International corporations trading Brazilian soy are keystone actors for water stewardship

The study investigates how transnational corporations (TNCs) influence the redistribution of global water resources through Brazilian soy supply chains and virtual water trade (VWT). While international food trade decouples consumers from local water pressures, prior VWT analyses have been largely at country scale, overlooking sub-national heterogeneity that drives unit water footprints (uWF). Moreover, the specific role of TNCs—identified as keystone actors in global production ecosystems—has been understudied in the context of corporate water stewardship. The paper addresses two gaps: incorporating sub-national environmental variability into VWT assessments and quantifying the role of TNCs in managing water use within soy supply chains. Brazil, with vast ecological diversity, significant freshwater reserves, and the world’s largest soy output, provides a critical case where soy expansion interacts with deforestation, hydrological alteration, and drought risks. The research aims to map spatially explicit, company-specific VWT from Brazilian municipalities to top importing countries, assess companies’ and countries’ uWFs and drought exposure, and evaluate trends from 2004 to 2018 to inform stewardship and sustainable supply chain strategies.

Previous research predominantly analyzed virtual water trade at national scales, missing intra-country heterogeneity affecting crop water use (uWF). Recent works began addressing inter-regional flows in large countries like China and India and emphasized the value of higher-resolution trade data to improve assessments. TNCs have been recognized as keystone actors with outsized biosphere impacts and are increasingly engaged in stewardship for climate and oceans, yet water footprint assessments of TNCs in agriculture remain limited. The Brazilian soy context has been widely studied regarding deforestation, biodiversity loss, and climate impacts on the hydrological cycle, but integrated, company-specific VWT linked to municipal-level conditions is lacking. This study builds on supply chain mapping advances (TRASE) and crop-water modeling to bridge these gaps.

Scope: Brazilian soy exports to the top-10 importing countries (China, Netherlands, Spain, France, Thailand, Germany, South Korea, Iran, Italy, United Kingdom) over 2004–2018. Company focus: top 9 TNCs dominating trade (covering ≥80% of each country’s imports in 2018). Data sources: FAOSTAT detailed trade matrices for country selection and volumes (primary soy, soy oil, soy cake converted to primary equivalent); TRASE for municipality–company–country flows; biophysical inputs from CRU TS climate, MIRCA2000, SPAM2010/MapSPAM, and literature crop parameters. Trade data processing: Selected top importers by cumulative 2004–2018 soy-equivalent tons. Built country-specific trade matrices with rows as producing municipalities and columns as trading companies (297 total, with focus on 9 dominant). Compared FAOSTAT and TRASE imports to quantify variation; obtained mean weighted variation of 5%. Unit water footprint (uWF) estimation: uWF at municipality-year is total actual evapotranspiration (ETA) during growing season divided by actual yield (ton/ha), summing green (rainfed) and blue (irrigation) components. ETA modeled at 5×5 arc-min grid using FAO-56 approach with daily reference evapotranspiration and crop coefficients across growth stages; water stress coefficient varies with soil moisture in rainfed; set to 1 under irrigation. Rainfed and irrigated ET weighted by corresponding harvested areas. Planting/harvest timing from literature; rainfed/irrigated areas from MIRCA2000 (circa 2000) and MapSPAM/SPAM2010 for 2010-era conditions; aggregated to municipalities for 2004 and 2018. Missing municipality ETA values infilled using nearest neighbor municipalities. Yield computed as production divided by harvested area at municipality scale (from TRASE). Virtual water trade (VWT) construction: For each municipality–company–country triplet, VWT equals uWF multiplied by traded tonnage from TRASE, enabling country-level imports and company-level exports by summation across dimensions. Weighted barycenters: Computed VWT-weighted geographic barycenters for each company–country and for each country and company overall, to locate sourcing cores across Brazilian biomes. Drought exposure: Computed municipality-level probability of moderate-to-extreme drought using monthly self-calibrating PDSI (≤ -2) over 1958–2018; aggregated to country/company by VWT weighting. Error and uncertainty analysis: Compared municipality yields to SPAM2010 (2010) and ETA to WATNEEDS (2016), yielding CV of 8% for yield and 15% for ET. Propagated to uWF CV of 17%. Combined with trade CV (5%), obtained VWT CV of 18%. Sensitivity and uncertainty summaries provided; code available on request; data and derived datasets available via cited repositories.

• Scale and structure of flows: Identified 4429 virtual water flows linking 1620 Brazilian municipalities to top-10 importers (2018). Top nine TNCs handled ~100 Gm³ in 2018 (~70% of total), averaging ~7 million m³ per exporting municipality. - Countries vs companies: China dominated with 80 Gm³ (80% of traced imports) in 2018; Netherlands 5 Gm³ (5%); South Korea and Spain ~3 Gm³ each (3%). Single TNCs displaced more VW than single countries (except China). In 2018, Bunge exported 15 Gm³ (~14%), Cargill 14 Gm³ (~14%), ADM 12 Gm³ (~11%), Louis Dreyfus 11 Gm³ (~10%). Bunge alone exceeded Thailand’s import (>3× 4 Gm³); Louis Dreyfus exceeded Netherlands’ import (>2× 5 Gm³). - uWF heterogeneity: Country-average uWFs had narrower ranges (Germany 1340 m³/ton minimum; Italy 1560 m³/ton maximum) than companies (Louis Dreyfus ~1350 m³/ton minimum; Gavilon ~1800 m³/ton maximum). Bianchini and Gavilon exhibited wide uWF distributions with 90th percentiles >2400 m³/ton; maxima observed in Rio Grande do Sul municipalities Sentinela do Sul (4718 m³/ton) and Sant’Ana do Livramento (3850 m³/ton). - Spatial sourcing cores: VWT-weighted barycenters for countries fell predominantly in the Cerrado. Company barycenters were more dispersed, including Mata Atlântica and Pampa (southern Brazil) and edges with higher ET, higher uWF, and higher drought probabilities. Extension into Amazon associated with lower uWF via higher yields but potential long-term hydrological risks through deforestation. - Drought exposure: Companies Bianchini, COFCO, ADM, and Louis Dreyfus had drought occurrence probabilities around 25%; Amaggi ~14%. Country-level drought exposure was mitigated by diversification across companies. - Italy case study (2018): Imported 468 million m³ total, with 430 million m³ handled by seven dominant companies. Main contributors: Bianchini 120 million m³, COFCO 96 million m³, Cargill 60 million m³. About 40% of Italy’s volume came from ten municipalities. Italy’s relatively high uWF (1560 m³/ton) reflects reliance on suppliers with higher uWF (Bianchini ~1790; Gavilon ~1800 m³/ton). Despite some suppliers’ higher drought exposure (Bianchini 26%, COFCO 25%), Italy’s overall exposure was ~23% due to supplier diversification. - Trends 2004–2018: Total traced VWT rose from 43 to ~100 Gm³ (+133%). Growth was driven by 1230 new connections and a 44% increase in average flow per connection. Production expanded via extensification (602 new exporting municipalities) and intensification in stable municipalities, raising water demand by 48 million m³ in stable areas. Sorriso (Mato Grosso) increased VWT from 185 million to 2.3 billion m³, with yield rising 24% to 3.7 t/ha by 2018. Despite rising VWT, average uWF decreased ~40% (2566 to 1553 m³/ton) due to yields increasing from 2.2 to 3.3 t/ha (+36%). New flows emerged in drier northeastern Cerrado (MATOPIBA), yet overall drought exposure ranges slightly narrowed (countries: from 19–29% to 21–24%; companies: from 14–35% to 14–26%) due to expansion into less dry municipalities (e.g., western Cerrado, parts of Amazon like Rondônia). - Company dynamics: By 2018, COFCO rose to fifth-largest exporter. Gavilon and COFCO gained significance; Amaggi’s export declined 22%. ADM and Cargill tripled exports since 2004. Bianchini increased yields (+123%) and cut uWF (-61%) but reduced exports by 22%; Louis Dreyfus increased exports sevenfold with yield gains (~42%) and uWF declines (~30%). - Uncertainty: Propagated coefficient of variation for VWT ~18% (uWF ~17%, trade ~5%).

The findings show TNCs as pivotal actors in shaping virtual water flows and, consequently, the water footprint of importing countries. Importers’ uWF and drought risk are heavily influenced by companies’ sourcing portfolios across heterogeneous Brazilian biomes. While policy remains crucial, countries are structurally tied to firms’ sourcing choices, emphasizing the need for corporate water stewardship. Diversification across municipalities and companies can mitigate drought exposure, but trade-offs exist: firms linked to lower uWF through high-yield regions may coincide with higher CO2 emissions and biodiversity risks from deforestation, whereas southern sourcing can imply higher uWF and drought probabilities. The study underscores interdependencies among climate change, land-use change, biodiversity loss, and hydrological stability. It proposes a collaborative stewardship model where science, TNCs, countries, and society share data and co-develop tools: TNCs supply detailed trade data to enable robust water footprint and risk assessment; in return they receive tailored risk metrics to guide sourcing; countries leverage insights to design water-secure supply and policy; and consumers benefit from transparency (e.g., water footprint labels). Such coordination aims to conserve water resources and secure food supply chains, placing water centrally in public and private governance.

This work provides a high-resolution, company-specific mapping of Brazilian soy virtual water trade linking 1620 municipalities to the top-10 importers and quantifies how a handful of TNCs displace volumes of virtual water comparable to or exceeding most countries. It documents rising VWT despite declining uWF due to yield gains and shows pronounced sub-national heterogeneity in uWF and drought risk that shapes both company and country exposures. The study contributes data, methods, and a stewardship framework to align TNC actions and national policies for sustainable water use and resilient supply chains. Future research should integrate multiple environmental dimensions (water, carbon, biodiversity) to illuminate synergies and trade-offs, extend coverage beyond the top importers and companies, track dynamic responses to climate variability and deforestation, and develop decision-support tools for risk-aware sourcing and policy design.

• Scope limitations: Analysis focuses on top-10 importing countries and top-9 dominant trading companies; results may not represent the entire global network of Brazilian soy trade. - Data dependencies: Trade links rely on FAOSTAT and TRASE; discrepancies exist (mean weighted variation ~5%). Spatial production and irrigation data use MIRCA2000, SPAM2010/MapSPAM approximations for study years, potentially introducing temporal mismatches. - Modeling uncertainty: Actual evapotranspiration and yield estimates entail uncertainties (CV_ET ~15%, CV_Y ~8%) propagating to uWF (~17%) and VWT (~18%). Some municipalities required imputation from nearest neighbors. - Risk metric simplification: Drought exposure uses scPDSI thresholds as a single proxy and may not capture all dimensions of water risk (e.g., competition among sectors, infrastructure, governance). - Long-term feedbacks: While expansion into less dry areas reduced short-term drought exposure, potential long-term hydrological impacts from deforestation (e.g., altered rainfall, soil moisture) are acknowledged but not fully quantified within this analysis.