Background

This research project develops AI-based tools to remotely estimate shade-tree cover and carbon stocks in cocoa farms, enabling spatially explicit agroforestry recommendations.

Publications

Key figures

  • Country and region(s): Ghana & Côte d’Ivoire

  • Project duration: October 2020 – September 2023

  • Implementation partner: ETH Zurich, University of Queensland and University of Zurich

  • Further partners: International Center for Tropical Agriculture (CIAT), University of Oxford, University of Cambridge, KNUST and Joint Cocoa Research Fund of CAOBISCO and ECA

  • Budget: CHF 330’220 (59% of total project budget)

Main objectives

This project aimed to develop innovative methods for assessing the environmental benefits of cocoa agroforestry in Ghana and Côte d’Ivoire. Specifically, it focused on:

  • Mapping shade-tree cover on cocoa farms using satellite imagery

  • Estimating above-ground carbon stocks and the carbon sequestration potential of cocoa agroforestry systems.

  • Generating spatial data products to inform on-the-ground farming decisions and support a transition toward more sustainable, climate-resilient cocoa production systems

Results

  • Two machine learning models were developed using Sentinel-2 satellite imagery and field data from 828 cocoa farms.

  • The models produced high-resolution maps showing: Shade-tree cover at 10-meter resolution & Above-ground carbon stocks at 50-meter resolution for the year 2022

  • Produced a final report, policy brief, infographics, and held a dissemination workshop

  • Average shade-tree cover was only 13.2 percent, far below the 30 percent target promoted by many chocolate companies

  • Cocoa farms currently store 117.7 million tons of carbon

  • Increasing shade cover to 30 percent could unlock an additional 84 million tons of carbon storage – roughly equivalent to the carbon stored in the region’s remaining undisturbed forests