Nitrogen rate strategies for reducing yield-scaled nitrous oxide emissions in maize

As part of recent international climate change agreements, more than 100 countries have pledged to reduce non-CO₂ emissions from agriculture. While this marks significant progress in climate change mitigation policy, it is clear that greater efforts are needed to meet on-farm greenhouse gas (GHG) reduction targets (Davidson 2012, Wollenberg et al 2016). Soil nitrous oxide (N₂O) emissions have a radiative forcing potential roughly 300 times larger than CO₂ and agriculture contributes approximately 75% to total anthropogenic N₂O emissions in the US (Davidson et al 2014). Given that global N₂O emissions from agriculture are expected to increase by 35%–60% by 2030 (FAO 2003), region- and crop-specific N₂O mitigation strategies are needed, yet large uncertainties remain about the effectiveness of different approaches and potential tradeoffs with other sustainability indicators, including potential impacts on crop productivity (Decock et al 2015).

Inventories of agricultural N₂O emissions are often based on the strong relationship between direct N₂O emissions and fertilizer N inputs (Grace et al 2011, Philibert et al 2012, Shcherbak et al 2014, Gerber et al 2016). However, because N fertilizer is also important for sustaining crop yields, an integrated understanding of yield impacts related to changes in N fertilizer management is needed to evaluate options for simultaneously meeting future food security and climate change goals (Van Groenigen et al 2010, Abalos et al 2016). One method is to evaluate N₂O emissions in terms of crop productivity using 'yield-scaled' emissions rather than per unit area (Van Groenigen et al 2010). Nitrogen rate reduction strategies which simultaneously maintain yields while reducing N₂O emissions are lacking, particularly for some of the world's most productive cropping systems which receive high levels of fertilizer N inputs and contribute disproportionately to global N₂O budgets (Gerber et al 2016).

The US Midwest accounts for approximately 30% of global maize production on 16% of global maize production area (FAOSTAT 2016, USDA-NASS 2016). Large fertilizer N inputs in this region have led to significant environmental consequences, particularly related to air and water quality (David et al 2010, Turner et al 2015). Due to fertilizer-induced N₂O emissions representing roughly 60% of total global warming potential in these cropping systems (Grace et al 2011), carbon offset protocols have recently been developed for this region based on N fertilizer rate reductions (Millar et al 2010, 2013). Such protocols rely on an economic approach to N management using the maximum returns to N rate (MRTN) calculator described below. In these offset protocols, carbon credits are generated when N rates identified by the MRTN calculator are lower than baseline N rates currently applied by farmers, resulting in a corresponding decrease in predicted N₂O emissions.

Recent studies have highlighted the potential to reduce agricultural N losses through N rate reductions, yet these assessments are based on aggregated data and there is little information regarding yield impacts at the field-scale. Importantly, farmers are risk averse, thus if incentives or policies to reduce N losses are associated with crop yield penalties, the likelihood of them being adopted by farmers will remain limited. It has been suggested that state-level reductions in fertilizer N rate resulting from widespread adoption of the MRTN approach could reduce N₂O emissions in the US Midwest by more than 50% without impacting yields (Millar et al 2010). Similarly, an analysis of several generalized yield response functions found that N recovery efficiency (NRE) could be improved by approximately 30% without penalizing yields or economic profits based on alternative N sources (Zhang et al 2015a). Conclusions from global studies indicate that N rates can be reduced by matching N inputs with N exported in the harvested product, resulting in zero N surplus (Foley et al 2011). When performing assessments to inform policy at broad spatial scales, the difficulty in balancing N management goals is often acknowledged, yet few studies have quantified the degree to which N loss reductions can be achieved without sacrificing crop yields at the field-scale.

To aid the development of effective and scalable N₂O mitigation strategies for the US Midwest where current protocols are based on voluntary farmer adoption, N rate reduction approaches that maintain yields while reducing N₂O emissions as a function of yield (i.e. yield-scaled N₂O emissions) are required. The objective of this study was to evaluate the impacts of two economic and two environmental N rate reduction strategies on maize yield, economic returns, and yield-scaled N₂O emissions by combining an empirical N₂O emission function with yield results from 201 on-farm N response trials in Illinois.

2.1. On-farm N response trials

2.2. Estimated N₂O emissions

2.3. N rate reduction strategies

2.4. Data analysis

The response of maize yield to N rate varied widely across the 201 on-farm experiments (figure 1(a)). Average maize yields were 6.4 Mg ha⁻¹ without N fertilizer and reached 11.8 Mg ha⁻¹ at an average EONR of 171 kg N ha⁻¹. Average yield-scaled N₂O emissions remained below approximately 0.18 kg N Mg⁻¹ at N rates lower than 171 kg N ha⁻¹, but increased rapidly to more than 0.3 kg N Mg⁻¹ at 250 kg N ha⁻¹ (figure 1(c)). Average agronomic N use efficiency (ANUE; kg grain yield increase per kg N applied) decreased steadily per unit of N applied, declining from greater than 60 kg kg⁻¹ to less than 23 kg kg⁻¹ at 250 kg N ha⁻¹ (figure 1(d)). Average apparent NRE also decreased steadily per unit of applied N, reaching 33% at 171 kg N ha⁻¹ across all sites (figure 1(e)). Nitrogen rates below 107 kg N ha⁻¹ on average resulted in a negative N surplus (figure 1(f)), whereas N rates above this contributed to average N surplus values which continued to increase up to a maximum of 130 kg N ha⁻¹ at 250 kg N ha⁻¹.

Zoom In Zoom Out Reset image size

Across all sites, the average N rate for MRTN_high and MRTN_low was 207 and 178 kg N ha⁻¹, respectively. MRTN_low increased returns by a median value of 16 USD ha⁻¹ compared to MRTN_high, although there was substantial variability among sites (lower and upper quartiles were −10 and 25 USD ha⁻¹, respectively) (figure 2(a)). While MRTN_low had small negative impacts on yield at the majority of sites (figure 2(a) inset), it decreased N rates and N₂O emissions by median values of 15 and 18%, respectively (figure 2(b)). In addition, MRTN_low increased NRE by 14% and decreased N surplus by 32%, while significantly reducing yield-scaled N₂O emissions across sites (figure 2(c)). The median reduction in yield-scaled N₂O emissions within each site was 17%, with a low amount of variability occurring across sites (figure 3).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

When returns to N fertilizer were maximized within each trial using EONR, the average N rate was 171 kg N ha⁻¹. Compared to MRTN_high, EONR increased returns by a median value of 36 USD ha⁻¹ (first and third quartiles were 16 and 65 USD ha⁻¹, respectively) (figure 4(a)), yet yield impacts were negligible (generally within 1% of MRTN_high; figure 4(a) inset). EONR tended to decrease N rates and N₂O emissions, while increasing NRE and decreasing N surplus, with greater variability occurring across sites (figure 4(b)). Similarly, although EONR significantly decreased yield-scaled N₂O emissions across sites (figure 4(c)), there was much greater variability compared to MRTN_low. Within each site, EONR decreased yield-scaled N₂O emissions by a median value of 19% (figure 3), but decreases were only observed in approximately 75% of cases, whereas increases were observed in about 25% of cases (i.e. upper quartile).

Zoom In Zoom Out Reset image size

Increasing NRE by 30% above that occurring at the baseline N rate within each trial resulted in an average N rate of 143 kg N ha⁻¹ for T-30_NRE. Across sites, returns for T-30_NRE compared to MRTN_high were highly variable (figure 5(a)). Yield decreases with T-30_NRE were common, with yield reductions of more than 5% observed at 75 sites, more than 10% at 45 sites, and more than 20% at 15 of 201 sites (figure 5(a) inset). Within each site, T-30_NRE decreased N rates and N₂O emissions by 25 and 30%, respectively, while increasing NRE by (the goal of) 30% at the majority of sites and decreasing N surplus by 63% (figure 5(b)). Across sites, T-30_NRE significantly reduced yield-scaled N₂O emissions (figure 5(c)), with a median value of −29% observed within each site and relatively low variability occurring across sites (figure 3).

Zoom In Zoom Out Reset image size

Balancing N inputs with harvested grain N outputs to achieve ZNS within each trial led to an average N rate of 104 kg N ha⁻¹. Compared to MRTN_high, ZNS reduced returns by a median value of 50 USD ha⁻¹ (lower and upper quartiles of −196 and 28 USD ha⁻¹) (figure 6(a)). Negative yield impacts greater than 5, 10, and 20% occurred at approximately 61, 39, and 21% of sites, respectively (figure 6(a) inset). Meanwhile, N rates and N₂O emissions were reduced by median values of −48% and −49%, respectively (figure 6(b)). By definition, ZNS reduced N surplus by 100%, with a corresponding increase in NRE by a median of 48%. ZNS reduced yield-scaled N₂O emissions to the greatest extent across sites (figure 6(c)), with a median value of −41% within each site (figure 3).

Zoom In Zoom Out Reset image size

There is a critical need to develop agricultural intensification strategies which optimize crop productivity at N input levels that are economically and environmentally sustainable. Expressing N₂O emissions in terms of crop productivity rather than per unit area is a simple but informative tool for identifying N application rates needed to optimize yields while lowering N₂O losses per unit yield (Van Groenigen et al 2010). Over a wide range of soil and climate conditions in Illinois, we found that the minimum value for yield-scaled N₂O emissions occurred at 41 kg N ha⁻¹ which is well below the average EONR for this dataset (171 kg N ha⁻¹) (figure 1). This finding highlights the inherent tension between achieving high yields with minimum environmental impacts for maize production in this region. Still, with yield-scaled N₂O emissions accelerating as N rates increased above 41 kg N ha⁻¹, values further increased by 66% when moving from the economic optimum of 171–250 kg N ha⁻¹. The relationship between rapidly increasing yield-scaled N₂O emissions and excessive N rates is consistent with previous findings (Van Groenigen et al 2010, Grassini and Cassman 2012, Linquist et al 2012, Pittelkow et al 2014), providing further support for the concept that large increases in negative environmental consequences can be avoided by reducing excessive N inputs and integrating agronomic goals with environmental risks.

Despite the large variability in yield response across sites, our results indicate that MRTN_low is an effective strategy for modest but consistent reductions in yield-scaled N₂O emissions relative to MRTN_high. While MRTN_low generally had small negative impacts on yield, median returns increased slightly by 16 USD ha⁻¹, meaning this N rate reduction (15% median) decreased costs to a greater extent than yields in approximately 64% of cases, while also enhancing NRE and decreasing N surplus and N₂O emissions (figure 2). Given the increasing emphasis on N₂O mitigation in the US Midwest, including new N management credit programs funded by the Natural Resources Conservation Service, US Department of Agriculture, our finding that field-level economic returns can be highly variable relative to MRTN_high is noteworthy. Although median yield decreases were <1%, which is in agreement with recently developed carbon offset protocols based on the MRTN approach (Millar et al 2013), a wide range of both positive and negative returns occurred across sites relative to MRTN_high which may have implications for farmer adoption. An important limitation of the current study is the assumption of MRTN_high as the baseline N input level (see supplementary information for a discussion), highlighting the need for more information on field-level N management practices to accurately evaluate different options for sustainable intensification.

In contrast to the MRTN approach based on multi-site averages, there is great interest in developing soil tests, crop models, or in-season measurement tools to determine EONR earlier in the season when N management decisions are being made (Scharf 2001, Scharf 2015, Williams et al 2007, Zhang et al 2015a). In theory, such technologies would allow farmers to optimize N rates on a field-by-field basis each year, increasing sustainability by minimizing environmental N losses. Our analysis demonstrates the strong economic advantage of implementing EONR compared to MRTN_high, if it were to become possible in the future. Although EONR slightly decreased median N rates and N₂O emissions, there was a surprising amount of variability across sites, and EONR actually increased N₂O emissions, N surplus, and yield-scaled N₂O emissions in roughly 25% of cases (figures 3 and 4). This generally occurred when higher rates of fertilizer N were required to reach optimum yields compared to MRTN_high, leading to a corresponding increase in predicted N₂O emissions (see supplementary information for further discussion). Given the strong emphasis on developing in-season N management tools in the US Midwest, these results question the assumption that being able to apply EONR will necessarily reduce N rates and represent a more sustainable, site-specific N₂O emission mitigation strategy relative to the MRTN approach.

Improving NRE and applying N fertilizer in balance with crop demand are frequently proposed as environmental policy options for mitigating N losses (Tilman et al 2011, Zhang et al 2015b). Zhang et al (2015a) compared several generalized yield response curves obtained from different studies in the US Midwest to examine the economic and environmental consequences of technologies aimed at improving NRE. For example, enhanced efficiency N fertilizers were assumed to produce 15% higher yields with 9% less N fertilizer, increasing NRE by 30% without negatively impacting economic outcomes. However, whether farmers will adopt alternative N management technologies largely depends on how consistently they change yield response functions across a wide range of soil, weather, and management conditions while producing a positive return. For the target level of NRE improvement considered in this study (30%), we estimated that a 25% median reduction in N rates would be needed compared to MRTN_high. While this effectively reduced N₂O and yield-scaled N₂O emissions by approximately 30%, T-30_NRE also resulted in significant yield decreases and economic losses (37% of sites had a yield penalty >5% and returns were more negative than −84 USD ha⁻¹ at 25% of sites (figures 3 and 5). Similarly, achieving zero N surplus at each site required nearly a 50% median reduction in N rates compared to MRTN_high. As a consequence, yield-scaled N₂O emissions were reduced to the greatest extent with ZNS (41%), but this also resulted in the most significant maize yield losses and economic penalties (e.g. at more than 25% of sites, yields and returns decreased by more than 18% and 196 USD ha⁻¹, respectively) (figures 3 and 6). These findings contradict the notion that large reductions in N rate are possible without negative yield consequences.

Despite many farmers being concerned about environmental quality and N losses in the US Midwest, they are also faced with tough economic decisions (Reimer et al 2017). At present, N rate reduction strategies are voluntary, suggesting that even slight yield reductions could deter a farmer from considering enrollment in N loss mitigation programs, and that the economic decision-making process of individual farmers should be considered in top-down assessments. While we are not aware of prior studies assessing tradeoffs between maintaining economic returns and reducing yield-scaled N₂O emissions across a large population of field trials, another important area for future work using on-farm data would be to evaluate the variability observed across sites to search for win-win solutions, i.e. the ability to maintain yields while reducing N losses. The current database did not contain the necessary variables to support this type of analysis, but it is possible that sites with different soil types or crop management history may better withstand N rate reductions than others, suggesting that spatial reallocation of N inputs among a population of sites could increase overall N use efficiency without reducing total production (Bodirsky and Müller 2014, Lassaletta et al 2014). With this approach it would first be important to consider the inter-annual variability of N responses occurring within a site, given that consistent yield responses are necessary for spatial reallocation to be effective over time.

Our results regarding yield tradeoffs and potential economic losses may help identify barriers currently limiting farmer participation in N rate reduction programs in the US Midwest. Major obstacles preventing farmer adoption already exist, including perceptions of climate change, understanding of N fertilizer-induced GHG emissions, and knowledge of effective management practices to improve N use efficiency (Stuart et al 2014). To enable the development of more effective and scalable N loss mitigation strategies, we argue it is necessary to account for farm-level yield consequences. At the same time, policy discussions should also consider that a narrow focus on reducing environmental N losses may come at the cost of other cropping system functions necessary for sustainable production. For example, application of N fertilizer at agronomic optimum rates has recently been shown to be important for maintaining or increasing soil organic carbon in long-term maize-based rotations (Poffenbarger et al 2017), bringing into question whether the environmental strategies assessed here could sustain soil fertility levels in the future. To minimize tradeoffs, programs aimed at promoting sustainable N fertilizer use should account for multiple agronomic, economic, and environmental indicators where possible.

This study provides an improved evidence-base for assessing the variability of economic risks and developing insights into potential farmer adoption of N rate reduction protocols aimed at balancing agronomic and environmental goals in the US Midwest. By combining an empirical N₂O emission model with on-farm maize yield response data in Illinois, we found that two environmental N rate reduction strategies (enhancing NRE by 30% or maintaining zero N surplus) had greater potential to reduce N₂O emissions on an area- and yield-scaled basis compared MRTN_high than two economic strategies, MRTN_low and EONR. However, both environmental strategies were associated with large reductions in N rate which negatively impacted yields and economic returns. In contrast, results show that MRTN_low was a practical N rate reduction strategy that resulted in smaller yield reductions and increased economic returns in 64% of sites, while consistently mitigating N₂O emissions per unit yield relative to MRTN_high. By comparison, the site-specific EONR approach resulted in slightly greater reductions in median yield-scaled N₂O emissions, but with greater variability across sites. Important limitations regarding the policy implications of this study are the assumption of MRTN_high as current farmer practice and the need to interpret results within the specific context of all N application prior to planting. We conclude that potential decreases in field-level yields and economic returns should be accounted for in policy discussions to increase the likelihood of voluntary farmer adoption and support the scaling up of environmental N loss mitigation strategies in agriculture.

The authors gratefully acknowledge the participating farmers in Illinois who assisted in field trial management and data collection, with special thanks to Dan Schaefer, who coordinated some of the recent field trials. This material is based upon work supported by the National Institute of Food and Agriculture, US Department of Agriculture, Hatch project (No. ILLU-802–949), and the Youth Innovation Promotion Association, CAS (Member No. 2015249).