based on local weather (e.g., mean temperatures) were used to optimize the crop canopy, and make maximum use of regional environmental resources (e.g., light and temperature).Compared to the HY system, most farmers' fields used later sowing and used more seeds.Finally, in the HY system, Table 1.N application rate and wheat grain yield for different N application rates and the two systems.N application rate including no N as a control (0 N), 50 % of median N rate (50 % N), 100 % of median N rate (100 % N), 150 % of median N rate (150 % N), and 200 % of median N rate (50 % N).The systems included a conventional practice (CP) and a high-yielding system (HY).

Sites CP HY
N rate (kg N ha −1 ) Grain yield (kg ha −1 ) N rate (kg N ha −1 ) Grain yield (kg ha we improved sowing quality by careful management to foster strong individual plants and make them uniform, creating a lodging-resistant architecture in the crop canopy.Weeds were well controlled with the use of spray herbicides and manual pulling.Pest and disease stress were controlled using spray insecticide and fungicide before the stem elongation stage and after anthesis.No obvious water, weed, pest, or disease stress was observed during the wheat-growing season for both CP and HY system.A randomized complete block design was employed in three replications with plots measuring > 40 m 2 .All plots received approximately 90 kg P 2 O 5 ha −1 as calcium superphosphate (Ca(H 2 PO 4 ) 2 • H 2 O) and about 60 kg K 2 O ha −1 as potassium chloride (K 2 SO 4 ) before planting.
At maturity, three separate areas (each 2-3 m 2 ) were harvested manually.All plant samples were oven dried at 70 • C in a forced-draft oven to a constant weight, weighed, and yields were adjusted to 125 g kg −1 moisture content.

Farmers' survey
With the key winter wheat growing region of northern China from 2004 to 2009, approximately 2-8 typical townships were randomly selected in each county, and 4-6 typical villages were randomly selected in each township.Out of these, 8-10 farmers were randomly questioned regarding their choice of fertilizer, application rate, and grain yield in the past year.Data required included fertilizer production, N content, fertilizer application rate and grain yield.For grain yield and N application, only a few observations (< 5 %) fell outside the normally expected ranges of the entire data set.However, considering the great variation in each parameter among fields, we treated the upper and lower 2.5 percentiles of the data as outliers (Fig. S2).By considering all of the survey data and removing the top and bottom 2.5 % of respondents, a total of 2938 (39 counties in 5 provinces) were evaluated in this study.

Data analysis
For each experiment, the total GHG emissions, including CO 2 , CH 4 , and N 2 O during the whole life cycle of wheat production, were divided into three components: (1) those emitted during N fertilizer application, including direct and indirect N 2 O emissions, which can be calculated based on the empirical N loss model (see below); (2) those released during N fertilizer production and transportation; and (3) those emitted during the production and transportation of pesticides to the farm gate and diesel fuel use in farming operations such as sowing, tilling, irrigation and harvesting (Supplement Table S2).The impact of the GHG emissions was calculated as CO 2 eq.The 100 yr global warming potential (GWP) of CH 4 and N 2 O are 25 and 298 times the intensity of CO 2 on a mass basis, respectively (Forster et al., 2007).The soil CO 2 flux as a contributor to global warming potential was not included in our analysis, because net flux has been estimated to contribute < 1 % of the GHG emissions from agriculture on a global scale (Smith et al., 2007).The change in soil organic carbon content was also not included in our analysis because it was difficult to detect such a small magnitude of change over a short time (Conant et al., 2010).
We used values in the published literature to simulate the relationship between N loss and N application rate and to estimate GHG emissions from N fertilization.Total N 2 O emissions included both direct and indirect emissions.Indirect emissions were estimated using a method of the IPCC (IPCC, 2006), where 1 and 0.75 % of ammonia (NH 3 ) volatilization and nitrate (NO − 3 ) leaching are lost as N 2 O, respectively.The N losses were calculated based on an empirical model that employs the following equations from Supplement Fig. S3: Direct N 2 O emissions (kg N ha −1 ) = 0.33 exp(0.0054N rate), (1) NH 3 volatilization (kg N ha −1 ) = 0.17 N rate − 4.95, N leaching (kg N ha −1 ) = 2.7 exp(0.0088 N rate). (3) The system boundaries were set using scales in the life cycle from production inputs (such as fertilizers and pesticides), delivery of inputs to the farm gates, farming operations, and wheat harvesting.Using the emission factors for all agricultural inputs given in Supplement Table S2, we calculated total GHG per unit area, expressed as kg CO 2 eq ha −1 , and the GHG intensity, expressed as kg CO 2 eq Mg −1 grain.The relationship between wheat grain yield and GHG emissions at each of the 33 sites in the two cropping systems with either four or five N treatments was determined using the IPNI Crop Nutrient Response Tool (http://nane.ipni.net/article/NANE-3068) and the NLIN procedure in SAS (SAS Institute, 1998).We evaluated five models: quadratic, quadratic with plateau, linear with plateau, square root, and spherical with plateau.In most cases, all five models significantly fit the data (P < 0.01), and had similar coefficients of determination (R 2 ).Considering the continuity and smooth simulation, we chose the spherical with plateau model for all of the sites (Cerrato and Blackmer, 1990).We determined the minimum GHG emissions needed to achieve maximum grain yield as the inflection point of the curve (Cerrato and Blackmer, 1990).

Results
Considering all 33 locations, wheat grain yield averaged 5993 kg ha −1 in the median N treatments (201 kg N ha −1 ) of CP systems.For the HY system, grain yield averaged 8662 kg ha −1 (208 kg N ha −1 ), which was 45 % (∼ 2669 kg ha −1 ) higher than that of the CP systems.Correspondingly, grain yield with no N control in the HY system averaged 6244 kg ha −1 , which was 43% (∼ 1866 kg ha −1 ) higher than a grain yield of 4378 kg ha −1 from the CP system (Table 1).Although a large difference in grain yield was observed between the CP and HY systems, there were no differences in soil properties and soil type (Supplement Table S1).

Relationship between wheat yield and GHG emissions for different management systems
Pooling data from all 33 experimental sites receiving either four or five N treatments, the relationship between wheat grain yield and GHG emission fit a spherical-plateau model (P < 0.001; Fig. 1).The minimum GHG emissions needed to achieve maximum grain yield was 3555 and 3905 kg CO 2 eq Mg −1 for the CP and HY system.In contrast, the corresponding grain yield for the HY system was 8429 kg ha −1 , 39 % greater than the 6077 kg ha −1 for the CP system.The GHG emission intensity reduced by 21 % from 585 kg CO 2 eq Mg −1 for HY system to 463 kg CO 2 eq Mg −1 for CP system.Large site-specific variations in GHG emission and grain yield were observed across the 33 experimental sites (Table 2).Calculated minimum GHG emissions needed to achieve maximum grain yield for the CP system by spherical with plateau model ranged from 2736 (S11) to 5475 kg CO 2 eq ha −1 (S9), similar to the HY system, which ranged from 3055 (S12) to 5476 kg CO 2 eq ha −1 (S15) (Table 2).The corresponding maximum yield for the CP system ranged from 5012 to 7421 kg ha −1 , whereas in the HY system, these values ranged from 7314 to 9598 Mg ha −1 (Table 2).As a result, GHG emission intensity ranged from 456 to 998 kg CO 2 eq Mg −1 for the CP system and from 343 to 652 kg CO 2 eq Mg −1 for the HY system (Table 2).

Opportunity to produce higher yields with reduced GHG emissions
Based on a survey of farmers' practices for 2938 farmers, the N application rate averaged 284 kg N ha −1 and ranged from 77 to 573 kg N ha −1 ; the corresponding grain yield averaged 6050 kg ha −1 with a range from Table 2.The minimum GHG emissions needed to achieve maximum grain yield and the corresponding yields for a conventional practice (CP) and a high-yielding system (HY).

Sites CP system HY system
Mini.GHG Max.

GHG emission
Mini.GHG Max.GHG emission emission yield intensity emission yield intensity kg CO 2 eq ha −1 kg ha −1 kg CO 2 eq Mg −1 kg CO 2 eq ha −1 kg ha 3.44 to 8.31 Mg ha −1 (Fig. 2, Supplement Fig. S2).The calculated GHG emissions averaged 4783 kg CO 2 eq ha −1 (Fig. 2), of which 1183 kg CO 2 eq ha −1 was attributable to field management (e.g., irrigation, tillage, and harvesting), 1270 kg CO 2 eq ha −1 was from N fertilization, and 2330 originated from N production and transport.Calculated GHG emission intensity averaged 807 kg CO 2 eq Mg −1 .The GHG emissions ranged from 2106 to 10757 kg CO 2 eq ha −1 with a variance of 38 %, whereas GHG emission intensity ranged from 382 to 1795 kg CO 2 eq ha −1 with a variance of 39 % (Fig. 2).Compared to average farmers' practices (point A), the minimum GHG emissions needed to achieve a maximum grain yield for CP systems (point B) was reduced by 26 % from 4783 to 3555 kg CO 2 eq ha −1 without any losses in yield (pathway from A to B, Fig. 2).The GHG emission intensity of point B was 585 kg CO 2 eq ha −1 , which was only 74 % of current practices (point A).With the HY system, grain yield increased to 8429 kg ha −1 (or 39% compared to point A) with a GHG emission reduction of 18 % (∼ 3905 kg CO 2 eq ha −1 ) (pathway A to C, Fig. 2).As a result, the GHG emission intensity for point C reduced by 41 % from 807 kg CO 2 eq Mg −1 for point A to 463 kg CO 2 eq Mg −1 , for HY point C.
If food crop yields need to be increased by 100-110 % in the future (Tilman et al. 2011), a wheat yield of 12 Mg ha −1 will be necessary in China.This would require approximately 292 kg N ha −1 (Yue et al., 2012), close to the 284 kg N ha −1 total N rate used under current practices.This indicates that the target yield of 12 Mg ha −1 could be achieved using current N application rates if N losses can be controlled.Thus, GHG emissions from N fertilizer would be similar to or less than the level associated with current practices.A new level for productivity and environmental sustainability should be created for the pathway from point C to D in Fig. 4.

Discussions
While the concept of producing higher yields with less GHG emissions as a goal has been widely debated, studies on crop productively and GHG emission have been notably disconnected in the past (Tilman et al., 2002;Burney et al., 2010;Carberry et al., 2013).Generally, the increasing of N application rate cannot promise a substantial increase in crop productivity because of diminishing returns (Cassman et al., 2003) but increase GHG emission (McSwiney and Robertson, 2005;Hoben et al., 2011;Van Groenigen et al., 2010;Cui et al., 2013ab).Previous studies have focused on how to optimize N management (e.g., appropriate source, timing, placement, or product) to enhance crop recovery of applied N and reduce N losses and GHG emissions (Snyder et al., 2009; Millaret et al., 2010;Cui et al., 2013a, b).For winter wheat systems in China, an in-season root-zone N management strategy can reduce the N application rate by 61 % from 325 kg N ha −1 to 128 kg Nha −1 compared to current practices, resulting in an large decrease in GHG emissions from N fertilizer with no loss in wheat grain yield (Cui et al., 2013b).This result is represented by the pathway from point A to B in Fig. 2.Although these practices represent a large step forward, increasing rather than merely maintaining grain yield, they also present a fundamental challenge.
In intensive cropping systems, the more efficient cycling of N depends on environmental management interactions that influence the balance and rate of microbial processes (e.g., nitrification and denitrification) and transport among plant, soil and environments (e.g., air and water) (Robertson and Vitousek, 2009).When a high-yield system was adopted in a previous study, crop health, insect and weed management, moisture and temperature regimes, supplies of nutrients other than N, and use of the best-adapted cultivar or hybrid all contributed to more efficient uptake of available N and greater conversion of plant N to grain yield, therefore reducing reactive N losses and GHG emissions (Cassman et al., 2003;Cui et al., 2013b).
In the HY system of the present study, the better combination of adopted varieties, planting data, and planting quality was determined to optimize the crop canopy, and this maximized the use of regional environmental resources (e.g., light, temperature).Yields were increased by 39 %, and GHG emission intensity was reduced by 41 %, compared to current practices.Within the CP system, late sowing and the use of too many seeds often results in excessively large canopies and weak individuals, which lead to high susceptibility to lodging, low efficiency of light capture, small spikes, small grains, and consequently low yields (Xu et al., 2013).
To the best of our knowledge, this is the first on-farm study to report the relationship between wheat grain yields and total GHG emissions.Grain yield increased with increasing GHG emissions before reaching the maximum yield, with the lowest GHG emissions achieved when emission intensity decreased, indicating a trade-off relationship between high yields and GHG emissions (Figs. 2 and 4S).In this study, grain yield in the HY system increased by 39 % while GHG emissions increased by only 10 %, and GHG emission intensity was reduced by 21 %, compared to the CP system.This new paradigm for productivity and environmental sustainability is currently being extended to farmers throughout the cereal crop production area in China, but it also appears to be relevant for other high-yield cropping systems outside China.For example, in UK wheat production, GHG emission intensity is 313 kg CO 2 eq Mg −1 of grain, and grain yield is about 10 Mg ha −1 (Berry et al., 2008).Maize in central Nebraska achieves higher grain yields (13.2 Mg ha −1 ) with lower GHG emission intensity (231 kg CO 2 eq Mg −1 of grain) (Grassini and Cassman, 2012).
In the future, yields must be doubled to meet the growing food demands of an ever-increasing population, without further compromising environmental integrity; therefore, new frontiers for food and environmental sustainability must be created (from point C to D in Fig. 2).Most see this pathway being met by genetically modified crops (Phillips, 2010).Yet obtaining substantially higher yields without further depleting soils, destroying natural habitats, and polluting air and water will demand a comprehensive approach.(Zhang et al., 2013).In reality, pushing the boundaries of productivity will likely evolve from the synergies between novel plant genetics, innovative management technologies, and increasing soil fertility (Keating et al., 2010).Moving millions of smallholder farmers to new productivity and environmental protection paradigms will require research into, and the delivery of, new technologies that increase production at much the same level of investment.

Conclusions
The current relationship between wheat yield and GHG emissions due to farmers' practices can be reversed for highyielding systems using innovative management technologies, and a new paradigm of productivity and environmental sustainability can be created to produce higher yields while reducing GHG emissions.In this study, we increased yield by 39 % and reduced GHG emission intensity by 41 %, compared to current practices.In the future, there will need to be an eco-efficiency agricultural revolution, with large increases in grain yields complemented with reduced GHG emissions.A win-win outcome for agriculture and emissions will require eco-efficient solutions that create new productivity and environmental frontiers to achieve food and GHG security.

Fig. 2 .
Fig. 2.A stylized grain yield-GHG emission framework demonstrating three pathways to produce higher yields with less GHG emissions.The gray dots represent grain yields and GHG emissions for the 2938 farmers surveyed.The line of dashed line (1) and solid line (2) mean relationship between grain yield and GHG emission for CP and HY system, respectively.Point A is the average for all farmers; points B and C are the minimum GHG emissions for maximum grain yield with the CP and HY system, respectively (the details are shown in Fig.3); and point D represents the target of 12 Mg ha −1 of wheat grain yield in the future.