FTW Global — Prediction Confidence & Quality Layers (500 m)
Global 500 m raster layers that quantify the reliability of the PRUE global agricultural field-boundary predictions (Robinson et al. 2026), part of Fields of the World. Derived from the full-resolution (10 m) model outputs on a common global grid (EPSG:4326, ~0.004167 deg/px, -180..180 lon, -60..84 lat; each 500 m cell ~ 50x50 of the 10 m pixels). Five products: a modeled confidence score; field & boundary prediction density (unfiltered and confidence-filtered); model entropy; an external cropland-consensus count; and precision/recall against that cropland consensus. These layers let users assess where the field predictions can be trusted and filter the vector dataset accordingly. Temporal coverage inferred as the 2025 prediction year from the paper (the cropland-consensus layer is year-independent); pending confirmation from the data authors.
Spatial Coverage
- Bounding Box: [-180.0, -60.0, 180.0, 84.0]
- CRS: EPSG:4326
Temporal Coverage
- Start: 2024-01-01T00:00:00Z
- End: 2025-12-31T23:59:59Z
Files
STAC Metadata
- root:
../../catalog.json - parent:
../../catalog.json - self:
https://data.source.coop/ftw/global-data/predictions/confidence/collection.json
Source
https://source.coop/ftw/global-data
Processing Notes
500 m quality-indicator and confidence rasters computed from the full-resolution (10 m) PRUE model outputs on a common global grid (EPSG:4326, ~0.004167°/px, −180..180 lon, −60..84 lat; each 500 m cell ≈ 50×50 of the 10 m pixels). The modeled confidence layer is a Random Forest trained on model-internal indicators (entropy, prediction density) over the 24 FTW-labelled countries (leave-one-country-out mean AUC 0.842). Organised here as a STAC collection of five items, one per measurement; the two EPSG:3857 8-bit GeoTIFFs are web-display ("visual") assets.
Citation
Robinson, C., Muhawenayo, G., Khanal, S., Fang, Z., Corley, I., Tárano, A. M., Estes, L., Marcus, J., Jacobs, N., Kerner, H., Becker-Reshef, I., & Lavista Ferres, J. M. (2026). The first global agricultural field boundary map at 10 m resolution. arXiv:2605.11055.
Attribution
Fields of the World / PRUE (Robinson et al. 2026), CC-BY-4.0
License
Contact
Fields of the World — Taylor Geospatial isaac.corley@taylorgeospatial.org
Known Issues
Temporal coverage is inferred as the 2025 prediction year from the paper (figures and the confidence-retention curve are 2025) and is pending confirmation from the data authors; the crop_count_mean (cropland-consensus) layer is year-independent external data. The exact threshold/method of the field-density "_filtered" variant is pending author confirmation (its values differ from the explicit conf0.4/conf0.5 products). The confidence layer is conservative in regions whose visual signature differs from FTW training data (e.g. smallholder systems outside the training distribution), where real fields may receive low confidence — use the unfiltered density and continuous confidence rather than the default 0.4 threshold there.
Generated by Portolan from STAC metadata and .portolan/metadata.yaml