Earth

About this dashboard

Tracking deforestation
from forest to fork

Every year, millions of hectares of forests are cleared for agriculture and timber. We track where this deforestation happens and which commodities drive it — and connect these production footprints to the countries and consumers whose demand ultimately bears responsibility.

Dataset at a glance

180+ countries covered

180+ commodities tracked

2001–2023 time period covered

9,300+ unique deforestation-carbon footprints

Next update: Sept – Oct 2026

Overview

What is this dashboard?

This dashboard presents the outputs of the Deforestation Driver & Carbon Emissions (DeDuCE) model — a global dataset linking commodity production to forest loss — and integrated trade models. It is designed for researchers, journalists, policymakers, and companies who need to understand where deforestation happens, what drives it, and how demand in one country connects to forest loss in another.

The Deforestation tab shows annual deforestation and carbon emissions attributed to over 180 commodities across more than 180 countries for 2001–2023. The Trade tab shows how this production-side footprint translates into deforestation embedded in imports and consumption — connecting forest loss to the food and goods traded across borders — estimated using two complementary trade models for 2005–2023.

Methodology

Linking commodity production to deforestation

The DeDuCE model identifies deforestation — the permanent replacement of natural forest by cropland, pasture, or plantation — and attributes it to the commodity that caused it, then estimates the carbon dioxide released as a result.

Where satellite data on specific crops is available (soybeans, oil palm, cocoa, rubber), attribution is direct and spatially explicit. Where it isn't, the model uses a two-step statistical approach: broad land-use data divides clearing between cropland, pasture, and plantation; then national harvest statistics allocate the cropland share between individual commodities.

Carbon emissions are estimated using maps of forest carbon stocks — above- and below-ground biomass, including soil organic carbon, and statistically estimating emissions from dead wood and litter, minus the net of carbon sequestered by the replacing land use. Emissions from peatland drainage are estimated separately by overlaying deforestation with a global peatland map.

Detect tree cover loss

Satellite imagery identifies loss of trees globally at 30m resolution

Identify forest loss

Combination of multiple spatial and statistical datasets help identify loss of natural forests from other conversions

Attribute to specific commodities

Crop maps and agricultural commodity statistics determine what replaced the forest

Estimate carbon emissions

Forest and soil carbon stock maps quantify biomass loss and peat drainage emissions

Methodology

Connecting deforestation to trade and consumption

DeDuCE provides the production-side deforestation estimates — quantifying which commodity cleared which forest and how much carbon was released. Two separate trade models then use these estimates as inputs to link deforestation across supply chains, connecting production-side clearing to the countries and consumers whose demand ultimately drives it. Deforestation is amortised over five years before being linked to trade, reflecting that cleared land typically produces commodities for multiple years.

Model 1

Physical Trade Model

Uses FAOSTAT bilateral trade data in physical units to track agricultural commodities across international supply chains, accounting for processing steps and re-exports.

✓ Best for identifying first-tier supply-chain pathways and direct commodity flows

✓ Retains commodity-level resolution throughout the supply chain

! Limited coverage of processed products — tracing may not reach final consumption

Model 2

Hybrid MRIO Model

Combines physical trade data with economy-wide monetary flows (a multi-regional input–output framework) to track embodied deforestation all the way to final consumption.

✓ Captures full consumption-based footprints, including complex indirect links

✓ Retains commodity-specific footprints for consuming countries

! More aggregated demand-side resolution; indirect links can be harder to target through policy

The two models are complementary, not interchangeable. Use the physical model to identify supply-chain exposure; use the hybrid model for consumption-based national accounting.

Data quality

Understanding the quality index

Each country-commodity estimate from DeDuCE carries an Integrated Quality Index (IQI). Before interpreting any estimate, check the IQI — it tells you how the number was produced and how much confidence to place in it.

IQI > 0.6 — Direct land-use change (dLUC)

Attribution is based on commodity-specific satellite data. You can be reasonably confident that this commodity directly drove the clearing. The estimate reflects observed land-use change, not modelled inference.

IQI ≤ 0.6 — Statistical land-use change (sLUC)

Attribution relies on national agricultural statistics, sometimes of variable quality. Treat as a signal of deforestation risk, not confirmed direct clearing. Useful for ranking exposure, not confirming sourcing.

Scope & limitations

A few things to keep in mind

The data is built from the best available inputs, but every methodology has scope boundaries. Here are the three most important ones:

Footprints mix direct and statistical attribution

Most country-commodity footprints are estimated via a land-balance model at national level (sub-national for Brazil). Results combine direct land-use change (dLUC) derived from satellite data and statistical land-use change (sLUC) from agricultural statistics. The quality index tells you which applies to each estimate.

Trade figures reflect national averages, not specific sourcing

Deforestation embedded in trade and consumption uses national production averages and cannot indicate whether a specific importer sources from deforested or deforestation-free regions. Treat import estimates as deforestation risk exposure — a signal of where supply-chain due diligence is warranted.

The two trade models answer different questions

The physical model covers direct agricultural commodity flows but excludes highly processed products and indirect pathways. The hybrid MRIO captures these indirect links but sacrifices demand-side resolution. The models are complementary, not interchangeable — do not add or directly compare their numerical outputs.

Citations

How to cite this work

If you use data from this dashboard in research, reporting, or policy work, please cite both the dataset and the relevant methodological publication(s). Separate citations are needed because the dataset and the underlying models are independent intellectual contributions.

Cite the dataset

Dataset · Zenodo · 2026
Singh, C., Persson, U. M., Croft, S., Kastner, T., & West, C. D. (2026). Commodity-driven deforestation, associated carbon emissions and trade 2001–2023 (Version 2.1) [Dataset]. Zenodo. https://doi.org/10.5281/zenodo.18953516

Cite the methodology

Deforestation attribution model · Nature Food · 2026
Singh, C., & Persson, U. M. (2026). Global patterns of commodity-driven deforestation and associated carbon emissions. Nature Food, 7(2), 138–151. https://doi.org/10.1038/s43016-026-01305-4

View model code on GitHub →

Hybrid trade model · Journal of Cleaner Production · 2018
Croft, S. A., West, C. D., & Green, J. M. (2018). Capturing the heterogeneity of sub-national production in global trade flows. Journal of Cleaner Production, 203, 1106–1118. https://doi.org/10.1016/j.jclepro.2018.08.267

Physical trade model · Ecological Economics · 2011
Kastner, T., Kastner, M., & Nonhebel, S. (2011). Tracing distant environmental impacts of agricultural products from a consumer perspective. Ecological Economics, 70(6), 1032–1040. https://doi.org/10.1016/j.ecolecon.2011.01.012

Versions

Changelogs

A record of major updates to the dataset on this dashboard.

2001-2023 v2.1

Updated and integrated the latest FAOSTAT, IBGE, and MapBiomas dataset versions.
Replaced the Curtis et al. 2018 forest loss driver dataset with the newer Sims et al. 2025 dominant driver dataset.
Expanded coverage to four additional territories (French Guiana, Guadeloupe, Martinique, Réunion); note that deforestation attribution for these regions is only available up to 2006 due to data limitations.
Switched the MRIO framework from EXIOBASE to GLORIA for hybrid trade model.

Released March 11, 2026

zenodo.org/records/18953516 →

2001–2022 v2.0

Used DeDuCE estimates (2001–2022) for commodity-driven deforestation and associated carbon emissions, and tracked impacts embedded in global supply chains by adapting the methodology from Pendrill et al., 2019.

Released April 22, 2024

zenodo.org/records/10633818 →

FAQs

Frequently asked questions

Deforestation

What do you mean by “deforestation” (and how is it different from “tree cover loss”)? +

Tree cover loss is a remote-sensing signal (trees disappeared in a pixel), which can include temporary losses (e.g., rotational harvesting) as well as permanent conversion. Our data focuses on deforestation as permanent forest conversion linked to productive land uses (like cropland, pasture, and plantations). For our use case, we define tree-cover loss as areas with ≥25% canopy cover and integrate multiple datasets to separate natural forest loss from other types of tree-cover loss (e.g., plantations).

What does a high or low quality index (IQI) mean for deforestation estimates, and how I should use the data? +

IQI is a deforestation quality score: it rises when the estimate is built from explicit, high-resolution, commodity-specific datasets, and falls when it relies more on aggregated statistics or gap-filled inputs. On this dashboard we use 0.6 as a practical rule-of-thumb cutoff: above 0.6 = more spatially explicit commodity linkage; 0.6 or below = more statistical attribution, best used as indicative risk exposure.

Don't see your question answered? Send it to us directly.

Submit a question →

Legal

Terms of use

Data and visualisations are freely reusable — just cite us. Charts and maps are licensed under CC BY 4.0; datasets on Zenodo are governed by their individual release licences.

Visualisations Charts, maps, and graphics are available under Creative Commons Attribution 4.0 International (CC BY 4.0). You may reuse, adapt, and republish with attribution.

Downloaded data & code Governed by the licence stated with each Zenodo release or GitHub repository. Where a specific licence is stated, it takes precedence.

Attribution Any reuse should credit DeforestationFootprint.earth and include a link back to the site. Use the preferred citations provided above.

Third-party sources Model outputs incorporate third-party datasets under their own terms. Users are responsible for compliance with applicable third-party licences.

Disclaimer Data are provided "as is". We apply quality checks and document known limitations, but accept no liability for decisions made based on the site.

Corrections Spotted an error? Email us at deduce.env@chalmers.se — we take data quality seriously.

Acknowledgements

Funding

The DeDuCE model and other dataset on this dashboard have been made possible through support from the following funding sources:

ÅForsk Foundation — Grant 22-64 · Refining consumption-based estimates of Deforestation land-Use Change Emissions (ReDUCE)

Belmont Forum via FORMAS — Grant 2022-02563 · Building an evidence-base for deforestation-free landscapes (BEDROCK)

German Federal Ministry for Economic Cooperation and Development — Grant GS22 E1070-0060/029 · Global impacts of agricultural trade and consumption on ecosystems and biodiversity: exploring Germany's and China's roles as importing and consuming countries (GRADED)

UK Research and Innovation (UKRI) — Grant ES/S008160/1, Global Challenges Research Fund · Trade, Development and the Environment Hub

DEFRA & Joint Nature Conservation Committee (JNCC) — Support for the development of the Global Environmental Impacts of Consumption (GEIC) indicator

We are grateful to colleagues, partners, and users of the data — within and beyond academia — who have contributed data, insights, and feedback over many years of development.

Collaboration