Field Boundary Map

Catching an overlapping parcel boundary before it hits the registry

Old wooden fence post with dried vines at the edge of a harvested maize field in fog

An overlap that slips into a live registry doesn't stay small. One farmer's claimed parcel edge eats three square meters of a neighbor's, the subsidy calculation runs against both, and six months later someone in the appeals office is pulling vertex coordinates trying to figure out which boundary moved and when. Catching it before the load is cheaper than catching it after, by a wide margin.

Where overlaps actually come from

Most overlaps aren't drawing errors in the sense of someone being careless. They come from two boundary sources disagreeing: a new digitized layer against last year's registry geometry, or two adjacent parcels digitized independently and never snapped to a shared edge. A field gets re-surveyed, the new trace runs a meter inside the old fence line on one side, and now you've got a sliver gap on one edge and a sliver overlap on the other. Multiply that by a few thousand parcels in a region and some percentage will always disagree with their neighbor.

The other common source is a stale parent-child split. A parcel gets subdivided, the registry still holds the old single-polygon boundary, and the new sub-parcels overlap it until someone retires the parent geometry. If your intake process doesn't check new load against existing active geometry, that kind of overlap will sit there until an audit finds it.

Running a topology check before the load

A topology check is just a set of rules run against the polygon set before anything goes into production: no self-intersections, no polygon overlapping another polygon in the same layer, no gaps above a tolerance threshold, no duplicate geometry. Most GIS desktop tools and spatial databases have a topology or validation module built for exactly this, and it's worth running as a gate, not a spot-check. Set the tolerance to match what your registry actually enforces, run it against the full batch, and export the list of failing feature IDs rather than eyeballing the map.

The tolerance number matters more than it sounds like it should. Set it too tight and you'll flag thousands of trivial overlaps from rasterization and rounding that nobody cares about. Set it too loose and a real boundary dispute slides through. A tolerance in the range of your source imagery's ground sample distance is a reasonable starting point, tightened once you've seen a batch or two of real flags.

Slivers are the usual suspect

A sliver overlap is a thin strip, usually a few centimeters to a couple of meters wide, running along a shared edge where two boundaries were digitized from slightly different references. It's rarely a real claim dispute. It's two edges that should be the same line and aren't. The fix is almost always snapping the newer edge to the established one within tolerance, and ownership rarely needs re-adjudicating at all. Flagging these separately from genuine overlaps (where one polygon sits substantially inside another) saves your review queue from drowning in geometry noise that doesn't need a human decision.

Building registry load QA into the workflow

Registry load QA runs the topology check on every batch as a standing step on the way into the registry, not as a cleanup pass you do once and forget. A reasonable sequence: validate geometry (no self-intersections, no invalid rings), check overlap against the active registry layer, flag slivers below a set width for auto-snap and overlaps above it for manual review, then load. Keep a log of what got auto-corrected versus what a person signed off on, because an auditor will eventually ask.

If your boundaries are coming from manual digitizing, this QA pass is where a lot of the review time goes, because hand-traced edges rarely snap to a neighbor's edge on the first try. A boundary layer generated consistently from the same source imagery and the same process year over year tends to produce far fewer of these edge disagreements in the first place, since every parcel in the batch is traced against the same reference rather than whatever the previous digitizer eyeballed. That's the gap Field Boundary Map is built to close: a parcel boundary layer vectorized from high-resolution satellite imagery for your region and reference year, delivered edit-ready so your QA pass is checking real disputes instead of chasing snapping errors.

Run your next batch through a tolerance check before it ever touches the registry, and see how much of your review queue turns out to be slivers rather than real overlaps.

Get started

← Back to the blog