Decline in carbon emission intensity of global agriculture has stagnated recently

Significance

The agrifood system is responsible for 1/3 of global anthropogenic greenhouse gas (GHG) emissions. Yet, to fulfill the global demand for food in 2050, it needs to expand by 50%. This requires a sharp decrease in the agrifood system’s GHG emission intensity (GHG_i). There was a two-thirds reduction in the GHG_i from 1961 to 2019; however, in more recent years, we have observed stagnation or even an increase in emission intensity in some countries. This change in trend suggests that incremental improvements alone are insufficient to continue reducing the sector’s GHG emissions. Instead, concerted efforts and innovative strategies are required to address these challenges and ensure a more sustainable trajectory for the agrifood system’s contribution to food production and GHG emissions.

Abstract

Using global data for around 180 countries and territories and 170 food/feed types primarily derived from FAOSTAT, we have systematically analyzed the changes in greenhouse gas (GHG) emission intensity (GHG_i) (kg CO_2eq per kg protein production) over the past six decades. We found that, with large spatial heterogeneity, emission intensity decreased by nearly two-thirds from 1961 to 2019, predominantly in the earlier years due to agronomic improvement in productivity. However, in the most recent decade, emission intensity has become stagnant, and in a few countries even showed an increase, due to the rapid increase in livestock production and land use changes. The trade of final produced protein between countries has potentially reduced the global GHG_i, especially for countries that are net importers with high GHG_i, such as many in Africa and South Asia. Overall, a continuous decline of emission intensity in the future relies on countries with higher emission intensity to increase agricultural productivity and minimize land use changes. Countries with lower emission intensity should reduce livestock production and increase the free trade of agricultural products and improve the trade optimality.

Globally, greenhouse gas (GHG) emissions from agrifood system and land use and land use change were estimated to be 17 Gt CO₂ equivalent (CO₂eq) y⁻¹ in 2022 (1). This corresponds to nearly one third of global anthropogenic GHG emissions (2). Food demand is projected to further increase by 35 to 56% between 2010 and 2050, considering the increasing requirements from a larger and wealthier population (3). Expanding the food system without decreasing its environmental impacts jeopardizes global efforts to limit temperature increases within 1.5 °C above preindustrial level (4, 5). Resulting from the need of climate action and the pursuit of food security, the emission intensity from agriculture and associated activities must decrease in the future, either through reducing the total GHG emissions or by improving the agricultural productivity.

Previous studies have either focused on a specific element of agricultural production–related GHG emissions, e.g., those from land use (6), cropland (7), grassland (8), or livestock production (9), or they have used model simulations to explore measures for reducing GHG emissions from agriculture and associated activities (2, 10). Recent studies have highlighted the possibilities of reducing GHG emissions without diminishing food security, which means a significant reduction of GHG emission intensity (GHG_i) (11–13). Such conceived improvements also characterize scenarios of future developments (14, 15). However, it is hard to judge how realistic such a proposition is when comparing to the trend of actual changes in history.

Several studies have analyzed the GHG_i for particular products at either global or regional level, such as milk and beef (16–19). Additionally, research has assessed the variability of GHG_i across different regions for croplands, establishing a global mean GHG_i for cropland at 0.16 g CO₂eq kcal⁻¹ (7). Historical trends in GHG_i for certain products have been examined at the farm gate level by previous studies (20, 21). Nonetheless, a comprehensive analysis of the GHG_i for the whole global agricultural system is still missing and their driving forces are still unclear. This is especially relevant considering the current rapid transition of global agriculture production resulting from increased livestock production and potentially expanded trade of agricultural products (22, 23).

Trade in agricultural products has greatly increased in the past few decades due to globalization and improved international transportation systems. Currently, around 25% of agricultural products are designated for export globally, with large variations at the country level, for example, up to 60% in Brazil and 35% in the United States (1). Livestock production consumed around 40% of crops (24), and is responsible for almost two-thirds of emissions from the agrifood system at the global scale (9), again with large variations between countries (25). All of these may lead to substantial differences of GHG_i between net importing and exporting countries or between countries with different levels of livestock production. Despite these variations, there is no systematic analysis available yet. In addition, there might be large differences in determining GHG_i depending on whether exported products and feed products are excluded or not.

To systematically understand how GHG_i varies and changes among countries with different levels of livestock production and contributions to trade, we need to distinguish GHG_i at three different levels, namely: i) primary product emission intensity (GHG_i-primary), which includes all emissions associated with the production of primary protein domestically (both in animal and crop products); ii) semifinal product emission intensity (GHG_i-semifinal), which excludes locally produced animal feed protein (to preventing double-counting of locally produced feed protein that reappears in animal products within a country), and iii) final product emission intensity (GHG_i-final), which excludes all animal feed protein (to exclude the exported animal-feed protein in addition to the exclusion of semifinal products).

In addition, the trade of agricultural products may substantially impact global GHG_i, due to variations in GHG_i between importing and exporting countries. For example, exporting agricultural products from lower GHG_i countries to higher GHG_i countries could avoid unnecessary GHG emissions at a global scale. It has been found that trade has potentially improved global crop productivity and enhanced partial factor productivity of nitrogen and phosphorus fertilizer (26, 27). However, it is still unclear whether trade of agricultural products has contributed to reduction of GHG_i at the global level, especially when GHG_i is inversely related to productivity. Here, we applied a cumulative GHG_i distribution curve, alongside the trade concepts of optimality and functionality to illustrate how trade impacts the agrifood system–related GHG emissions at a global scale (26).

Results and Discussion

GHG Emissions and Categorization of Countries Into Different Groups.

GHG emissions included those from farming activities such as crop production, livestock production as well as energy use in agriculture, and those from land use changes such as net forestation, savanna fires, and drained organic soils. All GHG emissions data were sourced from the “Climate Change: Agrifood Systems Emissions” of FAOSTAT database, with related data sources listed in SI Appendix, Table S1. The GHG_i was defined as GHG emissions (CO₂ equivalent) per kg of protein production, and evaluated at three different levels of protein production.

In this study, we determine how GHG_i was impacted by different levels of trade and livestock production, considering whether the livestock production intensity was below or above the world average for livestock protein production per capita for the period of 1961 to 2019, and whether a country was a net importer or exporter of agricultural proteins. All countries were categorized into four groups, namely net exporters of agricultural products with high intensity of livestock production (ExH) or low intensity of livestock production (ExL), and net importers of agricultural products with high intensity of livestock production (ImH) or low intensity of livestock production (ImL) (Fig. 1).

Fig. 1.

Categorization of countries and territories into four main groups. Note: Countries and territories have been categorized based on two indictors, i.e., net protein trade and intensity of livestock production. If animal protein production per capita in a country was larger than global average from 1961 to 2019, this country was identified as a high-intensity livestock production country (i.e., High intensity noted in the figure). If not, the country was identified as a low intensity livestock production country (i.e., Low intensity noted in the figure). If a country was a net importer of protein embodied in agricultural products for more than 70% of the selected years (from 1961 to 2019), the country was classified as an importing country. If not, the country was classified as an exporting country. “Export & High livestock” refers to net protein exporting countries with high-intensity livestock production, which includes 18 countries. “Export & Low intensity” refers to net protein exporting countries with low intensity livestock production, which includes 21 countries. “Import & High intensity” refers to net protein-importing countries with high-intensity livestock production, which includes 33 countries. “Import & Low intensity” refers to net protein-importing countries with low intensity livestock production, which includes 85 countries.

Agricultural GHG Emissions and Protein Production.

Increase of GHG emissions and protein production varied between different groups of countries (Fig. 2). Globally, increases in GHG emissions occurred mostly from countries that import most of their protein and have low intensity of livestock production per capita (ImL) from 1961 to 2019 (Fig. 2A), an effect that is strongly driven by land use change in Africa and Southeast Asia where natural forest was converted for agricultural production (Fig. 2C) (6). Primary protein production increased more rapidly in exporting countries with high intensity of livestock production (ExH) (around threefold increase), and relatively slowly in importing countries with high intensity of livestock production (ImH) (less than 100% increase) during 1961 to 2019 (Fig. 3A). The final protein production increased from 0.11 to 0.37 Gt y⁻¹ (Fig. 3B) during 1961 to 2019, respectively. Final protein production was more rapidly increased in ExH and ImL countries, and slowly increased in ImH countries, similar to that of primary protein production (Fig. 3B).

Fig. 2.

Changes of GHG emissions and final protein production. (A and B) Changes in total GHG emissions from agriculture and associated activities and final protein production in the country groups. (C) Changes in cumulative GHG emissions from different groups of countries. Note: CHN, China; DEU, Germany; IND, India; USA, United States of America; BRA, Brazil; ARG, Argentina.

Fig. 3.

Trend of protein production from 1961 to 2019. Results are shown for four groups of countries and specified for six important countries (ARG: Argentina; BRA: Brazil; CHN: China; DEU: Germany; IND: India; USA: United States of America). (A) primary protein production; (B) final protein production; (C) locally produced feed protein only; and (D) exported feed protein only. Note: For full name of abbreviations, see Fig. 2.

Locally produced crop feed protein increased from 0.04 Gt to 0.17 Gt y⁻¹, while exported feed protein grew from <0.01 Gt y⁻¹ to 0.08 Gt y⁻¹ between 1961 and 2019 (Fig. 3 C and D). ExH countries accounted for >90% of total exported protein used as animal feed during 1961 to 2019 (Fig. 3D). Feed protein production for local use was uneven increased between different groups of countries, but still the rate of increase was high in ImL countries, such as China (Fig. 3C). Here, feed refers only to the main products from crop production, excluding crop residues and grass.

Change of Emission Intensity at the Global Level Over Time.

The GHG_i of primary products (GHG_i-primary) decreased by nearly two-thirds globally from 60 kg CO_2eq kg⁻¹ protein (with a range of 33 to 75 kg CO_2eq kg⁻¹ protein, 95% CI and the same for the following) in 1961 to 19 kg CO_2eq kg⁻¹ protein (with a range of 11 to 26 kg CO_2eq kg⁻¹ protein) in 2019 (Fig. 4A and SI Appendix, Fig. S1). We found that the GHG_i of final products (GHG_i-final) decreased in a similar way to GHG_i-primary at the global scale, decreasing by around two-thirds from 84 kg CO_2eq kg⁻¹ protein (with a range of 46 to 105 kg CO_2eq kg⁻¹ protein) to 32 kg CO_2eq kg⁻¹ protein (with a range of 18 to 23) kg CO_2eq kg⁻¹ protein in the past six decades (Fig. 4A and SI Appendix, Fig S1 K–O). Similar trends and rates of decline have been observed among three levels of protein production for different groups of countries, proving this phenomenon to be pervasive (Figs. 4 and 5). This phenomenon indicates that the rapid increase in livestock production has occurred in parallel with much stronger intensity improvements which—among other things—has also reduced agriculture’s GHG_i (Fig. 5B).

Fig. 4.

GHG_i of primary, semifinal, and final product. (A–E) GHG_i for global average and different groups of countries from 1961 to 2019. (F–J) Rates of annual GHG_i decline during three periods, globally and for different groups of countries.

Fig. 5.

Relative change of GHG_i between 1961 and 2019 for each group of countries for final protein production (A) and variation of GHG_i along with the variation of animal protein production per capita (B).

In addition, similar decreasing GHG_i trends for primary production were noted, based on different databases of GHG emissions (i.e. EDGAR, from 29 to 16 kg CO₂eq kg⁻¹ protein) in each group of countries between 1990 and 2019 (SI Appendix, Fig. S2 A–E). When protein used as feed derived from pasture, residues, food losses and waste were included in primary production, GHG_i also showed a decreasing trend then stagnate from 1961 to 2019 (SI Appendix, Fig. S2 F–J), which proved that historical variation of GHG_i in the present study is robust.

The decline in GHG_i has largely been slowing down for all three levels of protein production in the most recent decade (Fig. 4 A–J). The decline rate of GHG_i-final, which is defined as the average decrease in GHG_i per year, decreased from 1.7 to 2.2 kg CO_2eq kg⁻¹ protein per year from 1961 to 1979, down to 0.49 to 0.82 kg CO_2eq kg⁻¹ protein per year between 1980 and 2006, and then further down to about zero during 2007 to 2019 (Fig. 4F). This differentiation into three periods was jointly determined by analyzing the annual changes of GHG_i, total protein production, and impacts of trade on mitigating or increasing GHG emissions (which will be introduced in a later section) (Fig. 6).

Fig. 6.

Impacts of agriculture products trade on the GHG emissions. (A) Yearly based global protein trade embodied in agricultural products, GHG emissions intensity, and virtual reduction of GHG emissions through trade in terms of final (A) protein production from 1961 to 2019. (B) Total GHG emissions reduction or increases, GHG emissions reduction or increases per year and average GHG emissions reduction per kilogram of traded final (B) protein during each period. Note: Reduction is GHG emission reduction via global protein trade. Trade is the total amount of protein traded. Emission intensity is average GHG_i. Positive value means decrease of GHG emissions while negative value means increase of GHG emissions via global trade.

Spatial Differences in the Changes of Emission Intensity.

Different groups of countries exhibited different changes in GHG_i between different periods. There was a steady and consistent decline in GHG_i-final for ExH countries, representing by countries like Brazil, Australia, Canada, Türkiye, and the United States (SI Appendix, Fig. S3), across all three periods, with annual decline rates similar to the global average level from 1961 to 2019 (Fig. 4 F and G). In contrast, there were relatively small declines in GHG_i-final in ImH countries (Fig. 4 D and I), represented by European countries such as Germany, Denmark, and the Netherlands (SI Appendix, Fig. S5). However, there was a decade of GHG_i-final increases for ExL and ImL countries between 2007 and 2019, following sharp decreases in the two preceding periods (Fig. 4). This trend was particularly notable in Zambia, Tanzania, Philippines, and Thailand (SI Appendix, Figs. S4 and S6). The annual GHG_i-final increased by 0.44 to 0.65 kg CO_2eq kg⁻¹ protein and 0.42 to 0.77 CO_2eq kg⁻¹ protein for ExL and ImL countries from 2007 to 2019 (Fig. 4 H and J). In consequence, their emission intensities nearly reverted to the levels observed in 1990 (Fig. 4 C and E). The average GHG_i-final was similar between ExL and ImL countries, at around 40 CO_2eq kg⁻¹ protein, which was around two times of that in found ExH and ImH countries (SI Appendix, Fig. S7).

We also looked into trends on a continental level (SI Appendix, Figs. S7 and S8). There was a steady and consistent decline in GHG_i at all three levels of protein production in South and North America, and Oceania between 1961 and 2006. However, GHG_i stagnated in these continents between 2007 and 2019 (SI Appendix, Fig. S8). Overall, GHG_i decreased by around 80-90% in South and North American, and Oceanian from 1961 to 2019. GHG_i slightly increased in the most recent decade in Africa, Asia, and Europe countries, despite sharp decreases in the previous two periods (SI Appendix, Fig. S8). GHG_i decreased by around 50 to 60% in Africa, Asia, and Europe in the past six decades, which is relatively small compared with other continents (SI Appendix, Fig. S8). In addition, the differences in GHG_i between continents in the most recent periods were larger than those between different groups of countries. For example, the GHG_i-final of South American, Oceania, and Africa were generally 6 to 10 times of those in North America in the most recent decades (SI Appendix, Fig. S7). The differences in GHG_i were even larger at the country level, with the highest values observed in developing countries such as Africa countries, Mongolia, Peru, and Indonesia in 2019 (SI Appendix, Fig. S9). There were countries with a much stronger increase in GHG_i in recent decades, such as New Zealand, Czech-Slovak, Poland, Columbia, Iceland, Philippines, Thailand, Laos, Syria, and Kenya (SI Appendix, Figs. S3–S6).

In addition, there were larger variations of GHG_i reduction at the country level. Our analysis showed a reduction of GHG_i-primary by more than 200 kg CO_2eq kg⁻¹ protein in Brazil, whereas the reductions were less than 50 kg CO_2eq kg⁻¹ protein in the United States, China, and India (SI Appendix, Fig. S10 A–S). On average, GHG_i-primary decreased more in net exporting and high-intensity livestock production (ExH) than in other groups (SI Appendix, Fig. S10G). The relative differences in GHG_i-final and GHG_i-primary became larger in recent years. The difference was about 29% in 1961 to 1963, while it had increased to 38% during 2017 to 2019 in the 2010s (Fig. 4A). Large variations were also observed between countries. GHG_i-final was over 80% larger than that of GHG_i-primary for the United States, Canada, and Brazil, and less than 10% larger for many African countries (SI Appendix, Fig. S11 averages the period 2017 to 2019 to increase data stability), owing to large differences in crop production structure, livestock number and production structure, and domestic animal-feed availability between countries (SI Appendix, Fig. S12).

Key Driving Forces of the Change in Emission Intensity.

GHG_i is an outcome of protein production and related GHG emissions, so factors that impact on the production and emissions will inevitably affect GHG_i. Here, we mainly focus on productivity, land use change, and livestock production levels, due to their much stronger impacts on GHG_i than other factors, such as the contribution of rice cultivation area to total harvest area and the proportion of oil crops to total crop yield, as well as the economic development level (Table 1). This study highlights the complex interactions among agricultural productivity, protein sources, changes in land use, and their combined impact on GHG_i during different periods (SI Appendix, Tables S2–S4). If the effects of agricultural productivity and livestock production have been accounted for using country-specific panel data, land use change showed a strong positive correlation with the changes in GHG_i-final from 1961 to 1979. The rapid decline in GHG_i-final was attributed to the combined effects of productivity and land use change (SI Appendix, Table S2), during which productivity rapidly increased while land use change-related GHG emissions intensity decreased (SI Appendix, Fig. S13). Additionally, the stabilized animal-sourced protein ratio may have contributed to the decline in GHG_i-final in the period of 1961 to 1979 (SI Appendix, Table S2 and Fig. S13). Similar trends were found during the period 1980 to 2006 (SI Appendix, Table S3 and Fig. S13).

Table 1.

Regression results between protein production, total GHG emissions, and GHG_i of final protein production at global level from 1961 to 2019

	Ln production	Ln GHGs total	Ln GHGi
Ln productivity	0.995***	0.233***	−0.763***
	(0.00667)	(0.0277)	(0.0270)
Ln animal ratio	−0.00492	−0.155***	−0.151***
	(0.00639)	(0.0265)	(0.0259)
Ln LUCcrop ratio	−0.0297***	−0.674***	−0.645***
	(0.00270)	(0.0112)	(0.0110)
Ln GDP per capita	0.0389***	−0.00431	−0.0432**
	(0.00329)	(0.0136)	(0.0133)
Ln rice ratio	0.00973***	0.00365	−0.00608
	(0.00273)	(0.0113)	(0.0110)
Ln ruminant ratio	−0.0982***	0.169***	0.267***
	(0.00866)	(0.0359)	(0.0351)
Ln cereal&pulses&oilcrops ratio	0.102***(0.0116)	0.0754(0.0483)	−0.0264(0.0472)
_cons	1.986***	11.51***	9.522***
	(0.0733)	(0.304)	(0.297)
N	5,214	5,214	5,214
F	10331.3	570.7	854.8
r2_w	0.934	0.439	0.540

Note: Productivity represents the amount of protein production per hectare of harvest cropland area, in kg protein per ha (corresponded to each protein production level); Animal ratio refers to the amount of animal-sourced protein to total protein production; LUC_crop ratio indicates GHG emissions from land use change from forest to crop land per agricultural land area to total GHG emission, in kg CO₂eq per ha; GDP indicates gross domestic product per capita, in current US$ per capita; Rice ratio refers to the proportion of rice harvested area to total harvested area; Ruminant ratio refers to the percentage of protein from ruminant animal to total animal protein production; Cereal&pulses&oilcrops ratio refers to the proportion of harvest area of pulses & cereals & oilcrops to total harvested area.

SE in parentheses. *P < 0.1, **P < 0.05, ***P < 0.01.

In the most recent decades, the most influential factor impacting GHG_i-final has been protein productivity, which has shown stronger effects on GHG_i-final than other driving factors (SI Appendix, Table S4). However, the GHG_i-final reached a plateau after 2006 (Fig. 6A), despite a continued and accelerated increase in final protein productivity compared with the previous two periods (as evident in the differences of slopes shown in SI Appendix, Fig. S13). This plateau indicates that the increased livestock production and land use changes for crop production from 2007 to 2019 have offset the favorable effects of improved final protein productivity on the reduction of GHG_i-final. As a result, these developments have been responsible for the plateau of GHG_i-final in the last decade (SI Appendix, Table S4 and Fig. S13) and need to be well managed to continuously reduce GHG_i-final in the future.

Agricultural Products Trade Has Potentially Reduced GHG Emissions.

International agricultural trade is noted to have substantial benefits for climate change adaptation (28, 29). As coupling and decoupling of GHG_i between importing and exporting countries changed in different periods (SI Appendix, Fig. S14), contrasting effects on global agriculture-related GHG emissions through trade become visible. Bai et al. introduced a novel methodology for assessing how trade in agricultural products influences global agricultural productivity, alongside new indicators to measure the optimality of trade and the hypothetical resource savings it might engender (26). In this present study we applied an adjusted concept, which allows the inverse of the indicator (here: GHG_i) to be evaluated, relative to the productivity and efficiency indicator used by Bai et al. (26), to systematically quantify the impacts of trade on GHG emissions (Fig. 6 and SI Appendix, Fig. S15). In brief, we assumed traded protein production in importing countries would have occurred at the GHG_i in the importing countries, and calculated the difference of total GHG emissions from traded protein production in exporting countries as potential effects of trade on GHG emissions change. Given the constraints imposed by climatic conditions on the cultivation of specific crops, direct comparisons of individual crops across different countries are not feasible. Furthermore, focusing on single products fails to capture the broader adjustments in the structure of crop and livestock production prompted by trade. Consequently, this study eschewed a detailed trade matrix for each product and instead focused on the aggregate net trade in protein for each country.

Adjusted for the amount of protein traded, it turns out that trade has potentially increased GHG emissions by 4.6 to 7.3 Gt CO_2eq cumulatively during 1961 to 1979, depending on the level of protein production (Fig. 6 and SI Appendix, Fig. S15). However, thereafter, trade had a nearly neutral impact on reducing GHG emissions between 1980 to 2006 in terms of final protein production, but reduced GHG emissions by around 3.6 to 6.9 Gt CO_2eq for primary and semifinal protein production. Trade had a pervasive effect on reducing GHG emissions for the period 2007 to 2019 for all three levels of protein production. In 2017/2019, the annual potential GHG emissions reductions through trade of final proteins reached as much as 0.53 Gt CO_2eq y⁻¹, equivalent to 4.4% of the global annual emission from agricultural system during 2017 to 2019 (Fig. 6 and SI Appendix, Fig. S15). The reduction was 2.7% higher when cross-border freight transportation-related GHG emissions were considered separately for both exporting and importing countries, compared to the scenario where these transportation-related emissions were not considered (SI Appendix, Figs. S16 and S17). In summary, the international trade of final proteins during these years had a positive impact on reducing GHG emissions, especially when we included the transportation-related emissions in the calculations.

According to the data from 2007 to 2019, the annual reduction rate through trade was much higher when the protein production was at the primary level (1.2 Gt CO_2eq y⁻¹) and semifinal level (1.3 Gt CO_2eq y⁻¹), compared with the final protein production level (0.46 Gt CO_2eq y⁻¹) (Fig. 6 and SI Appendix, Fig. S15). Similar higher reduction rates were observed for primary (8.0%) and semifinal (8.7%) protein level, when cross-border freight transportation-related GHG emissions were included in our analysis (SI Appendix, Fig. S17). In the most recent decade, 9 to 13 kg CO_2eq was reduced when 1 kg of protein was traded, depending on level of protein production (Fig. 6 and SI Appendix, Fig. S15). Hence, in the absence of trade, the actual GHG_i might be higher than the current level.

The major nonmonetary factor behind this was probably the differentiation between countries in the varying need for food security (30). For major net importers like China, maintaining self-sufficiency in daily energy consumption is crucial. It’s possible that importing countries excel in producing food calories rather than protein. In addition, there may be large differences in the impacts that the intermediate use of imported protein has on the different protein production levels. Increasing trade of feed protein from lower GHG_i exporting countries, e.g. Brazil, to high GHG_i countries, e.g., China and other Asian countries for all three levels of protein production, is the main driving force behind possible reductions of GHG emissions in the most recent decade. Food protein has also been sourced from low GHG_i exporting countries and traded to high GHG_i importing countries in the most recent decade, which also contributed to the global GHG emissions reduction, because a better-optimized resource allocation has been realized (Fig. 6).

The optimality level of protein trade in reducing GHG emissions has seen a modest enhancement across all three levels of protein production, from less optimal VI to optimal III (SI Appendix, Fig. S18; Methods). This improvement is reflected in the reduction of GHG emissions and a more equitable distribution of trade between countries with low GHG_i (exporting countries) and those with high GHG_i (importing countries). The contribution of protein import by high GHG_i importing countries increased from 0.06 to 0.11 in 1960s to 0.15 to 0.16 in 2010s, which indicates that high GHG_i countries are importing more proteins, but there is still much room for improvement, since the optimal value should close to 1.0 (Methods). Ratio of protein exported by high GHG_i exporting countries increased from 0.19 to 0.31 to 0.33 to 0.37 between 1960s and 2010s depends on protein production level, which reduced the trade optimality level, since this value should be close to 0.0 in theory (Methods) (SI Appendix, Fig. S18). These findings underscore the substantial potential for lowering GHG emissions related to global agriculture through the refinement of trade practices.

The results presented in the main text focus only protein based GHG_i. For results related to calorie-based GHG_i, please refer to SI Appendix, Figs. S19 and S20. The similar decline trend has been found between the calorie- and protein- based results. Therefore, we did not duplicate the description of changes in GHG_i and potential impacts of trade for calorie-based production for the sake of brevity.

Future Implications.

Our analysis reveals a strong overall initial decrease in GHG_i during the early phase of our investigation period, primarily related to a rapid increase in productivity within the agricultural sector. However, in recent years, GHG_i has leveled off, largely due to the expansion of livestock production and land use change for crop production. The recent emergence of geopolitical conflicts has showed that mitigation of GHG emissions may be placed in a secondary position when global food security is at risk (31, 32). In order to meet the escalating food demands of a growing human population—estimated to require an additional 35 to 56% in production by 2050 compared to 2010 levels (3), the GHG emissions from agricultural systems would increase at a similar or even higher rate, if the current trend persists. This may cause an additional emission of 4 to 7 Gt CO_2eq into the atmosphere, hampering the achievement of the critical 1.5-degree target. This concern has been reported by many other studies (13, 33, 34). Hence, there is a need to reduce the absolute GHG emissions from agricultural systems, as the increased productivity alone could not solve the food and climate change dilemma in the past decade.

Many mitigation options are available in reducing GHG emissions, such as adopting diet changes with fewer animal-sourced products, enhanced efficiency fertilizer, and better nutrition of ruminant animals (4, 10, 35). However, the challenge exists in the implementation of these measures by countries and stakeholders. Until now, only a few countries have included the agricultural sector in their National Determined Contributions or national carbon neutral pledge, and even then, these commitments are still not as mandatory as those for other sectors (36). This is possibly due to the concerns of food insecurity risks. The policies aimed at mitigating GHG emissions from the agricultural sector, including land use change, need to be strengthened for every country in the future, to ensure a sustainable planet for all. It is also important to explore new and advanced mitigation technologies with maintaining, or even enhancing food protein production capabilities. To support developing countries or countries with food insecurity issues, it is essential to facilitate the transfer of mitigation technologies and knowledge, coupled with financial aid, such as the proposed annual commitment of $100 billion in climate funding assistance by developed countries (37). This approach will not only address the differences in resources but also increase global collaboration in the pursuit of a sustainable future.

Future protein trade from agricultural products may have further repercussions on global GHG emissions. Minimizing such emissions can be seen as a guiding principle. Adopting this principle would result in different priorities for improvement in different countries. Increase in total protein production and export would have the highest priority for low GHG_i exporting countries such as the United States and Brazil. Improvements should focus on technologies without land expansion, to avoid countereffects on GHG_i reduction (as has been observed in the recent decade). For example, intensification of farms could help to maintain or even increase protein productivity in already highly productive exporting countries under climate change in 2100 (38, 39). Increase of multiple cropping in the lower GHG_i exporting countries has also proven possible and has a relatively larger potential than in importing countries (40).

Increase of domestic protein production in the low intensity livestock production importing countries also shows considerable potential for countries such as the Netherlands, France, the United Kingdom (UK), and Germany. More protein could then remain in these countries and be available for import by the high-intensity importing countries, such as China and Indonesia. Possible technologies include building of large-scale greenhouses and investing in natural gas-based microprotein or food waste-based insect protein production technologies, which are costly yet affordable in these high-income countries (41, 42). In addition, maximizing plant-based protein supply, while optimizing nutrient availability to less well-supplied populations, optimization of water and nutrient supply, and toward increased cropping index, where possible, are also promising options (42, 43). In all cases, GHG emission reduction would greatly benefit from the removal of trade barriers between countries.

Limitations.

There are some limitations in the present study which are mainly related to data availability. When estimating primary protein production, we did not include feed from pasture, crop residues, and swill, as it is difficult to obtain accurate estimates for these factors. However, this exclusion may lead to an overestimation of GHG_i. Nonetheless, based on our general evaluation of feed input from animal products and manure excretion, we observed that GHG_i for primary products decreased by half from 1961 to 2019, although it has remained stagnant in recent years (Fig. 4). In our study, we assumed the net importing countries have enough resources to produce all the imported agricultural products, which allowed us to quantify the potential effects of trade on GHG_i. However, this assumption may overestimate the effects, as net importing countries may lack the necessary resources, such as land and water, to produce all the imported agricultural products. Global trade would lead to extra GHG emissions during international transportation of goods (44). It is reported that nearly 1.3 Gt CO₂eq emissions were generated during international trade of food in 2017 (44), which could largely offset the GHG emissions reduction from global trade estimated in present study. The trade may impact the potential reduction of GHG emissions by 3 to 9%, depending on the protein production level. However, due to lack of time series country-specific data, this overall effect over the past 60 y has not been considered in this study.

Conclusions

Based on past observations the GHG_i has been understood to continuously decline. This study reveals that such a decline indeed occurred since the 1960s. That trend, however, has halted in recent years, reaching a plateau or even increasing in some cases. The primary drivers of this reversal were the expansion of livestock production and changes in land use for crop cultivation. This rebound in emission intensities contrasts with the more optimistic projections in agricultural development and hence poses an extra challenge to global climate change mitigation efforts, especially as the demand for food continues to rise. Our findings in this study underscore the urgent need for targeted strategies to address the identified challenges. It highlights the importance of global cooperation and policy interventions for immediate actions in increasing land productivity and minimizing land use changes in countries with increasing or stagnant emission intensities. This has been shown particularly for developing countries, which may need knowledge and fiscal transfer from low-emission intensity countries. Countries with low emission intensity could focus on reducing GHG emissions through reducing livestock production. The potential for international trade to mitigate GHG emissions has been recognized. Thus, the free trade of agricultural products between countries should be enhanced to reduce GHG_i on the global scale. In summary, while acknowledging the initial progress in reducing agricultural GHG emission intensities, the recent stagnation is a clear signal that more concerted and innovative efforts are required to ensure a sustainable and climate-resilient global food system.

Methods

The main objective of this paper is to evaluate changes of GHG_i on a protein basis, and their differences in countries with different levels of protein trade and livestock production which were greatly enhanced in the recent few decades. We start with estimation of GHG emissions at the farmgate level from agriculture production and net forest conversion using data from FAOSTAT (1) and related literatures (6), which has been detailed in the following paragraph in this section. Then we quantify protein production by crops and livestock at three different levels (i.e., primary product, semifinal, and final product), to exclude double-accounting of feed protein that is not used for livestock production but not directly consumed by humans. To estimate primary protein production, we used data from the Production Sheet in FAOSTAT, which provided direct estimates of protein production for each product in each country. To determine the protein content of each product in each country, we referred to the work of Becker et al. (45). Semifinal protein production was estimated by subtracting the protein embodied in locally produced animal feed that is consumed domestically from primary protein production. Final protein production was evaluated subtracting the protein embodied in feed exports from semifinal protein production.

To estimate the GHG_i of different protein production levels, we divided the emissions associated with each level by the corresponding protein production. Moreover, countries are classified into four different groups, in terms of their role in trade (i.e., net exporter or importer), and level of livestock production (below or above global average livestock protein production per capita per year from 1961 to 2019), to evaluate their impacts on the trends of GHG_i. Cumulative GHG_i distribution curve, trade optimality, and functionality have been updated to quantify effects of trade on improving or halting global climate change actions (26, 27).

GHG Emissions.

GHG emissions included those from farming activities such as crop production, livestock production as well as energy use in agriculture, and those emissions from land use change such as net forest conversion, savanna fires, and drained organic soils. All GHG emissions data were sourced from the Climate Change: Agrifood Systems Emissions section of the FAOSTAT database, with related data sources listed in SI Appendix, Table S1. There were a few exceptions, such as emissions from energy use in agriculture, drained organic soils, and net forestation, due to the lack of consistent data from FAOSTAT between 1961 and 2019.

GHG emissions from drained organic soils have been available in the FAOSTAT database from 1990 onward. Emissions from drained organic soils from 1961 to 1989 were kept constant at the 1990 value for boreal and temperate peatlands, while emission data were interpolated from 0 in 1980 to the 1990 value for tropical peatlands in equatorial Asia (6, 46, 47). FAOSTAT has reported GHG emissions from energy use since 1970. GHG emissions from energy use during 1961 to 1969 were estimated using agricultural energy use from Ritchie and Roser (48), and we assumed that the agricultural energy consumption structure during 1961 to 1969 was the same as in 1970 from FAOSTAT. Then we calculated the energy-related GHG emissions based on the total energy consumption of each energy source and the corresponding emission factors (1, 49). GHG emissions from net forest conversion from 1961 to 2019 were directly derived from Hong et al. (6), since FAOSTAT only reports these emissions after 1990. Global warming potential values of methane and nitrous oxide were derived from AR5 (1), and then used to calculate the aggregated GHG emissions.

Protein Production.

Production data for around 170 major agricultural products, including 145 crop products and 28 animal products (SI Appendix, Tables S5–S7), in around 180 countries from 1961 to 2019 were derived from the Production Sheet in FAOSTAT. The protein content of each agricultural product was derived from Food Composition Tables in food balance sheets (45). Given the limited data availability, we used the same protein content for each product in different countries and in different years, by assuming no spatial differences and temporal changes in the protein contents (26). Protein production was estimated using production amount multiplied by protein content for each product in each individual country.

Three levels of protein production were used, to avoid of double-counting of locally used and exported animal-feed protein resulting from increasing international trade of agricultural products and livestock production in the recent decades, which are defined are below: 1) primary product protein production, which includes all produced primary protein (primary crop and livestock protein production); 2) semifinal product protein production, which excludes locally produced animal feed protein; and 3) final product protein production, which excludes both locally produced animal feed protein and exported protein used as animal feed.

Feed consumption was estimated based on the Food Balance Sheet for each country from FAOSTAT. Here, only the economic organ of crop products was included, except for maize, oil crops, and sugar crops, of which the processed by-products were usually used as animal feed. Therefore, the processed part of these items was also categorized as animal feed. Crop straws, residues, and grass have been excluded from the analysis due to data limitations, and their negligible role in agricultural product trade. In the present study, we estimated primary protein production using data from the Production Sheet in FAOSTAT and information on protein content for each product in each individual country (45). The calculation is shown below. To estimate semifinal protein production, we subtracted the protein embodied in locally produced animal feed consumed domestically from the primary protein production. The calculation process is showed as below. We estimated final protein production by subtracting the protein embodied in feed exports. The calculation is shown below. Locally produced animal feed consumed domestically and feed exports were estimated utilizing the following methodology.

Locally produced animal feed consumed domestically.

For net import product items, locally produced animal feed consumed domestically was estimated using national feed amounts derived from the Food Balance Sheet in FAOSTAT, multiplied by the self-sufficiency rate of the corresponding item. For the net export product items, locally produced animal feed was directly estimated using the feed amount derived from Food Balance Sheet from FAOSTAT (SI Appendix, Fig. S21). As some domestic use of by-products from crop products processing could be used as animal feed, including cottonseed cake, groundnut cake, maize cake, rapeseed, and mustard cake, palm kernel cake, soybean cake, sunflower cake, sesame seed cake and copra cake, the corresponding product items (i.e., soybeans, copra, cotton seed, groundnuts, maize, rapeseed and mustard seed, palm kernels, sesame seed, and sunflower seed) were also included as locally consumed animal-feed. These calculations were on yearly basis for each country.

$$\begin{array}{c}local\mathit{ }fee{d}_{r,y}={\displaystyle \sum _{i\_pn\_ex}}{\mathit{feed}}_{r,y,i\_pn\_ex}+{\displaystyle \sum _{i\_py\_ex}}({\mathit{feed}}_{r,y,i\_py\_ex}+{\mathit{processing}}_{r,y,i\_py\_ex})\hfill \\ +{\displaystyle \sum _{i\_pn\_im}}{\mathit{feed}}_{r,y,i\_pn\_im}\times \frac{{\mathit{Pro}}_{r,y,i\_pn\_im}}{{\mathit{Consum}}_{r,y,i\_pn\_im}}\hfill \\ +{\displaystyle \sum _{i\_py\_ex}}({\mathit{feed}}_{r,y,i\_py\_ex}+{\mathit{processing}}_{r,y,i\_py\_ex})\hfill \\ \times \frac{{\mathit{Pro}}_{r,y,i\_py\_ex}}{{\mathit{Consum}}_{r,y,i\_py\_ex}},\hfill \end{array}$$ [1]

where local feed refers to locally produced animal feed protein consumed domestically. feed and processing refer to product use as feed and processing in the Food Balance Sheet in FAOSTAT, respectively; Pro and Consum refer to production and consumption of product item, and consumption of the product item was estimated using production, import, and export of the corresponding item from the Food Balance Sheet in FAOSTAT; i, a, r, and y refer to country, production item, and year respectively; ex and im refer to net export and import, respectively; pn and py refer to by-product of the product item that could not be used as animal feed and could be used as animal feed. Therefore, i_pn_ex and i_pn_im refer to net export and import item, respectively, and their by-product cannot be used as animal feed; i_py_ex and i_py_im refer to net export and import item, respectively, and their by-product can be used as animal feed, which has been listed above.

For the product, whose by-product is used as feed (Eq. 2) and not used as feed (Eq. 3), calculation of the global total exported animal feed from crop products in a given year uses the following equations.

$$\begin{array}{c}Global\mathit{ }feed\mathit{ }expor{t}_{i\_py,y}=global\mathit{ }feed\mathit{ }impor{t}_{i\_py,y}\hfill \\ ={\sum }_{r\_m}({\mathit{feed}}_{i\_py,y,r\_m}\times \frac{net\mathit{ }impor{t}_{i\_py,y,r\_m}}{{\mathit{Consum}}_{i\_py,y,r\_m}}\hfill \\ +{\mathit{processing}}_{i\_py,y,r\_m}\times \frac{net\mathit{ }impor{t}_{i\_py,y,r\_m}}{{\mathit{Consum}}_{i\_py,y,r\_m}}),\hfill \end{array}$$ [2]

$$\begin{array}{c}Global\mathit{ }feed\mathit{ }expor{t}_{i\_pn,y}=global\mathit{ }feed\mathit{ }impor{t}_{i\_pn,y}\hfill \\ ={\sum }_{r\_m}({\mathit{feed}}_{i\_pn,y,r\_m}\times \frac{net\mathit{ }impor{t}_{i\_pn,y,r\_m}}{{\mathit{Consum}}_{i\_pn,y,r\_m}}),\hfill \end{array}$$ [3]

$$\begin{array}{c}Feed\mathit{ }expor{t}_{r,y}={\displaystyle \sum _{i\_pn}}\left(\frac{net\mathit{ }expor{t}_{i\_pn,r,y}}{global\mathit{ }net\mathit{ }expor{t}_{i\_pn,y}}\times global\mathit{ }feed\mathit{ }expor{t}_{i\_pn,y}\right)\hfill \\ +{\displaystyle \sum _{i\_py}}(\frac{net\mathit{ }expor{t}_{i\_py,r,y}}{global\mathit{ }net\mathit{ }expor{t}_{i\_py,y}}\times global\mathit{ }feed\mathit{ }expor{t}_{i\_py,y}),\hfill \end{array}$$ [4]

where notations in Eqs. 2–4 are same as Eq. 1. Global feed export and Global feed import refer to the global total exported animal feed from crop products and global total imported animal feed from crop products, respectively. Feed export refers to the total protein exported and used as feed in the partner country in a country in a given year. Net export refers to net export of corresponding primary product of the product item in country r in each given year. Global net export refers to the global net export of corresponding primary product of product item in each individual year; i_pn and i_py refer to product item, whose by-product could not be used as animal feed and could be used as animal feed, respectively; r_m refers to a country with net import for an individual product item i_pn or i_py. SI Appendix, Fig. S21 also illustrated the calculation of exported animal-feed protein.

Protein production of three production levels.

Protein production of primary, semifinal, and final product are as shown below.

$$CP\_pr{i}_{r,y}={\displaystyle \sum _{i\_c}}CP\_cro{p}_{i\_c,r,y}+{\displaystyle \sum _{i\_a}}CP\_anima{l}_{i\_a,r,y},$$ [5]

$$CP\_sem{i}_{r,y}=CP\_pr{i}_{r,y}-{\mathit{localfeed}}_{r,y},$$ [6]

$$CP\_fina{l}_{r,y}=CP\_sem{i}_{r,y}-{\mathit{feedexport}}_{r,y},$$ [7]

where notations in Eqs. 5–7 are same as above equations; CP_pri, CP_semi, and CP_final refers to protein production of primary, semifinal, and final product, respectively; CP_crop and CP_animal refer to directly protein production from crop products and animal products, respectively; i_c and i_a refer to crop product item and animal product item, respectively.

Protein production estimates for primary, semifinal, and final products at the global level were aggregated directly from the corresponding country data. These group of countries include those with net export protein and high livestock production levels, net export protein and low livestock production levels, net import protein and high livestock production levels, and net import protein & low livestock production levels,

GHG_i.

GHG_i in each country in each year was estimated based on national GHG emissions aggregated from all sectors within farm gate and from net forest conversion and national protein production for different category described in the last section of the methodology. In addition, estimation of GHG_i for different protein production levels was based on all GHG emissions including those from feed production and corresponding protein production. The calculation is as shown below.

$$EInt\_pr{i}_{r,y}=\frac{{\mathit{GHG}}_{r,y}}{CP\_pr{i}_{r,y}},$$ [8]

$$EInt\_sem{i}_{r,y}=\frac{{\mathit{GHG}}_{r,y}}{CP\_sem{i}_{r,y}},$$ [9]

$$EInt\_fina{l}_{r,y}=\frac{{\mathit{GHG}}_{r,y}}{CP\_fina{l}_{r,y}},$$ [10]

where EInt_pri, EInt_semi, and EInt_final refer to GHG_i for primary product, semifinal product, and final product, respectively. GHG refers to total GHG emissions within the farm gate and from the net forest conversion. Other notations refer to those in the previous equations. GHG_i for different groups of countries or at global level are estimated based on the aggregated GHG emissions and aggregated protein production for the corresponding group of countries or globally.

Traded Protein.

For primary product, net traded protein for each country in individual year was aggregated from the net traded protein of each product, which estimated based on net agricultural trade and protein content of the corresponding product for each product. Protein content of each product was derived from Food Composition Tables in food balance sheets (45). Net traded amount of each product for each country in an individual year was calculated based on import and export of the corresponding product, which was derived from Trade Sheet from FAOSTAT. Related calculations were showed as below.

$$Net trade\_pr{i}_{r,y}={\displaystyle \sum _i}[({\mathit{import}}_{i,r,y}-{\mathit{export}}_{i,r,y})\times CP\_conten{t}_i],$$ [11]

where Net trade_pri refers to net national import protein in an individual year for primary product, which is defined in the present study previously; import and export refer to import and export amount of each product for a country in an individual year, respectively. CP_content refers to protein content of each product. Other notations refer to those in the previous equations. Since semifinal product only excluded locally produced animal feed consumed domestically, which would not affect trade, the net traded protein for semifinal products in a country during a given year is the same as that for primary products. However, the total traded protein for final product does not include net traded feed protein.

Classification of Group of Countries.

All 180 countries were classified into four different groups of countries, in terms of their role in trade and level of livestock production, to evaluate the impacts of trade on the changes of agriculture GHG_i. If a country was a net importer of protein embodied in agricultural products for more than 70% of the selected years (from 1961 to 2019), the country was classified as an importing country. Otherwise, it was classified as an exporting country. Selection of >70% as the threshold value was mainly due to its similar classification of countries into net exporting and importing countries with the most recent decade. Countries were categorized as high-level livestock production countries or low livestock production countries depending on whether their animal protein production per capita is above or below the world average livestock protein production per capita during 2017 and 2019, respectively. Categorization of the countries is shown in Fig. 1.

Estimation of the Trade Optimality Level and Effects On Agricultural GHG Emissions.

We applied cumulative productivity distribution curve (26), which has been used to evaluate impacts of trade of agricultural products on improving or reducing global land and nitrogen use efficiency. However, the method needs to be improved to evaluate impacts on GHG_i, since the original method was developed based on positive indicators, such as resources use efficiency or agricultural productivity. GHG_i was rather a negative indicator compared to efficiency indicators, of which the larger the values the worse. Here, we started with building the cumulative GHG_i distribution curve, followed by developing the trade optimality and functionality analytic framework. Finally, we quantified the reduction or increase in the amount of GHG emissions through trade of protein between countries during different periods from 1961 to 2019.

Cumulative emission intensity distribution curve.

Countries were plotted on the X-axis in ascending order of GHG_i, while the share of total protein production for each country was plotted on the Y-axis (%). The curve divides the graph into two parts, area A lying between the Y-axis, the 100% contribution line and the same curve, and area B lying between the X-axis, the max GHG_i line (Max X), and the c curve (SI Appendix, Fig. S22). Then we quantified the relative concentration of production in high GHG_i countries (CPHE), which equal to Area A divided by Area A + B (26). A lower value indicates production was more concentrated in countries with lower GHG_i, which is an optimal situation for the entire world. CWPE is short for the concentration-weighted productivity representing a CPHE-corrected productivity for a given product on the unit of productivity or GHG_i. It is calculated using CHPE multiplied by the maximum GHG_i, which is positively correlated with average GHG_i (26). In the present study, the unit of CWPE was in kg CO₂eq kg⁻¹ protein. The larger CWPE corresponds to worse global agriculture-related GHG_i.

Trade optimality and functionality in terms of impacts on global agriculture and associated GHG_i.

We applied the improved concept to the net protein-importing and exporting countries and then developed an illustrative concept in SI Appendix, Fig. S22, following the guideline from Bai et al. (26). However, trade optimality and functionality levels in terms of GHG emissions were opposite to that of resource use (land use and fertilizer use) derived from Bai et al. (26). Based on the same scheme, we developed a framework to quantify the optimality and functionality of trade in terms of reducing or increasing GHG emissions. Trade was considered functional when CPHE of exporting countries (CPHE_ex) < 0.50 and CPHE of importing countries (CPHE_im) > 0.50, and trade was considered near-optimal when CWPE_ex/CWPE_im < 1.0 (SI Appendix, Fig. S22). These conditions indicate that more than 50% of products are exported by relatively low GHG_i countries, more than 50% of products are imported by relatively high GHG_i countries, and the exporting countries on average have a low GHG_i than importing countries. There are eight possible combinations of CPHE and CWPE for exporting and importing countries as presented in SI Appendix, Fig. S22. These eight combinations can be categorized into two groups: an “optimal” group (Level I to IV), and a “nonoptimal” group (Level V to VIII) (SI Appendix, Fig. S22).

Impacts of protein trade on potential for reducing or increasing global GHG emissions.

The effects of global trade on agricultural GHG emissions refer to the effects of global trade on land use illustrated in Bai et al. (26), which assumed that importing countries would have the resources to scale up production at will. Estimation of trade effects on GHG emissions in each year are showed as below.

$$\begin{array}{c}GHG reduction\_pr{i}_y={\displaystyle \sum _{r\_im}}\left(EInt\_pr{i}_{r\_im,y}\times net trade\_pr{i}_{r\_im,y}\right)\hfill \\ -{\displaystyle \sum _{r\_ex}}\left(EInt\_pr{i}_{r\_ex,y}\times net trade\_pr{i}_{r\_ex,y}\right),\hfill \end{array}$$ [12]

$$\begin{array}{c}GHG reduction\_sem{i}_y={\displaystyle \sum _{r\_im}}\left(EInt\_sem{i}_{r\_im,y}\times net trade\_sem{i}_{r\_im,y}\right)\hfill \\ -{\displaystyle \sum _{r\_ex}}\left(EInt\_sem{i}_{r\_ex,y}\times net trade\_sem{i}_{r\_ex,y}\right),\hfill \end{array}$$ [13]

$$\begin{array}{c}GHG reduction\_fina{l}_y={\displaystyle \sum _{r\_im}}\left(EInt\_fina{l}_{r\_im,y}\times net trade\_fina{l}_{r\_im,y}\right)\hfill \\ -{\displaystyle \sum _{r\_ex}}\left(EInt\_fina{l}_{r\_ex,y}\times net trade\_fina{l}_{r\_ex,y}\right),\hfill \end{array}$$ [14]

where GHG reduction_pri, GHG reduction_semi, and GHG reduction_final refer to GHG reduction via the global trade; net trade_semi and net trade_final refer to the net traded protein of semiproducts and final products, respectively; r _ im and r _ ex refer to the net importing country and exporting country, respectively. Other notations are the same as for the previous equations. In the present study, it was assumed that total exports aggregated for each country were equal to total global exports. Besides, the present study did not distinguish which countries directly involved in a specific trade but total import or export occurred in each country.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

Emission data, agricultural production data, and trade data are curated by the FAOSTAT, which are freely available from FAOSTAT at https://www.fao.org/faostat/en/#data/domains_table (1). Raw data of agricultural production, trade, and emissions from drained organic soils were accessed in December 2022. Raw data of emissions excluding that from drained organic soils were accessed in February 2023. All details of the data access have been listed in SI Appendix, Table S1. Land-use change emissions were obtained from the study of Hong et al. (6). The source data supporting the findings of this study are available within the article and its supplementary information files. The raw and final data will be shared in the Science Data Bank website (https://www.scidb.cn/s/3UBjUf) (50) after the manuscript is accepted for publication. Statistics were mainly performed using R (version 4.2.3). The spatial analysis was run in ArcGIS (version 10.8). The code for analysis will be shared in the Science Data Bank website (https://www.scidb.cn/s/3UBjUf) (50) after the manuscript is accepted for publication. All other data are included in the manuscript and/or SI Appendix.