Knowing exactly where one field ends and the next begins is fundamental to agricultural monitoring, crop assessment, and land management.
Traditionally, that boundary data has come from expensive high–resolution imagery or manual surveying. But it turns out satellites with much coarser resolution can do a surprisingly good job too — and at a fraction of the cost.
Parcel delineation is the process of automatically identifying agricultural field boundaries from satellite imagery. It’s usually associated with high–resolution data, but it can also be performed using Sentinel–2 imagery at 10 m resolution. The European Space Agency’s Sentinel–2 mission offers global coverage and frequent revisits, making it an attractive option for agricultural monitoring at scale.
Mallon colleague, Sita Karki has generated the following examples using the parcel delineation workflow from the Copernicus Data Space Ecosystem (CDSE) openEO community notebook.
Figure 1: NDVI images for a test area in Ireland at different acquisition dates, used as input for parcel delineation.
How It Works
Rather than tracing visible field edges in a single image, the workflow estimates parcel boundaries from how vegetation behaves over time. Multiple Sentinel–2 acquisitions are combined and analysed — primarily using the Normalised Difference Vegetation Index (NDVI) — to track vegetation development throughout the season.
Comparing observations from different dates reveals transitions between neighbouring fields far more reliably than a single snapshot can, since crops in adjacent parcels rarely grow, ripen, or get harvested in perfect sync. The trade–off is that this approach depends on having enough clear images to compare — something that’s not guaranteed everywhere. In cloud–prone regions like Ireland, gathering several cloud–free observations of the same area within a season can be genuinely difficult, which limits how well the temporal approach performs there.
Figure 2: Output of the parcel delineation workflow overlaid on a basemap, showing varying levels of accuracy.
Where It Falls Short
Sentinel–2’s 10 m resolution is the main constraint: it struggles to capture narrow boundaries, small parcels, and fine–scale field structures. Outputs are best treated as estimated parcel extents rather than exact cadastral boundaries.
The method’s reliance on multiple cloud–free acquisitions compounds this in cloud–prone regions, where a full, clean time series isn’t always achievable within a growing season.
Higher–resolution imagery can sharpen spatial precision, but acquiring it frequently enough across a season isn’t always feasible or cost–effective. Sentinel–2 sits in a useful middle ground — balancing revisit frequency, accessibility, and operational scalability.
For teams that need scalable, low–cost field boundary data and can tolerate some imprecision at the edges, Sentinel–2–based parcel delineation is a practical addition to the toolkit — not a replacement for cadastral–grade surveying, but a useful layer of Earth observation insight for agricultural monitoring at scale.
References
Copernicus Data Space Ecosystem (CDSE) openEO Community Example Notebook: Parcel Delineation
Further Information
For further information about the methods used to produce the study and images above, or to discuss your Earth Observation requirements, contact us below.
