Eating healthy or wasting less? Reducing resource footprints of food consumption

We map the food supply chain and its resource footprints using the Food and Agriculture Biomass Input–Output Model (FABIO; Bruckner et al 2019). FABIO is an environmentally extended multi-regional input–output (EE-MRIO) model covering physical flows and associated inputs of harvested biomass, land, and water for 130 food and agricultural products in 191 countries (Bruckner et al 2019). Using the latest available data for FABIO, we map the supply chains and biomass flows associated with German food consumption for the reference year 2013. We account for FLW along the supply chain from post-harvesting to final human consumption.

Footprints are commonly used indicators of human pressure on the environment (Galli et al 2013, Hoekstra and Wiedmann 2014) and generally refer to a supply-chain or consumption-based perspective (Giljum et al 2013, Tukker et al 2014). The biomass footprint shows the primary plant biomass harvested or grazed, according to the conventions of economy-wide material flow accounting (2001, 2013, OECD 2008). The land-use footprint refers to cropland, including both arable land and permanent crops (FAOSTAT 2019). The blue water footprint measures the consumption of freshwater (surface and groundwater), including both direct and indirect water use (Hoekstra and Mekonnen 2012). Following the EE-MRIO methodology, the footprints were calculated as $\boldsymbol{f} = \widehat {\mathbf{e}}\,{\mathbf{Ly}}$, where $\boldsymbol{f}$ is the total footprint of German food consumption, $\widehat {\mathbf{e}}$ is a diagonalized vector giving the environmental input of the addressed resource per unit of output of each product, ${\mathbf{L}}$ gives the Leontief inverse matrix, and ${\mathbf{y}}$ the final demand vector for Germany. Final demand, in the case of FABIO, comprises food use for 130 commodities and 191 source regions including those amounts of food that are wasted at the distribution and consumption stages. We implement the scenarios in the form of adapted final-demand vectors, applying the assumed changes to food consumption and waste (section 2.3). Due to a lack of data, the model does not cover environmental inputs for fish. This increases the uncertainty of the results, in particular for water footprints. We consider this uncertainty acceptable as the scenarios contain relatively low levels of fish. A discussion of the implications of this limitation is found in section 4.2.

The composition of the three dietary scenarios and the current average German dietary pattern are shown in figure 2. The current diet represents the average per capita food consumption pattern in Germany based on the demand vector available in FABIO (adjusted for FLW) and its related supply chains for the reference year 2013. The data is generally given in fresh weights of primary food product equivalents (e.g. oats, olive oil, milk) for 130 products. We use dietary recommendations or principles, specifying the amount of food consumed for different product categories (e.g. vegetables, fruit, and meat) to develop the final demand vectors for three reference dietary patterns. As the dietary recommendations forming the basis of our scenarios are given in quantities of food from different product categories, we translate these into primary food products (e.g. oats, onions, bananas, pork). We split product categories into primary food products according to the proportions of the product categories in the current German dietary pattern. The alternative dietary patterns are standardized to meet the average energy demand of 2796 kcal per day, i.e. 311 kcal less than the current diet. The average energy demand was estimated by Vásquez et al (2018) and is based on the demographic composition of the German population in the reference year 2013. To ensure that the scenarios are 'nutritionally viable' in terms of providing the population with an adequate supply of proteins and fats, the dietary scenarios are controlled against German official recommendations. As the study aims to assess general nutrition patterns and their main resource flows, we refrain from addressing individual diets, including for instance aspects of micronutrients, human uptake, fortified foods, supplements, or particular preparation practices. Information on the nutritional content of the analyzed diets and the recommendations of the German Nutrition Society (DGE) on nutrient intake can be found in the supplementary information (SI), table S1 (available online at stacks.iop.org/ERL/16/054033/mmedia).

Zoom In Zoom Out Reset image size

The guideline diet is based on the nutrition circle scheme recommended by the DGE, showing the proportions of different food groups (Oberritter et al 2013, DGE 2019). Due to consumer convenience, the proportions are given in terms of the weight of prepared food. However, as the form of preparation largely determines the weight, we assume fresh weights as given in FABIO. The assumed moisture contents are given in the supplementary dataset ('H. Moisture_content'). The sustainable diet is based on the reference diet suggested by the EAT Lancet Commission (Willett et al 2019). This diet is developed taking into account health, biophysical limits of the planet, and political targets for climate change limitation, which include indicators for greenhouse gas emissions, nitrogen and phosphorus application, consumptive water use, extinction rate, and cropland use. This reference diet provides more detailed product categories than the guideline diet and its composition is given in grams per day. Given the substantial scientific evidence of balanced vegetarian diets as nutritionally adequate, healthy, and beneficial to prevent certain chronic diseases (Sabaté 2003, Craig and Mangels 2009, Key et al 2006), we develop a low-dairy vegetarian diet as an additional illustrative scenario. Since the EAT Lancet reference diet reflects a generally balanced and low-meat reference diet, we adapt this scenario by excluding meat and fish products and assess protein and fat levels against German recommended levels (see table S1). As the resulting diet exceeds the recommended levels for proteins and fats, there was no reason to increase protein-rich or high-fat foods. We conclude that the dietary pattern reflects balanced vegetarian dietary habits, offering a broad variety of vegetables, fruits, whole grains, legumes, nuts, seeds, soy foods and oils (USDA 2015). Finally, the dietary composition is scaled to the same energy demand as the other scenarios. Similar to the other dietary pattern scenarios, this does not aim to represent an average vegetarian diet in Germany, but rather an appropriate explorative scenario to investigate possible resource use reduction potential. As this first analysis of German diets does not cover all nutritional factors, future studies need to delve into the specific micronutrient requirements of the plant-based scenarios to ensure their viability and adjust their food composition. As the DGE does not provide any recommendations for sugar, sweeteners, and alcoholic beverages, we assume this product group to comprise 10% ¹ of the total diet in the guideline diet. This corresponds to a halving compared to the current diet, but still allows the consumption of these product groups. The EAT Lancet Commission does not specify alcoholic beverages. Therefore, we apply the same quantity of alcohol products as in the guideline diet scenario for the sustainable diet and vegetarian diet scenarios.

The assessed dietary patterns represent variant diets with an increased plant base (figure 2). The current average energy intake is calculated based on the IO-data provided by FABIO whereas the other diets correspond to an intake of 2796 kcal a day as explained above. The guideline diet involves a significant decrease in meat products (98 g as compared to the current 249 g per day), which are partially replaced with dairy products. The sustainable diet replaces large shares of meat and dairy products with plant-based products, such as pulses, beans, and nuts as well as vegetable oils. This diet involves less potatoes and roots than the guideline diet. The vegetarian diet is identical to the sustainable diet except for the replacement of meat and fish with a proportional increase of other product groups. The latter two dietary patterns meet the energy requirements of 2796 kcal per day with about 500 g less of food per day than the guideline diet. The current diet corresponds to an average energy intake of 3107 kcal, with the uncertainty interval of 2646 kcal to 3567 kcal (supplementary dataset 'J. Uncertainty_SQ_Food_Intake'), because of uncertainties in FLW levels (see section 2.3). The dietary scenarios all exceed the national recommendations for protein and fats (not accounting for micronutrients). Given that recommended individual energy intakes for adults vary between 1700 kcal and 3100 kcal depending on age, sex and physical activity level (DGE 2019), the estimated status quo average energy intake of 3107 kcal per capita and day appears high. The plausibility and the implications of this estimation are further discussed in section 4.2. However, the difference of 311 kcal between the status quo scenario and the alternative dietary scenarios is within reasonable uncertainty levels, i.e. the resulting impact of food intake reduction (only) in our dietary scenarios stays within accepted uncertainty levels. All dietary patterns are provided in detail together with nutritional data per product in the supplementary dataset ('G. Diets_analysis').

For the assessments of a 50% food waste reduction, we follow the FLW reporting standard (WRI 2016) and the proposed interpretation of SDG 12.3 (Champions 12.3 2017) and define FLW as commodities that were intended for human consumption, including their non-edible parts. We use shares of FLW per product group at each step along global supply chains and applied these to all products in a respective product group. For supply chain processes within Germany, we use data from the national food waste baseline report (Schmidt et al 2019) and from a study by WWF Germany (2015). The latter is used to fill data gaps of sufficiently disaggregated data in consumption and distribution in the baseline report. For supply chain processes outside of Germany, we use data from FAO (2011), following the methodology outlined by SIK (2013). However, this data excludes inedible parts of plant-based products and eggs from the FLW estimations and is therefore likely to underestimate the FLW. To account for the uncertainties in post-harvest FLW data, an uncertainty analysis was conducted by assuming a range of variation of ±50% of the FLW ratios. In cases where we found estimates in the literature diverging by more than 50%, we use these values as the respective maximum divergence. For more details on the FLW data used, the uncertainty analysis and the considered literature, see the SI.

Our results show that dietary changes are particularly effective to decrease biomass and cropland footprints, whereas blue water footprints are less responsive to dietary change (figure 3). Halving food waste corresponds to a decrease in resource footprints of 11%–15% (shown as FW↓ in figure 3). The error bars indicate uncertainty related to the assumed food waste shares in the status quo. Therefore, scenarios with reduced food waste show a smaller uncertainty range.

Zoom In Zoom Out Reset image size

A change in dietary patterns to the sustainable diet, i.e. the EAT Lancet Commission's reference diet, would result in reductions of the biomass, cropland, and blue water footprints by 54%, 43%, and 7% respectively. For cropland and biomass footprint reductions, this scenario turns out to be three to five times more effective than halving food waste with no dietary changes. The latter has a biomass and cropland saving potential of 11% and 15%, respectively. For decreasing the blue water footprint, however, food waste reduction is more effective than dietary changes (14% compared to up to 7%, depending on dietary scenario). This is due to the particularly high consumption of blue water in fruit and vegetable growing, the proportion of which increases in the alternative dietary scenarios. We enlarge upon this in section 3.3. For the more conservative guideline diet, the water footprint even increases by 6% if not combined with food waste reduction measures. This dietary pattern reduces biomass and cropland footprints by 19% and 22%, respectively, compared to 11% and 15% for the food waste reduction scenario. With a vegetarian diet, the biomass footprint can be decreased by up to 61%, the cropland footprint by 48%, and the blue water footprint by 7%. This implies that a mainly plant-based diet reduces cropland and biomass footprints three to six times as much as the food waste reduction scenario. To reduce water footprints, however, reducing food waste is twice as effective as changing diets to the sustainable or vegetarian diets, and even more compared to the guideline diet.

While the total biomass and cropland footprints related to animal-based products plummet with dietary changes, those of plant-based products increase by up to 41%. The overall decrease of footprints, thus, is related to the reduction of animal-based products in diets, which are the most resource-intensive foodstuffs. Today, 72% and 82% of the total cropland and biomass footprints, respectively, are associated with the consumption of animal products. For water footprints, only 32% of the current total footprint comes from animal products (supplementary dataset, 'C. SQ_Analysis').

A look at the peculiarities of individual foods provides further insights. Figure 4(a) shows the relation between footprints per unit of mass and per unit of energy content (y-axes) and food waste shares (x-axes) for ten food categories. Meat has particularly high footprints per mass unit for all three resources, but most prominently for biomass and cropland use (figure 4(a)). For water, the difference between animal-based products and plant-based products is not as pronounced, which also explains the smaller difference for water footprints between the scenarios in comparison to biomass and cropland. Plant-based products generally show lower footprints per unit of energy content, with vegetables and fruit being the exceptions in terms of water footprints.

Zoom In Zoom Out Reset image size

Based on the food waste estimates for each product group underlying this study, we can expect interactions and even goal conflicts between the strategies of dietary change and food waste reduction. Vegetables, cereals, fruit, potatoes, and roots are associated with higher levels of food waste (figure 4(a)). Subsequently, an increased share of these product groups in the dietary pattern might cause the overall amounts of food waste to grow, which may hamper the possibility of reaching the political target of halving food waste. Figure 4(b) shows the composition and amounts of food waste for the assessed dietary scenarios and the current diet. Given that food waste shares of each product remain the same, this shows that changes according to the dietary scenarios would increase the amounts of food waste, in particular for the guideline diet.

The main question here is how this interaction affects the total footprints of food consumption.

Plant-based products that are associated with relatively high levels of food waste are particularly relevant in this context. If a dietary shift includes greater amounts of such products, there is a risk of increased amounts of food waste. At the same time, larger shares of these products decrease the overall footprints of food consumption, with the water footprint in the case of the guideline diet as the only exception. Thus, in terms of both total consumption footprints and food waste footprints, scenarios of dietary change would be beneficial. In combination with decreased food waste, in particular for fruit and vegetables, also water footprints decrease for all scenarios. Because of the simultaneous reduction in footprints and increase in food waste associated with dietary changes, total resource losses are not reflected in the amount of food waste. Resource footprint indicators, which embed the overall resource use and loss associated with final consumption (from food losses in agriculture to conversion rates in processing and food losses along the whole supply chain), would thus be better suited for political target setting. These losses are much higher than what is addressed as food waste (Shepon et al 2018).

Few studies have addressed the implications of the circumstance that cereals, potatoes, roots, fruit, and vegetables are associated with greater food waste shares but generally lower footprints (Conrad et al 2018). This study adds new evidence to the challenging trade-off between healthier diets and food waste. The importance of the resource efficiency of food products themselves in comparison to losses was underlined by Shepon et al (2018), showing that the 'conversion loss' of animal products by far exceeds total food losses. Our results confirm that including biomass footprints in resource efficiency assessments allows capturing all types of resource losses in one indicator and thereby reflects resource efficiency more adequately than the amount of food waste.

The results of this study substantiate other research findings regarding the high potential of dietary changes to decrease resource footprints (Aleksandrowicz et al 2016, Bryngelsson et al 2016, Behrens et al 2017, Conrad et al 2018, Vanham et al 2018). A literature review across countries (the majority European) shows average land savings of 20% for diets following dietary guidelines, 51% for vegetarian diets, and 55% for vegan diets (Aleksandrowicz et al 2016). This is very much in line with our results of 22% land savings for the guideline diet and 48% for the low-dairy vegetarian diet. For blue water footprints, previous studies show an average decrease of 37% for vegetarian diets and 22% for diets following the local dietary guidelines (Aleksandrowicz et al 2016, Vanham et al 2018). Our results show a smaller impact of dietary changes on the blue water footprint, ranging between +6% and −7%. Particularly the blue water footprint of diets varies among studies (Meier and Christen 2013, Aleksandrowicz et al 2016, Tom et al 2016, Rehkamp and Canning 2018). Our results differ from a study focusing on Germany that shows a tripling in blue water use for vegan and vegetarian diets, with more than two-thirds of the footprint being associated with nuts and oilseeds (Meier and Christen 2013). Although the authors do not further explain the deviation from other studies, the inventory data of 1425 l of blue water per kilogram of nuts and oilseeds (Meier and Christen 2013) indicate a particularly water-intensive composition of this product group (see Mekonnen and Hoekstra 2011). Differences in the reduction potentials among different countries can be explained by differences in consumption patterns, supply chain structures, and production methods (Kim et al 2020).

Limitations of this study mainly include the previously discussed uncertainty of FLW data and the resulting high average energy intake. Additionally, the lack of environmental data related to fish and the absence of other indicators such as a scarcity-weighted water footprint, pastures, nutrient loss, and greenhouse gas emissions leave room for improvement. Complementing nutrition studies need to delve into the nutritional viability of dietary patterns.

The current average energy intake of 3107 kcal appears high in comparison with survey-based data on German food consumption habits, which has several possible reasons. A comprehensive German survey showed an average energy intake of 2159 kcal per person and day from 2005 to 2013 (Gose et al (2016), average from table 4, converted to kcal). The product groups mainly contributing to the difference to the FAO data used in our study are meat and alcohol. The survey acknowledges an expected underreporting of energy intake of 11%–17%, which would imply a total energy intake of 2396–2526 kcal per person and day. Heseker et al (1994) came to a similar average energy intake of 2414 kcal (p 123, converted to kcal). The difference to our resulting average energy intake is about 650 kcal, with an uncertainty interval of 185–921 kcal. The most possible sources of error of our estimate involve uncertainties in FLW data and a detected discrepancy in the product energy content in FAO data. The high uncertainty in FLW data is mainly due to the lack of proper monitoring possibilities, as well as the exclusion of inedible parts of plant-based products and eggs as waste flows, which may contribute to an underestimation of FLW. Given the high uncertainty of FLW data, the uncertainty intervals are highly relevant for the interpretation of the results. Additionally, we noted a discrepancy in FAO energy content data. In this study, we used the energy content data given in the Food Balance Sheet (FBS) handbook (FAO 2001) (see supplementary dataset 'G. Diet composition'). However, our estimations are, on average, 12% higher than the energy content derived from the FBS for Germany (reported in kcal and kg per capita). Diverging data within FAO or possible moisture content inconsistency of product group reporting constitute ambiguities. We keep to the data provided by FAO (2001) while assessing the implications of possible data faults. With the maximum waste levels in our uncertainty range, this study estimates the current diet to consist of 2646 kcal (see supplementary dataset 'J. Uncertainty SQ Food Intake'). Assuming these higher levels of FLW and a more conservative estimation of current food intake, the effects of food waste reduction in the current diet would be 17% for biomass, 20% for cropland, and 21% for blue water (the lower bound of our uncertainty range). The possibly lower energy content of all diets does not alter the main conclusions of the study.

Fish is considered to be an environmentally friendly substitute for meat (Willett et al 2019). However, while fish is shown to be preferable over meat in terms of greenhouse gas emissions and land-use change, fish from pond aquaculture has a remarkably high blue water footprint (Kim et al 2020). The dietary patterns in this study only include small amounts of fish, the sustainable diet being the one with the largest share of fish. Thus, the water footprint for this diet may be slightly underestimated. However, the environmental relevance of water footprints depends on the water scarcity at the point of withdrawal (Ridoutt and Huang 2012), which is not accounted for in the present study. Scarcity-weighted national water footprints are relatively high in the case of nuts, fish, red meat, chicken, and olive oil (Clark et al 2019). Thus, a closer look into the scarcity-weighted water footprint of these products would be useful. Similarly, increased nuances of dietary assessments can be achieved by also accounting for pastures, land-use suitability, and availability. Based on this perspective, a study of US food consumption concludes that some balanced omnivore diets have a higher carrying capacity than vegan diets (Peters et al 2016). To get a more holistic view of the environmental impacts of different target scenarios, studies on carbon footprints and nutrient cycles is needed. Indirectly, these aspects are partly captured as the sustainable diet scenario is based on a reference diet that considers environmental requirements including nutrients and greenhouse gas emissions. Moreover, this study assumes as a baseline for all scenarios the German food supply chains of 2013. In reality, supply chains change to some degree when dietary patterns are altered and the product composition changes. Changing dietary patterns would also provide opportunities to change production patterns accordingly. For instance, reduced demand for certain crops in some parts of the world can ease the pressure on valuable ecosystem services in other parts, but at the same time, demand can increase the production of more sustainable alternatives in the country of consumption.

The potential of dietary changes for resource savings and public health opens up a broader scope of policy options. We reveal that dietary changes could be up to six times more effective in decreasing resource footprints than halving food waste. These findings urge policymakers to consider consumption-oriented political actions. Political instruments to implement targets of healthier and increasingly plant-based dietary patterns can comprise a mix of approaches. These include legal tools such as changing macro-economic policies, fiscal measures, and stricter marketing rules for unhealthy food, as well as promotion and supportive strategies such as nudging through changing store layouts, education, empowering community initiatives, and offering healthy and sustainable food in schools (Garnett et al 2015, UNEP 2016). At the same time, effective political instruments for food waste reduction are still not in place. So far, the German strategy mainly relies on a dialog forum for stakeholders, workshops, and information campaigns (BMEL 2019). Thus, even reaching the comparatively small potential to reduce the footprint by halving food waste is still in the distant future. Presumably, it is easier to suggest food waste reduction than dietary changes, as the latter invokes individual food consumption habits which are often a sensitive topic. However, research on food waste causes and reduction measures shows that food waste reduction also requires considerable behavioral changes, such as shopping and cooking habits (Stancu et al 2016, Hebrok and Boks 2017, Schanes et al 2018). However, the responsibility is perceived as more clearly shared with food supply stakeholders and the individual changes seem less intrusive. Nevertheless, the obscure necessity of behavioral change to decrease food waste may hinder successful implementation.

Based on our findings, we conclude that focusing solely on targets and indicators for food waste may be misleading in achieving the overarching goal of significantly reducing the resource footprints of food consumption. However, policies addressing dietary shifts would lead to an increased proportion of cereals, roots, potatoes, fruit, and vegetables in the diet—this will lead to an increase in food waste as these products have higher food waste ratios, although they will still reduce overall resource footprints.

We show that footprint indicators could help political prioritization and target setting. At present, footprint indicators are entirely lacking in policy strategies and target monitoring tools for the food sector. The SDG 12 indicators in the proposed list as of 2016 (IAEG-SDGs 2016) do include the material footprint, although not specifically for the food sector. Indicators that capture the trade-offs between resource use and food waste reduction are lacking. The EU action plan for the Circular Economy (EC 2015) has similar shortcomings. The first monitoring framework lacks footprint indicators and instead focuses largely on the quantification of waste (EC 2018). The basic deficiencies in this monitoring approach (EC 2018) are highlighted by Helander et al (2019). In the recently updated Circular Economy Action Plan (EC 2020), an update of the monitoring framework is foreseen, which intends to include indicators on resource use and material footprints. To monitor strategies related to the food sector particularly, we suggest including resource footprints of food consumption in the updated version of the monitoring framework. The possible trade-offs between goals revealed by this study emphasize the necessity of including such indicators if a circular economy is to help avoid the irreversible damages caused by the high rate of resource consumption, as the action plan suggests (EC 2015).