2.2 Tier 1 approach: single-issue functional units
In a study of Breton pâté production, Teixeira et al. (
2013) compared the carbon footprints of nine different production systems under mass-based (100 g of product), energy-based (kcal) and nutrition-based (protein) functional units. The systems were differentiated by farming practice (conventional, organic,
label rouge, a French Governmental certification based on organoleptic properties determined by sensory panels, and
Bleu-Blanc-Coeur, an initiative which promotes omega-3 fatty acid content through feeding regimes) and packaging types (tin can, aluminium can or glass jar). All pâté was produced from pigmeat. The system boundary was cradle-to-grave and included waste management at the end of the life cycle. The authors found that, on a mass basis, organic pâtés had higher carbon footprints than the other systems whilst conventional,
label rouge and
bleu-blanc-coeur products all had similar emission intensities. When considering the nutritional content, however, relative rankings were affected depending on which functional unit was adopted. For instance, when considering the carbon footprint in terms of g CO
2-eq/g protein, the organic system performed marginally better than one of the conventional systems, due to a higher protein content driven primarily by the cuts of meat used in the pâté. On the other hand, energy-based carbon footprints considered as g CO
2-eq/kcal suggested that the organic system once again fared least favourably, whilst relative rankings amongst the conventional and
bleu-blanc-coeur systems varied to a certain degree, depending on the calorific content of individual pâtés. The authors concluded by stressing the importance of functional unit selection in comparisons of food products which generates a considerable effect on research findings.
Tyszler et al. (
2014) proposed a framework to maximise information provided by single-issue functional units by utilising linear programming to create scenarios that replace individual food products with nutritionally equivalent alternatives. Using two weekly diet case studies as examples, the authors first replaced apples (
Malus pumila) in the fruit component of a baseline diet—consisting of 3.6 servings of apples, 1.2 servings of oranges (
Citrus maxima × Citrus reticulata), 1.2 servings of kiwis (
Actinidia deliciosa) and 0.9 servings of strawberries (
Fragaria × ananassa)—with an equivalent portion of oranges. As this change resulted in 4.8 servings of oranges and thus an excess intake of vitamin C, the authors used a constrained linear optimisation algorithm and removed the equivalent portions of kiwis and strawberries from the weekly diet. This substitution resulted in slightly higher carbon footprints compared with the baseline diet, energy requirements and land use. In the second case study, Tyszler et al. (
2014) replaced 2.2 servings of chicken (
Galus galus domesticus) and 0.8 servings of red meat with 3 servings of vegetarian burgers. Nutritionally, this replacement led to a deficiency in lysine, methionine and selenium. Under the constraint that livestock meat was excluded from the diet, the model then added 0.1 serving of salmon (
Salmo salar) and 0.2 servings of cod (
Gadus morhua) to meet the essential amino acids and selenium requirements. This time, the substitutions resulted in markedly lower carbon footprints, energy use and land use than the original diet. To circumvent the requirement for linear programming knowledge, the authors also developed a software package to allow other LCA researchers to perform similar studies. Despite the benefits of the approach, the authors point out that data requirements, in terms of nutritional quality and dietary habits, are highly intensive, which may restrict wider applicability.
In a comparison between conventional ultra-high temperature (UHT) milk and a nutritionally enhanced UHT milk in Spain, Roibás et al. (
2016) used the LCA framework to consider carbon and water footprints of both production methods in combination with health effects. The authors used a cradle-to-gate system boundary with a baseline functional unit set as 1 l of packaged UHT milk leaving the dairy factory; nutritional values of the final products were calculated externally to the LCA framework. The enhanced UHT milk was produced through cow-feed supplementation of linseed (
Linum usitatissimum), naturally high in omega-3 α-linolenic acid. As a result, it contained 1% less saturated fat but 82% more unsaturated fats, with the omega-6/omega-3 ratio dropping by 58% compared with the conventional UHT milk. The enhanced milk also had around three times more selenium than the conventional milk. Based on epidemiological studies, the authors asserted that improved conjugated linoleic acid, lower ratios of omega-6/omega-3 and higher levels of selenium could have benefits to human health in the form of reduced instances of atherosclerosis, certain types of cancer and contributions to normal thyroid and immune system functions. Regarding the environmental footprints, although the production of fodder in the enhanced system had higher greenhouse gas (GHG) emissions than the conventional fodder (mainly maize (
Zea mays) and soybean (
Glycine max) meal), methane emissions from manure management and enteric fermentation were lower in the enhanced system, making the total carbon footprint lower also. The water footprint, on the other hand, was marginally higher (2%) than the conventional system, but the authors concluded that when farm-level variation was included, no significant differences were drawn.
Motivated by differences in product-level supply of essential amino acids (EAA), Tessari et al. (
2016) compared land use and carbon footprints of 15 foods (including beans (
Fabaceae), cauliflower (
Brassica oleracea), beef, fish, maize, milk, peas (
Pisum sativum), potato (
Solanum tuberosum), quinoa (
Chenopodium quinoa) and rice (
Oryza sativa)) under three different functional units. The baseline functional unit was arbitrarily chosen as 100 g edible fraction of each product. The second and third functional units, on the other hand, were determined by the mass of the product required to provide (1) 13 g of total EAA, irrespective of deficiencies in certain amino acids, and (2) recommended quantities of all individual EAA, regardless of oversupply of certain amino acids, respectively, for a 70 kg male. Data on land use and carbon footprints were sourced from previously published studies whilst nutritional composition was obtained from Italian national nutrient tables. Under the first functional unit of 100 g of edible product, meat, fish and peas had considerably higher demand for land use than the other foods whilst beef and fish tended to have the highest carbon footprints. Switching to EAA-specific functional units, however, resulted in marked rank reversals. For instance, beans, peas, potatoes and rice required substantially more land to provide a human with the recommended intake of all EAA in comparison with 100 g of edible product. Regarding carbon footprints, beef, cauliflower and rice demonstrated the greatest changes across functional units, with EAA-based estimation resulting in markedly lower GHG emissions relative to the 12 other products. The authors point out that, when detailed protein requirements are accounted for, environmental gaps between livestock products and vegetables can be notably reduced, reinforcing the argument that mass-based comparisons are often an inappropriate use of functional units.
Schaubroeck et al. (
2018) conducted a sustainability scoring exercise for canteen meals offered at Ghent University, Belgium, using a functional unit of one meal regardless of its energy or nutritional content. Sustainability was assessed according to: ecological scoring, nutritional scoring, sustainability of suppliers and other information considered important by key stakeholders. Ecological scores were largely determined through LCA studies of the composite ingredients in each meal, or studies of entire meals themselves, and simplified to single score comparisons across meals based on a range of endpoint ecological footprint impact categories. Nutritional scores were based on meals’ provision of energy, protein, fat, saturated fat, carbohydrates, sugar and salt. These scores were expressed by integers based on the number of nutritional criteria met. Suppliers’ sustainability was assessed on readily available information such as adoption of water recycling. Lastly, additional information collected was determined based on a qualitative case study with producers and consumers, who identified issues such as the inclusion of genetically modified organisms in meals or consideration of animal welfare. Based on the information gathered from each meal, the authors suggested that meal providers could provide colour-coded indicators on posters or menus to indicate the level of sustainability under each of the four themes addressed. Schaubroeck et al. (
2018) also highlighted the limited sustainability information provided by LCA studies due to the sole focus on environmental issues, which are not the primary aspect of decision-making for some consumers.
2.3 Tier 2 approach: multiple nutrients within single functional units
Doran-Browne et al. (
2015) applied the concept of nutrient density scores (NDS) in the assessment of GHG emissions attributable to agricultural products typically found in southeast Australian diets. Four functional units were considered: mass of product (t); mass of protein (t); energy content (GJ) and NDS. The NDS of each product was determined according to the Nutrient Rich Food model (NRF9.3) originally developed by Fulgoni et al. (
2009), whereby higher contents of nine encouraged nutrients (protein, fibre, vitamins A, C and E, calcium, iron, magnesium and potassium) are associated with a higher score, and three discouraged nutrients (saturated fat, sodium and added sugar) with a lower score. The quantity of each nutrient present in a product was first divided by its recommended daily intake (RDI), or daily allowances (RDA) for discouraged nutrients, to obtain the percentage of RDI satisfied by the product, and subsequently converted to a weighted score according to the product’s relative importance as measured by energy value. The NDS for each product was then derived as the difference between the sum of these weighted scores associated with “positive” nutrients and the sum of the similar scores associated with “negative” nutrients. The food products assessed were beef (lean and untrimmed), lamb (lean and untrimmed), regular milk, reduced fat milk, wheat (
Triticum aestivum) flour and canola (
Brassica napus) oil. When using the standard mass metric (t CO
2-eq/t product), the authors found that wheat flour generated the lowest GHG emissions whilst milk and canola oil had similar levels of impacts. Meat products had the highest impacts, with the lean cuts having a higher CO
2-eq value than the untrimmed cuts. However, when the novel metric (t CO
2-eq/NDS) was applied, the lean cuts had considerably lower environmental impacts than the untrimmed cuts, and the gap between non-meat products and lean meat was substantially narrowed. Similarly, both regular and low-fat milks were found to have considerably lower impacts than canola oil, whereas wheat flour still had the lowest impacts. Although not covering the whole life cycle of products and stopping short of distinguishing between different compounds within each nutrient group, for example between polyunsaturated (PUFA) and monounsaturated fatty acids (MUFA), and amongst different EAA, the study proposes a useful technique for comparing different food groups based on their nutritional value.
Using industrial data and national nutritional statistics from France, Drewnowski et al. (
2015) explored interlinkages between carbon footprints and nutrient densities for 661 different foods and beverages. Carbon footprints were calculated under 100 g and 100 kcal functional units, with their relationship with NDS subsequently analysed using linear regression. Thirty-four food categories considered by the authors were split between five common food groups: meat and meat products, milk and dairy products, frozen and processed fruit and vegetables, cereals and sweets. Two density scores were calculated, with one accounting for six encouraged nutrients (ND-6; protein, potassium, magnesium, calcium, phosphorus and vitamin D) and the other accounting for 15 encouraged nutrients (ND-15: ND-6 nutrients plus fibre, vitamins A, C and E, iron, thiamine (vitamin B1), riboflavin (vitamin B2), niacin (vitamin B3) and folate (vitamin B9)). Grains and sweets were found to have the lowest carbon footprints of the major food groups regardless of functional unit but were also found to have low density scores. When reported per gramme, meat and meat products tended to have the highest carbon footprints across all food groups; when reported on an energy basis, however, processed and frozen fruit and vegetables leapfrogged meat and meat products and had the highest carbon footprints due to their low energy densities. The authors pointed out that nutrient-dense foods such as meat and dairy products typically have high carbon footprints, with the reverse also being true (foods with low nutrient density tend to have lower carbon footprints). Drewnowski et al. (
2015) also demonstrated the complexity of choosing suitable functional units in comparative LCA of food products with marked reversals in relative rankings in system-wise environmental performance, and posited that to formulate a truly sustainable diet, simply focusing on one metric, carbon footprints in this instance, is not an effective assessment method.
In an attempt to identify a functional unit suitable for capturing a wider measurement of the sustainability of food products, Masset et al. (
2015) examined environmental footprints of foods and drinks representative of a typical French diet under a number of different impact categories. The authors defined sustainable food products as low emitting, affordable and of high nutrient quality, determined collectively by a final score that takes the value: 0, 1, 2 or 3. Nutritional quality of food products was determined in one scenario using the French SAIN, LIM method, whereby five nutrients (protein, fibre, calcium, vitamin C and iron) are encouraged whilst three (saturated fat, added sugar and sodium) are discouraged. If a food product obtained more than 97% of its energy from fat (as is the case for nuts and oils), then vitamin E, MUFA and α-linolenic acid contents were also accounted for in the encouraged nutrient profile. A product’s overall sustainability score was then derived by comparing its performances in the aforementioned three areas against their respective median values; a product received a point if its GHG emissions and price were lower than the median, and if its nutritional score was higher than the median. Masset et al. (
2015) argued that mass and energy-based functional units are generally unhelpful in determining sustainable products. In particular, they demonstrated how functional unit manipulation can affect relative rankings across food products, with those of monogastric meat products and fruits/vegetables easily reversed between energy-based and nutrition-based computations of GHG emissions.
Building upon the NRF9.3 framework described above, Saarinen et al. (
2017) developed a novel nutrient index to specifically compare the overall quality of protein-rich foods. The authors used multiple functional units such as individual nutrients applied per mass of product (e.g. CO
2-eq/g calcium or /μg cobalamin (vitamin B12)), as well as a novel nutrient score specifically designed for protein-rich foods, which included MUFA, PUFA and vitamins B2 and B9, but excluded nutrients that are not typically provided in abundance by these food groups (e.g. magnesium and potassium). Global warming potential (GWP) was then estimated under both mass-based and nutritional score-based denominators. Nutrient contents of individual products as well as recommended intake values were sourced from public databases in Finland, whilst background LCA data were gathered through published literature. Twenty-nine food products ranging from cereals and pulses to dairy products, meat and seafood were considered. Saarinen et al. (
2017) demonstrated that the choice of functional unit can affect interpretation of results considerably. For example, beef had the largest GWP on a mass-based functional unit (100 g of product) but overtaken by cheese and lamb as the most burdensome food group when the functional unit was changed to the nutrient content included in 100 g of product. In general, animal-based products had higher environmental impacts than cereals and pulses regardless of the functional unit, although the authors did not consider contents of some important micronutrients such as vitamin B12 or account for bioavailability of nutrients (e.g. haem iron) when consumed in different forms of food.
In a study focusing on replacing refined wheat flour used to produce typically cereal-based products with Canadian yellow pea (
Lathyrus aphaca), Chaudhary et al. (
2018) examined the effects on nutritional and environmental performance. Three common products were considered: pan bread, breakfast cereal and pasta; and three functional units were considered: kg food, single serving and nutrient density in a single serving. Nutritional content was assessed according to a nutrient balance concept which consists of three metrics similar to those proposed by Fulgoni et al. (
2009): qualifying nutrients; disqualifying nutrients and the nutrient balance score (NBS). The qualifying nutrients were made up of 27 micro- and macronutrients including a wide range of minerals and vitamins as well as water, fibre, protein, α-linolenic acid and linoleic acid. The disqualifying nutrients were sugar, sodium, total fat, saturated fat and cholesterol. The three products were assessed according to Canadian recommended intake values (for qualifying nutrients) and maximum intake values (for disqualifying nutrients) under both typical ingredients and systems which included yellow pea. Carbon footprints were calculated where possible using Canadian-specific data; where this was impossible, EU-specific data were sourced from inventory databases. Nutritional values and mass-based carbon footprints were combined by dividing the NBS by the carbon footprint of each product related to typical serving sizes (75 g bread, 30 g breakfast cereal and 85 g pasta). The authors found that replacing refined wheat flour with yellow pea flour improved the NBS by 11, 70 and 18% for bread, breakfast cereals and pasta, respectively, whilst simultaneously reducing carbon footprints of 1 kg of each product by 4, 11 and 13%. Chaudhary et al. (
2018) concluded that pulses could play a pivotal role in improving human nutrition and reducing the food sector’s carbon footprint in line with the United Nations’ Sustainable Development Goals.
Modifying the framework devised by Saarinen et al. (
2017), McAuliffe et al. (
2018a) explored the effect of adopting nutritional functional units on relative rankings amongst the most common meats consumed in the UK (beef, chicken, lamb and pork). Using previously published carbon footprint and carcass yield data, the authors first compared carbon footprints of products based on mass functional units (100 g of deboned meat), and then examined effects of adopting nutritional functional units based on omega-3 fatty acid content and NDS. System boundaries were set as cradle to farm gate, and where appropriate, liveweight was converted to carcase weight and subsequently to meat ready for cooking. Secondary processing (e.g. pork made into sausages) was not considered. Data from nutritional tables were used to develop UK-specific NDS based on scores produced by Saarinen et al. (
2017) as well as a novel NDS which included zinc, selenium and vitamin B12. The authors reported that ruminant systems tended to have higher environmental impacts than their monogastric counterparts when compared on a mass basis. However, when the nutritional composition of meats was included in the carbon footprint analysis, emissions attributable to beef production, particularly from concentrate-finished systems, were found to be comparable or lower than pig and chicken production depending on management practices. The authors concluded that nutrient density of products provides important information to improve environmental performances of agri-food systems; equally, they observed that eating smaller portions of higher quality products may improve human nutrition whilst simultaneously reducing the global carbon footprint through potentially decreased food production.
Xu et al. (
2018) analysed a wide range of common local sources of carbohydrates in China. The authors used several functional units to examine differences between products; these consisted of mass, energy, protein, carbohydrate and two nutrient profile scores. From an environmental perspective, carbon footprints were calculated using China-specific data where possible. Nineteen products were evaluated in total, grouped into five categories of rice, wheat, maize, potato and pulses. The two nutrient profiles consisted of 11 (NU11: protein, carbohydrate, fibre, vitamins B1, B2 and B3, calcium, iron, zinc, magnesium and selenium) and 21 (NU21: all of the above plus fat, vitamins A, C and E, potassium, sodium, manganese, copper, phosphorus and cholesterol) nutrients, respectively, and were calculated similarly to formulae set out in Fulgoni et al. (
2009). Regardless of the functional unit, rice generated the highest carbon footprint due to large methane emissions resulting from anaerobic conditions. However, there were notable reversals to relative rankings amongst other products. For example, when compared on a mass basis, potatoes tended to have the lowest carbon footprints, but when the functional unit was switched to protein and NU11, the carbon footprint of potatoes became larger than maize kernel and sorghum (
Sorghum bicolor) flour, respectively. The authors concluded that food comparisons should always account for the nutritional composition of individual products and pointed out the opportunity for China to considerably reduce its agricultural emissions by switching from rice-based diets to those supported by other sources of carbohydrates.
Hallström et al. (
2019) investigated seafood products that should be recommended to consumers based on a combined assessment of nutritional qualities and environmental performances. Acknowledging the sensitivity of NDS to its mathematical formulation, the authors calculated seven different NDS for 37 Nordic commodities. Following a detailed comparison of NDS, Hallström et al. (
2019) concluded that the scoring method most appropriate for their goal was the 24-nutrient system, which incorporate the densities of 22 qualifying nutrients and two disqualifying nutrients. A mass-based reference flow (100 g edible product) was deemed sufficient as the base (denominator) of NDS calculation, as energy and water content of seafood products did not vary enough to warrant the inclusion in the formula. Despite the general perception that all fresh seafood is nutritious and climate-friendly, 37 products were found to differ considerably both in terms of environmental and nutritional performance; certain species such as shrimp (
Pandalus borealis) and plaice (
Pleuronectes platessa) scored significantly less overall than pelagic species such as herring (
Clupea harengers) and mackerel (
Scomber scombrus). The authors note, however, that further work is required to consider other environmental impacts of seafood production systems, for example toxins arising from water-based pollutants such as micro-plastics and mercury.
2.4 Tier 3 approach: commodity-level scores incorporated into diet-level analysis
Stylianou et al. (
2016) developed the Combined Nutritional and Environmental Life Cycle Assessment (CONE-LCA) method to empirically apply the conceptual framework devised by Heller et al. (
2013). CONE-LCA is a hybrid approach that utilises traditional midpoint LCA modelling but adjusts output values to predict endpoint metrics based on the nutritional quality of food products. The ultimate output of this novel approach is the impact of diet change not only on midpoint environmental measures (GWP and respiratory effects via particulate matter in this particular case) but also the human health effects of these impact categories represented as disability-adjusted life years (DALY). The authors carried out a case study whereby an extra serving of milk (244 g), acting as a de facto functional unit, was added to three dietary scenarios: no changes to the rest of the diet; removal of other food products with an equal caloric value (119 kcal), and removal of an equal caloric quantity of sugar sweetened beverages. The authors used epidemiological data to assess milk’s effects on human health, both positive and negative, as expressed by DALY. Whilst adding a serving of milk to the diet increased both GWP and respiratory inorganics at the midpoint, milk consumption was found to be beneficial for long-term health, as reduced risk of colorectal cancer and stroke outweighed increased risk in prostate cancer in all scenarios. Whilst Stylianou et al. (
2016) acknowledge that uncertainty associated with endpoint impact assessments is considerable, their framework is a significant contribution to the methodological advancement of nutrition-based LCA.
Sonesson et al. (
2017) developed a new index to account for differences in protein quality between food products under the LCA framework. They established a new functional unit, protein quality index-adjusted mass, for each food product studied (bread, chicken breast, minced pork, minced beef, milk and pea soup) based on contents of the nine EAA in that particular product as well as in the overall diet. Protein quality index was formulated in such a manner that, if a particular EAA was deemed deficient in a given diet, food products with higher contents of the said EAA scored higher (and vice versa). The performance of each food product was then evaluated within the context of three diets: an average Swedish diet; a lacto-ovo vegetarian diet; and a low meat diet. Under the average Swedish diet, meat products, particularly beef, scored poorly and, as a result, were judged to have high marginal environmental impacts. Conversely, under the low meat diet in which EAA such as leucine and lysine tend to be deficient, meat products scored favourably, and the results subsequently reversed. Through this example, the authors showcased the importance of considering the nutritional value of each food group in a wider picture of human dietary requirements and food availability.
Building upon the aforementioned protein-focused work published in 2017, Sonesson et al. (
2019) extended their nutrition-in-a-dietary-context method to 12 micro- and macronutrients by employing the NRF9.3 profiling score as a functional unit. In this study, the authors focused on carbon footprints of seven food products: bread, apples, tomatoes (
Solanum lycopersicum), milk, hard cheese, spread and chicken fillets. Echoing the argument from their 2017 paper, the authors posited that nutrient scores are only meaningful if they are accounted for in the context of an individual’s diet; therefore, they considered NDS applied to an average Swedish diet and a typical “unhealthy” diet. The diets were first differentiated by contents of individual nutrients, with the unhealthy diet shown to contain considerably more sodium, sugar and saturated fat than the average Swedish diet. Then, for each nutrient, the diet-level ratio between total content and RDI was used to determine context-specific commodity-level density scores. Using bread as the baseline product for comparison, the authors observed that apples and tomatoes had lower GWP whilst all other products had higher GWP when compared on a mass basis. When considering marginal nutrient requirements in the average diet, however, the gap between bread and other products was narrowed considerably. For example, in the mass-based comparison, chicken fillets had a GWP approximately six times higher than bread; whereas, when nutrient supply was accounted for, this difference reduced to approximately two times higher due to the higher nutrient density of chicken. In the unhealthy diet, the differences remained at around six times higher for chicken due to the relative oversupply of certain nutrients, in particular saturated fat.