The origin, supply chain, and deforestation risk of Brazil’s beef exports

How Are Cattle Export Supply Chains Structured?

Overall, exports made up 19.1% of Brazilian cattle production in 2017 (SI Appendix), though four states made a disproportionately large contribution to exports: Rondônia, Mato Grosso, São Paulo, and Mato Grosso do Sul. These four states each exported >25% of their cattle production (Fig. 2A), were home to 53% (81 in total) of exporting slaughterhouses, and were the source of 59.0% of exports between 2015 and 2017. The northeast of Brazil (shown in shades of orange in Fig. 2), on the other hand, was not well connected to export markets—supplying only 0.6% of exports, despite being home to 10.9% of cattle production. Divided per biome, 48.1% of exports originated from cattle in the Cerrado, 25.5% from the Amazon, 18.2% from the Atlantic Forest, 5.0% from the Pampas, and 1.9% from the Pantanal and Caatinga.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

The origin and destination of Brazilian beef. (A) Cattle production (Mtons/year) and the proportion exported per state in Brazil, shown for 2017. (B) Breakdown of the origin of sourcing per major market. “Other EU” refers to other EU markets and “Other halal” are other halal markets, as defined in SI Appendix. States are colored by region, states in the Legal Amazon are in shades of red (AC, Acre; AM, Amazonas; MA, Maranhão; MT, Mato Grosso; PA, Pará; RO, Rondônia; RR, Roraima; and TO, Tocantins), the central-west in purple (DF, Distrito Federal; GO, Goiás; and MS, Mato Grosso do Sul), the northeast in orange (AL, Alagoas; BA, Bahia; PB, Paraíba; PE, Pernambuco; PI, Piauí; RN, Rio Grande do Norte; and SE, Sergipe), the southeast in blue (ES, Espirito Santo; MG, Minas Gerais; SP, São Paulo; and RJ, Rio de Janeiro), and the south in green (PR, Parana; SC, Santa Catarina; and RS, Rio Grande do Sul), with unknown (UN) origin shipments shown in grey.

Exports were consolidated in the hands of a few companies—204 in total between 2015 and 2017. Of these, 39.2% (80 traders) operated in all 3 y, and three companies, JBS, Minerva, and Marfrig (and their subsidiaries), handled 71.7% of exports (SI Appendix). These three companies each operated slaughterhouses in the Amazon states of Mato Grosso and Rondônia (SI Appendix, Fig. S1), though JBS had a particularly strong presence in the Amazon, handling 40.3% of exports from the biome, with Minerva responsible for 19.4% and Marfrig 9.9%. In contrast, Marfrig controlled a large share of exports from the Pampas (located mostly in the state of Rio Grande do Sul), which was the origin of 21.5% of their exports and 68.3% of all exports from the biome.

For most export supply chains, export companies were vertically integrated—overall, 94.4% of beef and offal exports were handled by 55 companies who operate their own slaughterhouses. These companies have strong control over the origination of their products, at least to the slaughterhouse level. The remaining 5.6% of beef and offal exports were mostly handled by import-export businesses, who specialize in international trade of multiple commodities. The slaughterhouses supplying exporting businesses each sourced cattle from a wide area. Exporting slaughterhouses, on average, directly and indirectly sourced cattle from 60.4 municipalities (±54.2 SD) and 364 municipalities (±193 SD), respectively, though most purchases were more local with 80% of each slaughterhouse’s sourcing being from 18.1 municipalities (±16.7 SD), less than 279 km (±175 SD) away (SI Appendix, Fig. S2).

Where Do Major Markets Source from?

Between 2015 and 2017, the largest export markets for Brazilian beef, offal, and live cattle were China (mainland and Hong Kong), which purchased 30.2% of Brazil’s exports by volume (30.1% by value). They are followed by Egypt (12.4% and 10.2% by volume and value, respectively), Russia (10.4% and 8.2%), Iran (7.1% and 7.2%), the European Union (7.1% and 11.9%), Chile (4.4% and 4.8%), Venezuela (3.9% and 4.3%), and the United States (2.3% and 4.9%). These markets have quite distinct and dynamic sourcing patterns (Fig. 2), driven by differing product portfolios, logistics, and food safety requirements.

Though the three largest export markets, China, Egypt, and Russia purchased from a wide area in Brazil (across 16, 14, and 11 states, respectively), imports into the United States and the United Kingdom were concentrated in the states of Mato Grosso do Sul, São Paulo, and Rio Grande do Sul. The latter pattern arises because the United States and the United Kingdom almost exclusively (>90%) imported processed meat products, such as beef jerky and corned beef, which have historically been produced in facilities in the consolidated central-west and south of Brazil (SI Appendix, Fig. S3), near to Brazil’s major urban markets which they also serve. In contrast, 69.7% of Venezuela’s sourcing came from states in Brazil’s Legal Amazon, with 16.0% imports from the state of Pará. This is because they were a major buyer of live cattle, purchasing 58.5% of all Brazil’s live cattle exports in 2015, more than three-quarters of which were exported through the port of Barcarena in Pará, drawing on cattle from surrounding municipalities (25) (SI Appendix, Fig. S3). This particular trade pattern is driven in part by simple logistics—ports in the north have the shortest shipping distances to Venezuela. These trade patterns are, however, not static: Venezuela’s imports declined from $644.0 million in 2015 to $5.7 million in 2017, linked to the country’s economic recession (26), and the United States’ sourcing shifted notably between July 2016 and July 2017, when they temporarily lifted restrictions on imports of fresh and frozen beef from Brazil (Fig. 3). These restrictions were again lifted in February 2020, setting the scene for another boost in US beef imports and deforestation risk.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

US deforestation risks have fluctuated notably over time, driven by changes in the products imported. The UnitedStates has historically had low deforestation risk (430 to 850 ha/year) importing processed meat products originating from meat packing facilities in São Paulo (A, Left) and sourcing cattle from municipalities where there was little recent deforestation (B, Left). The United States has historically not permitted fresh meat imports from Brazil because of concerns over bovine spongiform encephalopathy (BSE) and food safety in Brazil; from July 2016 onwards, however, the United States did permit fresh meat imports, and this led to a doubling in the Brazil-US beef trade (A, Right), before the Carne Fraca meat scandal broke and imports were banned again in July 2017 (trade window demarcated by vertical dotted lines). The slaughterhouses which met this new demand for fresh meat in 2016 were located across the country—with 14% of exports originating from facilities in the Legal Amazon region. These exports carried a higher deforestation risk than the previous imports (B, Right). Note: this plot does not include deforestation risk linked to “unknown” sourcing, and excludes 13.7% of US sourcing, where the month of shipment was unknown. State abbreviations are listed in Fig. 2 legend.

European Union countries, on the other hand, accepted only fresh and frozen beef from facilities in 10 states in Brazil’s south, southeast, and central-west, some of the first to be designated as free of foot-and-mouth disease (FMD) (SI Appendix, Fig. S4). EU sourcing is, however, expanding northwards—in 2016, in recognition of Brazilian efforts to eradicate FMD, the European Union approved 14 additional states for exports of processed meat, including five states in Brazil’s Legal Amazon: Acre, Rondônia, Pará, Tocantins, and Maranhão. In 2017, Rondônia was the first of these to sell processed meat to the EU market, supplying 55 tons of processed offal to Denmark and Italy.

How Are Deforestation Risks Distributed?

Overall, we identified 73,000 to 74,700 ha/year deforestation risk linked to cattle exports each year, assuming a 1-y amortization period (Materials and Methods), out of a total of 480,000 to 520,000 ha/year of cattle-associated deforestation risk. Of the deforestation linked to cattle exports, 40,200 to 41,900 ha/year (55.0 to 56.6%) arose from municipalities in the Amazon, 30,100 to 32,200 ha/year (40.7 to 43.0%) in the Cerrado, and 100 to 130 ha/year (0.1 to 0.2%) in the Atlantic Forest.

Our approach reveals how exposure to deforestation risk is far from uniform—as is implicitly assumed in environmental footprint analyses or life cycle assessments which rely on national-level or representative footprints, (e.g., refs. 17 and 27) (SI Appendix, Fig. S5).

Among exporters, deforestation risks vary greatly depending on where companies operate slaughter and processing facilities. The meat packer Irmãos Conçalves, for example, has the fifth-highest total deforestation risk of all major exporters (2,100 to 3,500 ha/year), and the highest relative deforestation risk (Fig. 4A). Irmãos Conçalves operates a slaughterhouse in the Amazon state of Rondônia which had, as of July 2020, not made any commitment to monitor their suppliers for deforestation (see below Coverage of zero deforestation commitments). In contrast, Frisa Frigorifico Rio Doce (“Frisa”) and Pampeano Alimentos, a subsidiary of Marfrig, had low deforestation risks. Frisa sourced from facilities in Minas Gerais, Espírito Santo, and Rio Grande do Sul, where deforestation rates are lower, and Pampeano Alimentos also operated a processing facility in Rio Grande do Sul, in the Pampas—a nonforested biome (SI Appendix).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 4.

The deforestation risk of different actors’ purchases of Brazilian beef. (A) The relative deforestation risk of major exporters, with the market average risk for exports in 2017 shown as a dashed vertical line. (B) The deforestation risk of major markets (as a percentage of all export-associated deforestation risk), plotted against their market share (% of exports). (C) The deforestation risk of major markets (2015 to 2017) in the Amazon, Cerrado, and Atlantic Forest biomes. (D) The relative deforestation risk of different products—live cattle exports carry a disproportionately high risk. The total (E) and relative (F) deforestation risk linked to exports and the Brazilian domestic market. Error bars show 95% confidence intervals, accounting for key parameter uncertainty (Materials and Methods).

Among importing countries, we also see stark differences in deforestation risk (Fig. 4B and SI Appendix, Fig. S6). China, for example, carried the greatest exposure to deforestation: 15,900 to 23,000 ha/year (21.7 to 31.1% of all export-associated deforestation risk), with a notable difference in deforestation risk of imports arriving into ports in mainland China versus Hong Kong (Fig. 5), driven by disparities in their sanitary requirements and slaughterhouse approval processes (SI Appendix). These risks are also dynamic—in June 2019, China licensed more than 20 additional cattle slaughterhouses for export to the Chinese mainland; these slaughterhouses had double the deforestation risk of slaughterhouses previously supplying the mainland (70.5 ha/kton ± 81.8 SD vs. 31.0 ha/kton ± 36.1 SD), though these slaughterhouses already supplied 14.2% of exports to Hong Kong between 2015 and 2017.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 5.

The origin of Brazilian beef exports to Mainland China and Hong Kong. (A) Slaughterhouses are shown as points scaled by the volumes exported, with the percentage of exports originating per municipality in the background. (B) The state of origin of cattle supplying each market. (C) The deforestation risk of slaughterhouses exporting to China. There is a notable difference in deforestation risk between slaughterhouses exporting to Hong Kong and the Chinese mainland and between slaughterhouses previously exporting to the mainland and those recently licensed for export to the mainland. Among newly licensed slaughterhouses, the top five with deforestation risks >100 ha/kton are located in Mato Grosso (two), Tocantins (two), and Pará (one).

Different countries’ risks arise in different regions (Fig. 4C). International markets for halal products (including Egypt, Iran, and the United Arab Emirates) sourced 30.8 to 39.5% of their imports from municipalities in the Amazon, which were responsible for 60.9 to 64.5% of their deforestation risk (15,000 to 20,700 ha/year of 24,600 to 32,100 ha/year across all biomes). The European Union’s deforestation risk, in contrast, was concentrated in the Cerrado: 2,100 to 2,600 ha/year, 72.9 to 75.2% of their total deforestation risk (2,900 to 3,600 ha/year). There is no evidence that differences in deforestation risk between countries arise from “preferences” for low or high deforestation risk—before the data presented here were assembled, the deforestation risk of each countries’ sourcing was unknown, precluding any informed efforts by downstream actors to reduce their sourcing risks within Brazil. Instead, sourcing patterns have historically been determined by a combination of disease control requirements, product preferences, sanitary requirements, and logistics (see above). A major application of our analyses is therefore to provide a tool for downstream actors (companies and governments) to differentiate sourcing risks, as exemplified by the breakdown of slaughterhouses newly licensed for export to the Chinese mainland (Fig. 5C).

Our results also draw attention to Brazil’s export of live cattle. This trade in live cattle is comparatively understudied, despite its large size (more than 200,000 animals worth >$200 million are exported each year), implications for animal welfare given the necessary long-distance transport, and disproportionate deforestation risk—live cattle exports are concentrated in the Amazon state of Pará and make up 3.9% of Brazilian cattle exports by value, but 11.6% of deforestation risk (Fig. 4D). Seventeen companies handled live cattle exports, of these Minerva was the most important, handling 47.3% of all trade. Beyond Venezuela, mentioned above, other major markets for live cattle were Turkey, Lebanon, Jordan, and Iraq, where cattle are both used as breeding stock and raised for slaughter.

By spatially mapping cattle exports, we also elucidate the proportion of cattle-associated deforestation which is linked to consumption by the Brazilian domestic market. The Brazilian domestic market purchased 80.9 to 82.4% of Brazil’s beef across 2015 to 2017, incurring 85.8 to 86.8% of deforestation risk each year, while export markets purchased 17.9 to 19.1% of Brazil’s beef, shouldering 13.2 to 14.2% of deforestation risk (Fig. 4E). The lower relative risk of export markets (Fig. 4F) is not because they do not source cattle from the Amazon, as sometimes suggested (28, 29). Overall, 17.2 to 21.5% of the Amazon’s cattle production was exported each year. Rather, within each biome, exports tend to originate from relatively consolidated municipalities, which have well-developed agricultural sectors (higher agricultural gross domestic product [GDP]) and where deforestation is less recent (lower remaining natural vegetation cover), than municipalities supplying the domestic market (SI Appendix, Fig. S7).

Clarifications in the Attribution of Deforestation Risk.

The attribution of deforestation risk to actors and specific markets comes with some clarifications. First, we use the term deforestation risk, though our data include some nonstrictly forested vegetation types, particularly in the Cerrado, where we used the PRODES-Cerrado dataset on natural vegetation loss. This dataset captures the conversion of a gradient of natural ecosystems, including forest, shrub, and savannah into “anthropic” land uses. We intersect these (and other) data on natural vegetation loss with maps of pasture expansion to limit our analyses to cattle-associated land use change (SI Appendix).

When reporting deforestation risk, we included only the footprint of pasture expansion, not other sources of feed. Around 10 to 12% of cattle are finished in feedlot systems. These systems are expanding into frontier regions in Brazil (30), though they ultimately still represent a small proportion of total feed demand, considering that cattle spend at most a few months in feedlots before slaughter. We also do not account for illegality in meat processing and export, though we believe this has a negligible effect on our results (SI Appendix).

Each actor’s deforestation risk represents the risk of sourcing cattle products from recently deforested land, given their subnational origin, and does not mean that that amount of deforestation was directly “caused” by the production they are purchasing. Rigorous estimates of how much deforestation is caused by each actor or market would ideally account for indirect effects, which could be substantial. For example, although beef exports tend to originate from areas with lower deforestation risk than domestic consumption, exports grew 30% from 2010 to 2017 and may drive expansion in frontier regions by displacing domestic demand elsewhere (31). A spike in late 2019 in beef prices within Brazil, resulting in large part from a surge in demand by the Chinese market, lends support to the idea that changes in exports do have knock-on effects on the domestic market (32). Looking beyond the cattle sector, the pasture area in Brazil has been relatively stable since 2005, at approximately 180 Mha (33). Cropland, in particular soy and sugar cane, expanded by 19.2 Mha between 2005 and 2017 (34), with the majority of expansion occurring onto pasture, which, in turn, has expanded into forest. This complex dynamic makes the relative role of cropland and pasture expansion in driving vegetation loss in Brazil an area of active research and debate.

Overall, each year between 2000 and 2017, cattle-associated deforestation ranged between 0.26 and 1.8 Mha, which on average made up 41% (±3% SD) and 17% (±2% SD) of all deforestation in the Amazon and Cerrado, respectively (SI Appendix, Fig. S8). These estimates of cattle-associated deforestation are conservative for two reasons. Firstly, our approach, which intersects remote sensing data on deforestation, pasture expansion, and soy planting, leaves 62% (±47% SD) of deforestation in the Amazon and Cerrado unattributed to a subsequent land use. While some of this deforestation may be attributed to other drivers (e.g., mining, swidden agriculture, other cropland expansion), it also reflects uncertainties in the classification of pasture in remote sensing products. Secondly, we try to limit our calculation to deforestation that is motivated by cattle production, per se, rather than speculative clearing, where vegetation is removed to valorize land, in particular for selling it for soy farming later on. We therefore excluded areas which were planted with soy within 5 y, as well as pasture which was detected more than 5 y postclearance. The latter assumes that the clearance of deforested areas which are idle for 6 or more years before productive use are driven more by speculation, than directly for cattle ranching (even if they then go on to be used for cattle ranching).

Finally, though our results are consistent across different assumptions about cattle slaughter rates, carcass weights, and property sizes, our results are sensitive to the amortization period chosen. Using a 1-y amortization, we estimate 73,000 to 74,700 ha of deforestation risk each year; a 10-y amortization period, on the other hand, increases this to 120,100 to 160,100 ha of deforestation risk each year, because cattle-associated deforestation rates were higher in the early 2000s, and 10-y amortization allocates some deforestation in this period to current exports. Though the overall quantity of deforestation linked to imports in 2015 to 2017 increases substantially, the relative importance of different actors remains similar (SI Appendix, Fig. S9), giving confidence to the relevance of our results for differentiating the risk exposure of different markets and buyers.

Coverage of Zero Deforestation Commitments.

The deforestation risks embedded in the purchases of companies and countries can be mitigated through government and corporate efforts to regularize land use in the sector. There are two commitments made by slaughter businesses in the Brazilian cattle sector, both initiated in 2009: 1) the Terms of Adjustment of Conduct (TAC) are legally binding commitments signed by individual slaughterhouses to not purchase cattle from properties with illegal deforestation within the Legal Amazon (the nine states making up the Amazon basin); 2) the G4 is an agreement from the three largest meat packing companies, JBS, Minerva, and Marfrig, to not purchase cattle from properties in the Amazon biome who cleared land post-2009.

We find that, in total, 31.2% and 17.8% of Brazil’s cattle exports were covered by the TAC and G4 agreements, respectively. These proportions rose to 82.6% and 69.6% of exports from the Amazon biome. Despite the high coverage of these commitments, we found 123,200 ha of deforestation risk linked to exports from the Amazon biome between 2015 and 2017, with G4 companies (and their subsidiaries) linked to 75,600 ha of deforestation risk within the Amazon, and 147,700 ha nationwide. This mismatch between high zero deforestation coverage and considerable deforestation risk arises because of several factors.

First, these commitments are only partially implemented (SI Appendix). Second, deforestation commitments are implemented at the level of properties (ranches), but we map supply chains and calculate deforestation risk at the municipal level. While animal movement data can be used to trace cattle back to individual properties, these data were not available for all slaughterhouses (Materials and Methods); property boundary data with identifying information (required for making the link with movement data) are not available nationwide, and meat packing companies are not otherwise forthcoming about the origin of their cattle purchases (SI Appendix). Monitoring risks at the municipal scale also brings advantages. While committed companies can and do make efforts to avoid deforestation-linked cattle entering their supply chains (23), in practice they only monitor their direct suppliers; they therefore miss the bulk of deforestation associated with their sourcing which arises from their network of indirect suppliers—properties which rear cattle, sell them on to other properties, who may fatten them before sending them to the slaughterhouse. The municipal-level approach taken here captures these landscape-level risks and provides the most complete picture possible, using publicly available data, of their exposure to deforestation nationwide. Previous research indeed suggests that though committed companies have reduced their purchases from properties with post-2009 deforestation, this has not led to landscape-level reductions in deforestation (23, 24, 35). Ultimately the success of commitments needs to be judged against overall changes in deforestation rather than changes in the direct exposure of individual companies. Finally, irrespective of the resolution of the analysis, G4 companies carry considerable deforestation risk also because of the limited geographical scope of their commitment: 47.1% of these companies’ deforestation risk arose from sourcing cattle in the Cerrado (SI Appendix, Fig. S10), where the G4 does not apply, and 17.2% of their deforestation risk stemmed from sourcing cattle outside the Legal Amazon, where TACs are not in place.