Fields of the Planet: Field Boundary Mapping Beyond 10m

2026-07-05

AI Summary

Smallholder fields often span only a few 10-meter Sentinel-2 pixels, leading to insufficient accuracy in existing field boundary mapping. To address this limitation, this study constructs a 3-meter resolution PlanetScope dataset, FTP, aligned with Fields of the World, enabling the first systematic evaluation of field boundary extraction at the 3-meter scale. By employing a unified model architecture and training strategy, combined with vectorized predictions and multidimensional evaluation metrics—including size-stratified Panoptic Quality (PQ) and meter-level matched boundary error—the overall PQ improves from 21.0 to 35.5. Notably, PQ for fields smaller than 0.5 hectares increases from 5.8 to 15.7, and boundary error decreases from 18.6 meters to 7.4 meters, substantially enhancing the accuracy of smallholder field boundary delineation.

Abstract

Field-boundary maps support crop monitoring, irrigation planning, and yield estimation, but many smallholder parcels span only a few 10 m Sentinel-2 pixels. We introduce Fields of the Planet (FTP), a 3 m PlanetScope companion to Fields of The World (FTW) that pairs the same polygons, seasonal windows, and train/test splits with 133,168 co-registered PlanetScope patch-window targets across 24 countries. FTP evaluates field delineation as parcel recovery by vectorizing predictions before scoring panoptic quality (PQ), object F1, size-stratified PQ, and meter-scale matched-boundary error. Under matched architectures and training recipes, 3 m imagery raises PQ from 21.0 to 35.5, raises PQ on sub-0.5 ha fields from 5.8 to 15.7, and cuts matched-boundary error from 18.6 m to 7.4 m.

Problem

• field boundary mapping
• smallholder parcels
• high-resolution imagery
• parcel delineation
• remote sensing

Innovation

• field boundary mapping
• PlanetScope
• panoptic quality
• sub-hectare parcels
• high-resolution remote sensing