Measuring peat on a Scottish landholding, tested against the national cadastre and the national habitat map

Published 21 August 2026. Data retrieved 21 August 2026.

The Land Reform (Scotland) Act 2025 requires a land management plan for holdings of 1,000 hectares or more, and those plans are expected to address carbon. In Scotland that mostly means peat. So a reasonable question for anyone advising a large estate is: how much of this holding is bog?

The data to answer it is published and free. A national cadastre gives the parcels, a national habitat map gives the bog, and the calculation between them is an overlay that any competent GIS analyst would price at an afternoon.

We ran it. Three separate defects compound against the answer, and any one of them could be fixed without much trouble. What makes them matter is that they all land on the same calculation.

First: the register publishes no areas

The INSPIRE specification defines an areavalue attribute as the quantification of a parcel’s area. Across all 33 registration counties and all 1,564,345 parcels, Registers of Scotland publishes that field with the literal value UNPOPULATED. It is the only value the field takes anywhere in Scotland. Five further attributes behave identically. The detail is in our study of the cadastre itself.

So the area has to be computed from the polygons.

Second: computed area depends on the method

Overlapping and stacked polygons are a valid part of the Registers of Scotland data model, and hole polygons stay in when the data is published. Adding parcels up therefore counts some ground twice. Taking the geometric union counts it once.

The gap runs from 1.33 per cent in Nairn to 95.40 per cent in West Lothian, so there is no single correction factor to apply. Of 270 connected parcel groups that reach 1,000 hectares by summation, 20 fall below it by union.

Third: the habitat map does not declare its own gaps

NatureScot’s Scotland Habitat and Land Cover Map 2024 is published as two GeoTIFFs, EUNIS Level 1 and Level 2, each 90,000 by 90,000 pixels at 10 metres. The README shipped alongside them states plainly: NoData value: 0.

Neither file encodes it. Open either raster and it reports no NoData value at all, so every tool that honours the file’s own metadata treats 0 as a valid class rather than as absence. Measured across the grid, 7,205,895,872 of 8,100,000,000 pixels, 88.96 per cent, are background presenting themselves as data.

For a holding-level overlay this has one specific consequence. Unclassified ground inside a boundary gets counted as a class instead of being left out, so the denominator and every per-class share come out wrong, and nothing warns you. The fix is one line at write time.

What the habitat map does get right

Because the two habitat maps sit on an identical grid, and because EUNIS is hierarchical so a Level 2 code must sit beneath exactly one Level 1 code, the two can be checked against each other for every pixel in Scotland. We compared all 8,100,000,000 pixel pairs, routed through each product’s own value attribute table because the two encode classes with different integers.

That 1.32 per cent is 118,374 hectares. Its shape matters more than its size. Only four distinct violating class pairs exist, and every one of them is the same marine and coastal confusion, where the Level 1 map calls ground marine benthic and the Level 2 map calls it coastal dunes, rock cliffs or shingle, or the other way round. The intertidal zone is the hardest thing in a product like this to classify, and one coherent cause points at something fixable.

For this question the important result is a null. The peat class Q1 is involved in none of the violations. No pixel in either raster uses a code absent from its attribute table. No pixel is classified in one map but not the other. Away from the coast, the two maps agree with each other exactly, and the carbon-relevant classes are clean.

How much bog there is

Roughly one hectare in seven of classified Scotland is raised or blanket bog. Q4 deserves a footnote of its own. It occupies 84 pixels in the whole country, under a single hectare, so the map documents a class that barely exists in it.

What this adds up to

The habitat map is good, and our own hierarchy test mostly demonstrates that. But the Act now attaches a duty to an area threshold and expects that duty to engage with carbon, and the public record cannot support that calculation from end to end. The areas are not published, the geometry does not sum safely, and the raster does not declare its own gaps. An adviser who runs the obvious overlay gets a number with no warning attached and no way to say how far off it is.

The defects we found in NatureScot’s delivery, including the unencoded NoData value and a README that gives a Level 1 class code as O where the shipped attribute table gives OW, are written up as reproducible reports in the repository, and we are sending them to the publisher.

Method and scope

Every headline in the cadastral half is computed twice, set-based in Python and by SPARQL over an assertion graph, by code paths that share nothing, with a gate that exits non-zero on disagreement. The habitat comparison is a single co-occurrence over all 8.1 billion pixel pairs with no parsing in the path. The pipeline is public.

Scope, stated plainly: no ownership data is used anywhere, because resolving a cadastral identifier to a title requires a paid service. Registers of Scotland states that roughly 30 per cent of Land Register titles are absent from the parcel dataset and that the Land Register is still replacing the General Register of Sasines, so every area figure is an area of published parcels and never an area of Scotland. The habitat figures are areas of a modelled classification, not of surveyed peat, and peat depth, which is what carbon actually depends on, is not in these data at all.

github.com/fabio-rovai/scotland-land-register-ontology (opens in new tab)

Contains public sector information licensed under the Open Government Licence v3.0. © Crown copyright. Reproduced with the permission of Registers of Scotland. Contains OS data. © Crown copyright and database right 2026. Habitat data © NatureScot, created with Space Intelligence.