Trase methodology: A step-by-step introduction

Do Pasto ao Prato (DPaP) is Trase's sister initiative, a mobile application increasing transparency in the Brazilian domestic market for beef. It uses a bottom-up mapping approach to allow Brazilian consumers to identify the origin of the beef available in supermarkets. By utilising crowdsourced subnational data, the app connects meat retailers to the specific meat processing plants and municipalities where the cattle were raised. To assess these supply chains, the app evaluates each slaughterhouse based on five sets of indicators: the audited performance of their sustainable supply chain commitments, deforestation, and the occurrence of fires within their supply zones, their sanitary and animal welfare performance, and possible cases of slave labor among their suppliers.

Find out more about DPAP

We attribute deforestation and associated greenhouse gas emissions to companies and countries by linking their commodity sourcing regions identified in their supply chains to the environmental impacts of commodity production within those regions. Our assessments use the highest spatial resolution supported by the available data.

For example, when we have full traceability to plot-level or concession of production, and spatially explicit commodity attributed deforestation and greenhouse gas emissions data, we directly attribute the deforestation and emissions occurring on the production area to the companies and countries along the supply chain. Trase uses this approach for its wood-pulp supply chains.

More commonly, production can be traced to a jurisdiction (such as a municipality, district or province), or country rather than to individual farms or concessions. In these cases, we attribute the deforestation and emissions observed within that jurisdiction or country to those companies or countries in proportion to the volume they source from that jurisdiction. The resulting estimates are referred to as deforestation and emissions exposure, reflecting the environmental impacts embedded in supply chains rather than the directly observed impacts attributable to a specific actor. We use this approach across our soy, beef, palm and cocoa supply chain assessments, and for linking our DeDuCE data to global trade supply chain data.

Global impact metrics: At the global scale, we use the Deforestation Driver & Carbon Emissions (DeDuCE) model, offering a globally homogenous and comprehensive assessment of deforestation and carbon emissions linked to over 180 commodities across more than 180 countries. It combines tree cover loss, land use and commodity maps, as well as dominant tree cover loss data derived from satellite imagery with statistical land-use data to attribute deforestation to specific agricultural commodities. Aiming for global comparability, DeDuCE deforestation data builds on Global Forest Change datasets, offering global deforestation mapping.

Subnational impact metrics: At subnational level, we aim to use the best available regional data for each country-commodity context tailoring our impact metrics to maximize regional representation and contextual relevance. For example, for Brazil we combine PRODES and MapBiomas deforestation data leveraging producing countries own data products for the assessment. For Indonesia we use data developed by The Treemap, providing the most accurate and up-to-date estimates of commodity driven deforestation. For cocoa deforestation, we employ the Tropical Moist Forest (TMF) dataset, including forest degradation, to ensure representation of the impact of shade grown cocoa, often mapped as degraded forests, rather than deforestation.

Carbon is stored as biomass in natural ecosystems (e.g. forests, savannahs, grasslands), as well as in mineral soils and peatlands. Once the biomass of terrestrial natural ecosystems is cleared or peatlands are drained to convert to cropland, pasture or plantations, greenhouse gas emissions are released into the atmosphere, accelerating climate change.

Gross greenhouse gas (gross GHG) emissions refers to the amount of carbon removed (in CO2-eq per year) released following the conversion of natural ecosystems in cropland, plantation or pasture without accounting for the biomass accumulated by the new land use through carbon sequestration. In contrast, net GHG emissions consider the carbon accumulated in the new land use through sequestration which are subtracted from the gross GHG emissions.

The Greenhouse Gas Protocol released in early 2026 specifies standards for companies in the forestry, land and agricultural sector about how to account for greenhouse gas emissions from land-use change. One of the key characteristics of this guidance is the 20-year land-use change assessment period (Greenhouse Gas Protocol Land Sector and Removals Standard, 2026).

Differences in the definition of what constitutes a forest impacts the outcome of any deforestation assessment.

Global forest definition: Our global DeDuCE model defines a forest as woody vegetation exceeding five meters in height with at least a 25% tree cover threshold.

Subnational forest definition: Within our subnational contexts we use an inclusive definition, categorizing forest as all native vegetation, meaning it encompasses standard forests, savannahs and grasslands. We aim to use best regionally and commodity representative datasets like Prodes and MapBiomas for Brazil, data developed by The TreeMap for Indonesia, and Tropical Moist Forest data developed by the Joint research Center of the European commission for Côte d'Ivoire.

Attributing deforestation to a specific commodity is challenging because land is often cleared before planting, land use may only become visible in satellite imagery after a delay, and crop maps are not always available to directly observe conversion.

To address these challenges, we use peer-reviewed spatial and statistical attribution methods. Spatial attribution links deforestation to subsequent commodity expansion using deforestation and commodity maps, while statistical attribution allocates deforestation based on changes in crop area and production where spatial data are unavailable or incomplete. The approach and timeframe vary by commodity and data availability.

Global deforestation attribution: The DeDuCE model combines spatial and statistical attribution, prioritising spatial attribution where suitable data are available. Spatial attribution typically links crop expansion occurring within three years of a deforestation event to the clearing. Statistical attribution allocates deforestation according to observed crop expansion dynamics derived from agricultural statistics.

Subnational deforestation attribution:

• Palm and wood pulp: Deforestation is attributed within a one-year window following clearing. This reflects the rapid conversion of forest to plantation, which can often be detected in satellite imagery shortly after deforestation.
• Soy and pasture (cattle/beef): Deforestation is attributed within five years of clearing to account for the time required for land preparation, establishment of production and reliable identification of land use from satellite data.
• Cocoa (Côte d’Ivoire & Ghana): Deforestation is assessed by directly overlaying deforestation data with mapped cocoa-growing areas.

Commodity-driven deforestation and emissions: This describes the physical conversion of native vegetation into agricultural land. It is based on the year the deforestation event occurred and is often used by governments for prioritizing local policy interventions.

Commodity-attributed deforestation and emissions: This tracks the environmental impact embedded in the actual product being sold (tied to the harvest year). This factors in the time lag between when the land was cleared, when the crop was planted, and when it was finally harvested. Additionally, in our subnational mapping, we look at long-term historical impacts by reporting the five, and ten-year cumulative deforestation embedded in the harvest.

The lag times differ significantly by commodity:

• Beef: Immediate (we assume cattle can roam deforested land immediately).
• Soybeans: One-year lag.
• Palm: Three-year lag.
• Cocoa: Four-year lag.
• Wood-pulp: Six-year lag.

Amortization means spreading the impact of a deforestation event or GHG emissions over multiple years of commodity production. Because land that is cleared once continues to be agriculturally productive over multiple years, it is common to attribute a fraction of this deforestation to multiple harvests, rather than placing 100% of the impact on the very first harvest. Trase typically uses a five-year amortization period.

We assess companies using the same method as Global Canopy’s Forest 500. We assess companies using information published on their websites in the year for which they are being assessed. We report companies as having a zero-deforestation commitment if they have a commitment to ‘deforestation-free sourcing’, understood as no loss of natural forests anywhere. Companies only committing to zero net deforestation, understood as a commitment to offset forest loss through forest restoration or companies trading deforestation-free certified, without providing indication of their corporate commitment to zero deforestation do not score for this indicator. A company cannot only aim to ‘reduce or avoid deforestation’.

To create our facilities datasets, we compile geolocated records of facilities using a combination of official government registries, corporate disclosures and advanced satellite mapping:

• Brazil (soy and livestock): We merge data from federal and state sanitary inspection systems (such as SIF, SIE, SIM, and SISBI-POA) to see which slaughterhouses are legally authorized to sell and export products. To find unlisted facilities, we use machine learning tools applied to satellite embeddings, such as the AlphaEarth foundation model, which our analysts then verify.
• Côte d'Ivoire (cocoa): We use public disclosures from trading and manufacturing companies and external accountability maps. When geographic data is missing, we conduct online searches to match town or city names embedded within a cooperative's name to map it to a specific administrative department.
• Indonesia (pulp & palm): We compile pulp wood concessions using official Forest Utilization Licenses spatial dataset and government wood supply plans. Pulp mills are compiled using information from the government wood supply plans and corporate sustainability reports. The palm oil mill list is built off the Universal Mill List and updated using information disclosed by companies or the government, such as certification reports, mill supplier lists, plantation statistics documents and corporate sustainability reports. Palm oil refineries are mapped via desk-based research.

When official capacities are not publicly reported, we use data estimation techniques. For example, for Brazilian soy facilities without declared storage capacities, we calculate the average capacity from known facilities and apply it as a baseline. For Côte d'Ivoire cocoa, we estimate a cooperative’s minimum farmer base by adding up the sizes of disclosed supply links from corporate buyers, sometimes relying on historical data or Monte Carlo estimates to fill in missing years.

We associate individual facilities to their parent companies or corporate groups by reviewing corporate sustainability reports, audited financial documents and official corporate registries. We also use data enrichment AI tools to retrieve local tax IDs and resolve ambiguities between a facility's informal trading name and its formal corporate legal title.