Geolocation rules: points vs polygons.

The regulation draws one line: plots of 4 hectares or less may be located by a single point (a latitude/longitude pair); plots above 4 hectares must be located by a polygon — a boundary of connected vertices.

The threshold is per plot, not per farm: a large estate of small blocks still owes polygons only for the blocks above the line — though mapping the whole estate is usually easier.

All coordinates use the WGS84 reference system (the “GPS” one — EPSG:4326), in decimal degrees with at least six decimal places: 5.351234, -4.028391. Six decimals puts you within roughly 11 cm of true position.

Common trap: coordinates exported in a projected CRS (metres) or in degrees-minutes-seconds. Convert to decimal degrees WGS84 before anything else.

If your supply base is thousands of smallholders, collect once at registration, not per shipment: a phone GPS app walking the boundary, a cooperative intake template, or KML/GeoJSON export from a mapping tool the co-op keeps current.

Acceptance criteria that save rework: closed rings (first vertex equals last), WGS84 decimal degrees, no self-intersections. A one-page intake template with those three checks prevents most rejection cycles.

The usual suspects: self-intersecting polygons (a boundary that crosses itself — the figure-eight), wrong CRS (projected metres read as degrees), duplicate consecutive vertices, and unclosed rings. All are trivially detectable — if you look before filing.

Automate the check at intake. Screening tools, ours included, validate geometry before any dataset query, so a malformed ring never reaches the expensive part.

Screening intersects your point or polygon with deforestation datasets — Hansen GFC tree-cover loss, GLAD/RADD alerts, Whisp convergence signals. A point gets a buffer; a polygon gets per-area statistics inside the boundary.

The output is a signal summary — indicative, decision-support — that becomes part of your evidence pack. Better geometry in, better signal out.