Navigating Data Challenges and Compliance for Deforestation-Free Supply Chains

Last but not least, there is a lack of robust protocols and best practices to address ethical issues in data sharing. While information about farmers is often the first step of data gathering required for traceability systems, many farmers, particularly smallholders, are not in charge of their own data and are thus unable to reap the benefits of data sharing, raising concerns about data ownership and rights, privacy and security. This makes it critical to ensure farmers give free, prior and informed consent for their data to be collected and shared. Any data sharing process should account for potential unforeseen impacts on farmer livelihoods, land rights and legal prosecution implications.

Solving the geolocation data gap

As a result of the barriers to data sharing, geolocation data for commodities may not be available, accessible or usable.

The good news is that there are already multi-stakeholder collaboration initiatives underway to address these gaps, including efforts to centralize data collection and processing through asset registries, which are comprehensive databases that can play a crucial role in traceability. These registries provide a unified platform for recording information on land parcels, forest reserves and agricultural holdings, for example.

Photo by Kelly M Lacy

Artificial intelligence is being harnessed in agricultural management for automated field boundary detection, eliminating the time- and cost-intensive process of field verification. Agricultural tech companies are also developing mapping applications, tailored for smallholders, to aid traceability across supply chains.

To help scale these efforts, FDaP is convening stakeholders to ensure more coordination around universal asset registries, making data collection efforts more efficient and enabling the standardization of data.

Forest and Land Monitoring

Accurate and timely forest monitoring is essential to prevent commodity-driven deforestation and to support effective restoration efforts. For regulation compliance, supply chain actors need to harness a variety of global and local data sources and weigh tradeoffs between accuracy, precision, transparency and cost-effectiveness of available data.

Addressing two key challenges will help stakeholders in their monitoring efforts:

Aligning definitions and classification systems

When is a forest a forest? Where does a forest start and end? How do you define and quantify forest degradation? Current data on land cover and land use is produced using definitions that can differ significantly across countries, institutions and communities.

These divergent definitions not only lead to inconsistencies in assessment, but they can complicate the efficient implementation, monitoring and tracking of global progress on deforestation and restoration goals.

There are initiatives that have long sought to address this issue, like the Accountability Framework Initiative (AFi), which aims to establish alignment on key terms and definitions so that companies can use consistent definitions when setting goals, developing policies and procedures, managing supply chains and monitoring and reporting on progress. These definitions, reflecting the consensus of AFi’s broad civil society coalition, are grounded in international norms and best practices.

While the European Union Regulation on Deforestation-free products (EUDR) follows the FAO (the Food and Agriculture Organization of the United Nations) definition of forest, it may not align with those used by available forest cover data products. This is particularly true with national data products, which can be a useful source of information for risk assessment though they tend to use their own national definition of forest.

Photo by Waren Brasse

In addition, there is no standard land cover classification system, though increasingly there is streamlined use of Land Cover Meta Language (LCML) as an internationally recognized standard enabling the consistent creation and description of land cover classes. The LCML provides a common reference structure for the comparison and integration of data for any generic land cover classification system.

It is clear that these challenges need to be tackled collectively rather than in competition. This is why FDaP is convening organizations, governments and private sector partners to enable the development of a consistent geospatial data ecosystem that will allow all stakeholders to access open source, validated data that can be used for the monitoring, verification and disclosure of progress in reducing deforestation and advancing restoration.

Land use mapping

Democratized access to cloud computing and machine learning methods over the past decade has also led to the proliferation of land cover data monitoring products at national, regional or global scales, while land use maps are still rarely found. Yet most regulatory frameworks and tenure registries are based on land use concepts (like forest or agriculture) rather than land cover concepts (like tree cover or short vegetation).

The emergence of high spatio-temporal density imagery and deep learning methods provide an opportunity to shift from land cover to land use mapping, provided that high-quality data is used to train models. FDaP is building data-driven methods to create community commodity maps . The community modelling approach harmonizes training data from various stakeholders and leverages machine learning to produce openly available modes and data products. The first example is the community palm model, available on GitHub, viewable online, and in FDaP’s Google Earth Engine publisher data catalog, accompanied by a publication describing the methodology.

FDaP is also testing the community model approach to create models and data sets for other EUDR-covered commodities, including those grown in complex agroforestry systems such as shaded cocoa or coffee. The iterative approach means these models and data products evolve and improve over time as more data is contributed.

The community palm model, as viewed in FDaP’s Google Earth Engine publisher data catalog

Aligning for a shared data ecosystem

FDaP aims to address the challenges discussed above through a five-step framework of engagement, alignment, innovation, deployment and assessment. Since its initiation in 2021, FDaP has been engaging with relevant stakeholders across different sectors and commodities through webinars and workshops.

Through these stakeholder consultations, FDaP is aligning parties around key data sets in the ever-expanding landscape of forest monitoring data and identifying critical data gaps. Initial focal areas are cocoa in West Africa, palm oil in Southeast Asia and cattle and soy in the Amazon Basin, as well as restoration across these geographies.

This alignment process is important for fostering greater buy-in, trust and collaboration. The process will also engage local stakeholders, such as smallholder farmers and communities, to ensure inclusive representation.

Photo by Kyle Hinkson

While we have discussed potential data solutions that can help to tackle the challenges discussed, it is also important to see these as pieces of a much larger puzzle. Imagine that these pieces (data solutions) are being produced independently of one another so that many of them do not fit together to form a cohesive picture. This lack of interoperability is a critical roadblock to advancing traceability in an efficient manner.

FDaP, together with SAFE (Sustainable Agriculture for Forest Ecosystems) and the Linux Foundation, is working on a digital public infrastructure through the development and deployment of open-source building blocks like geolocation standards, protocols or models.