WRI, with support from the Walmart Foundation, is developing an innovative toolkit to aid commodity traceability and enable the protection of vital landscapes.

Field boundaries are critical in the agri-food sector to support monitoring, risk assessment and compliance with sustainability commitments and deforestation regulations. However, accurately detecting them remains a huge challenge and current approaches for building field boundary datasets are often inadequate.

Researchers from Land & Carbon Lab are leveraging the latest in artificial intelligence (AI) and geospatial monitoring to create an open-source AI toolkit for field boundary detection. It will be a scalable tool for monitoring field operations and for supply chain actors to realize sustainability commitments and comply with the European Union’s Deforestation Regulation (EUDR). WRI has partnered with Dr. Hannah Kerner's lab at Arizona State University (ASU) to implement, test, verify and iterate this field boundary detection model focused on soy and potentially other agricultural supply chains in Latin America.

The data will also be a valuable input to Land and Carbon Lab’s crop-mapping initiative that allows for a better understanding of trade-offs between nature and food, informing solutions for more sustainable agricultural production.

The toolkit will be made available as a public good to enable broad adoption and traceability at scale, unlocking opportunities for supply chain actors to drive progress toward reducing deforestation, forest degradation and ecosystem conversion.

Recent advances in AI, most notably the advent of large, foundational models for geospatial tasks, give us a new way to tackle the challenge of field boundary detection. Our vision is to build on this scientific and technical momentum to create a toolkit for field boundary detection that can be used by actors across supply chains and sectors.

The model will help to automate field boundary identification, saving countless hours of costly administrative and field work. Speeding up and lowering the cost of traceability efforts will enable more supply actors to deliver on their corporate sustainability commitments and comply with regulations like the EUDR.

The project has two primary objectives:

  • Develop a foundational, open-source AI model for field boundary detection that enables reliable field boundary delineation and disambiguation across multiple agricultural value chains and geographies.
  • Test the model in real world settings to verify accuracy, and refine it to meet user needs for monitoring soy and potentially other agricultural supply chains in Latin America.

Project Funder

Walmart Foundation Logo