Diet changes in food futures improve Swedish environmental and health outcomes

Methods overview

We quantified diets based on the four Swedish Food Futures narrative scenarios, Food as Industry, Food as Food Tech, Food as Culture, and Food Forgotten, described in detail in Gordon et al.²⁸. All four scenarios focused on the Swedish Food system and aimed at simultaneously improving the three goals of healthy diets, climate mitigation, and halting biodiversity loss, while at the same time reflecting the specificities of six key aspects of the Swedish food system: trade, governance/institutions, supply chains, production of crops, livestock, and blue foods, cultural values around food, and consumption. For each scenario, we used current intake trends and dietary recommendations for food group intakes to develop four different scenario diets (see Box 1). Reduction in animal source foods needed to reach especially climate targets was imperative in all four scenarios. Using indicators for health, climate impact, cropland use, and biodiversity impacts, we assess how far these consumption changes go in contributing to consumption-based goals for each indicator: climate, healthy/nutrition, and global biodiversity loss.

Strategy for scenario diet development

The goal was to develop future diets that follow the scenario narratives and contribute to fulfilling the multiple goals of health, climate mitigation, and biodiversity conservation. We wanted, to the extent possible, to reach targets with diets that adhere to the current local (national) food culture and food production. The reference diet (current food intake) was used as the point of departure and represents the current national food culture. Further, future diets should be healthy (i.e., follow nutritional guidelines) but at the same time allow for some unhealthy/luxury food items such as coffee and alcoholic beverages. As a last step, we also checked that the food intake in the respective food groups fell within the ranges of those in the EAT-Lancet diet for the diets to get closer to staying within environmental boundaries. However, for the diets to be culturally appropriate in terms of degree of change from the reference (current) diet, we did allow for the diets to diverge from the suggested ranges for some food groups (e.g., dairy, nuts/seeds, and starchy roots) in some of the scenario diets.

The scenario diets were developed in three steps:

1. 1.

   Develop diets that are consistent with the scenario descriptions and follow national food consumption patterns. This was done by:

   1. a.

      creating a reference diet based on current food intake

   2. b.

      changing the reference diet in line with the scenario narratives

2. 2.

   Check that the diets are healthy, following the NNR in terms of food group level consumption

3. 3.

   Check consumption at the food group level in relation to the ranges in the EAT-Lancet diet to increase the likelihood of the diets staying within environmental boundaries.

The three steps for developing the scenario diets are further explained below:

The first step was to build a reference diet from the scenario narratives and current dietary data in Sweden. Scenario narratives and their key differences for diets are described in Gordon et al.²⁸. The narratives were developed in accordance with expert stakeholder input and reflection on future food systems pathways for the Nordic context^13,28. Current intake data for foods at the ingredient level (n = 168 foods) were harmonized by intake in g/day and classified into 21 larger food groups. Source data for different food groups varied^82,83. We first used the data and food categories from FAO Food Balance Sheets (food products that are not consumed were removed). The FAO data were used when it was in line with national statistics and for the food categories where no national statistics were available. For some food categories, such as meat, we use national statistics of raw meat supply⁸². Food waste at the household level was also taken into consideration, and waste was removed from the final consumption data, raw weight⁸⁴. This means that, for the nutritional indicators, we only account for what is eaten (but not what is wasted), whereas for the environmental indicators, we account for total demand for food (i.e., both what is eaten and what is wasted). Full explication of the data source for each food item in terms of current consumption and waste adjustments can be found in SM Table 2.Product Database.

We also collected data on the current import share of each food item as consumed in the Swedish diet from Schwarzmueller & Kasnter and for dairy products SBA^82,85—see SM Table 5.Diets and Import Proportions. Once data on the current intake by food ingredient and group were collected, the current diet in Sweden (2020) served as the foundation for the development of intake and import share quantities for each scenario—see SM Table 6.Export Trends. Intakes and import shares were set for individual food ingredients by checking each scenario narrative for descriptive information on intakes to ensure each diet was consistent with the scenario descriptions and intentions²⁸. Full details of the dietary intakes and imports by food, group, and diet can be found in SI Fig. 6, SM Table 1.Scenario Diets, and key changes made across the scenarios can be found in SM Table 4.Key Swaps.

The second step was to develop the diets to be healthy and follow the recommended food group intakes. We iteratively developed the individual food intakes by comparing scenario diets and their estimated macronutrient intakes, and the food group level consumption with recommendations made for the Nordic context. NNR are developed as the scientific advisory documents for developing national dietary guidelines²⁹. The NNR gives both nutrient and food group guidance for healthy, and according to the latest version, environmentally friendly, intakes for Nordic populations²⁹. We followed their recommended intakes for dietary energy and proportions of energy from fat, protein, and carbohydrates for healthy adults when developing the scenario diets. It is recommended to consume between 2055 and 2811 kcal/day for both female and male adults between 18 and 65 years old, from low to highly active lifestyles. For total fat, daily recommended intakes range from 25 to 40 percent of total energy intake (E%) of the full diet, corresponding to 76–103 g/day of fat intake. Protein intake recommendations are for 10–20 E% of the current diet, totaling between 76–104 g/day of protein. Recommended carbohydrate intake is 45–60 E%, corresponding to between 266 and 363 g/day of carbohydrates. Fiber is recommended to be ≥30 g/day. We also used NNR guidelines for food group intakes for food group totals (Fig. 1). For example, in the Blue Foods group, the NNR gives the overall advice, “It is recommended to consume 300–450 g/week (cooked or ready-to-eat weight), of which at least 200 g/week should be fatty fish. It is recommended to consume fish from sustainably managed fish stocks”²⁹.

The third step was to compare and adjust food group intakes following the general guidelines of the EAT-Lancet diet¹. In addition to being designed for favorable health outcomes, it has been shown that in combination with food waste reductions and technical improvements, the EAT-Lancet Diet can meet global environmental targets for the food system^1,86. Therefore, adjusting the diets so that food intake of different food groups falls within the suggested range in the EAT-Lancet Diet should increase the probability of the diets remaining within the environmental boundaries. However, for the diets to reflect current dietary patterns in Sweden, many of the scenario diets diverged from the suggested EAT-Lancet ranges for several food categories. This includes, for example, dairy and starchy roots. Sweden has a high consumption of dairy products (also a historically high consumption) and high domestic dairy production, which is largely based on the cultivation of grass-clover leys that can be grown in large parts of the country (also in the northern parts where crop cultivation is more challenging). Starchy roots (potatoes) are historically a staple food in Sweden, and therefore, consumption was kept constant throughout all of the scenarios.

A more detailed elucidation of food group level changes from the current diet baseline and justifications for each scenario diet can be found in SM Table 3.Justifications. The final intake amounts and import share on the Swedish market for each food item in each scenario diet can be found in SM Table 1.Scenario Diets.

Food-item-specific data collection

To quantify the nutritional and environmental impact outcomes for the scenario diets, we needed to collect data on nutrient contents and environmental impacts at the individual food item level.

Nutrient content

Nutrient composition data were collected at the individual food ingredient level (e.g., wheat flour in breads, or sugar in all cooked or composite dishes) from the National Food Agency in Sweden (Livsmedelsverket). We use the Livsmedelsverket data as it provides the most up-to-date and comprehensive dietary intake data for Sweden⁸⁷ and as it matches the dietary consumption data for the Swedish adult population^82,83. The Livsmedelsverket data contains validated and unified data for 57 nutrients in units per 100 g of food. We collected data on raw food items to match the intake in g/day of foods at the ingredient level—see SM Table 2.Product Database. Following previous methods for diet modeling, we did not account for any potential differences in nutrient content from cooking and preparation for any food groups and across all scenarios^31,65.

Environmental impacts

Environmental footprints were collected from the SAFAD database³³ at the individual food ingredient level, harmonized with the dietary consumption data. Briefly, the SAFAD database contains data on environmental indicators of food items as consumed in several European countries, reflecting domestic production and imports from main import countries. SAFAD has a selection of products harmonized with the list of European Food Safety Authority FoodEx2 coded food items⁸⁸. Impacts of the whole food item, regardless of if they were produced in Sweden or imported, are accounted for in the SAFAD tool. It was thus also possible to calculate the proportion of impact for each food, food group, and scenario diet allocated from Sweden. To calculate the share of each impact attributed to Sweden in each food item, the inverse of import shares per food ingredient (i.e., amount not imported, or produced in Sweden) was multiplied by the impact of each respective food item per kg, and that value was then divided by the final impact amount per food ingredient. The totals of impact proportions were then summed per food group and scenario diet, respectively.

SAFAD expresses the environmental impact per kg or liter of food product (in edible bone-free weight for meat and fish). SAFAD includes data both for raw unprocessed commodities (e.g., apples, milk), processed basic foods like wheat flour and rapeseed oil, and ready-to-eat meals like pizza and lasagna. We used data for raw unprocessed commodities and processed basic foods in our calculation, and system boundaries include primary production (agricultural processes or fishing), food processing (e.g., milling of flour or pressing of oil), transport, packaging, and any food loss and waste during processing, at the retail, and at the consumer. Cooking impacts were hence not included; however, we assumed that the ingredients would generally require a similar amount (in terms of energy) for cooking across all scenarios. SAFAD uses economic allocation to divide emissions and resource use across products that come out of the same commodity (e.g., oil and cake from rapeseed). Meat and offal were treated as a single product, and emissions from livestock production were allocated based on the total quantity of meat and offal produced. We assume current cropland yields and animal production efficiencies in the Swedish food system and for import countries. Import share proportions per food ingredient were adjusted with the SAFAD tool based on the current Swedish food system, and then modifications were made to reflect the customized import share proportions for each scenario. For the carbon footprint, the SAFAD database contains the impacts disaggregated for the different greenhouse gas generating steps, e.g., energy use in primary production, manure management, and transport. To estimate the carbon footprint assuming a phase out of fossil fuels in the energy system, we only included emissions from energy-related processes.

To align with the targets from Gordon et al.²⁸, three environmental indicators were used in the assessment: carbon footprint (kg CO₂e per kg of food), biodiversity impacts from land use (E/MSY per kg of food), and cropland use (m²*year per kg of food). The carbon footprint measures the climate impact of dietary choices by calculating greenhouse gas emissions in carbon dioxide equivalents (CO₂e) using the Global Warming Potential (GWP) metric. GWP aggregates the effects of gases like carbon dioxide, methane, and nitrous oxide based on their impact over 100 years (GWP100), following international standards. The model uses recent IPCC values: CO₂ (1), biogenic methane (27.0), fossil methane (29.8), and nitrous oxide (273)⁸⁹. Biodiversity impacts from land use are evaluated using the method established by Scherer et al.⁴¹. This approach estimates how land management practices and land use intensities, like cropping or pasture, affect different species compared to natural habitats. Notably, biodiversity loss assessment is challenging and comes with major uncertainties. The cropland use metric calculates the land required (in m²*years) to produce diet components, important for understanding land resource demands. Cropland use is estimated by 1/yield, while livestock products include cropland for feed, reflecting the land intensity of dietary choices.

Scenario diet impacts and boundary quantification

We then calculated climate, cropland use, and biodiversity impacts for the current diet and each scenario diet using the above collated food-item-specific data and intake amounts. We used consumption-based environmental boundaries for the Swedish food system following the targets set in the scenario narratives report²⁸ and scaled for an equal per capita sharing of the global EAT-Lancet planetary boundaries, and their sensitivity interval, for the medium population growth scenario of world population by 2045^1,90—see SM Table 7.Boundaries and Benchmarking for the boundaries and calculations. For the biodiversity boundary, we also scaled the boundary to the number of taxa included in the source data of Chaudhary and Brooks⁹¹. We then calculated the difference between the scenario diet totals and the boundaries (Table 2).

Diet quality and health assessments

Diet quality

Here, we use the targets for Swedish food consumption identified by Jonell⁹². They outline diet quality metrics such as food intake level (e.g., adherence to food-based dietary guidelines, NNR, EAT-Lancet), diet quality scores, and diet diversity. Nutrient adequacy was assessed through balanced energy intake as well as intake levels of individual nutrients in relation to recommended values and nutrient quality scores. Dietary health effects such as deaths and DALYs caused by nutritional deficiencies, overweight and obesity prevalence, and metrics focusing on undernutrition were beyond the scope of this study⁹².

To ensure meeting health and diet quality targets (e.g., adherence to food-based dietary guidelines), we first incorporated food group guidance from the NNR and EAT-Lancet diets into the scenario development—see Methods Strategy for scenario diet development—and iteratively assessed by comparison with recommended food group intake levels. We compare the current diet and each scenario diet with recommended intake levels by food group from the NNR²⁹ and the EAT Diet¹ intakes (SM Table 11.Group Intake).

Diet quality was measured through a comparison of the NRD score⁹³. NRD or Food scores are the most common metric used in analyses of foods, nutrient density, and in subsequent nutritional Life Cycle Assessment, when environmental impacts of foods/diets are compared by their relative nutrient contribution⁹³. The NRD score (Diet specifically here, as we assess whole diets) assesses the nutritional content of diets in relation to the recommended dietary allowances (RDAs). In this study, the NRD_25.4 is composed of 25 qualifying Nutrient Rich (NR) nutrients and 4 nutrients to limit (LIM)⁹³. The NRD_25.4 was calculated by modifying Green et al.’s⁹³ formulas for NR_diet and LIM_diet; follows First, a total of 25 positive nutrients were included in the NR_diet as the data availability allowed for comparison with given recommendations from the NNR, given in Eq. (1):

Where n = the total number of nutrients with positive health association, i = value of individual nutrients with positive health association—normalized by comparison with the total energy in each diet (calories_j), j = scenario diet, kcal = total energy intake per day (kcal), RDA = Recommended Daily Allowance, and AI = Adequate Intake. Capping was used to prevent the disproportionate impact of overconsumption of any one positive nutrient on the NR_diet total; so, if NR_diet was >1, the value was capped at 1. NR_diet was calculated over 25 positive nutrients in the meals (see SM Table 12.NNR_RDAs_db for the list of daily recommended values of each nutrient).

Secondly, 4 nutrients to limit were included in the LIM_diet as the data availability allowed for comparison with given recommendations from the NNR, given in Eq. (2):

Where n = the total number of nutrients with negative health association, i = nutrients with negative health association, j = scenario diet, MRV = Maximal Reference Values. This was calculated over 4 nutrients to limit in the meals—sodium, and total polyunsaturated, monounsaturated, and saturated fatty acids, selected as they have set upper limit of recommendations in the NNR.

Lastly, the difference of NR_diet  and LIM_diet for each scenario diet was calculated , given in Eq. (3):

Several food-based dietary guidelines include recommendations for dietary diversity, due to both health and environmental reasons⁹⁴. The NNR encourage consumption of a variety of different types of legumes, fruits, vegetables, and fibrous foods, among others²⁹. To assess the relative dietary diversity of the scenario diets we used the similarity-sensitive diversity metric Rao’s Quadratic Entropy (HR)³⁶. Originally, HR is a measure of biodiversity and accounts for the following key aspects of biodiversity: abundance, richness, and similarity of functional role⁹⁵. Here, we assessed only relative dietary diversity across the dies, not absolute dietary diversity, since maximizing HR would theoretically be equal richness across all dietary items, with high intakes of all foods, and such intake would not necessarily equate to a healthy dietary composition. In the case of dietary diversity, HR is a measure of functional similarity when two food items are selected at random, here taking into consideration the (dis)similarity of nutritional content of all of the foods in the diets, their abundance and richness following Eq. (4):

Where S is the total number of food items, p_i and p_j are the relative abundances of food items i and j, respectively, and d_ij the dissimilarity between foods i and j measured by differences in nutritional composition—calculated from the 57 nutrients included in the nutrient content data per food item—via the Euclidean distance measure⁹⁵.

Nutrient adequacy

Intake of micro- and macronutrients was assessed in relation to their respective recommendations in the NNR (2023)²⁹. We assessed the percent of intake for each of the 25 positive (NR) and 4 nutrients to limit (LIM) as in the NRD_25.4 against their recommendation in the NNR for each of the current and scenario diets separately—see the full list of daily recommended values of each nutrient in SM Table 12.NNR_RDAs_db.

Protein adequacy does not only depend on sufficient intake of total protein, but also on the composition of amino acids and their digestibility in the human digestive system. To assess these protein quality factors, we used the method recommended by the FAO (2013)⁹⁶, namely calculating the DIAAS. Briefly, DIAAS measures a food’s content of the most limiting indispensable amino acid, adjusted for digestibility in the small intestine, relative to nutritional needs (see FAO, 2013 for details⁹⁶). A DIAAS score above 1 indicates sufficient supply of all indispensable amino acids for a person consuming the minimum safe intake of total protein as recommended by WHO⁹⁷. Conversely, a DIAAS score below 1 indicates a risk of insufficient supply of at least one indispensable amino acid, at the minimum safe protein intake. Note that a DIAAS score below 1 (i.e., low protein quality) can be compensated by consuming more total protein than the minimum safe intake, which for adults is estimated at 0.83 g protein per kg body weight per day (WHO)⁹⁷, or about 58 g protein per capita per day for a 70 kg adult.

To calculate DIAAS for the diets, we first collated data on the content of digestible indispensable amino acids (DIAA) from the sources referenced by Adhikari et al.⁹⁸. We calculated the average DIAA content per unit total protein (expressed as grams of digestible amino acid per kg protein) for each of the food categories listed by Adhikari et al.⁹⁸: beef, cereals, dairy products, eggs, legumes, nuts, and pork. Data on DIAA content of eggs was complemented using the results reported by Woyengo et al.⁹⁹. We then calculated the total DIAA intake by multiplying the protein intake from each component of the diets (wheat, peas, beans, milk, cheese, etc.) by the DIAA content of the corresponding food category. Due to a lack of DIAA content data for fish, poultry meat, and game meat, we assumed the average DIAA content of beef and pork. For protein concentrates, protein isolates, and novel foods, we used the DIAA content of the most similar corresponding whole food. The whole-diet DIAA intake was then normalized by the amino acid pattern for adults⁹⁶, and DIAAS was calculated as the lowest of the individual amino acids (see DIAAS by food and diet in SM Table 16.Protein Quality).

Waste reduction

As an additional “what-if” analysis, we examined how the outcomes of the scenario diets would differ if, across the entire food system, waste and loss were reduced by 50% from current levels. In addition to the main analysis of dietary changes, which differ in each scenario, we adjusted each food item in each scenario diet to have 50% less waste and loss from each of the production, retail, and consumer food life cycle stages using the SAFAD data, as outlined above. These methods follow previous food systems models for respecting planetary boundies with a halving of food waste¹⁰ as well as the United Nations Sustainable Development Goal (SDG) 12.5 to “substantially reduce waste generation”¹⁰⁰. The import share proportions and all food amounts in this sensitivity analysis of waste reduction remained the same as in the main analysis. See Supplementary Discussion for more details on waste reduction in the scenarios.

Reporting summary

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