A coalition of organizations has assessed how locally produced maps stack up against global open-access data sets to evaluate deforestation in the context of cocoa production. The assessment will be useful for cocoa producers as they work toward compliance with the EU anti-deforestation regulation (EUDR), set to take effect at the end of the year, the organizations note in a recent report.
The EUDR’s goal is to prevent tropical deforestation by requiring that companies prove their products, including palm oil, timber, soy, beef and cocoa, didn’t come from land deforested after December 2020. Initially set to go into effect on Dec. 30, 2024, the rule’s implementation was delayed by a year partly because commodity producers struggle to accurately prove their products meet EUDR standards. Without accurate data and maps, producers could be unfairly locked out of European markets, the researchers note.
EUDR verification has been especially challenging for smallholder producers, Louis Reymondin, a senior scientist with the nonprofits Alliance of Biodiversity International (ABI) and the Center for Tropical Agriculture (CIAT), told Mongabay by email.
To tackle this issue, researchers from ABI and CIAT teamed up with the World Cocoa Foundation, a U.S.-based industry group with members including Nestlé and Mars. They tested data sets from global open-access platforms like Global Forest Watch alongside locally produced maps and one commercial provider, Satelligence, to see how well each could distinguish between cacao plantations, forests and other natural areas in the cacao-growing regions of Ghana and Côte d’Ivoire.
Each data point was independently verified using high-resolution satellite imagery, detailed enough to see individual trees. Each high-resolution image was then reviewed by at least four interpreters to reduce bias.
The study found that global open-source data sets are generally more standardized and easier to access but struggle to accurately distinguish complex landscapes, like cacao and natural forests, which, from a distance, can look very similar. “As a result, they are not recommended for deforestation-free compliance in such contexts,” Reymondin said.
On the other hand, locally made maps, produced by national governments, projects or private companies, tend to be more accurate in complex landscapes. “The classification models behind them are typically trained using ground-truth data collected directly from the landscape in question,” Reymondin said.
However, they typically aren’t as standardized and can be harder to access and use.
“None of the global datasets we assessed have a clear strategy for incorporating local knowledge or farmer input, despite the fact that such insight is essential for producing locally accurate maps,” Reymondin said.
The report’s authors recommend a hybrid approach that includes local data “built upon globally agreed best practices,” Michael Matarasso, impact director and head of North America at the World Cocoa Foundation, told Mongabay by email.
“With EUDR compliance just a handful of months away, we hope that our study will support companies to make informed decisions and use the most accurate data and analyses available,” Matarasso said.
Banner photo: A cacao farm, courtesy of World Cocoa Foundation.