OS Open Greenspace

OS Open Greenspace is Ordnance Survey's national layer of publicly accessible green space across Great Britain - and, just as importantly, of the ways into it. Two feature types carry everything. GreenspaceSite holds the polygons, generalised automatically from Ordnance Survey large-scale mapping, covering ten kinds of ground under the FunctionValue classification: Public Park Or Garden, Play Space, Golf Course, Playing Field, Bowling Green, Tennis Court, Other Sports Facility, Allotments Or Community Growing Spaces, Cemetery and Religious Grounds. AccessPoint holds the entrance locations, normally positioned on the site boundary, each typed through AccessTypeValue as Motor Vehicle, Pedestrian, or Motor Vehicle And Pedestrian.

Scale first: the April 2026 release carries 165,978 Greenspace Sites and 355,705 Access Points across England, Scotland and Wales, with the release notes recording no known data issues. Scope second: Great Britain only - Northern Ireland sits outside Ordnance Survey's remit. Honesty third, because what the schema omits is deliberate design: there is no opening-hours attribute, no pricing, no facility-condition field anywhere in the product. It answers one question exhaustively - where is publicly accessible green space, and how do you enter it - and leaves everything else to datasets that specialise in it.

Rows come in two families, joined by identifier. A GreenspaceSite row announces what a piece of ground is and where its boundary runs; an AccessPoint row says which site it belongs to and whether you drive, walk, or both:

Nine attributes define the entire schema - small by design, and every definition below is verified against the published technical specification rather than inferred from rows.

Identity and classification lead: id, a mandatory 38-character identifier on every feature; function, the mandatory FunctionValue code deciding whether a polygon is a playing field or a cemetery; and up to four distinctiveName slots of 254 characters each, so Battersea Park keeps its official title while local names survive alongside. Geometry follows - GM_MultiSurface polygons for sites, GM_Point for entrances, published at two-decimal-place precision, light enough to render the whole country in a browser tab.

The last two attributes are the pair that make the model navigable rather than merely drawable. accessType types each entrance by permitted mode, and refToGreenspaceSite is the mandatory foreign key tying every AccessPoint back to its site - join on it and 355,705 entrance points resolve onto 165,978 polygons in one hop.

Additional fields on request: the second through fourth name slots, code-list reference tables, and geometry delivery detail all stay folded here until we lock columns against your pipeline during sample preparation. The specification fully documents them; they simply rarely earn their place until you are joining sites to entrances at scale.

• Geography: all of Great Britain - England, Scotland and Wales - in one consistent national frame. Every site and every entrance lands in the same coordinate system with the same classification, so a catchment analysis spanning Carlisle and Cardiff never stitches council maps together. Northern Ireland is out of scope.
• Time frame: a current-state national snapshot. Releases replace the previous picture wholesale rather than accumulating history, so the dataset describes the greenspace estate as it stands - and change analysis means comparing consecutive snapshots, which Datadory retains when syncing on your behalf.
• Granularity: one row per site polygon, one row per access point, linked by identifier. Not local-authority rollups, not hectare totals - actual boundaries and actual gate locations.

Against the wider catalog this note scores 9/10, comfortably above the 7.81 average across 1,744 tracked datasets, earned on field documentation verified from specification and a release cycle whose notes reported no known issues. Where to rank it inside the 16-note leisure-facilities pack depends on your question: nothing else covers all of Great Britain's green space with typed entrances; nothing here tells you what happens inside the fences. See where the whole pack ranks at best leisure facilities datasets.

API, files, or your warehouse. Daily, weekly, or hourly.

You pick the channel and the cadence; scope-setting, normalization and schema stability are our problem. Polygons arrive as full boundary geometry, simplified shapes, or flat centroid latitude and longitude - whichever your stack consumes - and the nine-field dictionary above travels unchanged across all three channels, with refToGreenspaceSite already resolved so sites and entrances arrive joinable out of the box. Cadence changes are a settings conversation, not a re-integration project, and a sample cut to your geography comes first either way.

Six catalog personas carry this dataset in their Leisure Facilities packs; five draw real value:

• Data scientists & ML engineers engineer greenspace extent, site counts and access-point density as spatial features in GB-wide models - house-price hedonics, health-outcome studies, walkability scores. The data scientists use cases page expands the workflow.
• Market researchers & consultants quantify provision per town, borough or catchment with one classification instead of a patchwork of council PDFs.
• Developers & data-product builders power nearest-park routing, greenspace finders and planning screens on boundaries that behave identically nationwide. Detail at developers builders use cases.
• Journalists, academics & students cite a surveyed national layer with published specification and release notes behind it when writing about access to green space.
• Competitive-intelligence & product teams score competitor locations against surrounding greenspace provision, comparing consecutive snapshots for change over time. More at competitive intel product teams use cases.

Data scientists and ML engineers sit highest - a complete, consistently classified polygon layer with typed entrances is raw material for features, not something to be assembled from four councils' GIS exports. Market researchers follow at relevance 2: provision-per-catchment questions get answered from geometry rather than survey estimates. Developers and builders share relevance 2 for anything proximity-shaped.

Know the neighbours. Sport England Active Places Power audits English sites attribute by attribute - far richer per record, but England-only and facility-centric, while this layer spans all of Britain and records cemeteries and allotments Active Places never will. OpenActive Open Opportunity Data owns everything on a timetable inside the fence line. The tradeoffs against Active Places get their own page at vs Sport England Active Places Power.

Three things, all knowable upfront. First, scope discipline: with 165,978 sites nationwide, decide the counties or cities you actually analyse before the sample is cut - a scoped extract evaluates faster than a national dump and answers the same questions. Second, geometry weight: full generalised polygons render GB-wide but still carry heft, so samples ship at whatever resolution your prototype needs, centroids included. Third, snapshot semantics: this is a current-state picture, not a time series, so if your model wants change over time, say so at sample stage and consecutive snapshots get retained from day one. Samples are re-verified at preparation time rather than quoted from a stale inventory.