The carbon footprint of your chocolate is probably wrong.

TMF shows that about 50% of cocoa plantations in Ghana are linked to deforestation, about 3 times more than indicated by Harris.

The carbon footprint of your chocolate is probably wrong — because your land conversion footprint for cocoa is probably wrong.

Land conversion emissions dominate the carbon footprint of cocoa systems. Getting the land conversion data right → is key to getting the carbon accounting right → which is key to getting the priorities right.

While we know that “all models are wrong”, there’s also a spectrum of “more wrong” to “less wrong” that gets us closer to or further from reality.

After some testing with our product Orbae*, what we see for cocoa is that

1. Shifting your approach from statistical land use change (sLUC) to jurisdictional direct land use change (jdLUC)** can significantly influence the footprint.
2. Common data sources used to detect land conversion paint a picture of reality that’s not fit for purpose for carbon accounting (i.e., your LUC emissions might actually be double what you think they are).
3. The subnational spatial patterns we observe drive home the need for traceability, as soon to be required by the EUDR.

Testing publicly available data sources

Cocoa is grown in smallholder systems and partially under shade trees, making deforestation and land conversion from its cultivation hard to spot from satellite imagery.

We wanted to see how different publicly available data sources performed at detecting land conversion from cocoa production in Ghana and Côte d’Ivoire over the past 20 years, and so we put three of them to the test using the ETH cocoa plantation mask (Kalischek et al. 2023):

• Forest GHG emission data from Harris et al. (2021), which is derived from the GFW tree cover loss dataset from Hansen et al. (2013) (TCL)
• The primary forest layer from Turubanova et al. (2018)
• The tropical moist forest (TMF) layer from Vancutsem et al. (2021)

TMF, fit for purpose

What we found is that neither the TCL layer nor the primary forest layer are adequate for carbon accounting. It’s the TMF layer that’s the best fit. Why?

✖ Using tree cover as a forest proxy does not distinguish between forest and cocoa plantations.

✖ Using TCL as a deforestation proxy can lead to false positives (e.g., cocoa plantation renewal, which is classified as tree cover loss) and false negatives (e.g., the TCL layer doesn’t detect tree cover loss for cocoa grown under forest cover).

✖ Using primary forest ignores other types of land conversion — primarily the conversion of secondary forests and degraded forests, which is typically a large part of the land conversion we see in West Africa.

TMF gives us a more consistent and comprehensive assessment of the land conversion associated with cocoa. It makes it possible to “get closer to reality” with publicly available, peer reviewed data.

✅ TMF can accurately observe “undisturbed forest” and any disturbance events since 1990, which likely represent land conversion (deforestation) events.

✅ Overlaying TMF with the ETH cocoa cropping mask shows that TMF accurately reflects the conversion of undisturbed forest into cocoa plantations.

✅ TMF can explain how cocoa production in Ghana and Côte d’Ivoire increased by 50% in the past 20 years. Land expansion is likely the main explanation for the increase in production.

Highlights of some of our findings

There’s a lot to digest here, and we’d be happy to talk it through with you. Feel free to send us a message.

*Orbae automatically produces high-fidelity data on land conversion for agricultural commodities across the globe, at any level of traceability — from farm to country. For cocoa, Orbae relies on the TMF layer. Find out more at orbae.eco.

**Jurisdictional direct land use change (jdLUC), which is used by Orbae, is land conversion within a jurisdiction that can be identified with a comprehensive crop mask (i.e., crop classification) on a resolution that is higher than a typical plot size.