High quality Cocoa and Coffee maps for European Union Deforestation Regulation preparedness in Ivory Coast and Kenya

The European Commission's Joint Research Centre (JRC) supports agricultural commodity-producing countries in strengthening their preparedness for the EU Deforestation Regulation (EUDR) by providing technical assistance on the production and validation of high-quality commodity maps. This support is delivered through two complementary initiatives. The first initiative, the Copernicus4GEOGLAM Validation Service, delivers independent and transparent validation of country-produced commodity maps, applying internationally recognised methodological standards. The Service has been implemented in Côte d'Ivoire and Cameroon and is currently ongoing in Honduras. The Second initiative is part of the Technical Facility on Deforestation-free Value Chains, led by the European Forest Institute (EFI) and funded by the European Union and the Agence Française de Développement (AFD). Within this framework, the JRC provides technical guidance for high quality ground data collection, EO data selection and processing, and the development of classification algorithms to produce reliable national commodity maps.

Governments, national institutions, and private sector operators in agricultural commodity-producing countries face a common challenge: companies must demonstrate compliance with the EU Deforestation Regulation (EUDR). In a nutshell, the regulation requires that commodities placed on the EU market, such as cattle, cocoa, coffee, palm oil, soya, rubber, and wood — are not associated with deforestation after 31 December 2020. However, in many producing countries, the spatial information needed to support this process is absent, outdated, or lacks independent validation, which limits their use in regulatory and decision-making contexts.

Côte d'Ivoire and Kenya — major producers of cocoa and coffee respectively — illustrate both the scale of this challenge and the potential of coordinated efforts between EO expertise and national institutions.

In Côte d'Ivoire, the Ministry of Water and Forests commissioned the Bureau National d'Études Techniques et de Développement (BNETD) to produce a national land-use map with a focus on cocoa production areas to comply with EU funded initiatives such as the Sustainable Cocoa Initiative, international fair-trade certifications (including ARS-1000) and the EUDR. The JRC subsequently provided independent validation through the Copernicus4GEOGLAM framework.

In Kenya, the government and national institutions, including the Directorate of Resource Surveys and Remote Sensing (DRSRS), the Kenya Forest Service (KFS), Kenya Space Agency (KSA), and the Regional Centre for Mapping of Resources for Development (RCMRD) — requested support from the Technical Facility on Deforestation-free Value Chains to strengthen EUDR preparedness in the area of geospatial information and EO. Within this framework, the JRC and EFI provide targeted technical support on reference data collection and the development of classification algorithms, with the aim of producing a high-quality national coffee probability map.

In Côte d'Ivoire, the JRC provided an independent validation of a country produced land use map through the Copernicus4GEOGLAM Validation Service — part of the Copernicus Land Monitoring Service. The map was generated using Sentinel-2 annual composites and field reference data in a random forest classification algorithm. The Copernicus4GEOGLAM validation used independent reference data, collected through field surveys and photo-interpretation of Very High Resolution (VHR) imagery (Pléiades and WorldView), to assess the map's accuracy, producing a publicly available validation report that documents its strengths and limitations across different land-use classes — giving operators, competent authorities, and national governments a transparent basis for deciding how and where to use it.

In Kenya, JRC supports the generation of a coffee probability map using deep learning classification with Planet imagery at 5m resolution. The map is expected to be finalised in 2026 and will represent one of the first publicly available, high-resolution national coffee maps in the continent. A key output of this collaboration is a set of methodological guidelines for collecting high-quality reference data using VHR satellite imagery — a resource designed to be replicable by national institutions across other producing countries and commodities.

The two initiatives rely on complementary EO and in-situ data. In Côte d'Ivoire, Sentinel-2 imagery provides the primary input for the generation of the land-use map, which is based on annual composites at 10m resolution. Independent validation relied on reference data gathered through a statistical stratified random sampling design for field data collection, complemented by photo-interpretation of VHR imagery.

In Kenya, Planet imagery at 5m resolution is the primary EO input for the detection of smallholder coffee farms at a spatial detail not achievable with Sentinel-2 sensors. Deep learning classification is applied to these images to produce a national-scale probability map of coffee extent.

Across both applications, VHR optical imagery from Pléiades (0.5m) and WorldView (0.5m) played a central role in reference data collection through image interpretation, enabling photo-interpreters to reliably identify crop types at the farm level and build high-quality training and validation datasets.

The validation of the BNETD 2020 land cover map for Côte d'Ivoire represents a significant advancement in transparency and regulatory confidence in West Africa. The map is the first public and freely accessible national land-use dataset in the region to be accompanied by an independent and public accuracy assessment — setting a new standard for how country-produced maps can be shared and scrutinised.

The validation results are of very high quality. The map achieves an overall accuracy of 91% using the IPCC 6-class land cover legend, and 69% at the more detailed 22-class nomenclature. The map distinguishes relevant EUDR commodity classes such as cocoa, palm groves, rubber, cashew and coconut groves. Some confusion is reported between natural forest and tree crop classes — which is expected at 10m resolution — and the validation report documents this in a transparent way, giving operators and competent authorities a clear picture of where the map performs well and where additional caution is warranted.

Beyond the validation itself, the JRC has also published practical guidance for private operators on how to use national and global land use maps to conduct an initial assessment of deforestation risk under the EUDR. Using Côte d'Ivoire as a case study, the report compares a state-of-art global forest map produced by JRC and the BNETD national land cover map, highlighting their complementary strengths and limitations, and recommending their combined use as a first step to identify sourcing areas that may require further in-depth checks.

The BNETD map has been integrated into WHISP (WHat IS in that Plot?), an open-source tool developed by the Forest Data Partnership and FAO that helps operators, third-party verifiers, and government authorities assess deforestation risk for specific plots of land under the EUDR, making high-quality spatial information accessible to all actors involved in the supply chain, including smallholder farmers with no prior geospatial knowledge.

The relevance of the BNETD map for EUDR compliance has been independently recognised by the private sector: a joint assessment by the World Cocoa Foundation and the Alliance of Bioversity International and Alliance of Bioversity International and the International Center for Tropical Agriculture (CIAT) — involving major chocolate industry operators — evaluated the map against 21 datasets and explicitly highlighted it as a benchmark for other producing countries in terms of documentation, validation, and open data management.

The Kenya coffee probability map, once finalised, is also planned for integration into WHISP, where commodity-level information for Kenya is currently missing — making its completion an important step in supporting EUDR due diligence processes in one of Africa's leading coffee-producing nations. In addition, the coffee probability map is planned to be integrated into the national coffee information system hosted by the Kenya Coffee Directorate, ensuring it serves as a long-term national resource for coffee sector management and planning.

The collaboration between the JRC, EFI and national institutions is already delivering tangible results. Working together with DRSRS, KFS, KSA and RCMRD, the JRC has developed and published methodological guidelines for high-quality reference data collection using VHR imagery as a technical report. In December 2025, the JRC and the European Forest Institute jointly trained 8 national trainers representing the different partner institutions in the application of these guidelines and field data collection. As part of the training, two coffee-growing regions were visited to familiarise with local cultivation systems and agroforestry configurations, and to demonstrate how field sensitisation campaigns should be organised prior to field data collection and VHR image interpretation. A large-scale field sensitisation campaign is currently underway to cover additional coffee-producing areas across the country, with reference data collection proceeding in parallel. To date, more than 46,000 parcels of coffee planted area have already been digitised — bringing the national coffee probability map one step closer to completion.