We present results from the FAOSTAT emissions shares database, covering emissions
from agri-food systems and their shares to total anthropogenic emissions for
196 countries and 40 territories for the period 1990–2019. We find that in
2019, global agri-food system emissions were 16.5 (95 %; CI range: 11–22)
billion metric tonnes (Gt CO
Agriculture is a significant contributor to climate change as well as one of
the economic sectors most at risk from it. Greenhouse gas (GHG) emissions
generated within the farm gate by crop and livestock production and related
land use change contribute about one-fifth to one-quarter of total emissions
from all human activities, when measured in CO
Significant progress has recently resulted in the development of novel
databases with global coverage of country-level data on agri-food system
emissions (Crippa et al., 2021a, b; Tubiello et al., 2021a). Tubiello et al. (2021a), in particular, provided a mapping of emission categories of the
Intergovernmental Panel on Climate Change (IPCC) – used by countries for
reporting their national GHG inventories (NGHGI) to the United Nations
Framework Convention on Climate Change (UNFCCC) – unto internationally
accepted food and agriculture concepts that are more easily understood by
farmers and planners in countries, including in ministries of agriculture.
By providing a correspondence between IPCC and FAO terminology, we seek to
help countries to more adequately capture important aspects of food and
agriculture activities within existing climate reporting, so that they can
better identify effective climate actions across their agri-food systems
(Fig. 1, adapted from Tubiello et al., 2021a). Firstly, the correspondence
mapping expands the IPCC “agriculture” definition to include, in addition
to non-CO
Mapping of emissions across agri-food systems. Left: IPCC sectors and processes used in national GHG emissions inventories. Right: food and agriculture sectors and categories aligned to FAO's definitions.
We present herein and discuss results from the first agri-food system
emissions database in FAOSTAT. The new database covers, as in previous
versions (Tubiello et al., 2013), agriculture production activities within
the farm gate and associated land use and land use change emissions on
agricultural land. Importantly, it also includes estimates of emissions from
pre- and post-production processes along food supply chains, including
fertilizer manufacturing, energy use within the farm gate, food processing,
domestic and international food transport, retail, packaging, household
consumption, and food system waste disposal. The database provides emissions
data for four main GHG gases/categories (CO
Recent work (Rosenzweig et al., 2021; Tubiello et al., 2021a) helped to characterize agri-food system emissions into three components: (1) farm gate, (2) land use change and (3) pre- and post-production. Emissions estimates from the first two – generated by crop and livestock production activities within the farm gate and by the conversion of natural ecosystems to agriculture, such as deforestation and peatland degradation – are well established (IPCC, 2019). In particular, FAO disseminates annual updates in FAOSTAT (FAO, 2021a, b; Tubiello, 2019). This paper expands the available FAOSTAT data to include estimates of emissions from pre- and post-production processes, including energy use in fertilizer manufacturing, food processing, packaging, transport, retail, household consumption, and waste disposal.
The new FAOSTAT data are provided, for each country, in both IPCC and FAO classifications. Specifically, on the one hand, data can be downloaded using the following IPCC emissions categories: energy; industrial processes and product use (IPPU, henceforth referred to as industry); waste; agriculture; land use, land use change and forestry (LULUCF); and other. The total emissions from IPCC sectors are provided, as well as the portion directly related to agri-food systems. On the other hand, through the IPCC to FAO mapping discussed above and extending previous work (Tubiello et al., 2021a), data can also be downloaded in relevant FAO categories, covering emissions from farm gate, land use change, and pre- and post-production processes (Fig. 1).
The FAOSTAT emissions estimates follow the IPCC (2006) “territorial
approach”; i.e., they are assigned to the countries where they occur,
independently of production or consumption considerations. For example,
CO
FAO regularly disseminates emissions data for 15 sub-domains in relation to the farm gate and land use change components of agri-food system emissions, with published methodologies and results (i.e., Tubiello et al., 2021a). This paper relies in addition on new methods for computing emissions from pre- and post-production processes. Specifically, methods for emissions from energy use in fertilizer manufacturing, food processing, retail, and household consumption as well as refrigeration in retail are presented in Tubiello et al. (2021b), while Karl and Tubiello (2021a, b) presented methods for estimating agri-food system emissions in transport and waste disposal. Finally, emissions from on-farm energy use were developed by Flammini et al. (2022). We refer the interested reader to those original publications for full details, while for completeness we also provide a sufficiently detailed summary of methods and coefficients as the Supplement of this paper.
More generally, a step-wise approach was followed for the estimation of
agri-food system emissions, as follows.
The estimated emissions data expressed in CH
The processes covered herein do not span all processes attributable to agri-food systems. In particular, the scope of this work does not include, by design, upstream GHG emissions in the fuel chain, such as petroleum refining, as well as methane leaks during extraction processes and piping. These are expected to be not negligible if considered. While emissions from such sources can be estimated using a fixed fuel chain coefficient for certain fuel supply chains (see Crippa et al., 2021a), the authors do not consider such sources to be within scope of this work. GHG emissions attributable to electricity generation are included in the scope of this work, which itself excludes upstream GHG emissions in the fuel chain used to generate electricity (Flammini et al., 2022; Tubiello et al., 2021b).
Conversely, emissions of fluorinated gases (F-gases) from household
refrigeration and from climate-controlled transportation were not included
for lack of available country-level data for disaggregated cold chain
elements. However, one estimate suggests that the majority (over 60 %) of
global food-related F-gas emissions occur in the retail stage, which is
accounted for here in this work (International Institute of Refrigeration,
2021). Emissions from pesticide manufacturing were also not included due to
the paucity of information and methodologies for their estimation at the country
level, in contrast to advanced work in fertilizer manufacturing (Brentrup
et al., 2016, 2018; IFS, 2019). Bellarby et al. (2008)
estimated global emissions from pesticides manufacturing to be roughly 72 (range: 3–140) Mt CO2eq yr
Uncertainties in FAOSTAT farm gate and land use change emissions estimates have been characterized elsewhere and computed in line with IPCC (2000, 2006) guidelines as ranging 30 %–70 % across component processes. For the purpose of this analysis, we assigned uncertainties of 30 % and 50 % respectively to the farm gate and land use change components of the FAOSTAT agri-food system emissions, in line with previous work (i.e., Tubiello et al., 2013, 2021b). The uncertainties in the estimates of pre- and post-production activities described herein are by contrast less documented. On the one hand, uncertainties in underlying energy activity data and emissions factors are typically lower than for the other two components, ranging 5 %–20 % (Flammini et al., 2022). On the other hand, the relative novelty in estimating food system shares for a range of activity data across many processes makes our estimates more uncertain, with heavy reliance on literature results from a subset of countries and regions that are necessarily extended to the rest of the world (Karl and Tubiello, 2021a). For this reason, we assigned an overall uncertainty of 30 % to the pre- and post-production component. This is higher than the uncertainty of the underlying energy processes but more in line with values used in recent work (Crippa et al., 2021a). As shown below, considering a roughly equal, one-third contribution of the three components and their assigned uncertainties, an overall uncertainty of 40 % was estimated for the agri-food system emissions totals, applicable to countries and regional aggregates.
The above uncertainties are meant only as first rough estimates, useful to determine tentative 95 % confidence intervals for the overall agri-food system component of FAOSTAT emissions. Significantly more research is needed for further refinements in future studies, in particular on better characterizing sub-regional and regional activity data and emissions coefficients, given the diversity in agri-food system typology and their dependence on physical geography and national socio-economic drivers. These limitations nonetheless reflect the paucity of activity data available to describe agri-food system components and their trends, globally and regionally. While knowledge and data exist for regions and countries such as the EU, USA China and India, much remains to be done in terms of regional and country-specific coverage.
Work towards estimating agri-food system emissions at the country level can be advanced in several ways. The present approach could be expanded on by including other country- and region-specific studies that estimate trends in energy consumption across a range of similar activities as proxies – regardless of whether or not they are distinctly related to food. Furthermore, other data sources could help explain and estimate variations in agri-food systems between countries, such as GDP per capita, urbanization levels, proxies for infrastructure and industrial development, and geographic and climate considerations. The development of a methodology to estimate emissions from pesticides could be explored, as it would help complement the understanding of emissions associated with chemical use in agriculture, in addition to fertilizers. Emissions from machinery manufacturing and from upstream GHG emissions in the fuel chain could also be added to further refine the analysis. This work could be further expanded by focusing on specific food commodities – requiring an additional focus on international trade and on supply and demand patterns (Dalin and Rodríguez-Iturbe, 2016). Such analysis would ultimately enable consumers to understand the full carbon footprint of particular commodities across global supply chains, which can facilitate GHG mitigation actions taken at the consumer level (Poore and Nemecek, 2018). Furthermore, it would be also useful to further investigate the increasing role of bioenergy and renewables as important mitigation opportunities in the food sector (Clark et al., 2020, Monforti et al., 2015; Pablo-Romero et al., 2017; Wang, 2014).
The FAOSTAT dataset considered in this study estimates in 2019 total
anthropogenic emissions at 52 Gt CO
GHG emissions (Mt CO
World total GHG emissions from agri-food systems, 1990–2019. Color bars show contributions by emissions within the farm gate (yellow); land use change (green) and pre- and post-production along food supply chains (blue). Source: FAOSTAT (FAO, 2021a). Also shown are emissions per capita (authors' own calculations).
Emissions from within the farm gate and those due to related land use
processes, including details of their sub-components, have been discussed in
Tubiello et al. (2021a) and are regularly presented within FAOSTAT
statistical briefs (e.g., FAO, 2020a, 2021b). Here we provide a detailed
discussion of the components of agri-food system emissions from pre- and
post-production activities along supply chains and their relative
contribution to the food system totals (Fig. 3). Considering that the
uncertainties used above are rough estimates, we will not report
uncertainties in the following analysis. Our data show that in 2019
emissions from deforestation were the single largest emission component of
agri-food systems, at 3.1 Gt CO
World total 2019 GHG emission from agri-food systems, showing contributions on agricultural land (left panel) and from pre- and post-production along food supply chains (right panel). Net removals on forest land are also shown, for completeness. The sum of emissions from agricultural land and forest land correspond to the IPCC AFOLU category. Source: FAOSTAT (FAO, 2021a).
Finally, while emissions from agri-food systems increased globally by
16 % between 1990 and 2019, their share in total emissions decreased,
from 40 % to 31 %, as did the per capita emissions, from 2.7 to
2.1 t CO
Our results indicate significant regional variation in terms of the
composition of agri-food system emissions by component (Fig. 4).
Specifically, in terms of total agri-food system emissions in 2019, Asia
had the largest contribution, at 7 Gt CO
Total GHG emission from agri-food systems by FAO regions, 2019. Color bars show contributions by emissions within the farm gate (yellow), land use change (green), and pre- and post-production along food supply chains (blue). Source: FAOSTAT (FAO, 2021a).
Regional GHG emissions (Gt CO
The data show which pre- and post-production process was most important by
region (Table 2). In 2019, food household consumption was the dominant
process outside of agricultural land emissions in Asia (0.9 Gt CO
Our estimates show a marked variation among countries in terms of total
emissions as well as the composition of contributions across farm gate, land
use change, and pre- and post-processing components (Fig. 5). China had the
most emissions (1.9 Gt CO
Total GHG emission from agri-food systems by country, top 10 emitters, 2019. Color bars show contributions by emissions within the farm gate (yellow), land use change (green), and pre- and post-production along food supply chains (blue). Source: FAOSTAT (FAO, 2021a).
Top 10 country GHG emissions (Mt CO
The overall assessment of total agri-food system emissions found in this work confirms recent previous findings by the IPCC (2019) and Crippa et al. (2021a, b). With regards to pre- and post-production, the FAOSTAT estimates were consistent (Table 4) with previous findings (i.e., Crippa et al., 2021a, b; Vermuelen et al., 2012; Poore and Nemecek, 2018). In particular, emissions estimates for food transport, processing, waste and retail were consistent with EDGAR-FOOD (Karl and Tubiello, 2021b), and estimates for fertilizer manufacturing were in line with previous work by Vermeulen et al. (2012). Conversely, FAOSTAT estimates were higher than EDGAR-FOOD for household consumption and lower for food packaging, with the latter possibly linked to FAOSTAT estimates excluding indirect emissions from fuel supply chains, which were instead included in previously published estimates. Finally, our estimates of F-gas emissions from retail agreed well with those published in EDGAR-FOOD.
Overview of pre- and post-food production GHG emission estimates
from selected studies, Gt CO
The most important disagreement with previous work was observed in relation
to household consumption emissions. FAOSTAT estimates in this work,
1.2 Gt CO
One notable trend over the 30-year period since 1990 is the increasingly
important role of food-related emissions generated outside of agricultural
land, in pre- and post-production processes along food supply chains, at all
global, regional and national scales. Our data show that by 2019, pre- and
post-production processes had overtaken farm gate processes to become the
largest GHG component of agri-food system emissions in Annex I parties (2.2 Gt CO
Importantly, the FAOSTAT database presented here allows for an estimation of
the percentage share contribution of food system emissions in total
anthropogenic emissions, by country as well as at regional and global
levels, over the period 1990–2019. A number of important issues can be
highlighted to this end (Table 5 and Fig. 6). First, in terms of CO
Regional GHG emissions (Mt CO
Top panel: agri-food system emissions (Gt CO
An analysis on agri-food system impacts on total GHG emissions would not be
complete without a focus on component gases in addition to quantities
expressed in CO
World total and regional GHG agri-food system emissions shares
(%), 1990–2019, by single gas and CO
Finally, the data highlight a very large increase in agri-food system contributions of F-gas emissions, which went from near zero in 1990 to more than one-quarter of the world total in 2019 – with larger contributions in many regions. Such a marked increase is consistent with the growth in use of hydrofluorocarbons (HFCs) as refrigerants in the food retail and other sectors, following the banning of chlorofluorocarbons (CFCs) in 1990 (Fang et al., 2018; Hart et al., 2020; International Institute of Refrigeration, 2021; Tubiello et al., 2021b)). Our findings are furthermore consistent with the strong growth in F-gas emissions reported in recent studies (Minx et al., 2021; Park et al., 2021).
An important aspect of the dataset presented in this study is its provision of information mapped across IPCC and FAO categories alike. Specific IPCC sectors include agriculture and land use, land use change and forestry (LULUCF). The IPCC further considers the agriculture, forestry and other land use (AFOLU). While countries report their agriculture and food emissions to the UNFCCC within national GHG inventories, our findings highlight the importance of expanding that reporting to a fuller agri-food systems view, one that properly weights the contribution of food to the global economy. Indeed, our results show that agri-food system emissions in 2019 were one-third of total anthropogenic emissions. This important picture does not emerge from NGHGI reporting aligned to IPCC categories, according to which, for instance, LULUCF and AFOLU emissions contributed respectively 4 % and 15 % of the total.
The GHG emission data presented herein cover the period 1990–2019 at the
country level, with regional and global aggregates. They are available as
open data, with DOI
This paper provided details of a new FAOSTAT database on GHG emissions along the entire agri-food systems chain, including crop and livestock production processes on the farm, land use change activities from the conversion of natural ecosystems to agricultural land, and processes along food supply chains, from input manufacturing to food processing, transport and retail, including household consumption and waste disposal.
The data are provided in open-access mode to users worldwide and are available by country over the time period 1990–2019, with plans for annual updates. The major trends identified in this work help locate GHG emissions hotspots in agri-food systems at the country, regional and global level. This can inform the process of designing effective mitigation actions in food and agriculture. This work adds to knowledge on GHG emissions from agriculture and land use – generally well established in the literature – by adding critical information on emissions from a range of pre- and post-production processes. The new data highlight the increasingly important role that pre- and post-production processes along supply chains play in the overall GHG footprint of agri-food systems, globally and in most countries, providing new insights into food and agriculture development trends and future mitigation options.
The granularity of the dataset allows us, for the first time, to highlight specific processes of importance in specific countries or groups of countries with similar characteristics. The relevance of the information being provided cuts across several national and international priorities, specifically those aiming at achieving more productive and sustainable food systems, including in relation to climate change. To this end, the work presented herein completes a mapping of IPCC categories, used by countries for reporting to the climate convention, to food and agriculture categories that are more readily understandable by farmers and ministries of agriculture in countries. This helps better identify agri-food system entry points within existing and future national determined contributions. Finally, the methodological work underlying these efforts complements and extends recent pioneering efforts by FAO and other groups in characterizing technical coefficients to enable quantifying the weight of agri-food systems within countries' emissions profiles. The next steps in such efforts would need the involvement of interested national and international experts in compiling a first set of coefficients for agri-food systems as a practical agri-food systems annex to the existing guidelines of the Intergovernmental Panel on Climate Change, providing guidance to countries on how to better characterize food and agriculture emissions within their national GHG inventories.
The supplement related to this article is available online at:
FNT and CR conceived the ideas and together with GC and KK designed the methodology. LRS, AF and RQ provided pre- and post-production data. NW provided forest data. GOL, KK, XP, SYQ, HHH, RQ, LRS, AF and JG compiled and analyzed data. FNT led the writing of the manuscript with input from GC, KK, EMC and CR. PB commented on an earlier version of the manuscript. All authors contributed critically to the drafts and gave final approval for the publication. The views expressed in this paper are the authors' only and do not necessarily reflect those of FAO, UNSD and IEA.
At least one of the (co-)authors is a member of the editorial board of
The views expressed in this paper are the authors' only and do not necessarily reflect those of FAO, UNSD, UNIDO and IEA. Publisher’s note: Copernicus Publications remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
FAOSTAT is supported by FAO's member countries. We acknowledge the efforts of national experts who provide the statistics on food and agriculture, as well as statistics on energy use, that are the basis of this effort. All authors contributed critically to the drafts and gave final approval for the publication. We are grateful for overall support by the Food Climate Partnership at Columbia University.
This paper was edited by David Carlson and reviewed by two anonymous referees.