Back
Technology

Taylor Geospatial Releases Global Agricultural Field Boundary Map

View source

Global First: Taylor Geospatial Publishes Complete Map of Every Agricultural Field on Earth

This is the first global dataset of its kind, mapping agricultural field boundaries at a 10-meter resolution.

Taylor Geospatial has officially released a groundbreaking map detailing the boundaries of every agricultural field on Earth. Developed in collaboration with Microsoft's AI for Good Lab, the data is now publicly available.

Key Details of the Dataset
  • Innovative Technology: The map was created using a novel model and architecture specifically designed to infer field boundaries on a global scale.
  • Massive Infrastructure: The effort required significant cloud computing resources to process the immense volume of data.
  • High Resolution: The dataset provides field boundaries at a precise 10-meter resolution.
  • Collaborative Effort: Contributions came from researchers at Arizona State University, Washington University, Clark University, Source Cooperative, Wherobots, and the Taylor Geospatial Technical Fellows.
Background: Filling a Critical Gap

Previous attempts to create such a map failed, leaving a critical void for precision agriculture, food security monitoring, carbon accounting, and climate adaptation planning. This new dataset finally provides the foundational layer needed for these applications.

Expert Statements and Reception

Taylor Geospatial described the work as "one of the most ambitious GeoAI efforts" aimed at ensuring that GeoAI technology works effectively across the entire globe, not just in data-rich regions.

Early reception has been positive, with users praising the dataset for allowing analysis to shift from individual pixels to entire fields, and for enabling robust change detection. However, some users noted significant gaps in coverage for specific regions, including:

  • Eastern Ecuador
  • Parts of Finland
  • Parts of Sweden and Norway
Availability & Next Steps

The tools and research behind the map are fully open source.

  • The Model: The practical recipe, known as PRUE (Practical Recipe for Field Boundary Segmentation at Scale), is published as a preprint, with the code available on GitHub.
  • The Paper: A detailed paper on the global field boundaries, including evaluation metrics, has also been published.

Moving forward, Taylor Geospatial is partnering with NASA Harvest, the FAO (Food and Agriculture Organization), and other key organizations to distribute and refine this critical data.