Recent work has demonstrated the urgency of moving to a more plant-based diet combined with reductions in food loss and waste in order to keep within planetary boundaries [1
]. However, the concept of healthy sustainable diets includes more than just the environment and nutrition and it has been postulated by the Food and Agriculture Organization of the United Nations (FAO) and others that they should also be both affordable and culturally acceptable [3
]. Mertens and co-authors proposed the acronym SHARP to encompass environmental Sustainability, Health, Affordability, Reliability and Preference from the consumer [4
Food production and distribution has a significant contribution to global environmental impacts. For example, the annual greenhouse gas emissions (GHGE) arising from agricultural production have been estimated to range from 7.3 to 12.7 Gt CO2
-equivalent, or 14–24% of total global emissions [5
]. Additional environmental impacts in the form of energy use and greenhouse gas emissions arise from transport, processing and preparation of food. In addition to the harmful emissions to environment, food production is associated with the use of limited resources. For example, there are few opportunities to increase the land area available for agricultural production, despite the fact that the food requirement of growing global population is continuously increasing [6
]. Although food production is a necessary activity, it is possible to mitigate its environmental consequences, for example, by reducing its greenhouse gas emission intensity [7
]. Improving the efficiency of the production chain and directing consumption towards more environmentally friendly ingredients and production techniques provides an opportunity to considerably reduce the global environmental impacts associated with the food production chain [8
Mathematical modelling has already been used to assess the environmental sustainability of different diets. For example, Macdiarmid et al. [15
] applied modelling to generate combinations of foods which could form menus that are sustainable in terms of the environment, health, affordability and acceptability. However, the limitation of this work was that it used a database of only 82 food groups and did not account for GHGE beyond the primary production stage. Although such GHGE are more difficult to assess, estimates are available and can be incorporated. Saxe [16
] compared a New Nordic diet (which meets nutrition and health guidelines) with the Average Danish diet and estimated the environmental impact in a terms of other environmental impacts such as respiratory organics, land use and global warming potential (i.e., greenhouse gas emissions). The study took different scenarios into account in terms of transport and use of organically produced food.
In recent studies (e.g., [17
]), more detailed calculation methods have been applied to estimate the GHGE of actual and hypothetical diets. However, such analyses are still based on rather limited data on the emissions associated with specific food items, and the calculations are often based on rather generalized “food groups”. Although such analyses can provide reliable overviews of the environmental consequences of dietary choices, there is also a need for exploring the effects of more detailed small-scale changes in dietary patterns. To achieve this, novel modelling approaches that can better utilize the available data are needed.
In order to achieve the shift towards healthy and environmental diets, such diets need to be affordable to the consumer. There is a general assumption and some evidence [25
] that a higher quality diet costs more than a “normal” diet, but this is based on average current diets and the studies have not always fully explored possible examples where sustainable choices have resulted in a lower cost diet of adequate nutritional quality. The New Nordic diet was shown to be less costly than the average Danish diet [16
], thus demonstrating that it is possible to have a nutritionally and environmentally sustainable diet at a lower cost than the average western diet.
In general, much of the existing research evidence on the sustainability of current diets, although useful, has been based on population data, and generalisations and simplifications of the environmental impacts of standard food commodities and a limited number of product groups, as stated above [15
]. This is mainly analysed using publicly available “carbon footprint” information with only limited or no traceability, using data covering only a limited number of food items [18
] and excluding parts of the food supply chain. This has been productive in indicating the type of diets needed for both health and sustainability but limited to producing idealistic diets designed by statistical models which may not be culturally acceptable or affordable [24
]. Masset and co-workers [28
], using data from the French national dietary survey (INCA2), selected real individual diets exhibiting less than the median GHGE and higher than the median for their diet quality measure. However, although they measured cost as an outcome, they did not incorporate it when selecting their “More Sustainable” category.
Our aim was to develop an improved novel modelling framework to enable quantification of real household level food choices for health, affordability and environmental sustainability; while systematically accounting for the entirety of the food production, processing and supply chain, and using the actual disaggregated composition of food items rather than a limited number of food categories as a basis our analysis. Thus, our main objective was to develop a systematic, traceable, and comprehensive Life Cycle Assessment (LCA) framework to quantify the various dimensions of environmental sustainability of the main UK food items, taking into account the entirety of the food production, processing and supply chain. We also wanted to demonstrate how such methods can be used to assess the proportion and characteristics of households and/or individuals who purchase real diets (rather than idealistic diets that are sometimes produced by linear programming) that could be considered sustainable, healthy and affordable, and thus discover if such diets could be acceptable within the population. Hence, a further objective was to integrate LCA with measures of diet quality and cost of household food purchases and apply the framework to analyse a large-scale UK food purchase dataset in terms of environmental sustainability, healthiness and cost of household and individual diets. By doing this, we wanted to provide a method that could also be able to improve the evidence-based approach to assessing interventions and formulating sustainable dietary goals to improve the sustainability of household food consumption.
We showed that only a very small proportion of the UK population purchase a diet that is likely to be compatible with sustaining their own health or that of the planet. However, this proportion represented a range of household types and incomes and food was not restricted solely to items that would be considered healthy, with alcoholic beverages, cakes, sweets and soft drink being purchased in relatively small quantities to add variety to the diet.
Food and drink purchases and their cost over a 14-day period by representative households in the UK were combined with data on GHGE and land use to provide a workable framework from which to assess the sustainability of food purchase patterns. This was accomplished using a systematic methodology which could be extended to other scenarios and dietary data.
Several modelling frameworks aiming to quantify the environmental consequences of dietary choices have been presented in the literature. However, we believe that the novel approach to dietary LCA as presented in this study has several advantages compared to most of the earlier methods modelling the environmental sustainability of food [15
]. The current model was specifically developed for use in connection with food survey data, and the systematic approach and flexibility makes its application possible in a range of studies using such datasets, and also provides a tool that can be used in scenario analysis exploring alternative diets.
A practical advantage of the framework developed here is related to the detailed disaggregation of the purchased food items. In previous studies [15
], the environmental impacts of foods (mainly GHGE only) were usually based on “food groups”, not on individual products. Relying on such relatively coarse categories can in the worst case led to insufficient or even misleading conclusions. In reality, the composition of a single food group can be highly variable. As a simplified example, foods classified as “meat products” can include items such as whole meat, meat pies (containing mainly cereals) and meat soups (containing mainly water). Therefore, shifting dietary habits between such products can have a high impact on the GHGE associated with the diet, yet it can be observed only with detailed disaggregation of the food items. In general, the disaggregation approach allows for identification of the effects of much smaller scale changes in diets than major dietary shifts, e.g., from meat-based diet to vegetarian diet [8
]. However, it should be noted that the LCFS dataset applied in this study included some food categories that did not allow a detailed disaggregation (for example “complete meat-based ready meals”). However, a further advantage of our method is that such categories can be handled in systematic way, based on the weighted average of the actually consumed items belonging to that category (see Supplementary Material S1 and Table S2
). Therefore, we believe that our framework can handle both very detailed and less accurately specified food categories without bringing any bias to the results. Furthermore, the tool can handle unlimited combinations of raw materials in food items. Therefore, any number of new foods can be included in detailed calculations, as far as their “recipes” are known.
In the current study, the GHGE related to processing and cooking of different food items was based on rather simplified assumptions and generalizations [39
]. However, in future studies, the modelling framework can be utilised with much more item-specific processing data, if such data is available. In general, as the framework includes the whole food chain, it can be applied in future studies to explore scenarios with changes in different part of the chain, for example, raw materials produced either domestically or imported, organic vs. non-organic production, processed vs. non-processed food, the use of energy-efficient cooking methods, etc. In addition, since the framework allows a breakdown of the different components of the food chain, it can be used as an analytical tool when comparing existing diets; if there are differences between the GHG emissions associated with diets, the main sources of the differences can be identified, indicating the “hotspots” within the food chain.
The study used purchase data from a large representative sample of the UK population. It has been suggested that purchase data is less subject to bias than individual food diaries [31
], but as it combines data from purchase diaries within households, it is not possible to see the individual diets of household members or adjust for household composition or ages within the household. In addition, it is not possible to determine how much, if any, food and drink may have been purchased for friends and family outside the household, or whether the purchases were consumed within the 2-week recording period. However, a method for checking that the food purchased was an adequate amount for the household was provided and excluded about 20% cases where it was unlikely. Wastage was accounted for in calculation of the DQI using average figures from the Waste and Resource Action Programme (WRAP) [42
], but not in the calculation of GHGE and Land use as these will be appropriate for the actual food purchased.
The DQI used to determine the nutritional quality of the diet was constructed using widely accepted dietary guideline cut-off points at the time of the surveys, (for example those from the World Health Organization [50
] and others detailed in Supplementary Table S2
) and although there have been recent changes to recommendations for added sugars and fibre (with changes to definitions and cut-offs [51
]), it was not possible or considered appropriate to compare with guidelines that were constructed after the dates of the actual surveys.
Currently, there is no “recommendation” for the ideal GHGE from food for individuals to limit climate change, but this is also hampered by the fact that reported GHGE figures are variable and not standardised, as pointed out by Clune and co-workers [52
]. However, within our study, the figures were comparable with each other as a result of the systematic methods used. Results for different food items were fully consistent, traceable and transparent, and go beyond the “farm gate”. The method used was flexible (for example could be used to compare different scenarios) and could be applied to any food item (with known composition).
Our research confirms previous work that the average UK diet does not meet dietary guidelines [31
]. However, contrary to previous research, it would appear that it is possible to purchase a diet that is healthy and sustainable at a relative low cost. This would appear to contradict work that shows that the cost of a healthy sustainable diet is more expensive than a conventional diet [25
]. It should be noted that a relatively low-cost high-quality sustainable diet was purchased by a very small proportion of survey participants but this does show that there is potential for a carefully chosen diet to be both affordable, sustainable and healthy.
Using a systematic methodology in which purchased food items were disaggregated into their components with traceable environmental impact data, we found that there were households who were purchasing a diet that was both low cost and had a lower environmental impact, combined with a higher quality in terms of nutrition. Using the higher criteria for DQI the diet of the 100 more sustainable households was not unlike that recently proposed by EAT-Lancet Commission [2
]. The purchase patterns of the 100 households were not uniform and some purchased no red or processed meat or other types of animal protein. There has been criticism of the EAT-Lancet proposal on the grounds that it is not feasible nor practical [55
], nor fits within the UK situation [56
] but there seems to be no appreciation of the flexibility within the plan and that it represents an average. Some individuals and households choose to eat more of their protein from animal sources and others purely from plants. What is needed is a population shift towards lower meat and dairy consumption and higher consumption of wholegrains and fruit and vegetables—the point that nutritionists in public health have been making for over two decades.
There are several opportunities for further improvement of the methodology developed in this study. The results could be further enhanced with more detailed information on food purchased (including the exact origins of food items, their processing and cooking methods, etc.) and more accurate and specific information on GHGE, once available. This would allow a more detailed comparison of both actual diets and hypothetical diets in scenario analyses. Furthermore, although the current analysis on food cost did not take into account any environmental costs, such costs can be included in new versions of the framework. As the GHGE of different food items are already part of the model, it would be relatively straightforward to include carbon prices in the calculations to expand the analysis of the monetary effects of dietary changes. The carbon costs would also automatically handle both direct and indirect emissions arising from land use changes, due to the top-down methodology applied for land use changes applied in the modelling framework [35
]. This would provide a link between the land use estimates, GHGE and food costs, all of which are all already included in the current framework.