Main

Burgeoning food demands from growing and urbanizing populations, paralleled with increases in the consumption of animal-source foods (ASFs), drive an ever-larger pressure from food systems on the environment1,2. While causing one third of anthropogenic greenhouse gas emissions globally3, agriculture is also the leading contributor to the Earth system surpassing planetary boundaries in biodiversity loss and nutrient flows2. Concurrently, the additional burden of malnutrition, associated with poor or insufficient diets, further indicates that food systems are failing to meet health needs4. Such recent research has catalysed broad conclusions that urgently compel transitions towards sustainable diets5,6,7.

Many products, here termed ‘novel/future foods’ (NFFs), have the potential to reduce environmental impacts of diets while meeting essential nutritional needs in broader populations8. Novel foods are those produced from new production technologies or that are under novel regulatory frameworks such as cell-culturing technologies—cultured meat, eggs, milk, plants, algae, bacteria and fungi9. Future foods are those for which our production capacity has the potential to scale up and/or for which consumption may increase due to emerging climate change mitigation concerns, such as insects and spirulina; some foods may overlap in both the novel and future categories, such as mussels (Mytilus spp.) or chlorella (Chlorella vulgaris) produced with novel technologies8. Such NFFs may provide nutritious alternatives to ASFs while meeting multiple sustainability goals8,9. Compared with currently available plant-based protein-rich (PBPR) options such as legumes, pulses and grains, NFFs can have a more complete array of essential nutrients such as protein, calcium, vitamin B12 and omega-3 long-chain polyunsaturated fatty acids, they are more land- and water-efficient than current ASFs8. Additionally, alternative fortified food products can be developed, but the taste and texture of meat are key drivers in the development of cultured meat in particular10. In this paper, we combine novel and future foods into a selection of NFFs for which data on environmental impacts are available8.

Studies on alternative dietary pattern scenarios (for example, vegetarian, vegan or flexitarian)11,12,13 or currently consumed dietary patterns (for example, Mediterranean or New Nordic diets)14,15,16 confirm that large shifts from current diets towards more plant-based diets are needed. Vegan and flexitarian or partially omnivore diets, mainly reducing meat consumption, will be important diet shifts for synergistic benefits to health and environmental outcomes17,18. However, due to less favourable profiles in terms of some nutrients in plant-based options such as pulses and grains, diet-level comparisons with omnivore and plant-based diets are also needed to investigate the feasibility of including NFFs in future diets to meet nutritional needs with lower impacts. Additionally, studies comparing multiple environmental impacts of diets including NFFs are lacking, and a broadened understanding of the NFFs that best balance the trade-offs in impacts and nutrition can inform the development of sustainable options for future diets and recommendations18,19.

Here we estimated the prospect of reducing the global warming potential (GWP), scarcity-weighted water use (WU) and land use (LU) of current European diets (CDs). More specifically, we optimized the average European diet according to three diet types, which varied in their inclusion of ASFs, PBPR alternatives and NFFs. All NFF, omnivore (OMN) and vegan (VEG) diets were optimized to meet nutritional adequacy and feasible consumption constraints.

Results

Current average and optimized diets

Some food groups were consistently decreased in the optimized diets, irrespective of minimized objective function—notably, all beverages, dairy, meats, fish/seafood, animal fats, starchy roots/tubers and spices/condiments (Table 1). Large increases in (fortified) liquid PBPR alternatives were needed to meet calcium and vitamin D requirements in all modelled diets except the VEG minimum GWP. PBPR alternatives in the optimized diets increased many times over their intake compared with the CD (Supplementary Fig. 1). Similarly, increases in other PBPR options and vegetables were common among all optimized diets, except legumes in the NFF minimum LU diet.

Table 1 Composition of food groups in optimized diets

Certain food groups, however (such as grains, eggs, fruit, snacks, sugar and plant fats), had different directions and magnitudes of change depending on which impact was minimized when compared with the CD. For example, grains decreased in the OMN and NFF minimized LU diets but increased in the VEG minimum LU diet as well as in all minimized GWP and WU diets. Plant fats decreased in all minimized WU diets and increased in all minimum GWP and LU diets. Eggs were also included in the OMN minimum GWP diet. Liquid plant-based alternatives were not included in the VEG minimum GWP diet but instead were replaced by grains, fruits and plant fats. In the NFF minimum LU diet, NFFs increased with corresponding reductions in legumes/nuts and grains.

NFFs selected

In the NFF optimizations, different NFFs were selected for each minimized environmental impact. The primarily selected NFFs in the initial optimization—not the sensitivity analyses—were cultured milk (with an intake of 45–155 g per day depending on the minimized impact) and insect meal (34–113 g per day). The third most selected NFF differed by minimized objective function: microbial protein (0.02 g per day) was selected in the NFF minimum GWP diet, cultured meat was selected in the minimum WU diet (0.10 g per day) and mycoprotein was selected in the minimum LU diet (29 g per day) (see the optimized results by product, group and diet type in Supplementary Data 1).

Environmental impacts of optimized diets

In comparison with the CD, the optimized OMN and VEG diets reduced GWP, WU and LU by 81–84% (Table 2). NFF diets reduced GWP by 83%, WU by 85% and LU by 87% compared with the CD. NFF diets had 4–34% fewer overall impacts than the OMN and VEG optimized diets, with one exception: the VEG diet minimized for GWP had 8% less impact than the NFF diet minimized for GWP (see the impact ranges by food group in Supplementary Fig. 2).

Table 2 Diet environmental impacts and optimization constraints

Meats comprised most (>50%) of the GWP and LU in the CD, and they shared a large portion of the WU impacts with the ‘Other’ group, which includes sugars, all beverages (except liquid dairy and plant-based dairy alternatives), spices/condiments and snacks (Fig. 1). Not only do the optimized diets have over 80% lower environmental impacts than the CD, but in the optimized diets, the majority of the GWP and LU were from the PBPR alternatives (here denoting meat substitutes, tofu, plant-based milks and legumes/nuts), plant fats and other groups. The majority of WU impacts were from fruits, vegetables and grains in the optimized diets.

Fig. 1: Environmental impact by food group in different diets.
figure 1

Contributions of different food groups to the GWP, LU and WU of the CD (top) and optimized diets (bottom) with all nutrition and feasible consumption constraints. ‘Other’ includes snacks, sugars, juice, non-alcoholic beverages, alcoholic beverages and spices/condiments.

The main food groups contributing to macronutrient intake in each optimized diet depended on which environmental impact was minimized (Supplementary Fig. 3). It is important to note that for the VEG diet models, we needed to remove the vitamin B12 and D constraints, since no feasible solution was possible with these vitamin requirements. In the future, it is likely that PBPR alternatives and NFF products would be fortified with vitamins B12 and D and other micronutrients, and such diets would therefore include vitamins B12 and D from other sources.

Sensitivity analyses

In the first sensitivity analysis (NFF.1), we excluded insect meal and cultured milk, as they were the predominantly selected NFFs in the initial optimizations. Once insect meal and cultured milk were excluded, microbial protein was the primarily selected NFF in the minimized LU diet (111 g per day), cultured meat (29 g per day) was selected in the minimum WU diet and kelp was selected in the minimum GWP diet (17 g per day). Compared with the original NFF diets, the NFF.1 diets had 6% higher GWP, 16% higher WU and 18% higher LU, when each was minimized. Yet, all NFF.1 diets had over 82% fewer impacts than the CD. Additionally, even without all NFFs available in the optimizations, these sensitivity analyses show that optimized NFF diets have 82–85% fewer impacts than the CD (unoptimized), regardless of which environmental impact is minimized.

The second sensitivity analysis focused on ASF requirements for the OMN models. Since the initial optimized OMN diets became essentially almost entirely vegan—including only small amounts of dairy, offal, eggs and animal fats—we tested how the impacts would change if the models required no more than a ±80% change in the current intake of all ASFs. In other words, the OMN.1 sensitivity model was forced to include at least 20% of the current mean intake of meats, dairy, fish/seafood, animal fats and eggs (Supplementary Fig. 4). The OMN.1 optimization included the minimum allowable amount of all ASFs, except eggs in the minimum GWP and LU diets and dairy in the minimum WU diet, with greater impacts than the initial (almost vegan) OMN models: 42% higher GWP, 23% higher WU and 41% higher LU. Yet, in comparison with the CD, there were still large reductions in impacts: 70% lower GWP, 79% lower WU and 68% lower LU.

Lastly, the OMN-NFF model examined what the optimization would select if allowed to include both ASFs and NFFs. The OMN-NFF diets were subject to the original nutrition and feasible consumption constraints on intake of all food items, including between the 5th and 95th percentiles of all ASFs (Fig. 2). All OMN-NFF diets had less impacts than the CD (>83% for all categories) and the OMN.1 diet (28–62%). Additionally, the OMN-NFF diets had slightly less impacts than the original OMN diets (where NFFs were not allowed), with 4% less GWP, 7% less WU and 37% less LU when each was minimized. These results indicate that the inclusion of small amounts of ASFs could lower the impacts of diets that also include NFFs while meeting nutritional needs. In the minimum-GWP OMN-NFF diet, all ASFs were removed, and insect meal, microbial protein and cultured milk were the selected NFFs. When minimizing WU, the OMN-NFF diet included small amounts of meat offal, cream, cheese and mixed fats while also including a variety of NFFs—cultured meat, microbial protein, insect meal and cultured milk. In the minimum-LU OMN-NFF diet, cream and mixed fats were the only ASFs selected, and the NFFs selected included insect meal, microbial protein, cultured milk, mycoprotein and kelp. See the full sensitivity analysis results in Supplementary Data 1.

Fig. 2: Sensitivity analyses.
figure 2

a, Percentage change by food group from the CD in optimized sensitivity analysis OMN and NFF diets (OMN-NFF) with all nutrition and feasible consumption constraints, all NFFs and ASFs allowed, minimized for GWP, LU and WU. b, Total mass (average g per day) of NFFs selected in the optimization with all NFFs and ASFs available by impact minimized. Plant-based alternatives are removed since they were increased over 1,000% from CD intakes.

Discussion

This study identified diets that greatly reduce environmental impacts compared with CDs in Europe and include NFFs as replacements for ASFs17,20. Since we minimized diets for environmental impacts and assumed that large changes were needed from CDs, our models achieve higher relative impact reductions than other comparable optimizations18. Yet, in agreement with previous studies, our models show similar reductions in overall environmental impacts of optimized diets when compared with CDs17,19,21,22. Our models tended to have lower GWP impacts than previous reviews of optimized diets indicate (measured in greenhouse gas emissions)23. For example, many studies find on average 30–50% reductions in GWP of their optimized diets compared with baseline diets, but theoretical maximum decreases of 70%12 to 78%18. Our diet models (1.00–1.14 kg CO2 per day) are similar to such theoretical European diets minimized for GWP at 0.95 kg CO2 equivalent per day18. Such optimistically lower environmental impacts in our models are probably due to our wider consumption constraints and inclusion of a variety of novel products.

Allowing for ASFs to be replaced by NFFs resulted in notably lower impacts. Similar reductions across impacts are possible when ASFs are replaced by PBPR alternatives and plant fats. This is where our models differ from most diet optimization studies: our greater-than-5th-percentile-per-food-item constraint allowed most ASFs to be essentially eliminated from an OMN diet, whereas in other studies, the common objective function is to minimize the difference from currently consumed diets19,23. Yet, even in our sensitivity analyses, where we forced the optimization to retain at least 20% of the current average mass of each ASF item, the OMN diets showed 83–88% lower impacts than CDs. Meats account for a large portion of the impacts of CDs7,24. Even when allowed, livestock products are consistently reduced and often eliminated from optimized diets23, meaning that conventional ASFs are less environmentally efficient even when nutritional content is considered. Reductions in meats in particular are responsible for around 60% lower environmental impacts in optimized diets25. Our findings suggest that diets could be more land, water and carbon efficient if people would be amenable to more abstemious consumption. The findings also indicate that by adjusting CDs and/or including NFFs and even small amounts of selected ASFs, it is possible to reduce environmental impacts to similar levels as optimized VEG diets. The selection of mainly insect meal, cultured milk, microbial protein and mycoprotein by the optimization indicates that these NFF products have the best balance of trade-offs between nutritional content and environmental impacts given the current data. However, only a few NFFs were selected, and indeed the selection of relatively few types of products in all models may indicate the overspecialization common in linear programming, which limits the investigation of interdependent variables and the need for diversity in the diet7.

Since we optimized separately for three different environmental impacts, the question that remains is which of the diets should be followed. There were trade-offs and synergies among the diets optimized for different environmental impacts. Large increases in legumes (especially in the WU diets) and vegetables (especially in the LU diets) and large decreases in all ASFs and starchy roots/tubers were consistent across minimized impacts and diet types. Some food groups involve environmental impact trade-offs, as they may have more efficient resource use in one category than in others; for example, grains were reduced in the WU diets but increased in the GWP and LU diets. As with other models, we found that the diets tended to be similar in which foods were included and the relative amounts in each food group, and that most if not all ASFs were excluded even with the inclusion of NFFs or PBPR foods, which suggests synergies rather than trade-offs among minimized impacts22,23.

Given the substantial role of ASFs cross-culturally and myriad other functions of livestock in food systems, diets completely devoid of ASFs (such as those following our optimizations) may be difficult to realistically adopt at a large scale24,26. Additionally, concurrent reductions in micronutrients need to be managed through protein source replacement with nutritious options and carefully designed fortification and supplementation policies27. Indeed, even if optimization models yield encouraging results for replacing conventional ASFs with NFFs, the nuances of feasible consumption must be considered, driven by motives of taste, health, familiarity, attitudes, food neophobia or disgust, and social norms10. Although it varies among countries, the acceptance of PBPR alternatives is greater than that of cultured meat and insects10, with perceived naturalness and familiarity being main concerns in Europe28. Yet, the acceptance of NFFs can increase with positive information highlighting environmental, health (for example, micronutrient supplementation and antioxidant/anti-inflammatory properties) and animal-welfare benefits, though it still depends in large part on taste and price29,30. Some claim that NFFs provide additional possibilities for “dietary resilience” in the face of uncertain future climate change due to their prospect to provide essential nutrition through unforeseen disturbances29.

This study is limited by the sparse availability of life cycle assessment (LCA) data on NFFs, and these products constitute the limits of current data availability. We therefore recommend future research to expand the impact assessment for a full understanding of the environmental, socio-cultural and health implications when including NFFs in whole diets. Future studies should assess the capacity for producing these NFFs for entire populations, in Europe or at individual country levels. Further assessment of the food security and socio-cultural aspects of affordability, availability and cultural acceptability in future diet scenarios including NFFs is needed. Indeed, the affordability and viability of certain NFFs such as cultured meat are prerequisites to inclusion in future diets9. We also acknowledge that myriad other actions, which were not a focus of this paper (such as a combination of policy changes, education initiatives, sustainable production methods, closures of yield gaps and waste reduction), will also be required for more sustainable future food systems. Our study assesses the inclusion and impacts of NFFs in whole diets instead of assessing these products individually. The findings of this study demonstrate that including NFFs in whole diets and replacing conventional ASFs with PBPR and NFF alternatives have the potential to reduce GWP, WU and LU by more than 80%.

Hence, this study adds to the growing body of literature confirming that diet shifts towards the increased use of PBPR foods and investments into development, production and strategies for adoption of NFFs have great potential to reduce environmental impacts while providing nutritious options. NFFs may provide options for diversifying diets but require other, intermediate means for promotion and consumption, such as education on the similarity with familiar foods, market accessibility through lower prices for consumers and incentivizing procurement for institutional and corporate food businesses. Given the complexities and the acknowledgement that there is no such thing as a panacea, action is needed on all fronts to move towards such diets and sustainable future food systems.

Methods

Product database

Current diets

We followed the methods from the ‘SustTable Database’ of Gazan et al.31 to compile a database for the optimization of diets using multiple sustainability metrics for foods31. We obtained food consumption data on average (chronic) intake of food items in g per capita per day for the year 2013 from the European Food Safety Authority (EFSA) Comprehensive European Food Consumption Database compiled from 34 national food consumption surveys (n = 66,492 individuals) in 22 European Union countries32. We selected the food products at FoodEx2 Level 2 except where more detail was needed (here, only liquid PBPR alternatives and grains). FoodEx2 aggregates quantities (in grams) of food items into four levels: for example, grains and grain-based products (Level 1), grain milling products (Level 2), wheat milling products (Level 3) and wheat flour, durum (Level 4). The selected food items totalled 124 individual food products, aggregated into 18 food groups. We moved items such as butter and mixed fats to their own ‘animal fats’ group, and fine pastry wares (for example, cookies) from ‘grains’ to ‘snacks’. ‘Meats’, here a subcategory of ASFs (which also included fish/seafood, dairy, animal fats and eggs) denoted terrestrial animal flesh in whole or processed products such as beef, chicken, pork, meat offal and meat pastries. Though meats do have varying impacts by type of product, our categorization is for setting constraints and understanding the role of ASFs in overall diets. Mussels (Mytilus spp.), fish and crustaceans are included in the fish/seafood food group. See all products, food groups and data in Supplementary Data 1.

We linked each of the 124 individual food items from EFSA FoodEx2 Level 2 to data on product nutrient composition from the US Department of Agriculture (USDA) FoodData Central, chosen for its comprehensive inclusion of macro- and micronutrients and amino acids33,34. Data on iodine and omega-3 long-chain polyunsaturated fatty acids were not available. When matching EFSA data to USDA FoodData, we selected PBPR food alternatives, including tofu and plant-based milk replacements, which are typically fortified in some European countries with vitamins D, B2 and B12 and/or calcium. We focused on the European population, assuming healthy adults who are active and get most (if not all) of their nutrition from foods with no vitamin supplementation. The exception was vitamin D; we assumed that in addition to an intake of about five micrograms per day from food, vitamin D supplementation is needed to ensure adequate status. See Supplementary Table 1 for the full list of fortified foods and their added vitamins and minerals.

The environmental impacts of the foods were based on LCA studies35,36. Our system boundaries were from cradle to consumer—included cooking at the consumer stage, if necessary. The LCA inventory data for the 124 food items included in the CD were sourced from the AGRIBALYSE 3.1 LCA Database37 using the OpenLCA v.1.10.3 software38. AGRIBALYSE is a multi-indicator French life cycle inventory analysis database with data for over 2,500 products produced in France. We assumed that the French data reproduce similar relative transportation and production differences across Europe. AGRIBALYSE considers transportation emissions of products imported from outside of Europe. The ReCiPe 2016 Midpoint (H) method39 was used to calculate the GWP and LU, and the AWARE method40 was used to calculate scarcity-weighted WU of the food items. We matched the products in the AGRIBALYSE database with the EFSA FoodEx2 Level 2 coding of the food items.

NFFs

We selected eight NFFs to be included in the study because those products have the possibility to be produced in the future at scale with the nutrient profiles to replace conventional ASFs41. We also selected NFFs for which data on their production are currently available8,9. The NFFs included here were cultured meat, ovalbumin, microbial protein (hydrogen-oxidizing bacteria), microalgae (Chlorella vulgaris), insect meal (Hermetia illucens), cultured milk, cloudberry cell culture (Rubus chamaemorus), kelp (Saccharina latissima) and mycoprotein.

Environmental impact data for the NFFs were obtained from LCAs in recently published literature42,43,44,45,46,47. The microbial protein was assumed to contain 5% moisture and 65% protein42. The impacts of cell-cultured ovalbumin were given per kg of dried powder with an 8% moisture and 92% protein content46. Cultured milk was assumed to consist of 97% oat milk and 3% cultured milk protein (casein) by weight. For cultured milk protein, the environmental impacts were based on the same LCA data as what was used for cultured ovalbumin, since an LCA study for cultured milk protein indicates that the unallocated impacts are at the same level as those of cultured ovalbumin48; amino acid composition was assumed to be the same as that of liquid milk. The results for microalgae from Smetana et al.43 were originally calculated using the IMPACT 2002+ method. We therefore remodelled the product system with the SimaPro v.9.1.0.11 PhD software package49 using the inventory data for the scenario provided. This allowed us to recalculate the environmental impacts using comparable LCA methods, which additionally allowed for the modelling of uncertainties of the system. The environmental impact of cultured meat was calculated per kg of cultured meat with dry matter content of 30% and protein content of 19%. The cultured meat scenario used the same LCA methods as above47. The cloudberry cell culture LCA data used best-case scenario energy data and assumed dried product with 5% moisture and 19% protein content45. The impacts of mycoprotein are from aggregated reports of LCA for Swedish consumers50. Dried kelp, or sea belt, LCA impacts were calculated with the same methods from the AGRIBALYSE database. Electricity consumption for all products was modelled using the French electricity mix in the life cycle inventory, while the French non-irrigation characterization factor was used to assess the impact of WU. Since we were modelling European diets, we tested the sensitivity of the model to the use of French electricity data. We ran the same NFF optimizations with electricity from Europe (without Switzerland) and found that the use of the French data was a valid assumption. There was a difference of less than ±0.5% in all of the overall diet impacts and only slight changes to the amounts but not to the types of NFFs selected (‘Sensitivity analyses’).

Direct matching of LCA methods was not possible in some cases because the material provided in publications and by author correspondence was not sufficient for calculating the impacts with the same methods. For insect meal, we used Smetana et al.’s44 Hermetia illucens insect biomass attributional LCA with IMPACT 2002+ method51 mean data only35.

We added aspects of the life cycle not considered in the original LCAs of the NFFs to match the system boundaries—cradle to consumer—for the conventional CD products; these additional steps included transportation, packaging and retail impacts. For protein powder products, we added the required steps similar to those of dried nuts, for cultured milk we used the inputs of liquid milk, and for cultured meat those of minced meat. We obtained nutrition composition data on items or closely matching items to the NFFs from the USDA FoodData Central or from published studies (see Supplementary Table 2 for the data sources on nutrition and Supplementary Data 1 for the environmental impact calculations).

Diet optimization

Linear programming

We applied a linear programming optimization method using the lpSolve package in R version 4.1.0 (ref. 17,18). Optimization problems map a search space of decision variables into a solution space yielding optimized levels given the objective function52. Our objective function was to find diets that minimized three environmental impacts for three different diet types: OMN optional diets, VEG diets and NFF diets. We optimized the three diets by minimizing each of three separate impacts, GWP (kg CO2 equivalent), WU (m3) and LU (m2 arable land equivalent), resulting in nine optimized diets, plus three more diets resulting from sensitivity analyses. Each diet was constrained to fulfil nutritional requirements and feasible consumption constraints (see ‘Feasible consumption constraints’ and Supplementary Table 3 for the full list of constraints in each diet model). The three main diet models were differentiated by their excluded food groups: OMN diets excluded only NFFs, VEG diets excluded all ASFs and NFFs, and NFF diets excluded only ASFs. Furthermore, we estimated the environmental impacts of the CD for comparison with the optimized diets.

Nutrition constraints

We set optimization constraints for the diets to meet daily reference values of macro/micronutrients for EFSA/Nordic Nutrition Recommendations adult diets with the same energy intake as the CD (2,481 kcal per day)33,53. We used only the constraints provided by the boundaries of the recommendations. For some nutrients, only one boundary applies—for example, an upper and lower boundary on total polyunsaturated fatty acids but only an upper limit on total saturated fatty acids. Essential amino acid requirements were from FAO/WHO54 amino acid requirements for an adult, reference weight 70 kg from EFSA55. See Supplementary Table 3 for all nutrition constraints.

Feasible consumption constraints

We set optimization constraints for individual products to remain within the 5th–95th percentile of mass per product in CD consumption to ensure that the model diets stayed within feasible consumption limits. Following the methods of previous optimizations12 and conclusions of the EAT-Lancet Report7, we assumed that large shifts in diets will be needed, probably beyond what currently would be considered culturally acceptable. For the NFF and VEG diets, we set the ASF groups—meats, seafood, dairy, eggs and animal fats—equal to zero. Since most of the NFFs are not currently consumed, there are no consumption data and their cultural acceptability is not yet well understood, so we instead included feasible consumption constraints. We set feasible consumption constraints on the NFF products on the basis of replacement of designated proxies for ASFs (see the calculations in Supplementary Data 1). The NFFs, hypothesized to replace the ASFs, were given allowed intake constraints calculated to provide the same percentage of protein as proxy ASFs—liquid milk for cultured milk, meats for cultured meats and so on. The NFF constraints were set from 0 g per day to the current mean protein intake of the proxy product plus 0.5 standard deviation. Noted exceptions were microalgae, kelp and plant cell culture, known to have specific upper limits on safe daily intake56,57,58. Furthermore, since vegan diets in Europe significantly differ from non-vegetarian/omnivore diets, we also used proxy items as the feasible consumption constraints for individual PBPR alternatives—cheese for tofu, dairy for milk replacements and so on59. Additionally, since current liquid plant-based alternatives are the main PBPR alternatives that are fortified with vitamin D, vitamin B12, vitamin B2 and calcium, increases were expected in the optimized diets. A group constraint was therefore implemented for liquid PBPR alternatives to remain within realistic intake at mean plus 0.5 standard deviation of current dairy milk consumption (≤297 g per day). To maintain realistic food group intakes for recommended dietary diversity and daily consumption of a variety of fruits and vegetables, we set food group constraints for fruits and vegetables on the basis of the recommendations of the ‘planetary health diet’ from the EAT-Lancet Report, ≥200 g per day for vegetables and ≥100 g per day for fruits7. We set the total intake of alcoholic beverages to ≤20 g per day55.

Uncertainty analysis

We conducted an uncertainty analysis using Monte Carlo with 100 iterations per product when we ran the ReCiPe Midpoint(H)/AWARE LCA impact methods60. We ran the linear optimization on each product with 100 Monte Carlo iterations per impact, yielding the mean and standard deviation of impact calculated by food product of all optimized diets. Due to negative WU values for some products, we used only the mean water scarcity value resulting from the Monte Carlo iterations for all the NFFs. The cause of negative values resulting from Monte Carlo runs is more thoroughly discussed elsewhere42,61, so we decided to use the baseline results of the LCA for each optimization.

Sensitivity analyses

We ran three different sensitivity analyses to estimate how sensitive our models were to individual changes in the data and constraints. After we modelled the (initial) optimized NFF diets, in the first sensitivity analysis (NFF.1), we removed the NFFs that were selected in the highest amount in the model—here, insect meal and cultured milk. We then compared the difference in environmental impacts between this sensitivity analysis optimization and the initial NFF optimization, where all NFFs were allowed.

To understand how requiring the model to include some ASFs in OMN diets would affect the environmental impacts, we optimized an OMN diet (OMN.1) with limits on ±80% of the current mean intake (g per day) of ASFs—meats, dairy, eggs, fish/seafood and animal fats. Several previous diet models indicate that large reductions (≥80%) in ASFs are required for minimized environmental impacts23,62,63. Finally, to understand how NFFs may or may not be privileged over ASFs in the model, we ran a sensitivity analysis that allowed both ASFs and NFFs (OMN-NFF), subject to the original nutrition constraints and feasible consumption constraints on the intake of all food items. All NFFs and PBPR alternatives were allowed, and ASFs were allowed to vary from the 5th to the 95th percentiles of their intake per item in the CD.