Impact of nitrate and 3-nitrooxypropanol on the carbon footprints of milk from cattle produced in confined-feeding systems across regions in the United States: A life cycle analysis

It is estimated that enteric methane (CH 4 ) contributes about 70% of all livestock greenhouse gas (GHG) emissions. Several studies indicated that feed additives such as 3-nitrooxypropanol (3-NOP) and nitrate have great potential to reduce enteric emissions. The objective of this study was to determine the net effects of 3-NOP and nitrate on farmgate milk carbon footprint across various regions of the United States and to determine the variability of carbon footprint. A cradle-to-farmgate life cycle assessment was performed to determine regional and national carbon footprint to produce 1 kg of fat-and protein-corrected milk (FPCM). Records from 1,355 farms across 37 states included information on herd structure, milk production and composition, cattle diets, manure management, and farm energy. Enteric CH 4 , manure CH 4 , and nitrous oxide were calculated with either the widely used Intergovernmental Panel on Climate Change Tier 2 or region-specific equations available in the literature. Emissions were allocated between milk and meat using a biophysical allocation method. Impacts of nitrate and 3-NOP on baseline regional and national carbon foot-print were accounted for using equations adjusted for dry matter intake and neutral detergent fiber. Uncertainty analysis of carbon footprint was performed using Monte Carlo simulations to capture variability due to inputs data. Overall, the milk carbon footprint for the baseline, nitrate, and 3-NOP scenarios were 1.14, 1.09 (4.8% reduction), and 1.01 (12% reduction) kg of CO 2 - equivalents (CO 2 -eq)/kg of FPCM across US regions. The greatest carbon footprint for the baseline scenario was in the Southeast (1.26 kg of CO 2 -eq/kg of FPCM) and lowest for the West region (1.02 kg of CO 2 -eq/kg of FPCM). Enteric CH 4 reductions were 12.4 and 31.0% for the nitrate and 3-NOP scenarios, respectively. The uncertainty analysis showed that carbon footprint values ranged widely (0.88–1.52 and 0.56–1.84 kg of CO 2 - eq/kg of FPCM within 1 and 2 standard deviations, respectively), suggesting the importance of site-specific estimates of carbon footprint. Considering that 101 billion kilograms of milk was produced by the US dairy industry in 2020, the potential net reductions of GHG from the baseline 117 billion kilograms of CO 2 -eq were 5.6 and 13.9 billion kilograms of CO 2 -eq for the nitrate and 3-NOP scenarios, respectively.


INTRODUCTION
The dairy industry in the United States supplies nutritious food products and ingredients such as milk, butter, cheese, and whey protein, and contributes to the economy by creating job opportunities, funding government services, and supporting family businesses (IDFA, 2018).However, the dairy industry also has an impact on the environment through production of greenhouse gases (GHG) and reactive nitrogen, while utilizing natural resources such as land, water, and fossil fuel.According to Thoma et al. (2013), the US dairy sector produced about 2% of the total anthropogenic GHG in the country in 2007.As in many sectors of the economy, the dairy industry has made commitments to reduce its environmental impact and improve land and water stewardship of dairy production systems.A deterministic model-based estimation indicated that the US dairy farmers utilized 21% less land and 30% less water, and emitted 19% less GHG, to produce the same amount of milk in 2017 compared with 2007, due to adoption of new technologies and advancement of science and innovations (Capper and Cady, 2020).
Feed additives show great potential to reduce enteric methane (CH 4 ) emissions among the various CH 4 mitigation strategies examined (Table 1; Arndt et al., 2021).However, the effect of feed additives depends on several factors, such as animal type (beef vs. dairy), dose of feed additives, chemical composition of the diets fed to the animals (e.g., NDF content of TMR), feeding systems (grazing vs. confined systems), and expression of enteric CH 4 emissions [e.g., CH 4 production (g/d), CH 4 yield (g/kg of DMI), and CH 4 intensity (g/kg of fat-and protein-corrected milk, FPCM)].Effects of feed additives on animal performance and intake also vary between studies or between feed additives.For instance, supplementing the macroalgae Asparagopsis taxiformis tends to reduce DMI depending on dose of supplementation but improved feed efficiency (Kinley et al., 2020;Roque et al., 2021).Enteric CH 4 mitigation strategies need to be evaluated at the farm gate to account for trade-offs or interactions between components of the milk production systems (e.g., enteric, manure, and feed production-related emissions) and avoid misleading conclusions (Wattiaux et al., 2019;Uddin et al., 2020Uddin et al., , 2021)).However, impacts of feed additive supplementation at the farmgate level are not well studied under US production systems.The exception is one study that quantified the net reduction in GHG farmgate emissions of feeding 3-nitrooxypropanol (3-NOP) or nitrate in California milk production systems (Feng and Kebreab, 2020).Furthermore, the impacts of feeding additives on net GHG farmgate emissions might vary depending on geographic location and management practices.Therefore, our objectives were (1) to determine the potential benefits of nitrate and 3-NOP supplementation on the farmgate carbon footprint of milk produced under different geographic regions and management systems in the US and (2) to determine the variability of carbon footprint estimates across the country, since the average values of the input data were used to determine carbon footprint.

MATERIALS AND METHODS
A cradle-to-farmgate life cycle assessment (LCA) was performed, following the Food and Agriculture Organization of the United Nations (FAO) Livestock Environmental Assessment Partnership guidelines (FAO, 2016a,b) to determine the impacts of nitrate and 3-NOP supplementation on carbon footprint of milk across various regions in the US.The functional unit was 1 kg of FPCM [FPCM (kg/d) = milk production (kg/d) × (0.1226 × milk fat% + 0.0776 × true pro-tein% + 0.253)] (IDF, 2015).According to the Intergovernmental Panel on Climate Change (IPCC) fifth assessment, inclusion or exclusion of climate carbon feedback to calculate global warming potential (GWP) depends on user choice and goal (Gasser et al., 2017).The GWP100 metric may not accurately reflect the impact of short-lived GHG such as CH 4 , and, thus, other metrics such as GWP* may be preferred, depending on the question raised by the user (Allen et al., 2018).The authors suggested that additional studies are required to incorporate this into LCA methodology.Thus, we used the current 100-yr time horizon GWP to express CH 4 (28) and nitrous oxide (N 2 O; 265) emissions in carbon dioxide equivalents (CO 2 -eq; Myhre et al., 2013).Allocation of emissions between milk and meat was conducted based on the biophysical method following IDF (2015) recommendation.The allocation factor for milk was calculated as [1 − 6.04 × (kg of live weight sold/kg of milk produced)].The system boundary was cradle-to-farmgate because most GHG emissions (>70%) for milk consumption originate from on-farm sources (Thoma et al., 2013), and post-farmgate process are not affected by feed additive inclusion.The system boundary and processes included in this study are shown in Figure 1.

Foreground Data Collection and Mitigation Scenarios
Records on 1,355 confined dairy farms from 37 states across the US regions were obtained from voluntary evaluations conducted between 2017 and 2019 using the Farmers Assuring Responsible Management environmental stewardship program (FARM, 2021).The 37 states whose sample farm data were included in this study produce approximately 95% of milk in the US (USDA-NASS, 2020).Farm data included milk production, milk fat percentage, milk protein percentage, herd size, number of lactating and dry cows, number of heifers and calves produced either on-farm or off-farm, ingredient composition for lactating cow diet, on-farm and off-farm feed sources, BW and DMI of lactating cows, and fuel and electricity used for dairy activities (Table 2 and Supplemental Table S1; https: / / doi .org/ 10 .17632/55s6mczd9r .1;Uddin et al., 2022).The data set was then divided into 5 regions based on geographic and production management resemblances.The farms were grouped into Northeast (8 states), Southeast (5 states), Midwest (11 states), South (7 states), and West (6 states; list of states can be found in Table 2).Chemical compositions of regional lactating cow diets (Table 3) were calculated using NRC (2001) from the dietary feed ingredient compositions collected from the specific region.Chemical compositions of dry cow and heifer diets (Table 4) were also calculated using NRC (2001), using typical BW, DMI, and dietary ingredient compositions for heifers and dry cows (Thoma et al., 2013;Naranjo et al., 2020).These dietary ingredient compo-sitions were used to calculate total requirement for each feed ingredient for each model, which were later used for feed production-related GHG calculations.
Baseline LCA model scenarios for each region and across regions (national average) were developed, assuming no supplementation of nitrate or 3-NOP.Mitigation scenarios were then developed assuming supplementation of nitrate and 3-NOP within the baseline diets to lactating and nonlactating animals at recommended doses.The enteric CH 4 mitigation effects for 3-NOP (Dijkstra et al., 2018) and nitrate (Feng et al., 2020) were based on published meta-analysis results.Mitigation effects of nitrate and 3-NOP, including emissions associated with production and transportation of feed additives, were calculated using the equation (Supplemental Table S2; https: / / doi .org/ 10 .17632/55s6mczd9r .1;Uddin et al., 2022) and method described in detail by Feng and Kebreab (2020).

Emission Factors
Enteric Methane.Most livestock LCA studies calculate emission factors based on either IPCC (2006) or more refined models from the literature.In this study, enteric CH 4 for lactating cows was determined using the prediction model with the best fit for North America, reported by Niu et al. (2018), which requires BW, DMI, NDF, and milk fat information (Supplemental Table S1).Enteric CH 4 for nonlactating cattle was calculated using the IPCC (2019) tier 2 equation (Supplemental Table S1), which requires gross energy intake as in-   Manure management systems have significant influence on GHG emissions and vary across regions.Information on regional manure management systems was based on NAHMS (2014) survey results, which reported the latest in-depth information on manure management systems practiced across regions (Supplemental Table S3; https: / / doi .org/ 10 .17632/55s6mczd9r .1;Uddin et al., 2022).Calculation of volatile solids, which is an input to the IPCC (2019) equation for manure CH 4 calculation, was based on Appuhamy et al. (2018).The IPCC (2019) equation also requires methane conversion factors as input, which are influenced by both manure storage type and conditions (e.g., temperature).In this study, regional methane conversion factors were chosen based on regional manure management systems and average temperatures of cool and temperate regions.Manure N 2 O was calculated as described by Naranjo et al. (2020), where both direct and indirect N 2 O were included.
Feed Production.Dietary ingredients and chemical composition of dairy cattle diets vary across regions (Table 3).Most forages are produced on-farm, whereas inclusion of concentrates (grains and by-products) depends on availability in the region (Table 2).Thus, it is important to consider regional emission factors for each feed ingredient, which may vary due to differences in climatic conditions and production systems (Adom et al., 2012).Thus, regional emission factors for each feed ingredient were adopted from recent LCA studies, which determined carbon footprint for each feed in- gredient produced in the regions and accounted for all inputs (e.g., seed, fertilizer, insecticide), outputs, and associated indirect and direct emissions (Adom et al., 2012).If regional emission factors were not available for any feed ingredients, we determined the emission factors as described by Naranjo et al. (2020) or Uddin et al. (2021), which take into account all the inputs (e.g., seed, fertilizer, insecticide), outputs, and associated emissions (indirect and direct) needed to produce those ingredients (Supplemental Table S4; https: / / doi .org/ 10 .17632/55s6mczd9r .1;Uddin et al., 2022).Farm Energy.Amount and type of energy used by the farms for dairy activities across regions were collected as the primary data (Supplemental Table S4), which were converted into CO 2 -eq using conversion factors as shown in Supplemental Table S5 (https: / / doi .org/ 10 .17632/55s6mczd9r .1;Uddin et al., 2022).

Uncertainty and Sensitivity Analyses
A sensitivity index (ratio of changes in output due to changing inputs variables) was calculated for most input variables following the Rotz et al. (2020) method to determine the variables with the most influence on milk carbon footprint.A sensitivity index close to zero indicates that the parameter does not have influence on milk carbon footprint, whereas an index close to 1.0 indicates that the parameter has considerable influence on milk carbon footprint.
Site-specific and good-quality inventory data representing the production system under study are required for accurate LCA results (Meron et al., 2020).The use of average data might not account for variability between farms; thus, variability of carbon footprint was determined using a Monte Carlo simulation with 3,000 iterations.We determined 1 and 2 standard deviations (SD) of carbon footprint by varying input variables such as milk production, milk fat percentage, milk protein percentage, DMI, dietary NDF and CP concentrations, enteric CH 4 , manure CH 4 , and manure N 2 O.All input variables were assumed to be normally distributed, whereas SD of each variable for each region was obtained primarily from the original data set whenever available (Table 2).The SD of dietary NDF and CP were obtained from Niu et al. (2018), whereas SD of enteric and manure CH 4 and manure N 2 O variables were based on IPCC (2019).

Carbon Footprint of Baseline Scenario
The regional farmgate milk carbon footprints determined for the baseline scenario are shown in Table 5.The carbon footprints ranged from 1.02 (West) to 1.26 (Southeast) kg of CO 2 -eq/kg of FPCM, with a national average (across regions) of 1.14 kg of CO 2 -eq/ kg of FPCM.Contributions of each emission source were 39, 35, 21, and 5% for enteric, manure management, feed production, and farm energy-related emissions, respectively.Other studies also reported similar ranges of carbon footprints for California (Naranjo et al., 2020), although one should be cautious when comparing results across LCA studies, mainly due to differential assumptions.However, this study found lower carbon footprint for the Western region, which included California, compared with Naranjo et al. (2020; 1.02 vs. 1.12 kg of CO 2 -eq/kg of FPCM, respectively).This difference was mostly due to the differences in management systems considered in the region and the boundaries of the regions, since our study included data not only from California but from other Western states such as Oregon, Washington, and Idaho.The baseline data for our study were collected between 2017 and 2019, whereas Naranjo et al. (2020) used data collected in 2014.This difference in time horizon might explain the lower carbon footprint in our study compared with earlier studies, because production performances and efficiency of dairy production have improved over time.Capper and Cady (2020) also indicated improvements in environmental impact in 2017 compared with 2007.We determined enteric CH 4 from lactating cows using the Niu et al. (2018) equation for North America; however, using the IPCC (2019) equation, the carbon footprint was 4.5% lower across regions.In this study, off-farm heifers were not included due to large variability in heifer data between regions.However, heifers raised off-farm are often included in LCA studies (Alvarez-Hess et al., 2019).Inclusion of heifers raised off-farm in our model increased the carbon footprint by 12% on average, across regions, ranging from 8 (Southeastern region) to 12% (Southern region).

Impacts of Nitrate and 3-Nitrooxypropanol on Carbon Footprint
The impacts of 3-NOP and nitrate feeding on carbon footprint are shown in Figure 2.Both nitrate and 3-NOP scenarios had lower farmgate milk carbon footprints compared with the baseline scenario.Overall, supplementation of nitrate to all eligible animal categories reduced the carbon footprint 4.8% compared with baseline, mainly through a 12% average reduction in enteric CH 4 emissions.The range of reduction across regions was narrow, being lowest for the Southeast (4%) and highest for the West (5%).Feeding 3-NOP reduced the farmgate carbon footprint of milk production more than feeding nitrate.The overall reduction attributed Uddin et al.: FEED ADDITIVE EFFECTS ON MILK CARBON FOOTPRINT to 3-NOP feeding was 12% (ranging from 11% for the Southeast to 13% for the West) compared with the baseline carbon footprint.This reduction was mainly due to lowering enteric CH 4 emissions by an average of 31% across regions.Based on 2020 milk production in the United States, which is 101 billion kilograms, total national GHG emissions from milk production can be estimated at 117 billion kilograms of CO 2 -eq for the baseline scenario.Thus, nitrate and 3-NOP supplementation to all eligible animals (lactating and nonlactating) at the rate and dose recommended would reduce 5.6 and 13.9 billion kilograms of CO 2 -eq of annual GHG emissions from US milk production systems, respectively.Similarly, Feng and Kebreab (2020) reported 3.95 and 11.7% reductions in milk carbon footprint for California when nitrate and 3-NOP were supplemented, respectively, to either only lactating cows or the entire dairy herd.Although carbon footprint varied across regions in this study, our findings show a similar potential of net GHG reduction from milk production systems across US regions.This finding suggests that nitrate and 3-NOP supplementation strategies would have similar impacts on net reductions of GHG across all milk production regions in the United States.Furthermore, effects of feed additives on milk carbon footprints of grazing animals or organic production systems might be different due to large variations in dietary composition (e.g., NDF) and milk production between confined-feeding and grazing systems.Although feed additives might indirectly affect emissions by changing manure composition (Benchaar and Hassanat, 2019) and properties (Weber et al., 2021), we assumed that these feed additives do not affect manure managementrelated GHG emissions.
Overall, 3-NOP had no negative effects on animal performance and health across studies (Dijkstra et   2018).Urea can be replaced with an equivalent dose of nitrate as a mitigation strategy for enteric CH 4 from grazing cattle without adverse effects on animal health and productivity (Callaghan et al., 2014).Although nitrate was not economically feasible for dairy production systems in the Netherlands, it can potentially reduce net GHG by up to 5% (Arndt et al., 2021).Additionally, food safety aspects of feed additives must be evaluated carefully before using them at industrial scale (Vijn et al., 2020;Honan et al., 2021), although risk of toxicity depends on dietary composition, dose of nitrate, and feeding pattern or form of supplementation (Callaghan et al., 2014;Cottle et al., 2016).Furthermore, nitrate might lead to increased manure N 2 O emissions (Arndt et al., 2021).Honan et al. (2021) reviewed feed additives with antimethanogenic properties and reported that several of them have potential to reduce enteric CH 4 emissions.For example, macroalgae reduced enteric CH 4 by 80% in beef steers and up to 67% in dairy cows (Roque et al., 2019a(Roque et al., , 2021)).Therefore, this study warrants future research focused on evaluating various novel feed additives, which are expected to contribute to US dairy production to achieve carbon neutrality by 2050.

Sensitivity and Uncertainty Analyses
The sensitivity indexes for different input parameters are shown in Figure 3.Among tested parameters in this study, DMI and milk production had the most influence on milk carbon footprint, whereas dietary CP and NDF concentrations, farm energy, and feed productionrelated emissions had the least influence on milk carbon footprint.Increasing milk production or decreasing DMI reduces carbon footprint substantially, whereas changing dietary CP would have little effect on GHG emissions from milk production systems.However, in this study, variations in soil organic carbon stock due to differential feeding and forage systems were not considered.Additionally, local sensitivity analysis (i.e., changing one input parameter at a time to determine the impact on milk carbon footprint) did not account for interactions between 2 or more input parameters, which is a limitation of this analysis.
The uncertainty analysis revealed that carbon footprint results varied widely not only within region but also across regions.Overall, carbon footprints ranged from 0.52 to 3.62 (0.88-1.52 and 0.56-1.84kg of CO 2eq/kg of FPCM within 1 and 2 SD, respectively); however, the range was narrower for the Southeastern and Southern regions compared with other regions, probably due to relatively smaller data variability in these 2 regions.For instance, the SD for the 2 most influential input variables, such as DMI and milk production, were lowest for the Southeast and Southern regions compared with other regions, which might partly explain the narrower range of carbon footprints for those 2 regions compared with other regions.Details of distribution of milk carbon footprints across regions is given in Supplemental Figure S1 (https: / / doi .org/ 10 .17632/55s6mczd9r .1;Uddin et al., 2022).The vast variation in milk carbon footprints within and across regions suggest and reinforce the importance of using site-specific or even farm-specific data instead of average data for determining and monitoring carbon footprint of milk production.This variation might also mean that farms with comparatively lower carbon footprints adopted best management strategies and could be used as examples on how emission can be reduced by other farms with comparatively greater carbon footprints.

CONCLUSIONS
Our study demonstrated that 3-NOP had greater potential to reduce farmgate GHG emissions from milk production compared with nitrate.Although the baseline carbon footprints varied across US milk production regions, the impacts of nitrate and 3-NOP on net GHG reductions were similar across the regions.Our uncertainty analysis also revealed large variability in milk carbon footprint within and across regions, suggesting the use of site-specific (farm or regional level) inventory to determine milk carbon footprints.This study indicates that future research should focus on evaluating other potential feed additives and mitigation strategies to determine their impacts on net reduction of GHG from milk production systems.
Figure 1.System boundaries of the cradle-to-farmgate life cycle assessment study.
calves produced on farm. 2 Heifers and heifer calves produced off farm.put.Gross energy intake was calculated based on the Moraes et al. (2014) equation.Manure Management.Both manure CH 4 and manure N 2 O were calculated using IPCC (2019) equations.

Figure 3 .
Figure 3. Sensitivity indexes for different input parameters across US regions.

Table 1 .
Uddin et al.: FEED ADDITIVE EFFECTS ON MILK CARBON FOOTPRINT Empirical studies showing potential to reduce enteric methane emissions from ruminants fed feed additives 2 Reduction occurred at 12 wk of supplementation (Mootral).

Table 2 .
Dairy herd structure, animal performances, and farm characteristics across US regions represented in the data set used in this study Item

Table 3 .
Uddin et al.:FEED ADDITIVE EFFECTS ON MILK CARBON FOOTPRINT Ingredients and chemical composition (% of DM unless otherwise noted) of US regional lactating cow diets 1 1 Chemical composition of lactating cow diets were calculated based on NRC (2001) model, using ingredient compositions provided in the primary data set and NRC (2001) book values for chemical composition of each ingredient. 2 NFC = 100 − (CP + NDF + fat + ash). 2 NE L was calculated using NRC (2001) equation based on actual diet and cow performance data. 1 Nonlactating cattle diets were formulated as same across regions.Chemical composition of nonlactating cattle diets were calculated based on NRC (2001) model using NRC (2001) book value for chemical composition of each ingredient to meet NRC (2001) nutrient requirements for the nonlactating cattle diets.

Table 5 .
al., Uddin et al.: FEED ADDITIVE EFFECTS ON MILK CARBON FOOTPRINT Contribution of emission sources on the carbon footprint of milk produced in different US regions under each corresponding baseline scenario