## Thursday, March 25, 2021

### 3486. Food Systems Are Responsible for a Third of Global Anthropogenic GHG Emissions

• By M. Crippa
• E. Solazzo
• D. Guizzardi
• F. Monforti-Ferrario
• F. N. Tubiello &
• A. Leip

• ## Abstract

We have developed a new global food emissions database (EDGAR-FOOD) estimating greenhouse gas (GHG; CO2, CH4, N2O, fluorinated gases) emissions for the years 1990–2015, building on the Emissions Database of Global Atmospheric Research (EDGAR), complemented with land use/land-use change emissions from the FAOSTAT emissions database. EDGAR-FOOD provides a complete and consistent database in time and space of GHG emissions from the global food system, from production to consumption, including processing, transport and packaging. It responds to the lack of detailed data for many countries by providing sectoral contributions to food-system emissions that are essential for the design of effective mitigation actions. In 2015, food-system emissions amounted to 18 Gt CO2 equivalent per year globally, representing 34% of total GHG emissions. The largest contribution came from agriculture and land use/land-use change activities (71%), with the remaining were from supply chain activities: retail, transport, consumption, fuel production, waste management, industrial processes and packaging. Temporal trends and regional contributions of GHG emissions from the food system are also discussed.

## Main

Food needs to be farmed, harvested or caught, transported, processed, packaged, distributed and cooked, and the residuals disposed of. Each of these steps causes emissions of anthropogenic greenhouse gases (GHGs) and requires energy. Inputs such as fertilizers or energy need to be produced and made available at the right time and location1,2,3,4 with additional associated GHG emissions.

Major datasets of GHG inventories—including those with country coverage (National Inventory Reporting under the United Nations Framework Convention on Climate Change (UNFCCC)), regional or global coverage (for example, the Emissions Database of Global Atmospheric Research (EDGAR, https://edgar.jrc.ec.europa.eu/), GAINS (https://iiasa.ac.at/web/home/research/researchPrograms/air/GAINS.html) and FAOSTAT (http://www.fao.org/faostat/en/#data/))—provide detailed temporal and sectorial evolution of total GHG emissions. Yet, emissions from the food systems are scattered across many different source categories (Supplementary Fig. 1). Global estimates of the share of emissions associated with agriculture, which includes farm gate production and associated land use, have been produced5, and more recently emission estimates from the various stages of the life cycles of food products have also been made available6,7,8,9,10. Another recent estimate of global food-system emissions has been provided by the Intergovernmental Panel on Climate Change (IPCC) Special Report on Climate Change and Land11, attributing between 10.8 and 19.1 Gt CO2-equivalent (CO2e) emissions per year to the food system globally, corresponding to 21% to 37% of overall anthropogenic emissions11,12. Other studies report good agreement between ‘top down’ and ‘bottom up’ methods13,14 for Europe. The review of available resources for emissions from food systems shows how, overall, available data are based on detailed product-specific life cycle assessment studies8,14,15 or are using aggregated global data9,10. So far, however, studies encompassing global coverage of the whole food-system at country level are missing and, consequently, the total emissions and the total share of those emissions associated with food systems are largely unknown.

The global database of GHG emissions (CO2, methane (CH4), N2O, fluorinated gases (F-gases)) from food systems (EDGAR-FOOD) developed in this Article aims to fill this gap by using a consistent methodological framework. EDGAR-FOOD has been developed to aid the understanding of the activities underlying energy demand and use, as well as agriculture and land-use change, and emissions associated with the production, distribution, consumption and disposal of food through the various stages and sectors of the composite global food system. These data were complemented with data from the FAOSTAT database on GHG emissions from land use related to agriculture15. EDGAR-FOOD represents the first database consistently covering each stage of the food chain for all countries with yearly frequency for the period 1990–2015.

## Emissions from the food system

A third of global GHG emissions comes from the food system. Our estimate of the contribution of food systems to total anthropogenic GHG emissions was 34% (range 25% to 42%) for the year 2015. Global GHG emissions from the food system were 18 Gt CO2e yr−1 (95% confidence interval (CI) 14–22 Gt CO2e yr−1) in 2015, with 27% (or 4.9 (95% CI 3.7 to 6.4) Gt CO2e yr−1) emitted by industrialized countries (country definitions are regional groupings and are provided in Supplementary Table 2), and the remaining 73% (or 13 (95% CI 10 to 16) Gt CO2e yr−1) emitted by developing countries (including China) (Fig. 1). In 2015, 71% of global GHG emissions from the food system was associated with the land-based sector, defined herein as agriculture and associated land use and land-use change activities (the latter will be referred to as LULUC). In industrialized countries, the contribution of the downstream energy-related sectors (53%), which includes industry and waste, was larger than the land-based sector, while in developing countries agriculture and LULUC were the dominant fraction (73%) (Fig. 1).

In 2015, six top emitting economies (the term ‘economies’ is used to allow the European Union to be considered as a single entity) with individual contributions larger than 6% to the global total GHG emissions from the food system were responsible for 51% of our estimated global food-system total. They were: China with 2.4 Gt CO2e (13.5% of the global total), Indonesia with 1.6 Gt CO2e (8.8%), the United States with 1.5 Gt CO2e (8.2%), Brazil with 1.3 Gt CO2e (7.4%), the European Union with 1.2 Gt CO2e (6.7%) and India with 1.1 Gt CO2e (6.3%). Supplementary Table 6 reports country-specific contributions to global GHG food-system emissions in 2015.

While food-system GHG emissions increased from 16 (12 to 20) in 1990 to 18 (95% CI 14 to 22) Gt CO2e yr−1 in 2015 (an increase of 12.5%), the share of total GHG emissions decreased over time; that is, it was 10% higher in 1990 (44%) than in 2015 (34%) (Fig. 2a, solid lines). At the same time, global food production, taking cereals as proxy, increased by over 40%, indicating an overall decrease in the emission intensity of food during the same period. The temporal evolution of the share of food-system emissions differed significantly between groups of countries. The share was stable (around 24%) and relatively low for industrialized countries. On the contrary, in developing countries the share of food-system emissions significantly decreased, from 68% in 1990 to 39% in 2015 (Fig. 2a). This food emission trend was nevertheless different when we focused on specific countries: it grew by 41% in China and only by 14% in the rest of the developing countries. The global share of food systems to total emissions slowly decreased to approximately 25%, though a striking decreasing pattern was found for developing countries, where food-system emissions decreased from almost 70% of total emissions to ~40% (Fig. 2a). This sharp decrease, which is in line with previous analyses16,17, was due to the very high increases in non-food emissions over the period coupled to a significant reduction in land-based emissions, which was largely due to a reduction in deforestation (Fig. 2b).

## Emissions from LULUC

Almost one-third of food-system emissions comes from LULUC. According to FAOSTAT, emissions from LULUC associated with agricultural production (IPCC sector 3a) accounted for 5.7 Gt CO2e yr−1 in 2015, or about 32% of total food-system emissions. These emissions are mainly composed of carbon losses from deforestation and from degradation of organic soils, including peatlands17. Most of the LULUC emissions (5.0 Gt CO2e yr−1) occurred in developing countries, thereby substantially affecting the food-system emission share in this country group (Fig. 2a, dashed lines). Furthermore, estimates based on input–output models showed that much of these emissions was associated with food consumption in industrialized countries18,19.

## Energy use in the food system

The global food system is becoming more energy intensive, with almost a third of food-system emissions coming from energy-related activities. While our data confirmed that all life-cycle stages contributed substantially to GHG emissions, the production stages that bring foodstuffs to the ‘farm gate’ (including fishing, aquaculture and agriculture, plus emissions from the production of inputs such as fertilizers, but excluding LULUC) had the largest share of emissions, contributing 39% (or 7.1 Gt CO2e yr−1) to total food-system GHG emissions in 2015, followed by LULUC (32% or 5.7 (95% CI 2.8 to 8.5) Gt CO2e yr−1). Distribution (including transport, packaging and retail), processing, consumption and end-of-life disposal summed to 29% (or 5.2 (95% CI 3.2 to 7) Gt CO2e yr−1), a share that was higher in 2015 than in 1990 for both the developed and industrialized country groups.

CH4 emissions accounted for 35% of food-system GHG emissions (expressed in CO2e) (Fig. 3) consistently across developed and developing countries (32–37%) mainly due to livestock production, farming and waste treatment. (Non-CO2 GHG emissions (CH4, N2O and F-gases) are expressed as CO2e calculated using the 100 year global warming potential values (GWP100) used in the IPCC Fifth Assessment report (AR5), with a value of 28 for CH4 and 265 for N2O.) In most developing countries and globally, rice is one of the leading food crops and a principal source of CH4 emissions. Asian countries dominate global rice production, and the share of rice production to total food-system emission is 39% in Thailand and 40% in Bangladesh. To put things in perspective, China, India and Indonesia are the top rice-producing countries, followed by Bangladesh, Vietnam and Thailand20.

While food-system emissions of N2O were comparable across both groups of countries (9 to 14%), emissions from F-gases (2% of global food-system emissions), mostly linked to refrigeration in the retail stage, were predominantly from industrialized countries (8% of their overall GHG emissions).

Interlinkages between the components of the food-system GHG emissions, including the contribution of different gases to emission sectors and categories and their relation to food supply chain stages, are shown in Fig. 3a, which also shows the respective shares to total food-system emissions and emission trends. Globally, the share of CH4 slightly increased (an increase of 3% compared with the share in 1990) and total land-based emissions decreased (66% of the total for 2015, which is a decrease of 13% compared with 1990). The share of CO2 emissions from the energy sector increased (21% of the total in 2015, which is an increase of 31% compared with 1990). This was particularly true for the food processing and distribution stages, which include retail, packaging, transport and processing. All of these stages increased their shares by between 33% and 300% globally compared with the share in 1990. Also, the use of F-gases in industry has increased substantially—by more than 100% since 1990—due to their use in refrigeration. The increased importance of food supply chain emissions was more pronounced in developing countries (Fig. 3c) than in industrialized countries (Fig. 3b), where a particularly steep rise of F-gases for retail (quadrupling the 1990’s share) occurred. In developing countries, the share of emissions from agricultural production and LULUC within total food-system emissions dropped significantly (land-based sector shares decreased by 13%, LULUC by 26%), while the share of CO2 for energy increased by 78%. The share of GHG emissions from waste management, while decreasing in industrialized countries, increased in developing countries (an increase of 50%). Much of the increases in food-system GHG emissions from developing countries occurred in China; without China, emissions from energy were only 10% of total emissions, LULUC increased by 10% and packaging increased by 50%.

## Energy use at the farm gate

The overall energy use inside the farm gate, albeit small in its contribution to total food systems emissions, has increased substantially in the past 25 years, but has followed different paths across countries. Globally, our data show an increase of 15% in emissions from the use of energy (electricity, heat and fuels) in the agricultural sector compared with 1990, with the highest increase happening in developing regions (an increase of 50%) such as Africa, Latin America and Asia. In these economies, which generally remain strongly agricultural, emissions have increased because agricultural production has become more mechanized, and this includes increased use of fertilizers and pesticides. In addition, agriculture in some developing countries has expanded both to provide food to a steeply growing domestic population and also for export21. Conversely, the introduction of agronomic progress and environmental restrictions in agriculture22 has led to more efficient use of fertilizers and increased efficiencies in livestock production, and nowadays lower emissions from energy used in agriculture (which has decreased 28% compared with 1990) are found in industrialized countries23,24,25,26. At the same time, emissions associated with use of solvents, which are also used to produce pesticides, increased to 15 times the global level, while solvents used for the production of fertilizers increased by 24%. Significant portions of the increased production were for use in developing countries.

## Emissions from food distribution

GHG emissions from food distribution are on the rise but ‘food miles’ are less important than packaging. Our data show a global food system characterized by an increase in convenience and processed food, and an increasing globalization of the food supply chains everywhere, while at the same time huge differences exist in the distribution and availability of food27,28. To function, the food system requires materials and energy for processing, packaging, transporting and storage. Of these, packaging had the highest emissions. In 2015, packaging contributed about 5.4% (or 0.97 Gt CO2e yr−1) of total food systems emissions (Fig. 4a). Our estimate is higher than previous global estimates6,29 and might reflect either that our data is more recent or that upstream emissions—including emissions from input and energy production—are included.

However, not all food products and packaging materials are equal. We estimated major contributions from the pulp and paper industry (59.9 (52.7 to 70.7) Mt CO2e yr−1), aluminium production (29.9 (26.3 to 35.2) Mt CO2e yr−1), the metal industry (10.6 (9.3 to12.5) Mt CO2e yr−1) and use of glass (4.8 (2.2 to 7.7) Mt CO2e yr−1). This is consistent with the findings reported by Poore and Nemecek6, who show a significant share of packaging-related emissions for beverages (wine and beer, > 40%) and some fruit and vegetables (10–22%).

In view of the public and academic debate on ‘food miles’30,31,32, we estimate that transportation contributes 4.8% (or 0.86 (95% CI 0.30 to 1.5) Gt CO2e yr−1) to food-system GHG emissions, approximately the same as retail (4.0% or 0.72 (95% CI 0.21 to 1.44 Gt CO2e yr−1)). Our estimates indicate that due to the huge variations in energy needed per transported ‘food mile’ (from marine shipping at 10–20 MJ t−1 km−1, road transport at 70–80 MJ t−1 km−1 and aviation at 100–200 MJ t−1 km−1 (ref. 8)), the majority of emissions arise from local to regional transport via road (81%) or rail (15%), rather than navigation (3.6%) or aviation (0.4%). Urban policy and food logistic policies could thus play a significant role in improving the energy efficiency of food systems33. Transport-related GHG emissions are higher for heavy or easily perishable products, and some food products have a particularly high share of transport GHG emissions (>40% for bananas and beet sugar, according to Poore and Nemecek6). However, for road transport, detailed data were available only for Europe and the United States, and an average value based on these data was used for all other countries, and therefore these data are associated with large uncertainties. For shipping, aviation and railways, global average values, which may not adequately reflect the situation for individual countries, were used.

Globally, refrigeration has been estimated to be responsible for 43% of energy consumption by the retail/supermarket sector34. Our data suggest that GHG emissions from the retail sector increased by 4.2 and 3.6 times in Europe and the United States, respectively, between 1990 and 2015. An increase in the market share of supermarkets in the food distribution sector has been observed in all continents, including Africa, Asia and South America35,36,37. Supermarket refrigeration is not only energy intensive, but also generates leakage emissions of substitutes for ozone-depleting substances, although their contribution to food-system GHG emissions is estimated to be minor. This ‘cold chain’, including both industrial and domestic refrigeration, accounts for 5% of global GHG food-system emissions, but given that the number of refrigerators per capita in developing countries is about one order of magnitude lower than the number in developed countries, the importance of refrigeration to total GHG emissions is likely to increase38.

## Share of food-system GHG emissions

The share of food systems as a percentage of total anthropogenic GHG emissions varied significantly across countries and regions (Fig. 5), and varied between 14% and 92%. A high share of food-system GHG emissions can be a sign of a strong agri-food sector or of a weak economy. Large shares of food-related emissions found for selected countries in 2015 can be the consequence of different factors: (1) low-income countries where local industry and other economic sectors other than agriculture are relatively small (mostly in Africa and south-eastern Asia), (2) an important food exporting industry such as in Brazil, Argentina and other South American countries, or (3) high food-system emissions from LULUC, as found in Brazil, Indonesia and African countries. For example, in west Africa, to satisfy nutrition needs associated with a fast-growing population, the trend for the share of food-system emissions showed an increase from 69 to 79% (from 1990 to 2015, respectively). Conversely, in China the share fell from 51% in 1990 to 19% in 2015 due to the industrialization of the country and large trade flows of agricultural commodities39, which reflected a shift between land-based and energy-associated GHG emissions in the food system. Our results are consistent with the rapid transformation of food systems in developing regions40. Overall, the most energy intensive economies (for example the United States, Canada, Europe, China and Japan) showed the lowest contributions of food-system emissions to total GHG emissions.

In Brazil, GHG emissions from the food production sector decreased by about 30% from 1990 to 2015, primarily due to substantial decreases in deforestation rates. This decrease occurred despite continuous increases in emissions due to livestock production activities and an increase in the exports of beef and soybean by 720% and 530%, respectively41.

In sub-Saharan African countries, food production is still to a large degree realized by smallholder farms42, and eastern and western Africa show shares of land-based food-system GHG emissions of 88% and 69% in 2015, respectively (Fig. 5). Ricciardi et al.43 estimated that, globally, farms smaller than 2 ha produce 30–34% of the food supply on 24% of the gross agricultural area. To feed a population that doubled in size between 1990 to 2015, Nigeria increased rice production resulting in four times higher GHG emissions from this sector44, and in 2015 Nigeria emitted more than a third of food-system GHG emissions of the whole western African region. Livestock production emissions increased by 2.8 times compared with 1990 mainly due to the introduction of goats in addition to cattle. As a result of population growth, domestic waste water emissions increased by 3.4 times as well. In Ethiopia, the population doubled between 1990 and 2015, making the country the largest emitter of GHGs from the food system in eastern Africa. The Ethiopian food system contributes to 78% of anthropogenic GHG emissions, with high emissions from cattle which increased by 2.5 times over the same period45.

## Per capita emissions from the food system

Our data support the estimation of global food-system GHG emissions with broad coverage of emission sources, as well as with regional and temporal detail. On average, each person’s food-related emissions in 2015 were 2.4 (2.1–2.9) t CO2e. This represented a decrease relative to 1990 levels, when per capita food-related emissions were 3.0 (2.3–3.8) t CO2e. Table 1 reports the evolution of food-system emissions as a percentage of the national total GHGs and global food-system emissions from 1990 to 2015 by continent and region. Table 1 also shows the development of per capita food-system GHG emissions. These numbers are not to be mistaken for consumer GHG footprints, which are determined by the actual diet in a specific country and assign emissions occurring throughout the food supply chain to that country. Our data reflect the structure of the countries’ food system and economy. They are consistent with how GHG inventories are reported to the UNFCCC and can be used to benchmark national mitigation efforts to reduce GHG emissions from food systems. Effective policies to transform food systems towards sustainability need to be comprehensive and provide answers and adequate information46 to both the economy and its consumers. Policies also should address both food production and food consumption47,48,49,50.

## Conclusions

EDGAR-FOOD provides a picture of how an evolving world food system has responded to the evolution of world population in the last 25 years, which has coincided with changes in dietary habits and food-related technology. At the global level, the decoupling of population growth and food-related emissions is visible with emissions growing at a lower rate compared with population growth. The regional view is more diverse, with some areas rapidly increasing emissions due to domestic demand for either food or export.

Unlike overall GHG emissions, the food production sector is not overwhelmingly dominated by CO2 emissions from fossil fuels; land-based emissions are particularly relevant. Nevertheless, in line with the ongoing socio-economic development trends, food emissions are being increasingly determined by energy use, industrial activities and waste management. On the one hand, from the point of view of mitigation, such a trend suggests that the food sector will need specific sectorial energy efficiency and decarbonization policies. (For instance, the food industry generally requires lower heating temperatures than other types of industrial production, and those lower temperatures are more easily reached by non-fossil-based technologies.) On the other hand, the continuing predominant role of land-based emissions, within and outside the farm gate, shows that food production itself will continue to be a major source of emissions that will require dedicated mitigation policies.

The global food emissions database EDGAR-FOOD provides a broad level of geographical, temporal and thematic detail of national GHG emissions from the global food system and represents a milestone in our understanding of how the global food system has developed. With its detailed and consistent dataset of the emissions related to the various stages of the food system, it is possible to estimate the changes in food-system GHG emissions driven, for example, by consumer behavioural changes or technological evolution. Moreover, it is crucial to the anticipation of future changes in the overall food system and to the design of efficient mitigation strategies that avoid creating additional emissions in non-targeted sectors. Owing to its completeness and flexibility, our dataset is intended as a tool for the scientific community to allow researchers to focus on specific sectors or groups of sectors, freely aggregating and splitting data to design their investigations.

The completeness of the EDGAR-FOOD database is an important asset for effectively monitoring global food-system GHG emissions. This database is in line with the strategies that work with an integrated view of the food system, such as the new European Commission’s Farm to Fork Strategy46.

## Methods

We developed the new food emissions database (EDGAR-FOOD) using the EDGAR51 covering the IPCC sectors ‘Energy’, ‘Industry’, ‘Agriculture’ and ‘Waste’ (see Supplementary Fig. 1). We complemented the EDGAR-FOOD data with the FAOSTAT database52 for emissions and sinks in the LULUC sector. EDGAR-FOOD provides detailed, highly disaggregated and harmonized emission estimates covering all sectors and geographical areas using a rigorous and transparent process of data integration and verification. The EDGAR and FAOSTAT data are widely used for climate research and as a basis for global climate policies, and they are also used in the IPCC Fifth and Sixth Assessment Reports.

For EDGAR-FOOD, we complemented the EDGAR database with estimates of the food-system emission shares for each of the source categories. One of the challenges of compiling EDGAR-FOOD is to calculate the food-related portion of industrial and energy processes that impinge on other sectors (for example, the glass industry, energy production and waste, among others) for each country, while preserving global consistency. Details on the share of food-system emissions for each source category (value, method and uncertainty level) are given in Supplementary Table 1. Assumptions made and uncertainties are discussed below. Supplementary Table 1 also indicates the life-cycle stage and the food-system sector to which the source categories are assigned.

We distinguish six life-cycle stages in food systems: (1) LULUC: land use, land-use change; (2) production: primary production of food commodities; (3) processing: food processing; (4) distribution: food distribution including packaging, transport and retail; (5) consumption: food consumption including domestic activities of food preparation; and (6) end of life: end of life of food, including food residues management and management of non-food residues used in previous food-system stages.

EDGAR-FOOD allocates the extraction of raw materials and the production of inputs required for primary food production, and the provision of primary energy is assigned to each individual stage.

We distinguish the following food-system sectors: land-based sector (including crop and livestock products and LULUC), energy, industry and waste.

### Food-system emissions calculation at the global scale

The quantification of food-system GHG emissions is done using the EDGARv5.051 (https://edgar.jrc.ec.europa.eu/overview.php?v=50_GHG) as the underlying source of global GHG and air pollutants at country scale. Emissions are calculated using a bottom-up methodology complying with IPCC guidelines53, with high sectorial disaggregation54,55 (https://edgar.jrc.ec.europa.eu/index.php). GHG emissions are computed in EDGARv5.0 mostly following the IPCC Tier 1 approach, while a Tier 2 method is applied to estimate emissions from cattle enteric fermentation, rice cultivation, waste treatment and cement production. The 2019 IPCC refinement guidelines are not included in EDGARv5.0, and therefore they are not part of this work.

The EDGAR inventory has been coupled with shares describing the contribution of the food system to each emissions sector (Supplementary Table 1). Food-system GHG emissions are calculated by multiplying these sectorial food-system shares (SFSs) with the total emissions of each sector (equation (1)). Specifically, SFS equals 1 if the whole sector is part of the food system, such as food and beverage sector, while SFSs for sectors only partly representing the food system (for example, transportation) take a value from 0 to less than 1:

${E}_{i,c,t}^{Food}=\sum _{s=1}^{n}{E}_{i,s,c,t}×{\mathrm{S}\mathrm{F}\mathrm{S}}_{s,c,t}$
(1)

where i represents each greenhouse gas (fossil CO2, CH4, N2O and F-gases), c is each world country, t is each individual year (from 1990 to 2015), s are all emitting sectors in EDGAR, E is the emission of a certain sector, and SFS is the sectorial food-system share. SFS is defined per sector or subsector and can vary between countries and years, depending on the availability of data. In those cases when detailed information was not available, a global average share was adopted, though this does not reflect country-to-country variations, but still represents the best available knowledge to complete the global picture. Dependency of the shares on time, country and fuel type is indicated in Supplementary Table 1.

### Attribution of primary energy emissions to the food system

Energy consumption in the food system causes GHG emissions at different stages: (1) emissions caused by combustion in large and small scale industries, households, transport or other food-system actors; (2) emissions caused by centralized heat and electricity production that is consumed by food-system actors; (3) emissions caused by the ‘fuel chain’, including fuel extraction, transport and leakage; and (4) indirect GHG emissions caused by any of the above sources.

Emissions from 1–4 are available by fuel-type and sector.We developed an estimate for SFS for each sector with emission estimates available in EDGAR. We then calculated food-system emissions from centralized heat and electricity production, fuel chain emissions and indirect GHG emissions from energy consumption according to points 2 and 3.

#### Fuel combustion

Within the food-related power generation emission sectors, we considered the food production industry, fishing and food-related agriculture, packaging production, household and retail/services activities as those consuming electricity and heat for food-related purposes. We assumed that the total heat and electricity from fishing, food-related agriculture and food production contribute to food-system emissions. Shares reflecting the use of energy in the food system (for example cooking, refrigeration and so on) are calculated and applied to the packaging industry (for example, share of packaging used for food), household (for example, cooking, refrigeration and so on) and retail/services sectors (for example, food retail and grocery, restaurants and so on), as discussed below.

#### Heat and electricity production

Shares of energy used in the food system are based on International Energy Agency (IEA) electricity and heat consumption data56, detailed by country and subsector. The electricity shares by subsector are then applied to the EDGAR-FOOD emissions from power generation, in particular from public electricity and cogeneration plants as well as from auto producers of electricity and cogeneration. The shares of heat are applied to the emissions from auto produced heat plants and district heating plants. SFSs,c,t varies (minimum–maximum) between 4.4×10-5 and 0.13 depending on the year, country and subsector.

#### Fuel chain

Emissions from the fuel production sector have been calculated by allocating the emissions from fuel production, transformation and refineries, and determining the shares of each fuel that are used in the food system. This share is calculated as CO2 emissions from combustion of a specific fuel for each food-system category over the total CO2 emissions from combustion of the fuel in the country on an annual basis. These emissions represent the contribution of the fuel chain to the food system and allow quantification of the emission reduction through the full chain when reducing certain activities. SFSs,c,t varies (minimum–maximum) between 0.0003 and 0.87 depending on the year, country and fuel.

#### Indirect emissions

Indirect N2O emissions from ammonia (NH3) and nitrogen oxide gases (NOx) emitted by IPCC categories 1A (‘energy’), 2 (‘industrial processes and product use’) and 3 (‘agriculture, forestry and other land use’) related to the food system only are included. Indirect N2O emissions from leaching and runoff of nitrate are estimated from nitrogen input to agricultural soils. Indirect N2O emissions from atmospheric deposition of NOx and NH3 emissions from agricultural (crop residues, synthetic fertilizer, animals in pasture and manure input to soils) and non-agricultural sources (mainly fuel combustion and industrial processes) are estimated using nitrogen in NOx and NH3 from these sources as activity data. More details can be found in Janssens-Maenhout et al.55.

### Details of food-system emissions by life-cycle stage

#### LULUC

Agricultural land-use emissions involved in food production are those associated with carbon losses due to relevant LULUC. These include emissions from deforestation and from the degradation of organic soils (drainage and fires), which are derived by applying the relevant IPCC guidelines53 and are available in the FAOSTAT Emissions-Land Use domain52. They exclude forest removals in remaining forest land, as these are not typically related to crop and livestock production. While these associations to food systems are those employed in recent literature (for example, Tubiello et al.18, IPCC Special Report on Climate Change and Land11,52 and Rosenzweig et al.12), the main limitations to this approach arise from the assumption, in the absence of more detailed information, that all deforestation is associated with a conversion to agricultural land. However, we know that, globally, about 80% of deforestation is associated with agricultural expansion (IPCC, AR5 WGIII Ch 11)57. Furthermore, it should be noted that the FAOSTAT deforestation estimates are based on information about forest land area and carbon biomass, which countries report to FAO every ten years (on average) via the Forest Resources Assessment (FRA). For this reason, annual deforestation estimates are averages for the periods 1990–1999, 2000–2009 and 2010–2015, and may exhibit discontinuities across successive FRA reporting periods. We have not attempted to smooth out such changes because similar jumps can be observed in country reporting to the UNFCCC58.

Land-use emissions associated with the drainage and burning of organic soils, including peatlands59,60,61, are also included. While these data are well validated for Southeast Asia, there is significant uncertainty in national data outside of this region62,63. This is especially true in tropical central African countries, despite the fact that the agricultural expansion on organic soils is well documented in tropical peatlands around the world. Furthermore, non-CO2 emissions from the burning of biomass in humid tropical forests are included as an additional contribution to deforestation emissions63. Burning of biomass in tropical rainforests and tropical moist forests, as well as emissions from peat fires64, are assumed to be associated with deforestation events for agricultural purposes.

Finally, we are not considering in our estimates carbon removals on agricultural land that may arise from important, specific soil management techniques, for instance, reduced tillage or no tillage. While these actions have important mitigation consequences in relation to food systems (see IPCC Special Report on Climate Change and Land11), neither FAOSTAT nor any other available global dataset currently provides this information at the level needed for this assessment. Considering that the majority of emissions included in this study are from large-scale deforestation in some developing countries, and that most of the existing soil carbon sequestration techniques are currently in place in a small portion of the total cropland area of some developed and emerging economies, we estimate that the impact of including such soil carbon sequestration data—if it existed—would nonetheless be small compared with the results discussed herein.

Uncertainty in land-use emissions data is typically high. Uncertainty of the FAOSTAT emissions estimates was estimated at about 50% for deforestation data58 and over 100% for peatland degradation data65.

### Agriculture

Agricultural emissions contributing to the food system include cultivation of food and non-food crops and livestock production. Here we considered the following emissions from agricultural soils to entirely contribute to the food system (SFS = 1): animal waste as fertilizer, animals in pasture, cultivation of food crops, drainage of organic soils for crop cultivation, CO2 from urea fertilization, limestone and dolomite use, nitrogen-fixing crops, agricultural waste burning, manure management and enteric fermentation51. We quantified the use of fertilizers for non-food crops based on the FAO report66 on the use of fertilizers for different world regions as well as the US Department of Agriculture (USDA) data for the United States (https://www.ers.usda.gov/data-products/fertilizer-use-and-price/). We validated the reported values on the basis of our own calculations of GHG emissions using FAOSTAT commodity balances (http://www.fao.org/faostat/en/#data/BC) and nitrogen content data from Lassaletta et al.67 and Leip et al.68. Fibres are by far the most important non-food crops, with the export of nitrogen in cottonseed more than three times as important as tobacco, the second relevant non-food crop. Therefore, we focused our analysis on fibre crops. FAO quantified that, globally, 4.4% of fertilizers are used for fibre cultivation66. They provided more detailed regional values, reporting the highest use of fertilizers for fibre cultivation in India, Pakistan, Bangladesh and Sri Lanka. In the European Union, the use of fertilizers for non-food–related crops is also rather negligible69. For the United States, we apply a share of non-food use of fertilizers of 3%, based on USDA statistics (https://www.ers.usda.gov/data-products/fertilizer-use-and-price/). Overall, SFSs,c varies (minimum–maximum) between 0.922 and 1 depending on the country and region.

Indirect N2O emissions from NH3 and NOx emitted by all agricultural activities are also included. Emissions from combustion in food-related agricultural and fishing activities56 are entirely included in the food-system GHG emission calculations, while no emissions are allocated to non-food co-products such as straw, fish-oil drugs or pet food.

### Chemicals

Emissions from chemical production are available for chemicals related to plastic production (acrylonitrile, ethane/ethylene, methanol, vinyl chloride, adipic acid, caprolactam, glyoxal, calcium carbide) and to fertilizers70,71,72 (ammonia, SFS = 0.8; urea, SFS = 0.9; nitric acid, SFS = 0.7). For chemicals used for plastics, the same shares of plastic packaging described in the following are used.

### Solvents

Emissions from solvent production and use in pesticides and vegetative oil extraction are entirely attributed to food production (SFS = 1). No other solvents are relevant to the food-system emissions.

#### Processing

The contribution from combustion in the food and tobacco industry (including beverages) is entirely accounted to food-system emissions (SFS = 1), and it is based on IEA data56.

#### Distribution

The distribution stage includes emissions from food packaging, transport and retail.

### Packaging

We estimate emissions from food packaging for iron and steel, paper, aluminium, plastic, and glass containers and bottles.

• Iron and steel

Emissions from iron and steel production are associated with the production of tin and are calculated using country-specific statistics on the amount of iron and steel used for tin mill products, available from the World Steel Association73,74,75 for the whole time series. An average global share of tin mill products used for food packaging is available from the World Steel Association54 (SFSs,c,t minimum–maximum range: 0.0014–0.079).

• Glass

The fraction of glass used for food packaging is estimated based on European76 (with country-specific information) and global77 statistics (SFSs,c,t minimum–maximum range: 0.45–0.62). In addition, emissions from the use of soda-ash for glass production for food containers and bottles78 are included (SFSs,c minimum–maximum range: 0.225–0.31).

• Plastic

To calculate how much plastic is used for food packaging, we first evaluated how much plastic is refined and transformed from oil and gas fuels produced as a global average56,79. Then the fraction of plastic used for packaging for world countries or regions was taken from UNEP80. In the lack of more specific data, we applied a global average fraction of plastic packaging used in the food system81 (SFSs minimum–maximum range: 0.000018–0.00132).

• Paper

FAOSTAT82 reports data on total paper production and paper production for writing purposes. We assume that the remaining fraction is used for packaging. A global average share of paper used for food is applied only to ‘kraft paper’ and ‘other paper and paperboard’ components which are present in the EDGAR database and not strictly used for writing purposes (SFSs,c,t minimum–maximum range: 0.089–0.439).

• Aluminium

The Global Aluminium Flow model83 provides an historic time series (1971–2018) of aluminium production used for packaging (cans and foil) for some countries/regions (the United States, Canada, Europe, Russia, China, Japan, South Africa, Brazil, India and Australia) as well as for the rest of the world. We assume that the aluminium was entirely dedicated to food packaging (SFSs,c,t min-max range: 0.059–0.338). The same shares are also used to estimate food packaging GHG emissions from SF6 use in aluminium foundries.

Emissions from combustion in manufacturing industries related with the food system are computed based on IEA data56. An SFSs,c,t value that is less than unity, corresponding to the share of the food packaging production as a fraction of the total production, is assumed for industries that produce paper (SFSs,c,t minimum–maximum range: 0.089–0.439), plastic (SFSs minimum–maximum range: 0.000018–0.00132), non-metallic minerals (glass production, SFSs,c,t minimum–maximum range: 0.225–0.62) and non-ferrous metals (aluminium, SFSs,c,t minimum–maximum range: 0.059–0.338; iron and steel, SFSs,c,t minimum–maximum range: 0.0014–0.079) since these industrial processes are also important emitters of GHGs. The shares associated with the food packaging industries are applied both to the emissions from processes and combustion.

### Transportation

EDGAR includes detailed emissions from international shipping based on the work of Dalsøren et al.84. Emissions from reefers (in port and at sea) and fishing vessels (in port) have an SFS = 1. For general cargo vessels (carrying packaged items like chemicals, foods, furniture, machinery, motor and military vehicles, footwear, garments and so on) and for dry bulk carriers (carrying coal, grain, ore and other similar products in loose form), shares representing the contribution to the food system85 are applied (SFS minimum–maximum range: 0.022–0.029).

A fraction of road transport emissions associated with heavy duty, light duty and passenger cars is attributed to the food system based on data for Europe from Eurostat86 and from FAO87 for the rest of the world (SFS minimum–maximum range: 0.11–0.15).

A world average share of emissions related to the food system from international and domestic aviation (SFS = 0.005) as well as from inland waterways (SFS = 0.07) and railways (SFS = 0.155) is estimated from FAO data8.

### Retail

The share of electricity used in retail activities (for example, refrigeration, a quota of space heating, air conditioning and lighting for food-related activities) in Europe in the household sector as a fraction of the total consumption in that sector is computed using Eurostat data as summarized by Thomas88.

The electricity used for cooking and refrigeration for different types of retail activities (food sales and food services, retail (other than mall), enclosed and strip malls, warehouse and storage) is retrieved for the United States from EIA data89. For the rest of the world, the electricity consumption for cooking is calculated for China, India and Africa and as a world average based on region specific data90,91. These shares are applied to the electricity used in the retail sector and then applied to the energy emissions.

We also evaluated the amount of different fuels (for example gas, oil, solid) burnt in the retail sector for food-related activities (for example cooking, space and water heating) and calculated the corresponding share to be applied to retail combustion emissions. Eurostat88 and EIA89 data were used for Europe and the United States. For the rest of the world countries, the shares of the food system associated with retail are assumed to be the same as for the household sector based on the Eastern Research Group report91. Overall, SFSs,c,t for the retail sector varies (minimum–maximum) between 0.003 and 0.265.

In accord with the IPCC guidelines92, the main F-gases used for refrigeration are HFC-134a (in EDGAR we already have the fraction used for refrigeration), HFC-32, HFC-143 (entirely used for refrigeration, although a smaller share could be attributed to non-food refrigeration such as of pharmaceutical products) and HFC-125 (which is mainly used for refrigeration and partly for fire protection). Therefore, the SFS for this sector is 1.

#### Consumption

The share of electricity used for food-related activities (for example, cooking, food refrigeration, microwave ovens, coffee makers, toaster and so on) in Europe in the household sector over the total consumption in that sector is computed using the corresponding Eurostat data93. The same information (electricity used for refrigerators, freezers, cooking, microwave ovens and dishwashers) is retrieved for the United States from EIA data94. For the rest of the world, the electricity consumption for cooking is calculated for China, India, world average (Eastern Research Group report91) and Africa (Africa Energy Outlook from IEA90), using region specific data.

We also evaluated the amount of different fuels (for example, gas, oil, solid) burnt in the household sector for cooking relates purposes and calculated the corresponding share to be applied to the residential combustion emissions. Eurostat93 and EIA94 data were used for Europe and the United States, while for the rest of the world countries, individual country shares (for example, for China, India, Bangladesh and Uganda) as well as a world average share are computed to estimate the fraction of household emissions associated with the food system based on the Eastern Research Group report91. Overall, SFSs,c,t for the household sector ranges (minimum–maximum) between 0.003 and 0.265.

#### End of life

The end-of-life stage includes emissions from solid waste disposal and waste water treatment, as discussed in detail below.

### Solid waste disposal

Emissions from solid waste disposal from the food system are related to the incineration (without energy recovery) of biogenic waste, incineration of industrial solid, municipal solid waste, non-specified waste and of sewage sludge, waste disposal on landfills and composting (SFS = 1). The organic fraction of the municipal waste for each world country/region has been extracted from the World Bank What a Waste report95. We assumed that the organic biomass fraction in solid waste is predominantly associated with food systems, while the non-organic fraction is not predominantly associated with food systems. We used the low heating value (LHV) of the different components in solid waste as a proxy for the allocation of GHG emissions from waste incineration of the different waste fractions (for example, plastic, paper, organic, glass, other)95,96. Due to the lack of detailed data, we could not separate the fraction of food-related waste from the fraction of biogenic waste coming from non-food sources such as gardening and landscape maintenance. This assumption represents an overestimation of this component for industrialized countries in particular, while it can be considered rather reliable for developing regions where the garden and landscape maintenance collection of waste is less common. We considered the impact of our assumption in the uncertainty evaluation of this category, which varies from moderate to high (Supplementary Fig. 2). The shares for solid waste incineration are in the range (SFSs,c minimum–maximum) of 0.026–0.85, while SFSs,c for landfills varies between 0.26 and 0.88.

### Waste water

Domestic and industrial waste water almost entirely contribute to food-system emissions. Emissions from domestic waste water are computed separately for rural areas, urban-high-income and urban-low-income countries. The basic activity data, which is the total organically degradable carbon in waste water (TOW), was calculated for rural, urban-high, and urban-low-income populations. The share of rural population within the total population was estimated from the United Nations Department of Economic and Social Affairs (Population Division) data97. For urban populations, we used the information provided by the IPCC98 Supplementary Table 6.5 and data about population in slums from UNHABITAT and World Bank to distinguish between urban-low-income and urban-high-income. Industrial waste water emissions include the contribution from nitrogen-containing effluents (and sludge), alcohol refining, meat and poultry processing and raw sugar refining (SFS = 1). Industrial waste water emissions associated with pulp production are based on the paper used for food packaging with an SFSs,c,t minimum–maximum range of 0.089–0.439.

### Uncertainty analysis

As part of the yearly releases of the updated EDGAR database, comparison with national reporting for main emitters (Europe, the United States and China) is carried out in the context of an internal quality assurance protocol (for CO2 emissions in particular). More recently, EDGAR emission data for agriculture have been toughly assessed and compared against national reporting emissions during a verification exercise103 and for all sectors for CO2 (ref. 104) and for the complete GHGs105. To our knowledge, the calculation of food-system shares has never been done at the level of detail in this Article, so in-depth comparison with independent data was not possible. All shares used, however, were subject of scrutiny and sensitivity analysis (sense making) and were given an added uncertainty due to emissions, as detailed below.

Activity data (AD) and emission factor (EF) uncertainties for anthropogenic activities have been derived from IPCC guidelines53, while LULUC uncertainty is fixed at 50%58. A summary is provided in Supplementary Table 1. Total uncertainty is calculated by the sum of squares for the uncertainty of AD and EF for a given source. The uncertainty of the aggregation levels is calculated by assuming full correlation for fuel type (CO2) and for sector (CH4 and N2O), reflecting the more detailed information provided by IPCC53 on fuel type uncertainty for CO2 (assumptions taken, for example, in Bond et al.106, Bergamaschi et al.107, Petrescu et al.103, Choulga et al.104, and Solazzo et al.105). Moreover, developing countries are given higher uncertainties to account for underrepresentation in the data underlying our estimates53. A further source of uncertainty is the share factors adopted to account for the portion of GHG emissions stemming from the food system only. As described above, the food-related share of emissions from energy, transport, waste and some chemical and industrial processes is not readily available and assumptions have been made, either per country or for the whole world. Supplementary Table 5 reports these shares by sector and the associated uncertainty.

Supplementary Table 1 presents an overview of the sectors contributing to the food system considered in this work, as well as two sectorial aggregations representing the food chain and food-system structure. A confidence level for the assumed shares representing the food-system contribution to each sector is also provided. High confidence (H) means that the shares applied (for example, SFSs,c,t = 1) do not add further uncertainty to the original emissions for that sector. Medium (M) and medium–high (M–H) confidence mean that a rather small (10 to 20%) additional uncertainty is due to the application of the food shares. Low confidence (L) means that the GHG emission uncertainty will be enhanced by up to 100% due to the application of the food shares.

The share of the emissions from the packaging and chemical industries that are devoted to food is highly uncertain because accurate data were not available for this sector. A conservative estimate of 0% (assuming all emissions not related to food) to 100% (all emissions related to food) has been applied. Transportation (road, maritime, air) is also uncertain, because detailed country coverage is not available. Because regional shares were applied worldwide, spatial heterogeneity is not fully captured and an uncertainty of 35% to 75% is applied to account for the lack of representativeness of the share. Solid waste shares are also subject to an additional uncertainty of 5% to 10% due to the methodological assumptions and the LHV proxy discussed above. Supplementary Fig. 3 shows food-system GHG emissions by gas and by sector with the uncertainty estimates from Supplementary Table 1. Supplementary Table 5 reports minimum and maximum uncertainty values for food emission estimates by country and sector.

## Data availability

The data that support the findings of this study are available as Excel spreadsheets alongside the paper. Moreover, they are available on the EDGAR website and can be accessed at the following link: https://edgar.jrc.ec.europa.eu/overview.php?v=EDGAR_FOOD. When citing the EDGAR-FOOD dataset, please specify the following link108https://doi.org/10.6084/m9.figshare.13476666. All figures present in the manuscript are also available in figshare under the same doi as the EDGAR-FOOD dataset. Source data are provided with this paper.

## References

1. 1.

Zurek, M. et al. Assessing sustainable food and nutrition security of the EU food system—an integrated approach. Sustainability https://doi.org/10.3390/su10114271 (2018).

2. 2.

Monforti-Ferrario, F. et al. Energy Use in the EU Food Sector: State of Play and Opportunities for Improvement EUR 27247 EN – 2015 (Publications Office of the European Union, 2015).

3. 3.

Nutrition and Food Systems. A Report by the High Level Panel of Experts on Food Security and Nutrition of the Committee on World Food Security (HELPE, 2017).

4. 4.

Leip, A., Bodirsky, B. L. & Kugelberg, S. The role of nitrogen in achieving sustainable food systems for healthy diets. Glob. Food Secur. https://doi.org/10.1016/j.gfs.2020.100408 (2020).

5. 5.

Tubiello, F. N. & Conchedda, G. The Share of Agriculture in Total GHG Emissions. Global, Regional and Country Trends, 1990–2017 FAOSTAT Analytical Briefs Series (1) (FAO, 2020).

6. 6.

Poore, J. & Nemecek, T. Reducing food’s environmental impacts through producers and consumers. Science 360, 987–992 (2018).

7. 7.

Clune, S., Crossin, E. & Verghese, K. Systematic review of greenhouse gas emissions for different fresh food categories. J. Clean. Prod. 140, 766–783 (2017).

8. 8.

Energy-Smart Food for People and Climate (FAO, 2011).

9. 9.

Bajželj, B. et al. Importance of food-demand management for climate mitigation. Nat. Clim. Change 4, 924–929 (2014).

10. 10.

Vermeulen, S. J., Campbell, B. M. & Ingram, J. S. I. Climate change and food systems. Annu. Rev. Environ. Resour. 37, 195–222 (2012).

11. 11.

Mbow, C. et al. Food Security in Climate Change and Land: an IPCC Special Report on Climate Change, Desertification, Land Degradation, Sustainable Land Management, Food Security, and Greenhouse Gas Fluxes in Terrestrial Ecosystems (IPCC, 2019).

12. 12.

Rosenzweig, C. et al. Climate change responses benefit from a global food system approach. Nat. Food 1, 94–97 (2020).

13. 13.

Beylot, A. et al. Assessing the environmental impacts of EU consumption at macro-scale. J. Clean. Prod. 216, 382–393 (2019).

14. 14.

Sala, S. et al. Consumption and Consumer Footprint: Methodology and Results (Publications Office of the European Union, 2019).

15. 15.

FAOSTAT Agri-Environmental Indicators, Emissions Shares (FAO, 2020); http://www.fao.org/faostat/en/#data/EM

16. 16.

Tubiello, F. N. et al. The contribution of agriculture, forestry and other land use activities to global warming, 1990–2012. Glob. Change Biol. 21, 2655–2660 (2015).

17. 17.

Tubiello, F. N. in Encyclopedia of Food Security and Sustainability (eds Ferranti, P. et al.) 196–205 (Elsevier, 2019).

18. 18.

Wood, R. et al. Growth in environmental footprints and environmental impacts embodied in trade: resource efficiency indicators from EXIOBASE3. J. Indust. Ecol. 22, 553–564 (2018).

19. 19.

Bruckner, M., Fischer, G., Tramberend, S. & Giljum, S. Measuring telecouplings in the global land system: a review and comparative evaluation of land footprint accounting methods. Ecol. Econ. 114, 11–21 (2015).

20. 20.

FAOSTAT 2015 Data (FAO, 2015); http://www.fao.org/faostat/en/#rankings/countries_by_commodity

21. 21.

FAOSTAT Data (FAO, 2019); http://www.fao.org/faostat/en/#data

22. 22.

Kanter, D. R. et al. Nitrogen pollution policy beyond the farm. Nat. Food 1, 27–32 (2020).

23. 23.

Bora, G. C., Nowatzki, J. F. & Roberts, D. C. Energy savings by adopting precision agriculture in rural USA. Energy Sustain. Soc. 2, 22 (2012).

24. 24.

Pelletier, N. et al. Energy intensity of agriculture and food systems. Annu. Rev. Environ. Resour. 36, 223–246 (2011).

25. 25.

Beckman, J., Borchers, A. & Jones, C. A. Agriculture’s Supply and Demand for Energy and Energy Products EIB-112 (US Department of Agriculture, Economic Research Service, 2013).

26. 26.

State of the Art on Energy Efficiency in Agriculture. Country Data on Energy Consumption in Different Agroproduction Sectors in the European Countries (AgrEE, 2012); http://www.acrres.nl/wp-content/uploads/2018/05/AGREE_2.1-State-of-the-Art-of-EE-in-Agr.pdf

27. 27.

Oteros-Rozas, E., Ruiz-Almeida, A., Aguado, M., González, J. A. & Rivera-Ferre, M. G. A social–ecological analysis of the global agrifood system. Proc. Natl Acad. Sci. USA 116, 26465–26473 (2019).

28. 28.

Berners-Lee, M., Kennelly, C., Watson, R. & Hewitt, C. N. Current global food production is sufficient to meet human nutritional needs in 2050 provided there is radical societal adaptation. Elementa https://doi.org/10.1525/elementa.310 (2018).

29. 29.

Vermeulen, S. et al. Climate change, agriculture and food security: a global partnership to link research and action for low-income agricultural producers and consumers. Curr. Opin. Environ. Sustain. 4, 128–133 (2012).

30. 30.

Kreidenweis, U., Lautenbach, S. & Koellner, T. Regional or global? The question of low-emission food sourcing addressed with spatial optimization modelling. Environ. Model. Softw. 82, 128–141 (2016).

31. 31.

Schmitt, E., Dominique, B. & Six, J. Assessing the degree of localness of food value chains. Agroecol. Sustain. Food Syst. 42, 573–598 (2018).

32. 32.

Schmitt, E. et al. Comparing the sustainability of local and global food products in Europe. J. Clean. Prod. 165, 346–359 (2017).

33. 33.

Mundler, P. & Rumpus, L. The energy efficiency of local food systems: a comparison between different modes of distribution. Food Policy 37, 609–615 (2012).

34. 34.

Behfar, A., Yuill, D. & Yu, Y. Supermarket system characteristics and operating faults (RP-1615). Sci. Technol. Built Environ. 24, 1104–1113 (2018).

35. 35.

Bahn, R. A. & Abebe, G. K. Food retail expansion patterns in sub-Saharan Africa and the Middle East and North Africa: institutional and regional perspectives. Agribusiness https://doi.org/10.1002/agr.21634 (2019).

36. 36.

Weatherspoon, D. D. & Reardon, T. The rise of supermarkets in Africa: implications for agrifood systems and the rural poor. Dev. Policy Rev. 21, 333–355 (2003).

37. 37.

Reardon, T., Timmer, C. P. & Minten, B. Supermarket revolution in Asia and emerging development strategies to include small farmers. Proc. Natl Acad. Sci. USA 109, 12332–12337 (2012).

38. 38.

James, S. J. & James, C. The food cold-chain and climate change. Food Res. Int. 43, 1944–1956 (2010).

39. 39.

Hubacek, K. & Feng, K. Comparing apples and oranges: some confusion about using and interpreting physical trade matrices versus multi-regional input–output analysis. Land Use Policy 50, 194–201 (2016).

40. 40.

Reardon, T. et al. Rapid transformation of food systems in developing regions: highlighting the role of agricultural research & innovations. Agric. Syst. 172, 47–59 (2019).

41. 41.

Lapola, D. M. et al. Pervasive transition of the Brazilian land-use system. Nat. Clim. Change 4, 27–35 (2014).

42. 42.

Fanzo, J. From big to small: the significance of smallholder farms in the global food system. Lancet Planet. Health 1, e15–e16 (2017).

43. 43.

Ricciardi, V., Ramankutty, N., Mehrabi, Z., Jarvis, L. & Chookolingo, B. How much of the world’s food do smallholders produce? Glob. Food Secur. 17, 64–72 (2018).

44. 44.

Terwase, I. & Madu, A. The impact of rice production, consumption and importation in Nigeria: the political economy perspectives. Int. J. Sust. Dev. World Policy 3, 90–99 (2014).

45. 45.

Africa Sustainable Livestock 2050 - Country Brief: Ethiopia I7347EN/1/06.17 (FAO, 2017).

46. 46.

Shaddick, G., Thomas, M. L., Mudu, P., Ruggeri, G. & Gumy, S. Half the world’s population are exposed to increasing air pollution. npj Clim. Atmos. Sci. 3, 23 (2020).

47. 47.

Sanz-Cobena, A. et al. Research meetings must be more sustainable. Nat. Food 1, 187–189 (2020).

48. 48.

Springmann, M. et al. Options for keeping the food system within environmental limits. Nature 562, 519–525 (2018).

49. 49.

Aleksandrowicz, L., Green, R., Joy, E. J. M., Smith, P. & Haines, A. The impacts of dietary change on greenhouse gas emissions, land use, water use, and health: a systematic review. PLoS ONE 11, e0165797 (2016).

50. 50.

Willett, W. et al. Food in the Anthropocene: the EAT–Lancet Commission on healthy diets from sustainable food systems. Lancet 393, 447–492 (2019).

51. 51.

Crippa, M. et al. Fossil CO2 and GHG Emissions of All World Countries - 2019 Report EUR 29849 EN (Publications Office of the European Union, 2019); https://doi.org/10.2760/687800

52. 52.

FAOSTAT Land Use Emissions – Land Use, Forest Land (FAO, 2019); http://www.fao.org/faostat/en/#data/GF

53. 53.

IPCC Guidelines for National Greenhouse Gas Inventories (Institute for Global Environmental Strategies, IPCC-TSU NGGIP, IGES, 2006).

54. 54.

Crippa, M. et al. Gridded emissions of air pollutants for the period 1970–2012 within EDGAR v4.3.2. Earth Syst. Sci. Data 10, 1987–2013 (2018).

55. 55.

Janssens-Maenhout, G. et al. EDGAR v4.3.2 Global Atlas of the three major greenhouse gas emissions for the period 1970–2012. Earth Syst. Sci. Data 11, 959–1002 (2019).

56. 56.

Energy Balance Statistics for 1970–2015 (IEA, 2017); http://www.iea.org/

57. 57.

Federici, S., Tubiello, F. N., Salvatore, M., Jacobs, H. & Schmidhuber, J. New estimates of CO2 forest emissions and removals: 1990–2015. For. Ecol. Manage. 352, 89–98 (2015).

58. 58.

Tubiello, F. N. et al. Carbon emissions and removals by forests: new estimates 1990–2020. Earth Syst. Sci. Data Discuss. https://doi.org/10.5194/essd-2020-203 (2020).

59. 59.

FAOSTAT Land Use Emissions – Cropland (FAO, 2020); http://www.fao.org/faostat/en/#data/GC

60. 60.

FAOSTAT Land Use Emissions – Grassland (FAO, 2020); http://www.fao.org/faostat/en/#data/GG

61. 61.

Tubiello, N. F., Biancalani, R., Salvatore, M., Rossi, S. & Conchedda, G. A worldwide assessment of greenhouse gas emissions from drained organic soils. Sustainability https://doi.org/10.3390/su8040371 (2016).

62. 62.

Prosperi, P. et al. New estimates of greenhouse gas emissions from biomass burning and peat fires using MODIS Collection 6 burned areas. Climatic Change 161, 415–432 (2020).

63. 63.

FAOSTAT Land Use Emissions – Burning Biomass (FAO, 2020); http://www.fao.org/faostat/en/#data/GI

64. 64.

Rossi, S. et al. FAOSTAT estimates of greenhouse gas emissions from biomass and peat fires. Climatic Change 135, 699–711 (2016).

65. 65.

Conchedda, G. & Tubiello, F. N. Drainage of organic soils and GHG emissions: Validation with country data. Earth Syst. Sci. Data Discuss. 2020, 1–47 (2020).

66. 66.

Fertilizer use by crop. In FAO Fertiliser and Plant Nutrition Bulletin Ch. 17 (FAO, 2006).

67. 67.

Lassaletta, L. et al. Nitrogen use in the global food system: past trends and future trajectories of agronomic performance, pollution, trade, and dietary demand. Environ. Res. Lett. 11, 095007 (2016).

68. 68.

Leip, A. et al. Impacts of European livestock production: nitrogen, sulphur, phosphorus and greenhouse gas emissions, land-use, water eutrophication and biodiversity. Environ. Res. Lett. 10, 115004 (2015).

69. 69.

The Promotion of Non-Food Crops IP/B/AGRI/ST/2005-02 (European Parliament, 2005); https://www.europarl.europa.eu/meetdocs/2004_2009/documents/dv/studynon-foodcrops_/studynon-foodcrops_%20en.pdf

70. 70.

Glibert, P. M., Harrison, J., Heil, C. & Seitzinger, S. Escalating worldwide use of urea—a global change contributing to coastal eutrophication. Biogeochemistry 77, 441–463 (2006).

71. 71.

Production of Ammonia, Nitric Acid, Urea and N-fertilizer (Environment Agency Austria, 2017).

72. 72.

73. 73.

Steel Statistical Yearbooks 1978 to 1999 (World Steel Assocation, 1999); https://www.worldsteel.org/steel-by-topic/statistics/steel-statistical-yearbook.html

74. 74.

Steel Statistical Yearbooks 2000 to 2009 (World Steel Assocation, 2009); https://www.worldsteel.org/steel-by-topic/statistics/steel-statistical-yearbook.html

75. 75.

Steel Statistical Yearbooks 2010 to 2020 (World Steel Assocation, 2019); https://www.worldsteel.org/steel-by-topic/statistics/steel-statistical-yearbook.html

76. 76.

Analysis of the Industrial Sectors in the European Union. (EU-Merci, 2018); http://www.eumerci-portal.eu/documents/20182/38527/0+-+EU.pdf

77. 77.

Nangini, C. et al. A global dataset of CO2 emissions and ancillary data related to emissions for 343 cities. Sci. Data 6, 180280 (2019).

78. 78.

USGS Soda Ash Statistics and Information https://www.usgs.gov/centers/nmic/soda-ash-statistics-and-information (2016).

79. 79.

British Plastics Federation https://theconversation.com/the-world-of-plastics-in-numbers-100291 (2018).

80. 80.

Ryberg, M. W., Laurent, A. & Hauschild, M. Mapping of Global Plastics Value Chain and Plastics Losses to the Environment (UNEP, 2017) http://wedocs.unep.org/bitstream/handle/20.500.11822/26745/mapping_plastics.pdf?sequence=1&isAllowed=y

81. 81.

Unwrapping the Packaging Industry, Seven Factors for Success (EY, 2013) http://ifca.net.in/pdf/Management-stories-EY-Unwrapping-the-packaging-industry.pdf

82. 82.

Forestry/Forestry Production and Trade till 2016 (FAOSTAT Statistics Division of the Food and Agricultural Organisation of the UN, 2018); http://www.fao.org/faostat/en/#data/FO

83. 83.

Global Material Flow Model (World Aluminum, 2018); http://www.world-aluminium.org/publications/?search=food&year=

84. 84.

Dalsøren, S. B. et al. Update on emissions and environmental impacts from the international fleet of ships: the contribution from major ship types and ports. Atmos. Chem. Phys. 9, 2171–2194 (2009).

85. 85.

Andersen, O. et al. CO2 emissions from the transport of China’s exported goods. Energy Policy 38, 5790–5798 (2010).

86. 86.

87. 87.

Food Wastage Footprint & Climate Change (FAO, 2015); http://www.fao.org/3/A-BB144E.PDF

88. 88.

Thomas, S. Drivers of Recent Energy Consumption Trends Across Sectors in EU28 (Publications Office of the European Union, 2018); https://ec.europa.eu/energy/sites/ener/files/energy_consumption_trends_workshop_report-september_2018.pdf

89. 89.

Commercial Buildings Energy Consumption Survey (CBECS) (US Energy Information Administration, 2018); https://www.eia.gov/consumption/commercial/

90. 90.

Africa Energy Outlook (OECD/IEA, 2014) https://www.iea.org/publications/freepublications/publication/WEO2014_AfricaEnergyOutlook.pdf

91. 91.

Comparative Analysis of Fuels for Cooking: Life Cycle Environmental Impacts and Economic and Social Considerations (Global Alliance for Clean Cookstoves, ERG, 2017); https://www.cleancookingalliance.org/assets-facit/Comparative-Analysis-for-Fuels-FullReport.pdf

92. 92.

2019 Refinement to the 2006 IPCC Guidelines for National Greenhouse Gas Inventories: Volume 3, Industrial Processes and Product Use Ch. 7 (IPCC, 2019); https://www.ipcc-nggip.iges.or.jp/public/2019rf/index.html

93. 93.

EUROSTAT. Energy products used in the residential sector. https://ec.europa.eu/eurostat/statistics-explained/index.php?title=Energy_consumption_in_households#Energy_products_used_in_the_residential_sectorIPCC (2019)

94. 94.

Residential Energy Consumption Survey (RECS) (US Energy Information Administration, 2015); https://www.eia.gov/consumption/residential/data/2015/index.php?view=consumption&src=%E2%80%B9%20Consumption%20%20%20%20%20%20Residential%20Energy%20Consumption%20Survey%20(RECS)-b1#undefined

95. 95.

Hoornweg, D. & Bhada-Tata, P. What a Waste. A Global Review of Solid Waste Management (World Bank, 2012).

96. 96.

Fraunhofer-Gesellschaft zur Förderung der angewandten Forschung e.V. Waste Profiling (Waste 2 Go, 2014).

97. 97.

World Population Prospects: The 2015 Revision (UN Department of Economic and Social Affairs, Population Division, 2015).

98. 98.

Guidelines for National Greenhouse Gas Inventory. Volume 5: Waste (IPPC, 2006); http://www.ipcc-nggip.iges.or.jp/public/2006gl/vol5.html

99. 99.

Global Urban Indicators Database (UNHABITAT, 2016a).

100. 100.

World Atlas of Slum Evolution, 2015 (United Nations Human Settlements Program, UNHABITAT (2016b).

101. 101.

Population Living in Slums (World Bank, 2016) http://data.worldbank.org/indicator/EN.POP.SLUM.UR.ZS

102. 102.

Slum Population as Percentage of Urban, Percentage (UN, 2015); http://mdgs.un.org/unsd/mdg/SeriesDetail.aspx?srid=710&crid=

103. 103.

Petrescu, A. M. R. et al. European anthropogenic AFOLU greenhouse gas emissions: a review and benchmark data. Earth Syst. Sci. Data 12, 961–1001 (2020).

104. 104.

Choulga, M. et al. Global anthropogenic CO2 emissions and uncertainties as prior for Earth system modelling and data assimilation. Earth Syst. Sci. Data Discuss. https://doi.org/10.5194/essd-2020-68 (2020).

105. 105.

Solazzo, E. et al. Uncertainties in the EDGAR emission inventory of greenhouse gases. Preprint at Atmos. Chem. Phys. Discuss. https://doi.org/10.5194/acp-2020-1102 (2020).

106. 106.

Bond, T. C. et al. A technology-based global inventory of black and organic carbon emissions from combustion. J. Geophys. Res. Atmos. https://doi.org/10.1029/2003jd003697 (2004).

107. 107.

Bergamaschi, P. et al. Top-down estimates of European CH4 and N2O emissions based on four different inverse models. Atmos. Chem. Phys. 15, 715–736 (2015).

108. 108.

Crippa, M. et al. EDGAR-FOOD data. figshare https://doi.org/10.6084/m9.figshare.13476666 (2021).

## Acknowledgements

We are grateful to the EDGAR team (M. Crippa, D. Guizzardi, G. Oreggioni, E. Schaaf, M. Muntean, E. Solazzo, F. Pagani) for the work needed to publish the EDGARv5.0 dataset (https://edgar.jrc.ec.europa.eu/overview.php?v=50_GHG). We appreciated the contribution of LULUC data by FAO through its FAOSTAT database (G. Conchedda and F. Tubiello), and the entire manuscript revision by J. Wilson. The views expressed in this publication are those of the author(s) and do not necessarily reflect the views or policies of FAO.

## Author information

### Contributions

M.C. and D.G. designed and developed the EDGAR-FOOD database; E.S., M.C. and F.M.-F. worked on the definition of food-system shares for all GHG emitting categories; A.L. designed the project, revised the methodology and identified the key messages of the manuscript; F.N.T. provided the FAO data and supported the discussion of the LULUC component; all authors helped in drafting the manuscript.

### Corresponding authors

Correspondence to M. Crippa or A. Leip.

## Ethics declarations

### Competing interests

The authors declare no competing interests.

Peer review information Nature Food thanks Tasso Azevedo, Luke Spadavecchia and Berien Elbersen for their contribution to the peer review of this work.

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

## Supplementary information

### Supplementary Information

Supplementary Figs. 1–3.

### Supplementary Data 1

Supplementary Tables 1–8.

## Source data

### Source Data Fig. 1

Statistical source data.

### Source Data Fig. 2

Statistical source data.

### Source Data Fig. 3

Statistical source data.

### Source Data Fig. 4

Statistical source data.

### Source Data Fig. 5

Statistical source data.