A new England inventory of hedgerows, copses, and stone walls, built by Google Research and the University of Oxford, gives planners a mappable surface for restoration and carbon accounting.
Across the British countryside, hedgerows stitch field to field, copses cluster in field corners, and stone walls line roads that have carried farmers for centuries. For decades, those small woody features have slipped beneath the notice of national forest inventories, which were built to count whole woodlands from coarse satellite pixels. A new release from Google Research and the University of Oxford's Leverhulme Centre for Nature Recovery finally turns them into a planning-grade inventory for England.
The work, called Vectorized Farmscapes 2020, is the first large-scale, high-resolution map of hedgerows, copses, shelterbelts, and stone walls across the country. Where the original Farmscapes 2020 release gave researchers a pixel-level raster image, the new vectorized version traces each feature as a polygon, with a length, an area, and a location accurate enough to drop into a farmer's field plan or a council's restoration scheme.
The shift from pixels to polygons matters because a raster image tells you a patch of green exists. A polygon tells you where a hedgerow starts and stops, how long it is, and which field it borders. That is the kind of data a restoration group, a planning authority, or a carbon accountant can actually work with.
The model behind the map is a deep learning framework trained to detect fine-scale features that standard satellite passes miss. National forest inventories were not designed for objects that small. Hedgerows, often just a few meters wide, vanish in 10-meter or 30-meter satellite pixels, the resolution most government land-cover products use. The new framework recovers them, then traces them into the geometric primitives that planning software and carbon registries expect.
That engineering step is the story. It is what turns a research image into a tool. A farmer in Devon can now pull a parcel-level count of the hedgerows on their land. A local authority mapping tree-cover gain can include the features the old inventory excluded. A conservation group writing a landscape recovery bid can point to a measured baseline instead of an estimate.
The dataset also makes a specific bet about land use. Fine-scale woody features, the team argues, can store carbon and support biodiversity without displacing the crops they border. That is the constructive thread running through the announcement: a way to expand tree cover in working farmland, rather than trade food production for restoration.
It is a plausible bet, and a useful one for the climate-biodiversity tension that hangs over UK land policy. But it is still a bet. The dataset enables the claim. It does not prove it. Whether those hedgerows actually sequester as much carbon as the model assumes, and whether farmers and councils will adopt the inventory in their plans, are open questions the release does not answer.
What the release does provide is a measurable surface. A hedgerow that exists as a polygon in a planning database is a hedgerow a council can protect, a farmer can be paid to maintain, and a carbon registry can include in a credit. None of those things were possible when the same feature was a smudge in a satellite image.
For now, the inventory is England only, drawn from 2020 aerial and satellite imagery, and published as an open dataset for researchers and practitioners. The next test is uptake. If local authorities and landowner groups fold the polygons into their planning, the upgrade from pixels to vectors will be the kind of unglamorous infrastructure change that quietly reshapes how a country counts its countryside.