How to size a monitoring polygon for a cocoa supply shed

Cocoa buying stations inherit a mapping problem nobody chose on purpose. A cooperative hands over a few thousand farm-boundary points collected by field agents with handheld GPS units, a concession map shows a line drawn in an office somewhere, and someone in compliance has to turn that into a polygon worth watching every week. Get the shape wrong and you're either paying attention to pasture that was never going to be cocoa, or missing the forest margin where the actual clearing risk sits.

Start from the catchment, not the concession line

A supply shed is the area a buying station, cooperative, or first-mile aggregator draws volume from. That's rarely the same shape as a government concession boundary or a tenure map. Concession lines get drawn for licensing and land-rights purposes; they can run straight through land nobody there has ever farmed, or stop short of a cluster of smallholder plots that sell through the same scale house.

Pull your farm-gate or cooperative delivery records first. Plot where tonnage originates using delivery records. Permit maps show where cocoa is legally allowed to grow, a different question from where it's grown and sold. In most sourcing areas this gives you a cluster, sometimes two or three separate clusters around different buying points, not one tidy shape. Draw the polygon around the cluster, then check it against the concession boundary as a second layer, not the first.

Buffer the farm boundaries, don't just trace them

Farm-level GPS points from cooperative surveys are useful for traceability but they're a bad fit for a monitoring polygon on their own. A point marking a farm center tells you nothing about where that farm's edge meets the forest. Trace only the surveyed plot lines and you'll miss clearing that happens just outside them, on the expansion margin where farmers push into secondary forest or fallow land adjacent to an existing plot.

The common fix is a buffer, typically somewhere in the range of a few hundred meters beyond the surveyed farm edges, run around the whole cluster rather than around each farm individually. Buffering each polygon separately produces a Swiss-cheese shape full of gaps that are often exactly where new clearing shows up, because that's the land between farms that nobody surveyed and nobody is formally responsible for.

If your farm data is old (surveyed more than two or three seasons back) widen the buffer. Expansion since the last survey is the whole reason you're monitoring in the first place.

Size it to what a weekly pass can actually cover

A sourcing-area polygon that tries to cover an entire regional landscape dilutes the signal. Every clearing event gets buried under forest-cover noise from land your buying stations don't even source from, and whoever reviews the weekly list starts tuning it out. Keep the polygon scoped to the supply shed plus its buffer, and if you source from several distinct catchments, build separate polygons for each rather than one sprawling outline that covers the gaps between them too.

At the other end, don't scope it down to individual farm plots either. A polygon that size misses the landscape-level expansion pattern, where a cooperative catchment's clearing shows up as a dozen small cuts spread across adjacent farms rather than one obvious scar. The shed-level view is usually the one that catches that pattern early enough to act on it.

There's no fixed size that's right for every cocoa landscape. A tight cooperative catchment in a smallholder zone and a large estate-adjacent concession need different outlines, and the only way to know yours is scoped right is to watch what the weekly alerts actually flag once you've drawn it. Deforestation Alerts runs a cloud-free scan against the polygon you define and routes anything new back to your team the week it happens, so a boundary that's cutting off a real risk area shows up fast as a gap rather than months later as a surprise in an annual land-cover update.

Draw the shed, buffer the edges, and let the first few weeks of alerts tell you where the polygon needs adjusting.