Main

As the world transitions to an economy with net-zero greenhouse gas (GHG) emissions, those from the agriculture sector will be among the most difficult to abate. In contrast to sectors where emissions are dominated by fossil fuel combustion, emissions from crop production are dominated by biological processes, including carbon dioxide from land use change, methane from rice cultivation and nitrous oxide (N2O) from soil management. In regions that are no longer converting land to agriculture, cropland GHG emissions are dominated by methane and N2O (refs. 1,2). For upland crops, including soybean, N2O from soil management is responsible for the vast majority of emissions3. A life cycle assessment of soybean production indicated that 67% of emissions are attributable to N2O (Supplementary Figs. 6 and 7)4.

Because N fertilizer inputs are a major source of N2O emissions5,6 and a routine component of crop management, strategies to reduce N2O emissions focus on improving N fertilizer management7,8. Optimum placement, timing, type and rate of N fertilizer can result in substantial N2O emissions reductions. Of these factors, the type and rate of N fertilizer input have the most consistent effects on N2O emissions. Enhancing the efficiency of the use of N fertilizers can reduce N2O emissions by >30% (ref. 9). Moreover, applying the right N fertilizer rate is critical to limiting N2O emissions because N2O emissions increase with N fertilizer rate, especially when N fertilizer input exceeds crop demand10,11. As a result, rotating between crops that do and do not require N fertilizer can be a highly effective strategy for reducing emissions (Supplementary Fig. 1).

Nevertheless, crops that do not receive N fertilizer and the periods of a cropping cycle that are far removed from the timing of N fertilizer inputs are important sources of N2O (ref. 12). Consistent with these observations, N fertilizer input is not the sole factor affecting N2O emissions from annual cropping systems13. Emissions are also promoted by low oxygen availability and high soil organic matter14. High soil water content increases the microbial mineralization of soil organic matter, which consumes oxygen and releases inorganic N, which is the substrate for N2O production. In the main crop-producing region of North America, soils often contain >10,000 kgN ha−1 of organic matter and mineralization exceeds 100 kgN ha−1 y−1 (ref. 15).

An important portion of global maize- and soybean-producing regions have long fallow periods when inorganic N release from the microbial mineralization of soil organic matter exceeds plant N demand and rainfall exceeds evapotranspiration, thereby promoting the production of N2O (ref. 12). In these systems, N2O emissions are elevated when soils are warm and wet but crops are not growing12,16,17. These environmental conditions occur in most maize–soybean production systems in the world, including North and South America, Asia and Europe. In these systems, high N2O emissions are observed before, after and early during soybean growth16,17,18,19. Given the lack of N fertilizer inputs to soybean and the timing of N2O emissions, which are relatively evenly distributed across fallows and soybean growth, strategies to reduce GHG emissions during maize production cannot be applied to address emissions during soybean production (Supplementary Fig. 6). Therefore, an approach that integrates several interventions to address both the crops and the fallow periods between them is necessary to reduce emissions from this cropping system.

Crop genetics and agronomic management can be used to develop interventions that reduce N2O emissions by synchronizing soil N mineralization from organic matter and plant N demand. Advancing the soybean planting date and shifting to soybean genotypes with longer maturities can extend the time and amount of N uptake. This strategy warrants testing because there is an immediate opportunity to implement this strategy in North America, where maize planting is prioritized and, as a result of logistical limitations, soybeans are typically planted two weeks later than possible20,21,22,23,24. Indeed, the economic opportunity of earlier planting is becoming appreciated. Studies suggest that a large region of US soybean production would respond positively to earlier planting24. It is important for farmers and crop insurance programmes that set the earliest allowable planting dates to respond to opportunities for earlier planting because N2O emissions from fallows could increase as a result of the warmer springs associated with climate change25.

Over longer time frames, the breeding of cold-tolerant soybeans, which is an active area of research26, could further advance the soybean planting date beyond what is currently possible due to logistical constraints associated with maize prioritization23. However, breeding for cold tolerance must address several factors that reduce the yield under cold conditions, including frost tolerance and blossom drop during cold spells at flowering27.

Another agronomic option that reduces fallows and increases plant uptake of nitrogen from mineralized soil organic matter is planting cover crops. Cover crops, which are grown between maize and soybean, primarily to reduce erosion and generate plant N uptake during fallows, have been widely tested as an N2O emission-reduction practice. The current literature shows that legume cover crops generally increase N2O emissions with respect to fallow but that non-legume cover crops can reduce emissions28,29. Nevertheless, optimum cover-crop management for N2O emissions abatement has not been well characterized. Cover-crop management strategies that maximize growth, such as early sowing by aerial application into a standing maize crop, can increase N uptake and potentially augment N2O emissions reductions.

Here we address two questions: (1) What proportion of N2O emissions from the maize–soybean rotation occur during the unfertilized soybean phase of the system? (2) Can changes to soybean genetics and management, which are widely available and easily implemented by farmers, make substantial reductions in N2O emissions during the soybean phase of the rotation?

To quantify the proportion of N2O emissions from the soybean phase of the system, we conducted a systematic review of in situ measurements of N2O in side-by-side plots with rotated maize and soybean crops across the United States and Canada. To estimate the potential of agronomic strategies to reduce N2O emissions during the soybean phase, we used published experimental data and the Agricultural Production Systems Simulator (APSIM) across 15 locations in four main soybean-producing states in the US Midwest30. We simulated the effects of shortening fallows by planting longer-maturity soybean genotypes earlier in the spring and sowing non-legume cover crops immediately following maize physiological maturity but before harvest and then terminating them just before soybean planting. Finally, we performed a life cycle analysis with the Feedstock Carbon Intensity Calculator (FD-CIC)4 to estimate the impact of our strategies on carbon intensity scores for soybean production.

Results

N2O emissions during soybean production

We identified 16 studies comprising 131 paired datasets of N2O emissions from maize and soybeans in the same year. All studies measured N2O during the frost-free season (planting to harvesting), and seven studies (47 pairs) considered the full year. Of the 131 paired observations, 54 used management practices that the authors described as ‘conventional’ or ‘business as usual’ (BAU). These included recommended N fertilizer rates and other management practices typical for the location. The remaining 77 paired observations used management practices that the authors identified as ‘alternative’ or non-BAU, such as N fertilizer rates above or below local recommendations or with differences in source, time of application, use of organic fertilizers, tillage system, cover cropping or crop-residue management.

Under BAU management in the maize–soybean cropping system, mean emissions from maize were higher than soybean (3.9 ± 0.36 versus 2.0 ± 0.24 kgN2O–N ha−1 yr−1, n = 54, P < 0.001; Fig. 1). About 34% of the N2O emissions from the two-crop rotation occurred in the soybean phase. In three long-term experiments that had N2O emissions data for ≥10 years but only one crop phase of the rotation present in each year, the contribution of the soybean phase was 35–38% (Fig. 2). When considering only paired observations from BAU systems, no available experimental or management factors explained the variation in the proportion of emissions during each crop phase.

Fig. 1: Annual N2O emissions from 54 pairs of maize and soybean crops (site–treatment–year cases) grown in the same experiment under BAU management conditions in the United States and Canada.
Fig. 1: Annual N2O emissions from 54 pairs of maize and soybean crops (site–treatment–year cases) grown in the same experiment under BAU management conditions in the United States and Canada.
Full size image

The lower, mid and upper hinges of the box correspond to the 25th, 50th and 75th percentiles and whiskers extend to 1.5 times the interquartile range. The triangles show the mean values.

Fig. 2: Contribution of soybean crop to cumulative N2O emissions in the maize–soybean rotation as a function of maize fertilizer rate under BAU and non-BAU management.
Fig. 2: Contribution of soybean crop to cumulative N2O emissions in the maize–soybean rotation as a function of maize fertilizer rate under BAU and non-BAU management.
Full size image

BAU and non-BAU were defined for each study based on the authors’ description of a conventional (BAU) and alternative (non-BAU) practice. The 50% dashed line indicates that both crops contribute equally to the 2-year-rotation emissions. The black line and grey area show the bilinear model (equation (1)) and its 95% confidence intervals. Red dots represent values averaged across years from three long-term experiments that were not included in the analysis and model selection as they did not have both phases of the rotation each year.

However, when considering both BAU and non-BAU paired observations, the N fertilizer rate for maize partly explained the proportion of emissions during each crop phase (Fig. 2). In these observations, the non-BAU data included N fertilizer rates above and below the recommended rate of input (ranging 0–225 kgN ha−1 yr−1). The best model for predicting the contribution of the soybean phase was a bilinear function (equation (1)), where y indicates the soybean contribution (%) and x the maize fertilizer rate (kgN ha−1 yr−1). At zero N input to maize, soybeans accounted for 48% of emissions. As N inputs to maize increased, the proportion of N2O emissions from soybean decreased at a rate of 0.59% per 10 kgN input to maize up to a model breakpoint of 153 kgN ha−1 where 38.6% of the emissions were attributable to soybean. Beyond this point, the proportion of emissions from soybean decreased at a rate of 2.4% per 10 kgN input to maize.

$$y=\begin{cases}47.6-0.059x, & \text{if}\,x\le 153,\\ 75.6-0.242x, & \text{if}\,x> 153.\end{cases}$$
(1)

N2O mitigation strategies for US soybean production

To test the performance of our proposed emission-reduction strategies for soybean, we conducted simulations with a highly calibrated cropping-system model across 15 locations from Indiana to Iowa, 30 years of historical weather data (1991–2020), two crops, and four genetics × management scenarios generating 3,600 annual datasets. From these simulations, we extracted grain yield, annual N2O emissions and annual NO3− leaching. There were four genetic × management scenarios: (1) BAU, (2) cover-crop planting of cereal rye (Secale cereale L.) 30 days before the maize harvest, (3) early planting of soybean and (4) both cover crop + early planting. The planting date was set to the 10th percentile of the maize planting date for the same location or the earliest planting date allowed by government insurance programmes when the 10th percentile of maize planting was earlier. On average, this advanced the planting date by 27 days (21 May versus 24 April) and enabled the use of soybean genotypes with a longer growth period (maturity group 3.5 versus 2.5).

Across all locations and cropping-system scenarios, the mean maize yield at zero moisture was 13.6 Mg ha−1 (range 10.4–15.4 Mg ha−1). As expected, planting cover crops following maize and early planting of soybean had little effect on maize yield (<0.1 Mg ha−1; Supplementary Table 1). In contrast, early planting had large effects on the soybean yield owing to the longer growing season and maturity variety. Soybean yield in the BAU scenario was 3.55 Mg ha−1, whereas in the early planting and cover crop + early planting scenarios, yields were 4.12 and 4.11 Mg ha−1, respectively.

Early planting reduced N2O emissions from soybean on an areal basis by 0.56 kgN2O–N ha−1 yr−1 or 17% compared to BAU (Fig. 3a). The cover crop reduced N2O emissions by 0.66 kgN2O–N ha−1 yr−1 or 20%. When early planting and the cover crop were combined, the reduction in N2O was partly additive leading to a reduction in emissions of 0.93 kgN2O–N ha−1 yr−1, which is a 28% reduction (Fig. 3). Early planting of soybean led to a sustained decrease in emissions throughout the growing season. In contrast, the reduction in N2O emissions from cover crops was predominantly observed in early spring before soybean planting, and this decrease more than offset a small increase in N2O emissions from cover crops during maize growth (Fig. 4). The small increase in N2O emissions was driven by inorganic N release from decomposing cover crops whereas the reduction in the N2O emissions was driven by a reduction in soil NO3− concentration following maize through the early stages of soybean growth (Supplementary Fig. 3).

Fig. 3: Change in N2O emissions during the soybean phase of the rotation with respect to BAU management.
Fig. 3: Change in N2O emissions during the soybean phase of the rotation with respect to BAU management.
Full size image

a,b, Change expressed by area (a) and yield-scaled (b) basis. Each box represents simulations with the APSIM model for 30 years of maize–soybean rotation in 15 sites in the US Midwest (n = 450). The lower, mid and upper hinges of the box correspond to the 25th, 50th and 75th percentiles and whiskers extend to 1.5 times the interquartile range. The triangles show the mean values.

Fig. 4: Simulated daily N2O emissions during the maize–soybean rotation for Sutherland, IA.
Fig. 4: Simulated daily N2O emissions during the maize–soybean rotation for Sutherland, IA.
Full size image

a–d, Black lines and shaded bands represent the average and 95% confidence intervals across 30 years. Coloured areas indicate the mean growing season of the crops (from planting till harvesting or termination of the cover crop). In the simulations, the planting and harvesting dates changed each year based on USDA survey data at state level. a, Emissions under BAU management. b–d, Change in the N2O emissions from including cover crops (b), advancing the soybean planting date (c) and doing both together (d) with respect to BAU management.

The 16% increase in soybean yield with early planting combined with the reduced area-scaled N2O emissions produced a large decrease in yield-scaled N2O emissions of 0.34 kgN2O–N per Mg grain, which represents a 30% reduction compared to BAU. Planting of the rye cover crop before the maize harvest also reduced yield-scaled emissions by 0.12 kgN2O–N per Mg grain because the cover crop reduced area-scaled N2O emissions without affecting the soybean yield (Fig. 3b). Together, the effects of early planting and cover crops were partly additive and reduced yield-scaled emissions by 0.44 kgN2O–N per Mg grain or 39%.

The cropping-system scenarios had little effect on NO3− leaching in maize but significantly decreased NO3− leaching from soybean. Across all locations, mean NO3− leaching was 24.7 kgN ha−1 yr−1 in BAU, 22.3 kgN ha−1 yr−1 with early planting, 16.4 kgN ha−1 yr−1 with cover crops and 15.9 kg N ha−1 yr−1 with both cover crops + early planting. The reductions were the result of better synchrony between plant N uptake and soil N mineralization.

Reductions in NO3− leaching are expected to produce further downstream reductions in N2O emissions, as the Intergovernmental Panel on Climate Change (IPCC) tier 1 emission factor approximates that 1.1% of NO3− losses are emitted as N2O. Early planting of soybean produced a moderate increase in NO3− losses from maize owing to greater N in crop residues (+2.56 kgN ha−1 or 13%; Supplementary Table 1). In contrast, early planting and cover crop had negligible effects on N2O emissions from maize (<0.10 kgN ha−1; Supplementary Table 1).

Carbon intensity of grain production

We used the FD-CIC to estimate total GHG emissions on a carbon dioxide equivalent (CO2eq) basis from soybean production in the four cropping systems4. The FD-CIC accounts for GHG emissions that are not output by the APSIM process model, including fuel use, embedded emissions in fertilizers and pesticides, and tier 1 estimates of downstream N2O emissions from NO3− leaching. In addition to the modelled N2O emissions, management scenarios increased fuel use for cover-crop management and altered indirect N2O emissions associated with changes in NO3− leaching. In the FD-CIC, cover-crop management increased diesel fuel use by 4.68 L ha−1 and 1.1% of leached NO3− was emitted as N2O. We assumed the management practices had no effect on soil carbon stocks.

The substantial reduction in NO3− leaching due to cover crops increased the efficacy of this practice for GHG emissions reductions. The reduction in CO2eq from reduced NO3− leaching more than fully offset the increase in CO2eq owing to the fuel use required for cover-crop management. Averaged across all locations, cover crops + early planting reduced the life cycle GHG emissions of soybean production by 33% (Table 1).

Table 1 Life cycle assessment of GHG emissions from the soybean phase of the maize–soybean cropping system for the four management scenarios
Full size table

Discussion

N2O emissions across the maize–soybean cropping system

At rational N fertilizer rates, total N2O emissions from the maize–soybean cropping system are substantial. Approximately 40% of those emissions occur during the soybean phase of the system (Fig. 1). Moreover, reducing N fertilizer inputs below rational rates has little effect on the distribution of emissions across the maize and soybean phases of the system (Fig. 2). Together, these results emphasize that a systems approach integrating a variety of genetic, environmental and management factors is required to reduce emissions from the maize–soybean cropping system.

The relation between the proportion of N2O emissions occurring during the soybean phase of the system and the N fertilizer input to maize can be explained by the exponential response of N2O emissions to the N fertilizer rate in maize10. Previous studies have shown that N2O emissions during the maize phase of cropping systems increase with N rate and that the rate of increase grows when N inputs exceed crop demand11,31. This result has important implications for the abatement of N2O emissions in maize- and soybean-producing regions around the world. In regions with large N surpluses, such as China, improvements to N use efficiency remain an important step towards emissions reductions5. However, in regions where N fertilizer inputs are already at or below the agronomic optimum, such as South America, emission-abatement strategies will require greater focus on the soybean phase of the system32,33.

Our literature synthesis identified a breakpoint of 153 kgN ha−1, at which the proportion of N2O emissions attributable to maize increased fourfold from 0.59% to 2.4% per 10 kgN ha−1 (Fig. 2). This N fertilizer rate is rational although on the lower end of recommendations for North American maize–soybean cropping systems34, which is not surprising given that partial factor productivity of nitrogen fertilizer decreases with increasing N inputs (production per unit N input). Hence, reductions in N fertilizer input below this rate would create a substantial risk of yield loss and negative N balances that lead to the mining of soil organic matter. North American maize yields routinely surpass 11 Mg ha−1 at approximately 1.2% N, resulting in a N harvest of 132 kgN ha−1. Substantial evidence also indicates that, in the US Corn Belt, N outputs in a soybean grain harvest typically exceed N inputs from biological N fixation35,36, which emphasizes the need to consider sustainable levels of N fertilizer input.

The sizeable N2O emissions at rational N fertilizer inputs and their timing across both the maize and soybean phases of the crop system demand emission-reduction strategies beyond N fertilizer management. Soil carbon sequestration could partially offset some of these emissions but has limited potential. Under BAU management, which included recommended N fertilizer rates, average annual emissions were 2.95 kgN ha−1 or 1.38 MgCO2eq ha−1 (Fig. 1). For reference, soil C sequestration of 373 kgC ha−1 y−1 would be required to offset these emissions, yet this rate is greater than rates typically observed in maize–soybean cropping systems37,38.

Strategies for reducing N2O emissions must explicitly address all factors affecting emissions, including N fertilizer use, crop residues and the mineralization of soil organic matter. Long fallows between warm-season annual crops are increasingly recognized as key drivers of environmental N losses. High N mineralization rates can occur during these times13,17,39. In regions with high organic matter soils and high precipitation, fallow periods are particularly large contributors to environmental N losses. The notable lack of difference in NO3− losses between the fertilized maize and unfertilized soybean phases of the cropping system40 highlights the importance of managing fallows for reducing N losses. Indeed, soil can lose considerable amounts of N as N2O and NO3− despite a negative nitrogen balance, meaning that N losses are not solely dependent on new N inputs.

In this region, soybeans are net N users (N exported in grain exceeds N imports from fixation)35. Nevertheless, N2O emissions from soybean far exceed that of native perennial grasslands41. All crops, regardless of N balance, including soybean, affect emissions through their biophysical effects on the mineralization of soil organic matter. For example, the amount, timing and quality of crop-residue production affect soil oxygen, N, C and temperature, which in turn affect N2O emissions42.

Conventional strategies based on 4R fertilizer management have limited capacity to address N2O emissions during fallows because they are temporally disconnected from N fertilizer inputs. Similarly, it is challenging for emissions accounting schemes to isolate emissions from individual crops or explicitly account for fallows or interactions between crops43. To address the effects of fallow periods and interactions between crops in rotation, a holistic approach to emissions abatement and accounting is needed.

A systems approach

We have designed and evaluated immediately available opportunities for contemporary crop genetics and agronomic management to increase yield and reduce environmental N losses from the maize–soybean cropping system. Earlier soybean planting with longer-maturity genotypes and the planting of a third, cold-tolerant cereal cover crop increased plant N uptake while reducing soil NO3− availability during fallows. Importantly, cover crops and early planting increased the amount of inorganic N uptake by both increasing the size of the N sink and improving the timing of plant N uptake relative to soil NO3− availability.

Greater soybean yields achieved with earlier planting are a critical factor driving farmer adoption. Despite the importance of biological N fixation, earlier planting of longer-maturity soybean genotypes is effective at reducing N losses because longer-maturity varieties grow longer and because soybeans use large amounts of soil N. Biological N fixation contributes 50–60% of the ~100 kgN required to produce 1 Mg of soybean grain. Soil N contributes most of the soybean N uptake until the reproductive growth stages that occur later in crop growth35. This timing of N use from the mineralization of soil organic matter versus biological N fixation has important implications for N loss reductions with earlier planting because most losses occur early in the growing season when precipitation exceeds evapotranspiration thereby promoting N losses to N2O and NO3−.

Earlier soybean planting is immediately available because current planting dates are later than optimum24. Insufficient availability of mechanized planters and labour leads farmers to prioritize maize planting over soybean because maize has greater input costs and, hence, greater risk. Climate change may exacerbate the delayed planting of soybean if farmers are slow to adapt to warmer springs.

In the longer term, plant breeders and geneticists could reduce the risks associated with earlier planting of maize and soybeans by selecting for cold tolerance. Breeding for cold tolerance in both maize and soybean is one of the most effective and scalable approaches to improving N use efficiency in the maize–soybean cropping system because greater grain yields from longer-maturity crop varieties would drive rapid adoption. Note that breeding for cold tolerance is an important challenge27, but one worth undertaking as climate change and increasing weather variability will increase the value of cold tolerance traits.

Implementation

Adoption is a central challenge for any new intervention, as farmers must weigh the benefits versus the costs in their own environment. Recognizing that there are risks associated with early planting27, the benefits to the environment and farmer profitability may merit modifications to government insurance programmes to encourage earlier soybean planting. In the United States, crop insurance programmes preclude maize and soybean planting before dates set by the Risk Management Agency of the US Department of Agriculture (USDA). These programmes are used in more than 90% of the US maize and soybean planting area44.

Historically, the earliest soybean planting date allowable under federal crop insurance programmes was later than the potential date conferred by modern genetics. In 2023, the earliest soybean planting dates in much of the United States were moved earlier due to the recognition that earlier planting of soybeans would have yield benefits. In some locations, the earliest allowable planting date was moved forward by as much as 21 days. However, the adjustments to allowable planting dates did not consider potential environmental benefits. By recognizing the environmental benefits of earlier planting, the USDA may be willing to accept more risk to reap the environmental rewards.

The USDA could also incentivize earlier planting of cover crops to realize the full effect of cover cropping. Cover crops are generally applicable for government cost sharing through programmes that regulate the latest possible planting dates to ensure effective growth. Demand for cost sharing often outstrips availability. Hence, the USDA could target cost sharing at earlier plantings. Cover crops are already planted using unmanned aerial vehicles, which allow earlier planting into standing maize or soybean crops.

Earlier maize and soybean planting could similarly benefit from autonomous planters that relieve equipment and labour limitations. Lighter planters are being developed to allow earlier field trafficking. In addition, advances in crop-residue management could benefit early planting by improving seedbed conditions and crop establishment45.

As the agriculture sector continues to innovate, there are many near-term and long-term opportunities for reducing fallows through genetics and management. Managing the N fertilizer rate is a necessary first step in limiting N2O emissions from the maize–soybean cropping system but by itself is insufficient (Figs. 1 and 2). At rational N rates, remaining emissions are substantial, and those emissions are distributed across the fertilized maize crop and unfertilized soybean crop (Fig. 2). We have demonstrated how a systems approach that integrates immediately available opportunities in genetics and management to shorten fallows could promote higher yields while significantly lowering total GHG emissions from soybeans on a land area basis (−22% CO2eq) and a productivity basis (−33% CO2eq) (Table 1) without increasing emissions from maize (Supplementary Table 1).

Looking ahead, the warmer and more variable springs associated with climate change could further promote emissions from fallow periods25. Farmers, scientists and policymakers could cooperate to shift the production system to anticipate these changes. The holistic approach we describe herein represents an opportunity for agriculture to mitigate climate change and maintain productivity while adapting to changing weather patterns. We anticipate that early planting and cover cropping are only the first of many examples of mitigation opportunities that will come from broadening the focus of innovation from single crops to cropping systems.

Methods

Soybean contribution to systemic N2O emissions

We searched for all articles in the USDA National Agricultural Library containing the terms: (‘corn’ OR ‘maize’) AND ‘soybean’ AND (‘nitrous oxide’ OR ‘N2O’). From the results, we kept only those that (1) reported a cumulative value of N2O emissions for the whole year or at least the growing season, (2) were measured in field experiments (no greenhouse, growth chamber experiments or simulations), (3) had a fortnightly or higher sampling frequency during the growing season, (4) were for experiments that included both phases of the maize–soybean rotation in the same year and (5) were within the United States or Canada. Within articles, we classified observations between BAU and non-BAU management based on what the authors considered BAU or ‘conventional’. Non-BAU management included, for example, higher and lower N rates, splitting the fertilizer application, use of cover crops and residue removal. We also analysed separately three studies based on long-term experiments (lasting between 10 and 20 years) but with only one phase of the rotation each year. Our database is available in a public repository46.

The 16 studies included in our analysis reported N2O emissions based on experiments spanning 2 to 4 years. Our analysis used annual values rather than averaging across years. Three of the 16 studies did not explicitly state annual N2O emissions in the main text or supplementary information; however, we calculated these based on daily fluxes available in published databases using a linear interpolation between sampling dates (area under the curve). The criteria for annual emissions varied across studies. Nine studies measured N2O fluxes during the growing season or the snow-free period (generally from April to October), whereas seven studies extended measurements closer to the entire year, covering over 250 days per year. Two studies used the calendar year (January to December), four studies accounted from April to March and only one considered annual emissions from mid-October to mid-October of the following year.

To analyse the contribution of the soybean phase to the N2O emissions of the maize–soybean rotation, we calculated the log response ratio (LLR):

$${\rm{LRR}}=\log({{\rm{N}}}_{2}{{\rm{O}}}_{{\rm{soybean}}}/{{\rm{N}}}_{2}{{\rm{O}}}_{{\rm{maize}}}),$$
(2)

where N2Osoybean and N2Omaize represent the annual emissions of the soybean and maize phases of the rotation, respectively.

We tested linear and nonlinear models to evaluate the relation between LRR and nitrogen fertilizer rate in the maize phase of the rotation using the nlraa package in the R statistical language47. The models tested were linear, logistic, plateau-linear and bilinear and were compared with the Akaike information criterion (Supplementary Tables 2 and 3). Logistic and declining logistic models were considered, but these were not suitable. The bilinear model had the lowest Akaike information criterion value. We performed the statistical analysis using LRR as the response variable and then reported the results in terms of the contribution of the soybean phase to the total N2O emission as a percentage. The uncertainty associated with the bilinear model was assessed by estimating the regression and confidence bands through two different methods: bootstrap and leave one out. Both produced similar results (Supplementary Fig. 2). We attempted to account for the model residuals with a variety of potential explanatory variables, such as location, precipitation, temperature, and frequency and duration of N2O measurements, but no explanatory variables improved the model fit. We used MS Excel and the R statistical language48 for data curation and management and the ggplot2 package for generating figures49.

Simulations

The management-based mitigation strategies were analysed using APSIM (Next Generation v.2022.12.7130.0)30. APSIM is a process-based model and has been widely tested and used in the US Midwest. It has attained good performance accuracy in estimating maize and soybean yields and environmental aspects15,50,51. It simulates soil N2O emissions through nitrification and denitrification processes52. Here we further calibrated and evaluated the performance of the model to simulate N2O emissions and the effect of different management systems in Minnesota, Iowa, Illinois and Indiana. The evaluation included a comparison of simulated with observed data derived from 13 published field experiments comprising a total of 248 treatment–site–year cases of annual N2O emissions (in kgN–N2O ha−1 yr−1). The simulations adopted the specific soil, weather and management data for each experiment. After running the simulations, we extracted the crop yield as well as daily and cumulative N2O emissions and compared these with field data. We also extracted the soil temperature, water content and mineral N concentration, if they were available in the database measured. We tested the performance of the model in predicting the annual N2O emissions using the metrica R package53. The model predicted the annual emissions in those 248 cases with a root-mean-square error of 1.76 kgN ha−1 yr−1 and R2 = 0.64. The intercept was 0.13 kgN ha−1 yr−1 and the slope 1.08 (Supplementary Fig. 4). A comprehensive account of the model’s evaluation can be found in the Supplementary information (Supplementary Table 4 and Supplementary Figs. 4 and 5).

Following the evaluation, we conducted a simulation experiment across 15 distinct sites where the APSIM model had been calibrated previously for Iowa (six sites) and Illinois (five sites)51 and for Minnesota (two sites) and Indiana (two sites)50. This simulation encompassed a 30-year maize–soybean rotation, with both phases of the rotation present each year. Real weather data from 1991 to 2020 were obtained from the closest weather station. The planting and harvesting dates were calculated for each of the 30 years of the simulations as the 50th percentile from a survey by the National Agricultural Statistics Service of the USDA at the state level. Thus, we used the corresponding planting and harvesting dates for each year and state. The mean planting dates averaged across states and years were 7 May for maize and 21 May for soybean. Mean harvesting dates were 19 October for maize and 9 October for soybean. The simulated N fertilizer rate for maize was derived from official state extension service recommendations34. The rates were 160, 165, 200 and 220 kgN ha−1 for Minnesota, Iowa, Illinois and Indiana respectively. Consistent with local practices, soybeans received no N fertilizer. This baseline management was considered to be BAU.

The improved management strategies were cover crop, early planting and combined cover crop + early planting. These were simulated as follows. The early planting date for soybeans was set equal to the 10th percentile of maize planting dates in the National Agricultural Statistics Service survey for each year in each state (the average planting date across states was 24 April). A university extension recommends the same planting dates for maize and soybean20,21,54. In 80% of these cases (sites × years), the 10th percentile of maize planting was after the earliest planting date permitted and did not invalidate the crop insurance date. When the 10th percentile of maize planting was earlier than permitted for crop insurance coverage, we set the planting date to the required date. This occurred only in Illinois and Minnesota. For cover-crop management, we simulated the planting of the cereal rye as a winter cover crop 30 days before maize harvest with herbicide application terminated 15 days before soybean planting.

Life cycle assessment

To estimate total GHG emissions from soybean production, we used FD-CIC, which is a derivative module of the Greenhouse Gases, Regulated Emissions, and Energy use in Technologies model4. Both were developed at Argonne National Laboratory, US Department of Energy. The FD-CIC uses primary literature and publicly available data to estimate GHG emissions associated with the field-level production and harvesting of crops. The output is total emissions in units of CO2eq per unit crop production. The model contains crop-specific default inputs for crop yield, energy, pesticides and fertilizers. Nitrous oxide emissions were estimated from emissions factors. We substituted the default grain yield and the emission factor-based N2O emissions with outputs from the APSIM model simulations. The N2O emissions included simulated emissions from the soil surface plus 1.1% of simulated NO3−–N leaching, which was assumed to be emitted as N2O based on the IPCC tier 1 emission factor55. In scenarios with cover crops, we augmented the diesel fuel consumption to account for planting and termination of the cover crop. Otherwise, we used default inputs for energy, fertilizers and pesticides.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.