To make reliable estimates of the carbon footprints associated with agricultural commodities requires capturing a wide diversity of conditions along global supply chains. However, many such calculations rely instead on impact factors that are uniform at national or even global averages.

This paper presents a new bottom-up approach for quantifying the greenhouse gas (GHG) emissions embedded in the production and trade of agricultural products with high spatial resolution. It integrates life cycle analysis (LCA) principles with enhanced physical trade flow analysis.

Applying the new approach the authors estimates the carbon footprint (as tonnes of carbon dioxide equivalent per ton of product) of Brazilian soy exports over the period 2010–2015, differentiating ~90,000 individual traded flows of beans, oil and protein cake from the municipality of origin through to international markets.

The findings reveal an extremely large variability in carbon footprints of soy depending on the sourcing area, importing country and sub-stages along the supply chain.

The largest carbon footprints are associated with municipalities in the MATOPIBA states (Maranhao, Tocantins, Piaui and Bahia), and Pará, where soy is directly linked to natural vegetation loss. Importing soy from these states entailed up to six times greater emissions per unit of product than the Brazilian average (0.69 t t−1).

Soy imports to the European Union (EU) had the largest carbon footprint per ton (0.77 t t−1), and the largest total deforestation-related footprint.

The total GHG emissions associated with Brazilian soy exports in 2010–2015 were estimated at 223.46 million tons, of which more than half were imported by China, the number-one importer of Brazilian soy.

The new approach can contribute data for enhanced environmental stewardship across supply chains at the local, regional, national and international scales, while informing the debate on global responsibility for the impacts of agricultural production and trade.