Data sources
FAOSTAT (Crop and Livestock Products, https://www.fao.org/faostat/en/#data/QCL) provides numbers (Producing Animals/Slaughtered), production (Production Quantity) and slaughtered weight yield (Yield) of each livestock3. We used the protein content of livestock products from Global Livestock Environment Assessment Model (GLEAM)6. Human protein demand for all products was from FAOSTAT (Food Balances)3. Proportion information for specific livestock systems, such as pig production systems containing backyard, intermediate and industrial systems, was obtained from GLW (Gridded Livestock of the World, https://dataverse.harvard.edu/dataverse/glw/).
The yield and production quantity of feed crops were derived from FAOSTAT data3. For grass yield, we acquired the data from the literature33. The nitrogen content of each crop and grass was from GLEAM6. We estimated synthetic fertilizer consumption of each crop and grass from FAO3 (https://www.fao.org/faostat/en/#data/RFB), IFA34 and the literature35. Nitrogen deposition rates on cropland and grassland were derived from the literature36,37,38. The irrigation nitrogen rates were obtained from Lesschen et al.39 and the irrigation water use data was from AQUASTAT (https://www.fao.org/aquastat/en/). For the natural biological fixation from grass and crop biological nitrogen fixation, we used data from Lassaletta et al.35 and Zhang et al.37. Global land cover data and land use data were from the GLC-SHARE database (https://data.apps.fao.org/catalog/dataset/global-land-cover-share-database) and FAOSTAT data3.
Counterfactual analysis
We compare the environmental impacts between ruminants and monogastric livestock by estimating and comparing the Nr and GHG emissions from current ruminants and monogastric livestock replaced by the ruminants according to the standard of protein equality. Using this approach, we can analyse the huge land use change risks for growing feed crops and evaluate the contribution of ruminants and monogastric livestock to humans from a new perspective. The allocation of pigs and chickens among monogastric livestock is based on the human protein requirements of pigs, chicken and eggs (from FAOSTAT, Protein Supply Quantity)3.
Nitrogen loss from livestock supply chains
Feed production emissions
To estimate Nr emissions from each livestock supply chain, we mainly used GLEAM. A detailed description of the GLEAM model can be found in the literature6. For feed production, we first estimated the dry matter feed by food conversion ratio (FCR)4 and crop nitrogen content to calculate grass (N grass feed ), crop (N crop feed ) and crop residues (N crop residue feed ) feed nitrogen, respectively, following the feed ration percentage from the GLEAM model6, and then Nr emissions from cropland and grassland were calculated.
We calculated cropland Nr emissions from synthetic fertilizer, recycled manure and decomposed crop residues. First, we used the CHANS model40 to calculate all cropland NUE (NEU crop , without differentiating between feed and ration cultivation) across global countries (equation (1)). Then, we used the NUE crop to calculate the nitrogen input of crop feed, especially fertilizer and manure nitrogen input which were used to estimate nitrogen loss. We calculated the amount of nitrogen from decomposed crop residues returned to the field following equations in the GLEAM model, and used its removed fraction of above-ground residues of the cropland to estimate the crop residue feed quantity (N crop residue feed ). The available manure nitrogen recycled on cropland from each livestock system was estimated following Uwizeye et al.1 Finally, the Nr emissions from synthetic fertilizer, manure and crop residues were calculated following equations in the GLEAM model.
$${{NUE}}_{\rm{crop}} = \frac{{N_{{\rm{crop}}\;{\rm{products}}}}}{{N_{\rm{CBNF}} + N_{\rm{fertilizer}} + N_{\rm{manure}} + N_{\rm{irrigation}} + N_{\rm{deposition}}}}$$ (1)
where the nitrogen inputs of cropland consist of crop biological nitrogen fixation (N CBNF ), fertilizer nitrogen (N fertilizer ), manure nitrogen (N manure ), irrigation nitrogen (N irrigation ) and deposition nitrogen (N deposition ), and the nitrogen outputs of cropland are crop products (N crop products ).
We calculated grassland Nr emissions from synthetic fertilizer and manure deposited on pastures. The manure deposited on pastures from all livestock was calculated based on Uwizeye et al.1 The nitrogen loss from grassland was calculated in the GLEAM model.
Livestock raising emissions
The Nr emissions from manure management systems were estimated as livestock raising emissions6. We calculated nitrogen excretion following IPCC methods41,42 and the total ammoniacal nitrogen (TAN) from Uwizeye et al.1, and used the fraction of manure management system (MS S ) and EF (emission factor) to calculate Nr emissions at the livestock raising stage.
Nitrogen balance
Feed production stage
The nitrogen inputs of cropland consist of crop biological nitrogen fixation (N CBNF ), fertilizer nitrogen (N fertilizer ), manure nitrogen (N manure ), irrigation nitrogen (N irrigation ) and deposition nitrogen (N deposition ), and nitrogen outputs of cropland are crop (N crop feed ) and crop residues (N crop residue feed ) as feed, other crop residues (N other crop residue , not as feed for this livestock system, as feed for other livestock or for other uses) and Nr emissions (N emission , including NH 3 -N, NO x -N, NO 3 −-N, N 2 O-N). NUE in a cropland system (NUE cropland ) is calculated as equation (2). The nitrogen inputs of grassland contain natural biological nitrogen fixation (N NBNF ), fertilizer nitrogen (N fertilizer ), deposited manure nitrogen (N manure ) and deposition nitrogen (N deposition ), and nitrogen outputs of grassland are grass feed (N grass feed ) and Nr emission (N emission ). NUE in a grassland system (NUE grassland ) is taken from equation (3).
$$NUE_{\rm{cropland}} = \frac{{N_{{{\rm{crop}}\;{\rm{feed}}}} + N_{{{\rm{crop}}\;{\rm{residue}}\;{\rm{feed}}}}}}{{N_{\rm{CBNF}} + N_{\rm{fertilizer}} + N_{\rm{manure}} + N_{\rm{irrigation}} + N_{\rm{deposition}}}}$$ (2)
$$NUE_{\rm{grassland}} = \frac{{N_{{{\rm{grass}}\;{\rm{feed}}}}}}{{N_{\rm{NBNF}} + N_{\rm{fertilizer}} + N_{{{\rm{deposited}}\;{\rm{manure}}}} + N_{\rm{deposition}}}}$$ (3)
Livestock raising stage
Crop (N crop feed ), crop residues (including N crop residue feed and N other crop residue feed from the production of human rations or crop feed for other livestock), grass feed (N grass feed ), swill (N swill ) and other feed (N other feed , including synthetic amino acids and fishmeal) are the nitrogen inputs, and the nitrogen output contained livestock products (N livestock products , including meat, eggs and milk), Nr emission and manure nitrogen recycling to croplands and grassland. The NUE in the livestock system (NUE livestock ) is derived based on equation (4).
$$\begin{array}{l}NUE_{\rm{livestock}} =\\ \frac{{N_{{{\rm{livestock}}\;{\rm{products}}}}}}{{N_{{{\rm{crop}}\;{\rm{feed}}}} + N_{{{\rm{crop}}\;{\rm{residue}}\;{\rm{feed}}}} + N_{{{\rm{grass}}\;{\rm{feed}}}} + N_{{{\rm{other}}\;{\rm{crop}}}}\;_{{{\rm{residue}}\;{\rm{feed}}}} + N_{{\rm{{swill}}}} + N_{{{\rm{other}}\;{\rm{feed}}}}}}\end{array}$$ (4)
Whole livestock production chain
We defined NUE whole chain based on the total livestock supply chain, including feed production and livestock raising stages (equation (5)). N BNF contains N CBNF from croplands and N NBNF from grassland. N other manure is the manure nitrogen recycling to the cropland from other livestock, for instance, monogastric livestock that require more crop feed could not produce enough manure of their own and need manure nitrogen from other livestock.
$$\begin{array}{l}NUE_{{\rm{whole}}\;{\rm{chain}}} =\\ \frac{{N_{{\rm{livestock}}\;{\rm{products}}}}}{{N_{\rm{BNF}} + N_{\rm{irrigation}} + N_{\rm{fertilizer}} + N_{\rm{deposition}} + N_{{\rm{other}}\;{\rm{manure}}} + N_{{\rm{other}}\;{\rm{crop}}\;{\rm{residue}}} + N_{\rm{swill}} + N_{{\rm{other}}\;{\rm{feed}}}}}\end{array}$$ (5)
GHG emissions from livestock supply chains
The major GHG emissions from livestock systems in the GLEAM model are: (1) CH 4 emissions from enteric fermentation in ruminants and pigs; (2) CH 4 emissions arising from manure management; (3) N 2 O emissions released from manure management, done in the calculation of Nr emissions; (4) CH 4 emissions from rice production; (5) CO 2 emissions from fertilizer manufacture; (6) CO 2 emissions from field operations; (7) CO 2 emissions from feed blending, processing and transport; (8) CO 2 emissions from land use change. Items (1)–(7) were calculated by the following methods described in Supplementary Table 1 and item (8) was calculated following equations (6) and (7).
We calculated the cropland area by monogastric livestock more than ruminants and take the cropland area as the relative land use change with monogastric livestock, as shown in equations (6) and (7).
$${\mathrm{Land}}_c = \frac{{{\mathrm{DM}}_c}}{{{\mathrm{DMYG}}_c \times {\mathrm{FUE}}_c}} \times \frac{{{\mathrm{EFA}}_c}}{{{\mathrm{MFA}}_c}}$$ (6)
where Land c is the land area of feed c, DMYG c is the dry matter yield of feed c, in kg ha−1, and is calculated based on crop yield according to the GLEAM method, FUE c is feed use efficiency of feed c, and MFA c and EFA c are the mass fraction and economic fraction, respectively, and are derived from GLEAM.
$${\mathrm{LUC}}_{\mathrm{monogastric}}=({\mathrm{Cropland}}_{\mathrm{ruminant}}-{\mathrm{Cropland}}_{\mathrm{monogastric}}\times{\mathrm{LUC}})$$ (7)
where Cropland ruminant and Cropland monogastric are the areas of cropland required for ruminant and monogastric feed, respectively, and LUC is the land use change value43, representing annual GHG emissions released from forest to cropland, in tCO 2 ha−1 yr−1. The LUC value takes into account the long-term effects and was discounted to an average value for each year43. There was a large uncertainty in the calculation of GHG emission changes from forest conversion to cropland and reforestation (Uncertainty analysis).
Scenario analysis
The baseline scenario has been established as BAU, assuming the total amount of ruminant and monogastric protein produced in 2019 is maintained at a constant level. Three optimized scenarios were designed to assess the impact on mitigating livestock nitrogen and GHG emissions, including SYS, FED and ALL scenarios.
BAU scenario
The BAU scenario assumes the amount of protein produced by ruminants and monogastric livestock to be 7.06 and 6.64 TgN-protein in 2019, respectively. Currently, there are partially underutilized grassland areas (Supplementary Fig. 2) and sustainable unused crop residue resources.
SYS scenario
In this scenario, we model the effect of maximizing ruminant production and reducing monogastric production accordingly while keeping total livestock protein production constant (13.7 TgN-protein, calculated from FAOSTAT) in 2019. The SYS scenario represents a switch of 12.3% of global livestock production from monogastric to ruminant livestock. The resource constraints for maximizing ruminant protein production are current maximum production of total cellulose while considering the carrying capacity of the grass and the total amount of crop residues. As for the maximum available value of grass nitrogen, we take into account the degradation conditions (grass degradation adjustment rate (DAR))44 of grazing grassland to adjust the utilization efficiency (UE)6 of grassland. For DAR, we set mild degradation to 80%, slightly mild degradation to 65%, moderate degradation to 50% and severe degradation to 30% (Extended Data Fig. 10). The country’s average grassland cover share was calculated from the GLC-Share database45 to reflect the extent of grassland degradation, and using 3/4 value, 1/2 value and 1/4 value quadrature into four intervals, the adjustment factors were set to 80%, 65%, 50% and 30%, respectively. In addition, grassland degradation is not considered for non-grazed grasslands. The maximum available value of grass nitrogen (Nmax grass ) is calculated as shown in equation (8).
$$\begin{array}{l}N{\mathrm{max}}_{\mathrm{grass}}= {\mathrm{Production}}_{\mathrm{grassN}} \times R_{\mathrm{grazing}} \times {\mathrm{UE}} \times {\mathrm{DAR}}\\ \qquad \qquad \quad + {\mathrm{Production}}_{\mathrm{grassN}} \times (1 - R_{\mathrm{grazing}}) \times {\mathrm{UE}}\end{array}$$ (8)
where Production grassN is the total grass nitrogen production and R grazing is the grazing ratio of ruminants.
The maximum crop residue nitrogen removed from croplands was calculated from crop nitrogen (Production cropN ), crop residues to crop ratio (R residue–crop )6 and the proportion of crop residues removed (R removed )6, as shown in equation (9).
$$N{\mathrm{max}}_{\mathrm{crop}\;\mathrm{residues}} = {\mathrm{Production}}_{\mathrm{cropN}} \times R_{\mathrm{residue} - \mathrm{crop}} \times R_{\mathrm{removed}}$$ (9)
The maximum cellulose nitrogen production was obtained from the sum of Nmax grass and Nmax crop residues , dividing by the current amount of cellulose utilized by the ruminants and getting the maximum available ruminant production multiplier. Meanwhile, in the context of the Nr emission intensity of ruminants being lower than that of monogastric livestock, we can obtain the proportion of ruminants that maximizes cellulose utilization.
SYS2 scenario
This scenario is an extreme variant of the SYS scenario; it assumes the crop residues are not returned to the cropland and all removed crop residues are used to produce feed for ruminants, that is, R removed = 1. In SYS2, it is also a prerequisite that the Nr emission intensity of ruminants is lower than that of monogastric livestock. The SYS2 scenario reflects a switch of 20.7% of global livestock production from monogastric to ruminant livestock. To fully realize the benefits of this scenario it would be necessary to account for the fact that cropland may be deprived of nutrients from recycled crop residues; however, the deficit could be supplemented by the ruminant manure. This scenario would release more croplands and feed more people than the SYS scenario.
FED scenario
In this scenario, the production of ruminants and monogastric livestock remained was kept consistent with the BAU scenario. All Nr emissions (nitrogen from NH 3 , NO x and NO 3 −) from feed production and livestock raising are reduced to the global average, and those countries that are already below the global average remain unchanged. The FED scenario was designed to produce substantial emission reductions and could be achieved through targeted abatement measures on cropland, grassland and livestock system (Supplementary Table 2), but no additional croplands would be released.
ALL scenario
The ALL scenario is a combination of the SYS scenario and the FED scenario to achieve both an optimal livestock production ratio and emission levels. In this scenario, maximizing ruminant production (ruminant protein production is consistent with the SYS scenario) and targeted abatement measures at all stages are needed. The ALL scenario could maximize the benefits of food security, environmental protection and climate mitigation. This is the scenario advocated in this study.
Cost–benefit analysis
Implementation cost
The implementation cost under the SYS scenario is considered to be equal to the change in protein quality of all livestock P j (where j represents different livestock systems) multiplied by their unit product cost (PPrice j , in US$ per kg protein), as shown in equation (10). Here PPrice j is the regional animal producer price and is derived from the FAOSTAT database with regional producer prices.
$${\mathrm{Cost}}_{\mathrm{SYS}} = {\sum} {P_j} \times {\mathrm{PPrice}}_j$$ (10)
For the FED scenario, we used the Greenhouse Gas and Air Pollution Interactions and Synergies (GAINS) model (https://gains.iiasa.ac.at/models/index.html) to calculate the abatement costs from cropland (Cost FED–cropland,k ), grassland (Cost FED–grassland,k ) and livestock (Cost FED–livestock,k ) for each country. A detailed description of the GAINS model can be found in Klimont et al.46 The implementation cost under the FED scenario is calculated in equations (11)–(13).
$${\mathrm{Cost}}_{{\mathrm{FED - cropland}},k} =
abla E_{{{\mathrm{N - cropland}}},k} \times C_{\mathrm{{cropland}},k}$$ (11)
$${\mathrm{Cost}}_{{\mathrm{FED - grassland}},k} =
abla E_{{{\mathrm{N - grassland}}},k} \times C_{\mathrm{{grassland}},k}$$ (12)
$${\mathrm{Cost}}_{{\mathrm{FED - livestock}},j} = N_{i,j} \times C_{\mathrm{{livestock}},j} \times {\mathrm{AR}}_k$$ (13)
Where ∇E N–cropland,k and ∇E N–grassland,k are the Nr emission reduction from cropland and grassland in country k, respectively, C cropland,k , C grassland,k and C livestock,k are the unit abatement cost of the most appropriate mitigations (shown in Supplementary Tables 2 and 3) to reduce cropland nitrogen, grassland nitrogen and livestock nitrogen loss modified for the specific farming practices of country k, respectively; C grassland,k is set at one-fifth of C cropland,k . AR k is the calculated abatement rate for country k.
For the ALL scenario, the abatement costs from cropland and grassland are assumed to be equal to those in the FED scenario, and the abatement costs from livestock are the sum of the FED and SYS scenarios.
Societal benefits assessment
The societal benefits of optimizing global ruminant production in this study are defined as the sum of avoided damage costs for ecosystem health (E benefit ), human health (H benefit ) and climate change mitigation (C benefit ), as shown in equation (14).
$$SO_\mathrm{benefit} = E_\mathrm{benefit} + H_\mathrm{benefit} + C_\mathrm{benefit}$$ (14)
The E benefit is assumed to be the benefit of Nr mitigation on the ecosystem, which is also equal to reducing the avoided damage costs. We assume unit Nr damage cost in Europe and the United States is also applicable to other countries after adjusting for differences in the regional willingness to pay (WTP) and purchasing power parity (PPP) for the ecosystem services, as shown in equation (15).
$${\mathrm{UE}}_{{\mathrm{benefit,Nr}},k} = \partial _{\mathrm{EU}} \times \frac{{{\mathrm{WTP}}_k}}{{{\mathrm{WTP}}_{\mathrm{EU}}}} \times \frac{{{\mathrm{PPP}}_k}}{{{\mathrm{PPP}}_{\mathrm{EU}}}}$$ (15)
where ∂ EU is the estimated unit ecosystem damage cost of Nr emissions based on the literature45,47; the value of UE benefit,Nr,k can be found in Supplementary Table 4.
Then, the E benefit is summed according to equation (16).
$$E_{\mathrm{benefit}} =
abla E_{\mathrm{N}_2\mathrm{O}} \times {\mathrm{UE}}_{\mathrm{benefit},\mathrm{N}_2\mathrm{O}} +
abla E_{\mathrm{NH}_3} \times {\mathrm{UE}}_{\mathrm{benefit},\mathrm{NH}_3} +
abla E_{\mathrm{NO}_{x}} \times {\mathrm{UE}}_{\mathrm{benefit},\mathrm{NO}_{x}}$$ (16)
where \(
abla E_{\mathrm{N}_2\mathrm{O}}\), \(
abla E_{\mathrm{NH}_3}\) and \(
abla E_{\mathrm{NO}_{x}}\) are the calculated reduction in N 2 O, NH 3 and NO x , and \({\mathrm{UE}}_{\mathrm{benefit},\mathrm{N}_2\mathrm{O}}\), \(\mathrm{UE}_{\mathrm{benefit},\mathrm{NH}_3}\) and \(\mathrm{UE}_{{\mathrm{benefit}},{\mathrm{NO}}_x}\) represent the unit ecosystem benefit of N 2 O, NH 3 and NO x emission reduction, respectively, in US$ kgN−1 (values are listed in Supplementary Table 4).
The human health benefit (H benefit ) refers to the benefit of prevented mortality derived from PM 2.5 mitigation caused by animal Nr abatement. We derived the national-specific unit health damage costs of Nr emission from Gu et al.48, which connected the economic costs of mortality per unit of Nr emission with the population density, GDP per capita, urbanization and nitrogen share. The calculation of health benefits in this study is shown in equation (17).
$$H_{\mathrm{benefit}} =
abla E_{\mathrm{Nr}} \times H_{\mathrm{benefit,Nr}}$$ (17)
where ∇E Nr is the Nr emission reduction in specific scenarios and H cost,Nr represents the unit health benefit of Nr reduction in US$ kgN−1.
The climate-related benefits of optimizing ruminant production (C benefit ) are considered to be the sum of the GHG mitigation benefits and the Nr (NH 3 and NO x ) mitigation impact on climate, as shown in equation (18).
$$\begin{array}{l}C_{\rm{benefit}} =
abla E_{\rm{GHG}} \times {\rm{UC}}_{{{\rm{benefit}},{\rm{GHG}}}}\\ - (
abla E_{{{\rm{NH}}}_{3}} \times C_{{{\rm{benefit}},{{{\rm{NH}}}_{3}}}} +
abla E_{{\rm{NO}}_x} \times C_{{\rm{benefit}},{\rm{NO}}_x})\end{array}$$ (18)
where ∇E GHG is the GHG emission reduction in a specific scenario, in kgCO 2 e, and the GWP 100 for CH 4 and N 2 O are 27.9 and 273 kgCO 2 e, respectively. UC benefit,GHG represents the monetary climate benefit due GHG mitigation, which is assumed to be the carbon price, of US$40–80 tCO 2 −1 (ref. 49). \(C_{{\mathrm{benefit,NH}}_3}\) and \(C_{{\mathrm{benefit,NO}}_x}\) represent the monetary climate impact due to changed NH 3 and NO x emissions, which is associated with the cooling effect of NH 3 and NO x on the global climate based on previous studies.
Cropland recovered for human food production
We first calculated the recovered or ‘saved’ cropland area (Land sys and all ) under the SYS and ALL scenarios (no cropland is released in the FED scenario). Next, we calculated crop nitrogen yield per unit of cropland (Yield cropN ) by dividing all crop nitrogen production (Production cropN ) using the total cropland area (Cropland total , from FAOSTAT), which is multiplied by Land sys and all to obtain the total value of saved crop nitrogen production. We estimate the additional number of people that could be sustained under the assumption of a purely vegetarian diet by extra food production from saved croplands (Population saved ) by dividing the total saved crop nitrogen production by unit nitrogen nutrition requirements (Protein unit /6.25, where 6.25 is the protein to nitrogen conversion ratio and Protein unit is the per-capita protein requirement, in kg protein per capita per year, calculated from FAOSTAT). The calculation is depicted in equations (19) and (20).
$${\mathrm{Yield}}_{\mathrm{cropN}} = \frac{{{\mathrm{Production}}_{\mathrm{cropN}}}}{{{\mathrm{Cropland}}_{\mathrm{total}}}}$$ (19)
$$\mathrm{{Population}}_{\mathrm{saved}} = \frac{{{\mathrm{Yield}}_{\mathrm{cropN}} \times {\mathrm{Land}}_{\mathrm{sys\;and\;all}}}}{{{\mathrm{Protein}}_\mathrm{{unit}}}}$$ (20)
Uncertainty analysis
In this study, we estimated the uncertainties of nitrogen losses, GHG emissions, costs and benefits for each scenario in 166 countries using 10,000 Monte Carlo simulations. The 95% confidence intervals for all results were calculated. The coefficients of variation (CVs, %) of activity data and parameters are shown in Supplementary Tables 5 and 6, and the uncertainties of the final simulation results are shown in Supplementary Data.
Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.