Main

Global total livestock production has strongly increased in the last decades; in particular, the production of ruminant meat has more than doubled since 1961 (ref. 8). Current livestock production systems, especially ruminant-based farming systems, have substantial environmental consequences in terms of greenhouse gas (GHG) emissions, land use, terrestrial acidification, eutrophication and freshwater withdrawals1. The global food system is responsible for one-third of the global anthropogenic GHG emissions, with livestock production being a major contributor in particular because of methane (CH4) emissions from the digestive processes (enteric fermentation) of ruminants9,10. Land use for livestock production is particularly high, accounting for 80% of global agricultural land if pasture land for grazing and cropland for animal feed production are considered11,12. Moreover, it is estimated that the production of livestock feed accounts for 41% of total agricultural water use, with ruminant meat production being the single largest water consumer13. Further increases of livestock production are projected for the coming decades, specifically in present middle-income countries, driven by population growth and dietary shifts towards animal-based products due to increasing average individual incomes14,15.

A gradual shift towards diets with less animal-farmed protein, in particular ruminant meat, in favour of plant-based protein sources, as suggested by the flexitarian diet of the EAT-Lancet Commission, would be healthier for people and more sustainable for the planet2,16. Adoption of the EAT-Lancet planetary health diet in high-income nations alone could yield a substantial double climate dividend due to GHG emission reduction and carbon sequestration17. However, the question is how such a fundamental behavioural transformation could be achieved at globally relevant scales, considering that key barriers for the substitution of meat with plant-based protein sources include the sensory experience of eating meat, the taste and subjective concerns about the risk of protein deficiency18.

Alternative protein sources

An alternative to largely plant-based diets is to substitute meat by analogues that mimic the taste and texture of animal-farmed products19. Meat analogues can be broadly categorized into three groups: plant-based meat substitutes (for example, soybean burger patties), cultured meat (animal cells cultured in growth medium) and fermentation-derived microbial protein (MP) (microbial biomass produced in bioreactors, also known as single-cell protein)5,7,20. Plant-based meat analogues primarily rely on agricultural crops (for example, soybean) grown on cropland (roughly comparable to plant-based diets). By contrast, commercially available MP for human consumption (mycoprotein), is derived from fungal mycelium cultivated in heated bioreactors using sugar as feedstock6,21. The fermentation process largely decouples the production of edible MP from local biophysical conditions, which might become especially relevant under climate change. However, cropland is still needed for growing sugar crops21. Edible MP produced by methanotrophic or hydrogen-oxidizing chemosynthetic bacteria, which rely on methane or hydrogen and CO2 instead of sugar as energy source, is currently under development and not yet commercially feasible22,23. In a similar fashion, the cultivation of animal cells in a growth medium to produce cultured meat could be largely decoupled from traditional agriculture5. However, cultured meat is still in an early development stage with many unknowns, particularly with regard to the composition and costs of the growth medium7. In this study, we focus on sugar-based MP produced via biological fermentation, which is available commercially today in grocery stores in multiple countries. Biological fermentation has been applied at industrial scale for the production of mycoprotein since the 1980s4,24. Mycoprotein is microbial biomass with meat-like texture and high protein content4,6. The protein quality of mycoprotein, measured by essential amino acid content and digestibility, is equivalent to ruminant meat6,24. Moreover, mycoprotein has been generally recognized as safe by the US Food and Drug Administration since 2002 (refs. 4,25).

The environmental benefits and trade-offs of mycoprotein have been analysed in life cycle assessment (LCA) studies, suggesting substantially lower GHG emissions (approximately 80%), water use (>90%) and land use (>90%) for each unit of ruminant meat substituted with mycoprotein3,7. However, LCA studies also indicate that the replacement of other livestock products such as pork and chicken with mycoprotein would not result in substantial environmental benefits7,25,26. However, many effects of large-scale substitution of animal-farmed products are likely to be non-linear and cannot be scaled up on the basis of static LCA footprints of current production systems. The substitution of livestock products with fermentation-derived analogues has not been studied so far in a dynamic system model accounting for future population growth, food demand, land-use dynamics, agricultural intensification or international trade. Only a single study estimated the total global land savings of alternative protein sources on the basis of population and food production systems of the year 2011, without quantifying the associated GHG emissions and environmental impacts27.

Future scenarios of sugar-based MP

In this study, we analysed the environmental effects of partially substituting ruminant meat with sugar-based MP in global forward-looking scenarios between 2020 and 2050. In line with previous studies, we assumed that biological fermentation for single-cell protein production requires sugar cane grown on cropland as feedstock (Methods)28,29. We limited the substitution of livestock products to ruminant meat, for which previous LCA studies estimated the largest environmental benefits (in contrast to pork and chicken)7. To this end, we used the global multiregional Model of Agricultural Production and its Impact on the Environment (MAgPIE) 4 open-source land-use modelling framework30,31. The MAgPIE framework has been used earlier to study the impacts of replacing animal feed with MP. We built on this previous research and used the middle-of-the-road shared socio-economic pathway (SSP2) scenario, which features increasing population, income and livestock demand (Extended Data Fig. 1 and Supplementary Fig. 2), as our reference scenario (SSP2-Ref-MP0)15,28. In three alternative scenarios, we assumed that 20% (MP20), 50% (MP50) and 80% (MP80) of the per-capita protein consumption from ruminant meat is replaced with sugar-based MP in each model region by 2050 (Fig. 1a and Extended Data Fig. 2). To mimic the typical adoption of new technologies and products by consumers, the fade-in of MP follows an S-shaped curve from 2020 onwards, reaching the target in 2050. The scenario-specific per-capita consumption of ruminant meat and MP is multiplied with the corresponding population to obtain total demand, which is used as the driver in the model (Fig. 1b). In summary, all scenarios are driven by the same overall demand for food crops, feed, livestock products and bioenergy but differ in the substitution targets of ruminant meat with MP (Extended Data Fig. 3 and Supplementary Figs. 2–4).

Fig. 1: Future scenarios of ruminant meat and MP as protein sources in human diets.
figure 1

a, Global per-capita protein consumption of ruminant meat and MP between 2020 and 2050 for the reference scenario (MP0) and three MP scenarios, in which 20, 50 and 80% of the per-capita protein consumption from ruminant meat is substituted with sugar-based MP by 2050. The substitution is phased in in each model region from 2020 onwards following an S-shaped adoption curve (see Extended Data Fig. 2 for regional numbers). b, Total ruminant meat and MP demand, used as driver for the MAgPIE simulations, obtained by the multiplication of scenario-specific per-capita consumption with the corresponding population (Extended Data Fig. 1). Food losses along the supply chain between demand and consumption are accounted for. Historical data from FAOSTAT8. c, Overall feed and feedstock demand for livestock production and MP fermentation. Pasture-based feed demand includes grass for feeding livestock. Cropland-based feed/feedstock demand includes crops for feeding livestock and sugar cane for MP fermentation.

Source data

Full size image

Land-use dynamics

Land-use change, as projected by the MAgPIE model, differs substantially between the reference and MP scenarios. In the reference scenario (MP0), cropland and pasture both increase at the cost of forest and non-forest vegetation between 2020 and 2050 at the global level (Fig. 2). The increase of cropland (175 Mha) and pasture (96 Mha) by 2050 is driven by SSP2-based demand for food crops, feed and livestock products (Supplementary Figs. 2 and 4). The global loss of forest (178 Mha) and non-forest vegetation (92 Mha) by 2050 is largely driven by demand from sub-Saharan Africa and Latin America (Extended Data Fig. 7). In the MP20 scenario, global loss of forest between 2020 and 2050 is much lower (78 Mha), largely because pasture area, in contrast to the reference scenario, does not expand. At the same time, the increase in global cropland by 2050 is similar in both scenarios. The reason for the pasture dynamic is that the 20% per-capita substitution of ruminant meat with MP by 2050 results in rather static total global ruminant meat demand from 2025 onwards (Fig. 1b), which (notably) is sufficient to largely offset future increases of overall pasture feed demand at the global level (Fig. 1c). For cropland, mainly two counteracting processes cancel out each other in MP20: crop-based feed demand for ruminant meat production is reduced, while sugar cane demand as feedstock for MP fermentation is increased (Fig. 1c and Extended Data Fig. 5).

Fig. 2: Global land-use change between 2020 and 2050 for major land types.
figure 2

In the reference scenario, cropland and pasture expand at the costs of forest and non-forest vegetation (part of other land). In the MP scenarios, comparable cropland expansion causes much less deforestation and conversion of non-forest vegetation due to increasing pasture-to-cropland conversion, which is facilitated by lower feed demand from pastures (Extended Data Fig. 5). See Extended Data Figs. 6 and 7 for regional results and validation data.

Source data

Full size image

Higher substitution targets of ruminant meat with MP in the MP50 and MP80 scenarios enhance the land saving effects observed for the MP20 scenario. Further reductions of pasture-based feed demand (Fig. 1c) result in declining global pasture area between 2020 and 2050 (Fig. 2). Thus, cropland increasingly expands into those freed up pasture areas, saving forest and non-forest vegetation from conversion. In the MP80 scenario, there is almost no loss of forest and non-forest vegetation between 2020 and 2050 at the global level (Fig. 2). In comparison to the reference scenario, deforestation and loss of non-forest vegetation is especially reduced in the Congo Basin, Central America and the Amazon Basin (Supplementary Fig. 5).

Non-linear substitution effects

The substitution of ruminant meat with MP reduces several food-related environmental pressures, which can be mapped to sustainable development goals (SDGs). SDGs are aspirational goals with global coverage towards 2030. In this study, we used the following set of environmental indicators, partly adapted from a recent study on SDGs in which MAgPIE was contributing in a multi-model framework approach2: deforestation (SDG15: life on land), carbon dioxide (CO2), CH4 and nitrous oxide (N2O) emissions from agriculture and land-use change (SDG13: climate action), agricultural water use (SDG06: clean water and sanitation) and nitrogen fixation (SDG15). For consistency and for the analysis of relative effects, all environmental indicators reflect annual values. As the scope of MAgPIE is limited to agriculture and land use, we did not account for energy requirements and energy-related GHG emissions of MP production in this study (see Discussion for the implications).

In the reference scenario, global annual deforestation increases from 3.7 Mha year−1 in 2020 to 4.8 Mha year−1 in 2030 and 8.4 Mha year−1 in 2050 (Fig. 3a), mainly driven by forest-to-pasture conversion for animal grazing in sub-Saharan Africa (Supplementary Fig. 4 and Extended Data Fig. 7). In the MP20 scenario, these global annual deforestation rates are roughly halved, resulting in 3.7 Mha year−1 in 2050. A further increase of ruminant meat substitution to 50% by 2050 (MP50) again roughly halves global annual deforestation, resulting in 1.5 Mha year−1 in 2050. The same trend continues in the MP80 scenario, resulting in 0.6 Mha year−1 in 2050. Hence, the substitution of ruminant meat with MP supports the achievement of SDG target 15.2 of halting deforestation.

Fig. 3: Global development of environmental indicators mapped to SDGs.
figure 3

a–f, Absolute values for scenario projections for deforestation (a), CO2 emissions from land-use change (b), CH4 emissions from agriculture (c), N2O emissions from agriculture (d), agricultural water use (e) and nitrogen fixation (f) until 2050, complemented by historical data for validation (see Extended Data Figs. 8 and 9 for regional results and sources of validation data). g, Relative difference of environmental indicators, compared to the reference scenario, in 2050 as the function of scenarios with increasing ruminant meat substitution. Most indicators, including methane emissions, show a linear relationship. However, both deforestation and related net CO2 emissions are reduced by 56%, already at 20% per-capita substitution of ruminant meat with MP, followed by a saturation effect at higher substitution rates. The reason for this non-linear effect is that deforestation, and hence net CO2 emissions, depend on structural change in agricultural production. In contrast, the indicators with linear relationships largely depend on the level of production.

Source data

Full size image

Our results show a non-linear relationship between different levels of ruminant meat substitution and annual deforestation (Fig. 3g). The reason for the non-linear relationship is that land-use change typically does not depend on the level of production but on structural change in agricultural production. In the absence of land degradation, changes in land management or any other disturbing effects, no additional cropland or pasture is needed to maintain agricultural production at the same level. However, to increase the production more land and/or higher yields are needed. Likewise, a reduction of land-based production could decrease managed land or reduce land-use intensity. In our scenarios, the substitution of ruminant meat with MP strongly reduces the demand for animal feed from pastures. In the MP20 scenario, global feed demand from pasture is rather constant from 2020 onwards, in contrast to an increasing trend in the reference scenario (Fig. 1c). Therefore, no increase of global pasture area is needed in MP20 by 2050, which explains the strong reduction of deforestation relative to the reference scenario (56%). However, the forest-saving effect saturates with higher substitution targets in MP50 (82%) and MP80 (93%), in which the global pasture feed demand decreases compared to the reference scenario (Fig. 1c).

CO2 emissions from land-use change are strongly driven by changes in forest cover and hence follow the same non-linear pattern as observed for annual deforestation (Fig. 3b). The CO2 emissions reported in this study reflect net CO2 emissions as they account for carbon losses through deforestation and conversion of non-forest vegetation as well as for carbon gains from afforestation and regrowth of vegetation on abandoned agricultural land. In the reference scenario, global net CO2 emissions from land-use change decrease from 3,957 Mt CO2 year−1 in 2020 to 3,048 Mt CO2 yr−1 in 2030, followed by an increase to 5,496 Mt CO2 year−1 in 2050. The global increase of net CO2 emissions is largely driven by two counteracting regional dynamics. From 2020 onwards, CO2 emissions in Latin America decline but strongly increase in sub-Saharan Africa, both driven by differing socio-economic developments in terms of population and food demand (especially for ruminant meat; see the supplementary information for the regional details). In the MP20 scenario, net CO2 emissions amount to 2,392 Mt CO2 year−1 in 2050, which correspond to a relative reduction of 56% compared to the reference scenario (Fig. 3g). In line with the non-linear relationship for deforestation, the reduction of net CO2 emissions from land-use change shrinks with higher substitution targets. In MP50 and MP80, net CO2 emissions amount to 951 and 734 Mt CO2 year−1 in 2050, respectively. These levels correspond to relative reductions of 83% and 87% for MP50 and MP80, respectively. Hence, the substitution of ruminant meat with MP could strongly reduce net CO2 emissions from land-use change. Such emission reductions can be considered to support the targets of SDG13, although there are no quantitative targets for sectoral emission reductions.

Linear substitution effects

The substitution of ruminant meat with MP also reduces CH4 and N2O emissions from agriculture (SDG13), agricultural water use (SDG06) and nitrogen fixation (SDG15) (Fig. 3c–f). In contrast to land-use change and associated net CO2 emissions, these indicators largely depend on the level of production. Hence, each unit of ruminant meat replaced with MP yields about the same reduction of environmental pressures, indicating a rather linear relationship (Fig. 3g).

Agricultural CH4 emissions, which are largely caused by enteric fermentation in the rumen of cattle (belching), increase in the reference scenario from 208 to 282 Mt CH4 year−1 between 2020 and 2050 at the global scale. The reduced number of cattle in the MP scenarios results in lower CH4 emissions, which amount to 250, 210 and 172 Mt CH4 year−1 in 2050 for MP20, MP50 and MP80, respectively. In relative terms, these numbers reflect reductions of 11, 26 and 39% by 2050 compared to the reference scenario. Similarly, N2O emissions from agricultural soils (fertilizer application) and animal waste management, increase in the reference scenario from 9.5 to 13.4 Mt N2O year−1 between 2020 and 2050. The reduced number of cattle in the MP scenarios lowers the increase of global N2O emissions to 10–12.4 Mt N2O year−1 by 2050, which corresponds to relative reductions of 7–25%. Hence, the substitution of ruminant meat with MP has distinct effects on CH4 and N2O emissions from agriculture. At the regional level, sub-Saharan Africa, Latin America and Asia show the strongest reductions of CH4 and N2O emissions, which is largely driven by the scale of total ruminant meat substituted with MP (Extended Data Figs. 3 and 9).

The effects of ruminant meat substitution on agricultural water use (SDG06) and nitrogen fixation (SDG15) are rather small. Global agricultural water use for food and feed crops increases in the reference scenario from 3,057 to 4,200 km3 year−1 between 2020 and 2050. Reduced demand for animal feed crops in the MP scenarios limits the increase of agricultural water use to 3,868–4,137 km3 year−1 by 2050, which corresponds to relative reductions of 1–8%. However, this will likely not be sufficient to achieve SDG target 6.4 (sustainable withdrawals and supply of freshwater) as already today water withdrawals in many parts of the world tap into environmental flow requirements32. Similarly, nitrogen fixation, a proxy for nitrogen losses to the environment and hence ecosystem degradation, increases from 172 to 234 Mt N year−1 between 2020 and 2050 in the reference scenario. Reduced demand for animal feed crops in the MP scenarios limits the increase of nitrogen fixation to 212–227 Mt N year−1 by 2050, which corresponds to relative reductions of 3–9%. However, these levels are substantially above the target value of 90 Mt N year−1 for SDG 15.5 (ref. 2).

Discussion

In this study, we present the first analysis of substituting ruminant meat with sugar-based MP in forward-looking land-use scenarios. For our model-based projections with global coverage until 2050, we used the spatially explicit land-use model MAgPIE. Our scenarios are based on SSP2, a middle-of-the-road scenario for future population, income and food demand. We quantified the environmental benefits of substituting 20, 50 and 80% of per-capita ruminant meat consumption with MP by 2050 in each model region. Notably, the reduced animal feed demand in the 20% case (MP20) is sufficient to offset future increases of global pasture area, which translates into 56% less deforestation and 56% less net CO2 emissions from land-use change by 2050, both compared to the reference scenario. In the 50 and 80% case, deforestation is further reduced, resulting in relative reductions of 82 and 93% by 2050, respectively. Similarly, net CO2 emissions from land-use change are reduced by 83 and 87% in the 50 and 80% case, respectively. The reason for this non-linear substitution effect is that land-use change, and hence net CO2 emissions, depend on structural change in agricultural production, as opposed to the level of production. The substitution of ruminant meat with MP also reduces non-CO2 emissions from agriculture, agricultural water use and nitrogen fixation. However, these environmental indicators largely depend on the level of production, and hence decrease rather linearly with increasing substitution targets. In particular, global agricultural CH4 emissions are reduced by 11, 26 and 39% at per-capita substitution targets of 20, 50 and 80% by 2050, respectively.

Previous LCA studies estimated substantial environmental benefits of MP derived from fungal mycelium (mycoprotein) over ruminant meat at the product level3,7. In this study, we assessed the consequences of large-scale substitution of ruminant meat with sugar-based MP in global forward-looking scenarios on a set of environmental indicators. Owing to methodological differences, our results cannot be compared directly to existing LCA outcomes. However, our results complement existing LCA studies on the substitution of ruminant meat with MP. First, our study provides an estimate of the absolute and relative reductions of food-related environmental pressures for different substitution targets until 2050, globally and for 12 geopolitical regions. Second, our study shows that the large-scale upscaling of MP as substitute for ruminant meat results in a non-linear saturation effect on land-use change and associated net CO2 emissions—an effect that cannot be captured with the method of static LCA. Similarly, environmental pressures are context-dependent and are not reduced equally around the globe, depending on the development of socio-economic factors such as population dynamics, diet patterns and international trade. This underpins the importance of using a dynamic system model rather than static LCA for estimating the environmental benefits of MP as substitute for ruminant meat.

At the same time, the use of forward-looking modelling tools for analysing the substitution of ruminant meat with MP suggests that the quantified environmental benefits depend on scenario assumptions. In this study, we analysed the substitution of ruminant meat with MP in the context of an SSP2-based scenario broadly characterized by the continuation of current demographic, environmental, technological and societal trends into the future33. However, our results would likely differ under a more sustainable setting such as SSP1 (sustainable development), which is characterized by slower population growth, increased environmental awareness and reduced consumption of livestock products33. Under this setting some environmental benefits of replacing ruminant meat with MP are likely smaller because of (1) overall lower pressure on the land (lower population and dietary change) and (2) improved regulation of externalities such as deforestation. This could especially affect the two indicators for which we identified non-linear substitution effects—deforestation and associated net CO2 emissions—both of which depend on structural change in agricultural production. For instance, global forest cover is estimated to be rather constant throughout the 21st century under an SSP1 setting (SSP1-NDC), in contrast to declining forest cover under a comparable SSP2 setting (SSP2-NDC)2. Hence, the relative reduction of deforestation and net CO2 emissions attributable to the substitution of ruminant meat with MP is likely smaller under SSP1 compared to SSP2. On the contrary, the environmental benefits of substituting ruminant meat with MP might be stronger under a more pessimistic background setting such as SSP3 (regional rivalry), which is characterized by high population growth in low-income countries, low priority for addressing environmental problems and resource-intensive diets33. However, the use of biotechnology for solving environmental problems is inconsistent with the overall SSP3 narrative.

Further factors influencing the scenario set-up and thus the outcome include assumptions about land-based climate change mitigation measures (for example, bioenergy, forest protection and afforestation) and climate change impacts on land (for example, crop yields and carbon stocks in ecosystems). In this study, we deliberately focused on analysing the basic effects of substituting ruminant meat with MP under an SSP2 reference scenario without further assumptions on land-based mitigation and climate change impacts. We did, however, account for existing national polices on forest protection, afforestation and bioenergy (Supplementary Fig. 3). In addition to climate protection measures, future national policies in support of the transition towards a bioeconomy might increase the demand for biomass grown on agricultural land. In our results, the substitution of ruminant meat with MP reduces deforestation through increased pasture-to-cropland conversion. Alternatively, pasture areas no longer needed for livestock grazing could be partly repurposed to biomass cultivation. However, depending on the scale, the production of additional biomass might offset the environmental benefits of MP, especially with regard to deforestation and associated net CO2 emissions. To avoid such trade-offs, policies promoting biomass cultivation should be complemented by forest protection policies34.

Our study is limited to the replacement of ruminant meat with sugar-based MP that is currently commercially available for human consumption (mycoprotein)4. Edible MP produced by methanotrophic or hydrogen-oxidizing chemosynthetic bacteria (power to food) is an emerging technology that, in contrast to mycoprotein, does not rely on biomass as an energy source22,23. Therefore, the land-use requirement of power to food is considerably smaller compared to mycoprotein22, unless the hydrogen or methane itself is being produced using biomass28. The climate impacts of MP produced via power to food are estimated to be lower compared to mycoprotein but strongly depend on the use of low-emission energy sources22,23. Cultured meat is another technology that might play an important role in replacing animal-sourced protein in the future5,7,20. LCA studies indicate that cultured meat production might require smaller quantities of agricultural inputs and land than ruminant meat production26,35,36. However, those benefits could come at the cost of higher energy requirements, which might undermine the GHG emission savings of cultured meat production, depending on the availability of decarbonized energy generation35,37. Precision fermentation is a further future technology relevant to the alternative protein space, which could be utilized to produce milk protein (as ingredient for dairy analogues) or egg white38,39. However, at the time of writing, no public data for inclusion in our modelling framework on land-based feedstock requirements of cultured meat and precision fermentation were available. Nevertheless, our results for the substitution of ruminant meat with MP can be interpreted as a proxy for the large-scale substitution of ruminant meat or dairy products with other biotechnology-enabled alternatives such as cultured meat or fermentation-based milk analogues.

Our study covers several environmental indicators, including deforestation, GHG emissions from agriculture and land-use change, agricultural water use and nitrogen losses. However, we do not account for the environmental consequences of sugar-based MP production beyond the land-use sector. Especially, our modelling framework is not capable of tracking the energy requirements and energy-related GHG emissions of MP production, which is of key importance for assessing the sustainability of MP production. On the basis of LCA studies, it has been estimated that mycoprotein production has about the same energy requirements as conventional ruminant meat production7. However, this proxy should be interpreted with care because the energy requirements for mycoprotein and ruminant meat production have been calculated with different methods26,35. Moreover, the type of energy needed for MP and ruminant meat production differs. For ruminant meat, animal feed production is a major energy consumer (for example, diesel for tractors and natural gas for synthetic nitrogen fertilizer production)35. By contrast, in cell-cultured food production the whole idea is that bioreactors replace animals6,20. Instead of feeding animals, the feedstock is processed in bioreactors, which use electricity to regulate the temperature and other functions of the bioreactor. Therefore, the land-related GHG emission savings of sugar-based MP shown in our study need to be contrasted with energy-related GHG emissions for assessing the net effect. To avoid that GHG emission savings in the land-use sector are offset or even exceeded by GHG emissions from the energy sector, a large-scale transformation towards cell-cultured food, as assumed in our scenarios, would need to be complemented by a large-scale decarbonization of electricity generation. It is anticipated that recent technological advancements and cost reductions in solar photovoltaics, wind and battery storage could make renewable energy cost-competitive compared to carbon-based fuels in the near future and that considerably higher electrification shares across different sectors are possible40.

Moreover, if MP were to replace ruminant meat at large-scale, as assumed in our scenarios, this transformation would likely reduce the provision of non-food animal by-products such as hides and skins for leather products, organs for pet food, fat for chemicals, bones and blood for fertilizers as well as non-food services from animal husbandry such as traction and insurance, the latter being especially relevant in low-income countries41. These non-food by-products, which are often by-products of meat production, would need to be replaced by alternatives such as synthetic leather, synthetic fertilizer or plant-based fats, causing additional GHG emissions and other environmental impacts that are not considered in our assessment. Partly, non-food by-products could be replaced in the future by fermentation-enabled alternatives such as fungi-based leather42. However, in analogy to MP this could result in higher energy-related GHG emissions, depending on the sustainability of energy production.

Future research could address some of the identified gaps by studying the impacts of meat and dairy analogues in an integrated assessment model, which accounts for energy demand and production including GHG emissions, and economy-wide impacts, next to a detailed representation of land-use dynamics. In addition, this would allow us to analyse the role of meat and dairy analogues as part of a portfolio of climate change mitigation options.

Methods

MAgPIE land-use model

MAgPIE was developed and used to assess the competition for land and water and the associated consequences for sustainable development under future scenarios of rising food, energy and material demand30. The model version we used in this study is MAgPIE v.4.3.4 (ref. 31) (see the data availability statement at the end of the article for details). MAgPIE combines economic and biophysical approaches to simulate spatially explicit global scenarios of land use within the 21st century and the respective interactions with the environment (Supplementary Fig. 1). The MAgPIE framework has been used to simulate mitigation pathways for different SSPs15 and contributed to several reports of the Intergovernmental Panel on Climate Change43,44.

MAgPIE is a global multiregional partial equilibrium model of the land-use sector45. The model integrates regional economic conditions such as demand for agricultural commodities, technological development and production costs as well as spatially explicit data on biophysical constraints into an economic decision-making process based on the concept of recursive dynamic cost optimization. Geographically explicit data on biophysical conditions are provided by the Lund-Potsdam-Jena managed land model46,47 on a 0.5-degree resolution and include, for example, carbon densities of different vegetation types, agricultural productivity such as crop yields and water availability for irrigation. Owing to computational constraints, all model inputs in 0.5-degree resolution are aggregated to simulation units for the optimization process on the basis of a clustering algorithm48. Available land types in MAgPIE are cropland, pasture, forest, other land (including non-forest vegetation, abandoned agricultural land and deserts) and settlements. Cropland (rain-fed and irrigated), pasture, forest and other land are endogenously determined, whereas settlement areas are assumed to be constant over time. Cropland covers cultivation of different crop types (for example, temperate and tropical cereals, maize, rice, oilseed, roots), both rain-fed and irrigated systems and two second-generation bioenergy crop types (grassy and woody). International trade is based on historical trade patterns and economic competitiveness. Food demand is derived on the basis of population growth and dietary transitions, accounting for changes in intake and food waste, the shift in the share of animal calories, processed products, fruits and vegetables as well as staples.

In this study, we derived the following environmental indicators from MAgPIE (see Extended Data Table 1 for a structured overview), of which most have been used in previous studies2,34,49,50,51. Annual deforestation (Mha year−1) was calculated on the basis of the differences in forest area between time steps. As the calculation was based on changes of forest area, annual deforestation may have varied substantially between time steps (stock and flow problem). To avoid that our results are biased by the values of single years, we calculated in a post-processing step an average value of annual deforestation by applying a function (low-pass filter) that distributes values of annual deforestation over time, while making sure that the time integral over the modelled period are the same. Similarly, annual net CO2 emissions (Mt CO2 year−1) from land-use change were calculated on the basis of changes in carbon stocks of vegetation and therefore may have varied substantially between time steps (stock and flow problem). To avoid biased results, we applied the low-pass filter function also on annual net CO2 emissions from land-use change. Carbon stock changes in vegetation are subject to land-use change dynamics such as conversion of forest into agricultural land34. In case of afforestation or when agricultural land is set aside from production, regrowth of natural vegetation absorbs carbon from the atmosphere (removals). N2O emissions (Mt N2O year−1) from agricultural soils (fertilizer application) and animal waste management were estimated on the basis of nitrogen budgets for cropland, pastures and the livestock sector49,51. CH4 emissions (Mt CH4 year−1) from agriculture include emissions from enteric fermentation, animal waste management and rice cultivation, which were estimated on the basis of feed demand, manure and rice cultivation area, respectively49,50. Nitrogen fixation (Mt N year−1) is a proxy for nitrogen losses to the environment and hence ecosystem degradation. Nitrogen inputs on cropland via industrial (for example, production of inorganic fertilizers) and intentional biological fixation were calculated on the basis of a nitrogen budget approach2,51. Agricultural water use (km3 year−1) depends on the water requirements of crops, the available water for irrigation, irrigation efficiency and irrigation infrastructure, which can be extended endogenously on the basis of cost-effectiveness52. For more information on the MAgPIE modelling framework, we refer the reader to the model source code and documentation (see the data availability statement).

MP in MAgPIE

Fermentation-based MP production was implemented in an earlier version of MAgPIE to study the impacts of replacing animal feed with microbial protein28. Building on this previous research, we included a refined implementation of MP production into MAgPIE v.4.3.4 (ref. 31) to study the impacts of replacing ruminant meat with MP in human diets. In line with the literature on MP for human consumption, we assumed a dry matter (DM) protein content of 45% in microbial biomass (based on mycoprotein)4,6. For the production of MP, we assumed that sugar cane, grown on cropland, was needed as feedstock. Using the results of Pikaar et al.28, we assumed that 4.3 ton of sugar cane would be needed to produce 1 ton of microbial biomass, all on a DM basis. This suggests that approximately 0.2326 ton DM microbial biomass can be produced from 1 ton DM sugar cane. Assuming that 1 ton DM sugar cane yields 0.3363 ton DM sugar, we get 0.69 ton DM microbial biomass from 1 ton DM sugar, which is well within the range of 0.42–0.87 ton DM microbial biomass per ton DM sugar published in Lapeña et al. (Table 1)29. Sugar cane cultivation is largely limited to tropical and subtropical regions. Therefore, in our modelling framework, temperate and boreal regions partly rely on imports of feedstock for MP production. For ruminant meat, we assumed a food protein content of 33% in DM (own calculations based on FAOSTAT8 using a DM content of 41%). The DM food protein content of 33% reflects an average value across different ruminant meat products including beef, ground beef and processed meat. The corresponding fresh matter food protein content of 13.5% is comparable to other estimates for the average food protein content of beef products53,54. We used the DM protein content for the per-capita substitution of ruminant meat with MP. Together with the DM protein content of 45% in microbial biomass, this suggests that 1 ton DM ruminant meat is replaced by 0.73 ton DM MP. With regard to costs, we assumed that each ton DM MP costs US$789, using Supplementary Table 9 in Pikaar et al.28. The costs account for energy, oxygen, nitrogen and phosphorus requirements. Feedstock costs were excluded to avoid double accounting as MAgPIE has its own feedstock costs. We did not account for the environmental consequences of MP production beyond the land-use sector. In particular, we did not account for the energy requirements and energy-related GHG emissions of MP production.

Scenario assumptions

The reference scenario (SSP2-Ref-MP0) is based on SSP2 with regard to population, income, diets, land-use regulation and trade. The MP scenarios (SSP2-Ref-MP20, SSP2-Ref-MP50 and SSP2-Ref-MP80) differ from the reference scenario only with regard to the per-capita substitution of ruminant meat with MP in human diets. The consumption of per-capita protein summed over ruminant meat and MP is the same (Fig. 1a). In the MP scenarios, we assumed that 20, 50 and 80% of the per-capita ruminant meat consumption is substituted with MP by 2050 in each model region. The fade-in of MP follows an S-shaped curve to mimic the typical adoption of new technologies and products by consumers. In our modelling framework, livestock commodities (ruminant meat, whole milk, pork, poultry meat and eggs) are produced in five animal food systems (beef cattle, dairy cattle, pigs, broilers and laying hens). The production of ruminant meat is allocated to beef cattle and dairy cattle systems according to historical shares. However, the substitution of ruminant meat with MP in our scenarios only aims at reducing ruminant meat from beef cattle. Dairy production is largely unchanged, even at high MP substitution rates (Extended Data Fig. 4). Our scenario set-up with relative substitution rates (20, 50 and 80%) by 2050 in each model region is designed to enable straightforward comparison of environmental indicators between scenarios and regions. However, this suggests that low-income countries would cut ruminant meat consumption with the same level of ambition as high-income countries, neither accounting for the overall share of livestock products in diets and the likelihood of adopting new diets nor addressing the economic and cultural context in which a substitution of ruminant meat with MP would take place.

Reporting summary

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