The Wood and the Trees

Picture a timber lorry on the single-track road below Glenmeadhanach, pulled into a passing place to let the school bus by. It is carrying Sitka spruce from Compartment 7 of the estate's forest, planted in 1989 and felled this autumn. The logs are bound for a sawmill. Some will become fencing and pallets, some will be chipped, and some will be sawn into the studs that make up the frame of a house, possibly one of the new houses the community trust hopes to build at the foot of the glen. Behind the lorry, the hillside is stumps and brash for now, waiting for whatever is planted next.

Now ask some ordinary questions of that load:

• Who planted those trees, and why then?
• Who owns the hillside, and who decides what happens to it?
• Was the felling allowed, and what has to grow there next?
• How much carbon did the trees hold, and has anyone sold it?
• What lived in that forest, and what will live there now?
• Who cut the trees, and can they afford to live in the glen?
• Can a buyer in Rotterdam be sure the timber was harvested legally?

Each question has an answer, and as with the food system, each answer is held by someone different: the estate's forest plan, Scottish Forestry (which grants permission to fell), the certification body that audits the estate, the registry that holds its carbon units, NatureScot, the local deer management group, the haulier and the sawmill, the council's housing team, Registers of Scotland and the community trust. Each keeps its own records in its own vocabulary, often in a system bought to do one job. The hillside is the same hillside in all of them, but nobody holds the whole hill.

This is the fourth article in a series on knowledge graphs, after pieces for investigators and analysts (Joining the Dots), on the land question (Who Owns the Glen?) and on the food system (Field to Fork and Back Again). Forests are where the threads meet. Ownership, supply chains, carbon and nature all sit on the same ground, and people live among them. So this piece ranges more widely than the others, and goes further on three ideas that have run through the series: building around the data rather than the applications, dividing a graph into holons that each look after their own part, and checking records against the requirements of certification. Scotland is the focus, but many countries have tried parts of this, and some of what they have learned is worth borrowing.

A short history of a long crop

A century ago, only around 5% of Scotland was wooded. Today the figure is 20%. Forest Research put Scotland's woodland at 1.52 million hectares in March 2026 (cdn.forestresearch.gov.uk). Much of the increase is conifer plantation, a great deal of it Sitka spruce, a tree from the Pacific coast of North America that does well in Scotland's wet, windy climate. The Scottish Government wants 21% cover by 2032, and plans to raise new planting to 18,000 hectares a year by 2029-30, at least 40% of it native (gov.scot). The latest figures show how far there is to go. In 2025-26 Scotland created 7,220 hectares of new woodland against a target for the year of 10,000, though that was still 54% of all the new woodland in the UK, and 4,240 hectares of it was native (forestry.gov.scot).

The sector matters to the economy. A 2024 study for Scottish Forestry (forestry.gov.scot) valued forestry and wood processing, together with forest recreation and tourism, at £1.1 billion a year, supporting more than 34,000 jobs.

It also has a history of getting things badly wrong. Through the twentieth century, grants and tax incentives encouraged planting on the deep peat of the Flow Country in Caithness and Sutherland. In the RSPB's words (rspb.org.uk), "there were even tax incentives to grow forestry plantation on a bog!" Nearly a fifth of the Flow Country was drained and planted. The tax treatment that made it pay ended with Nigel Lawson's Budget of March 1988, a story Andy Wightman tells well (andywightman.scot). Since then more than 2,600 hectares of plantation have been removed and the drains blocked so that the bog can recover, and in July 2024 the Flow Country became the first peat bog anywhere to be made a World Heritage Site.

The lesson is not that forestry is bad. It is that "the right tree in the right place" is a question about data: about soils, peat depth, water, species, birds and people. The decisions on the Flow Country were taken without that data joined up, and it is taking half a century to undo them.

Forestry also runs on long cycles. A Sitka crop is usually felled after forty years or so. A Scots pine can stand for centuries. Carbon projects make promises for decades. The trees on that lorry have outlived every computer system that ever recorded them, and the house they become should outlive the next few. I will come back to that, because it turns out to be the most important point in this article.

One hillside, five questions

The themes of this piece are the ones any argument about land in Scotland comes back to: climate, the economy, nature, and people, with a roof over their heads and a living to be made. Each is a different way of asking what a hillside is for.

Climate. A forest stores carbon in its trees and its soils, and for a long time it was assumed that northern forests would go on soaking it up. That is no longer safe. In Finland, the land sector as a whole has become a net source of emissions, and in 2023 forest land itself turned into one: the carbon taken up by the trees no longer covered the emissions from the soils, according to Finland's Natural Resources Institute (luke.fi). In Germany, the national forest inventory published in October 2024 found that forests had emitted more carbon dioxide than they absorbed between 2017 and 2022, mainly because drought, storms and bark beetles destroyed so much spruce (oeko.de). Closer to home, the fires that began near Carrbridge and on Dava Moor on 28 June 2025 burned some 11,800 hectares of moorland and woodland (en.wikipedia.org), the largest wildfire Scotland has recorded and the UK's first designated megafire. And Scotland's Plant Health Centre judges the risk that the eight-toothed spruce bark beetle, the insect behind much of central Europe's spruce dieback, could establish here to be moderate for now, and higher in the south (planthealthcentre.scot).

Carbon is now a product too. The Woodland Carbon Code (woodlandcarboncode.org.uk) certifies the carbon captured by new woodland in the UK, and only new woodland: the land must not have been wooded for the previous 25 years (woodlandcarboncode.org.uk). Compartment 7 does not qualify, and nor will its replanting after this autumn's felling, because the land has been woodland for decades. The Code's version 3.0 was launched in August 2025, and the transition period for projects under the previous version ended in June 2026.

A project chooses a duration of between 40 and 100 years, and sets out how the woodland will be managed over that time and beyond, including whether it will ever be clearfelled. That choice matters, because the carbon that can be claimed is capped at the long-term average stock the woodland will hold (woodlandcarboncode.org.uk). For a woodland that is never clearfelled, that is at least what it holds at year 100. For one that is felled and replanted, it is the average over successive rotations, which is lower. So a promise not to clearfell earns more credits, at the price of the timber. The credits come in two stages. At validation, the project is issued Pending Issuance Units for its predicted growth, which can be sold but not used to report against emissions. At each verification, in year five and at least every ten years after that, the units for growth that has been confirmed become Woodland Carbon Units, which can be used that way (woodlandcarboncode.org.uk). A fifth of every project's units goes into a shared buffer against losses to fire, wind, pests and disease.

A carbon unit is therefore a claim built on claims: that the land had not been wooded for 25 years, that it would not have been planted without carbon finance, that it will grow at a modelled rate, that it will be managed as promised for up to a century, and that someone independent has checked. Each needs evidence, and each can change. The promise also outlasts the owner, since the landowner must tell any future owner about it, so the carbon graph has to join the ownership graph of the land article. Knowledge graphs can hold statements about statements (RDF 1.2 makes this standard), so a unit can carry its model, its measurements, its verification and its owner, and a fire, a sale or an application to fell shows up in the same graph as the credit.

The economy. Most of the wood the UK uses is imported: 81% of all the wood supplied in 2025, by Forest Research's count (cdn.forestresearch.gov.uk). Scotland builds more than 90% of its new homes with timber frames, but, as Confor's manifesto for the 2026 election points out, most of that timber is imported (confor.org.uk). Meanwhile a modern harvester records every stem it cuts in a common format called StanForD (skogforsk.se), so the data to follow wood from a compartment to a sawmill already exists. What is missing is the join to everything else: the felling permission, the certificate, the carbon project and the house.

Land itself has become an investment. As the land article noted, a 2022 study for the Scottish Land Commission (landcommission.gov.scot) found carbon and natural capital values to be an increasing influence on land prices, alongside timber, and the press has a name for the new buyers: "green lairds". Who benefits from a forest, and who carries the risk, is an economic question with a social edge. Answering it needs the ownership graph from the land article joined to the carbon and timber graphs.

Nature. Scotland's rainforest, the Atlantic oak, birch and hazel woods of the west coast, covers about 30,000 hectares, and 12,000 hectares of it needs clearing of invasive rhododendron and protecting from overgrazing (forestry.gov.scot). The capercaillie, the great grouse of the pinewoods, was down to an estimated 532 birds in the last national survey, in 2021-22 (doi.org). This spring's lek counts in the Cairngorms found 184 displaying males, up from 143 in 2025, which is something (cairngorms.co.uk). Forestry and Land Scotland estimates that there are now more than a million deer in Scotland, populations having doubled since 1990 (forestryandland.gov.scot), and deer eat young trees.

The Natural Environment (Scotland) Act 2026 (legislation.gov.uk) is the latest response. It provides for statutory targets for improving biodiversity, and overhauls deer management, from deer management plans and control schemes to a national deer management and venison plan, which ministers must prepare within five years. Targets and plans like these run on data: species records, habitat condition, cull returns, damage assessments. Some of that data is sensitive. Nobody wants the location of a capercaillie lek or an eagle's nest published, and that turns out to matter a great deal for how a graph is divided.

People: a roof and a living. Many Highland glens were emptied once, and some are emptying again for quieter reasons: a lack of affordable houses, jobs and services. The Scottish Government published an action plan to address depopulation in 2024 (gov.scot), alongside a rural and islands housing action plan (gov.scot). Forests can be part of the answer, by offering land for houses and livelihoods. A woodland croft puts woodland under crofting tenure, and crofting, community woodland and housing bodies have been promoting them together for more than a decade (crofting.org). Through Forestry and Land Scotland's community asset transfer scheme, a community in Glengarry has bought 66 hectares of Ardochy Forest for woodland crofts and repopulation (scottishhousingnews.com), and a trust in Lochalsh has been funded to buy land at Ratagan and Saraig for affordable housing and woodland crofts (forestryandland.gov.scot).

Sustenance belongs here too. Venison is about as local as meat gets, and the new deer plan must cover the venison industry and its supply chains as well as the deer. A forest also provides firewood and mushrooms and, in Scotland, a right of responsible access for everyone. And when its first section comes into force, the Land Reform (Scotland) Act 2025 (legislation.gov.uk) will require owners of holdings over 1,000 hectares to have a publicly available land management plan, drawn up with the local community and reviewed every five years, and to consider community requests to lease land or to create crofts. Our fictional Glenmeadhanach, at well over that size, will need one.

Every one of those five questions is asked of the same ground. Each is answered by different people, with different data, under different rules. That is the problem a knowledge graph is for.

Sentences about a forest

As in the earlier articles, the basic parts are short sentences of subject, verb and object, written in a standard called RDF. A slice of Glenmeadhanach's forest might read:

Compartment 7 — is part of —> Glenmeadhanach Forest

Compartment 7 — was planted with —> Sitka spruce

Felling permission FP-2026-0412 — covers —> Compartment 7

Lot GM-0731 — was harvested from —> Compartment 7

Batch K-118 — was sawn from —> Lot GM-0731

Batch K-118 — is part of —> the frame of 3 Station Cottages

(All invented.) The sentences come from the estate, Scottish Forestry, the harvester's computer and the sawmill. They join up because Compartment 7 has one identifier in all of them.

The food article introduced Semantic Arts' gist (github.com), a deliberately small upper ontology for business, and its careful distinctions between the ways one thing can belong to another. A forest uses all of them:

• Parts (gist:isPartOf). A compartment is part of a forest, and the forest part of an estate.
• Members (gist:isMemberOf). Glenmeadhanach Estates Ltd is a member of the local deer management group, which does not make its neighbours members of anything the estate belongs to.
• Substances (gist:isMadeUpOf). A stud is made up of spruce wood. It is not part of any one tree, just as the cheese in the food article was made up of milk rather than part of any one farm.
• Places (gist:isGeoContainedIn). Compartment 7 lies within a river catchment, the catchment within a regional land use partnership's area, and all of them within Scotland, so a question about the catchment finds the compartment.

And gist keeps a company and its land apart. Glenmeadhanach Estates Ltd, registered in Jersey and controlled by a trust, owns (gist:owns) the estate. When the estate is sold, the forest does not change, but the owner does.

Forestry has vocabularies of its own, though no single one covers everything:

• INSPIRE, Europe's framework for spatial data, defines common models for land cover, protected sites, habitats and the distribution of species.
• StanForD, maintained by Skogforsk, Sweden's forestry research institute, is the de facto standard for data from harvesters and forwarders.
• Darwin Core is the standard for species records, used by GBIF and by the UK's National Biodiversity Network.
• Forest inventories as linked data. Researchers have published inventories with ontologies for plots, trees and measurements, in Spain (frontiersin.org) and in Maine's Digital Forest Knowledge Graph (digitalcommons.library.umaine.edu).
• Ecosystem accounts. The UN's System of Environmental-Economic Accounting gives a framework for recording a habitat's extent, its condition and the services it provides, and the ONS uses it for its woodland natural capital accounts (gov.uk).

A forest's history is a series of events: planting, beating up (replacing the trees that failed), thinning, felling, restocking, surveys, fires and culls. As in the food system, it pays to record events rather than overwrite records, so that the graph can say not only what the forest is but how it came to be that way.

The forest outlives the software

Here is the point I promised to come back to. Compartment 7 was planted in 1989. Since then it has been described under several editions of the UK Forestry Standard, a succession of grant schemes, the EU Timber Regulation and its UK successor, several versions of the Woodland Carbon Code, and, most likely, several generations of the estate's own systems, from paper maps to spreadsheets to a forest management package or two. The trees did not change because the systems did. The facts about them (where they are, what they are, when they were planted, who owns them) ought not to have changed either. But every system kept its own copy, in its own format, and every change of system meant a migration, with losses.

Dave McComb of Semantic Arts has a name for the habit behind this: application-centric. In Software Wasteland (2018) and The Data-Centric Revolution (2019) (technicspub.com), he argues that organisations have let each application define its own data, so that the same facts are captured again and again in incompatible schemas, and much of the cost and complexity of their systems goes into reconciling them. His alternative is a data-centric architecture. The Data-Centric Manifesto (datacentricmanifesto.org), which he has championed, puts it bluntly: "Data is the center of the universe; applications are ephemeral." Among its principles:

• Data "is self-describing and does not rely on an application for interpretation and meaning."
• Data "is expressed in open, non-proprietary formats."
• "Applications are allowed to visit the data, perform their magic and express the results of their process back into the data layer."

It is worth being precise, because "data-centric" is easily confused with "data-driven", which is about using data to make decisions. McComb's definition is about architecture: "A firm is data-centric to the extent that its applications are designed and built to a single, simple data model" (brcommunity.com). Simple means small. He reckons a typical enterprise model needs 300 to 500 concepts, where most single applications have thousands in their schemas. And it does not mean one big database: "The data itself may continue to reside in existing databases, as long as the model and the query mechanism can access them with reasonable latency."

Forests make the case better than most businesses can, because they outlive everything. A forest plan written today will be read in the 2060s by people using software that does not yet exist. The only things that can sensibly last that long are the data and its meaning: an identifier for each compartment, species and permission; a small shared model that says what they are; and statements in an open standard that any future system can read. RDF and an ontology like gist were built for exactly that.

The data-centric view also changes what a scheme or a standard is. In the application-centric world, each scheme is a system: a grant portal, a felling application, a certification audit file, a carbon registry entry, an EUDR due diligence statement, and soon a land management plan. Each asks for the same facts about the same compartments, re-keyed and slightly different each time. In a data-centric world, the facts are recorded once, by whoever is responsible for them, and each scheme visits. It brings its own rules and its own questions, checks the data, and writes back its own conclusions (a permission granted, a certificate issued, a unit verified) as new facts for others to use. Schemes still come and go. The UK Forestry Standard was revised in 2023, the Woodland Carbon Code in 2025, and the UK Woodland Assurance Standard is due for review from July 2026. But a new edition means a new set of rules visiting the same data, not a new database.

That raises two practical questions. If the data stays with whoever is responsible for it, how is it divided up? And if a scheme is a set of rules that visits the data, how are the rules written? The answers are holons and shapes, and they are the subjects of the next two sections. The data-centric idea itself deserves a piece of its own.

Many ways to cut a hill

There is a proverb about not being able to see the wood for the trees. A graph of land has to do better than that. It must see the tree, the stand, the compartment, the forest, the estate, the catchment, the region and the nation, all at once and each for what it is.

Take the hillside the lorry has just left. It is divided in a dozen ways at once:

• by title, in the Land Register;
• by compartment, in the estate's forest plan;
• by woodland management unit, for certification;
• by carbon project boundary, in the carbon registry;
• by catchment, for flooding and water quality;
• by designation, such as a Site of Special Scientific Interest, a Special Protection Area for capercaillie or a national park;
• by council and community council;
• by deer management group, whose boundaries follow the deer rather than the fences;
• by crofting township and common grazing, where there are crofts;
• by regional land use partnership;
• and, for the EU buyer, by the polygons drawn round each plot the timber came from.

Some of these nest neatly, and gist's geographic containment handles them. Many do not. A carbon project can straddle two titles, a catchment can cross three estates, and a deer management group has no regard for any of them. Forcing them all into one hierarchy is how databases go wrong. A graph does not have to. Nesting can be stated where it is true (gist:isGeoContainedIn), and overlap can be computed from the boundaries where it is not, using GeoSPARQL, the geographic extension to SPARQL that the land article introduced.

The food article borrowed Arthur Koestler's word holon for something that is a whole in itself and part of something larger. For land, the useful form is the plural. There is not one hierarchy of holons but several, overlapping on the same ground, and the same compartment belongs to all of them. The question is how to divide the data so that each part has a clear keeper. Five things decide it:

1. Custody. Who is responsible for the truth of a fact? The estate for its compartments, Scottish Forestry for its permissions, the certifier for its certificates, NatureScot for its designations. The data lives with its keeper, and others refer to it by its identifier rather than copying it.
2. Rules. Each holon has its own boundary, the conditions its data must meet. An estate's forest plan, a carbon project, a felling permission and a deer management plan each come with rules of their own.
3. Audience. Each holon shows itself to others through projections: published views designed for a particular reader, such as the buyer, the regulator or the community.
4. Sensitivity. The locations of leks and nests, and the personal details of crofters and tenants, need different handling from compartment areas and species.
5. Time. Forests change slowly and then suddenly, after a felling, a fire or a storm, and every change needs recording.

I should declare an interest again, because I have been building a triplestore around this idea. HOLOS (pypi.org) is an RDF 1.2 store with SPARQL 1.2, and its holons fit the hill closely:

• A holon is a set of named graphs that travel together: a scene holding the current state, a boundary of SHACL shapes and rules, and an event log. The estate's forest, a carbon project or a deer management group's counts could each be one.
• Every change is checked against the boundary before it lands, and recorded in the log, accepted or refused, with provenance for each statement. So the bill of materials for every fact about Compartment 7 is kept as a matter of course.
• Access is checked as data is read, so a query only ever counts what its reader may see. Graphs can also carry classification labels with a level and a set of what HOLOS calls compartments, which in a forest is an unfortunate coincidence. That is how the location of a lek could be open to the NatureScot officer and the estate's ranger but not to a casual query, without a second copy of the data.
• Projections let the estate show a buyer the geolocation and legality of a batch of logs without its timber prices, or show the community its land management plan without its tenants' details.
• Branches let a holon be copied to test a change against real data: what if deer numbers halved, or a third of a catchment were planted?
• Geography. Its GeoSPARQL reads both British National Grid and GPS coordinates. That matters here, because Scottish land data is mostly held on the national grid, while the EU wants the geolocation of each plot in latitude and longitude.

In fairness, the limits matter more in this article than in the last. The holon layer is a working skeleton. It validates around 165 changes a second against a holon of 300,000 statements, but it cannot yet query past versions, and holons can so far be reached only from program code. Its federated queries run only against endpoints registered in the same process. Querying another organisation's database over the web is deliberately switched off for now, so a federation across organisations needs a layer above it. And checks that need arithmetic across a whole holon cannot run inside a commit, which, as the next section shows, is less of a limitation than it sounds. The idea matters more than the implementation, and other implementations would be welcome. The W3C's Holon Graph Community Group (w3.org), which began meeting in June 2026, is working towards a common model.

Certification as shapes

Most of Scotland's forests are certified. In March 2026, 896,000 hectares of Scottish woodland were certified under the Forest Stewardship Council or PEFC schemes, or both, 59% of the total (cdn.forestresearch.gov.uk). In the UK both schemes certify against one standard, the UK Woodland Assurance Standard, or UKWAS (ukwas.org.uk). Its fifth edition, in force since December 2024, is full of requirements a computer could check, given the data:

• at least 15% of a woodland management unit is managed with the conservation and enhancement of biodiversity as the primary objective;
• at least 10% is semi-natural habitat;
• at least 1% is natural reserve, and at least 1% is long-term retention or managed under low-impact systems;
• in semi-natural woodlands over 10 hectares, no more than 10% is felled in any five-year period.

There is also a lovely rule in the guidance. An area may fall into more than one category, but it "can only be counted once towards the 15%".

Beneath UKWAS sits the UK Forestry Standard, which grants and felling permissions must meet. Its 2023 edition cut the most that any single species may make up from 75% to 65% (forestrycommission.blog.gov.uk), after diseases such as Phytophthora ramorum showed how exposed forests that depend on one species can be.

Checks like these are what SHACL, the Shapes Constraint Language, was made for. The goods-in inspection of the earlier articles becomes the auditor's checklist, written once and run against the data. Some checks look at one record at a time: every compartment has an area, a species and exactly one primary objective, and every felling permission says which compartments it covers and when they will be restocked. In SHACL, the last of those reads:

fs:FellingPermissionShape a sh:NodeShape ;
    sh:targetClass fs:FellingPermission ;
    sh:property [ sh:path fs:covers ; sh:class fs:Compartment ; sh:minCount 1 ] ;
    sh:property [ sh:path fs:restockBy ; sh:datatype xsd:date ; sh:minCount 1 ;
        sh:message "A felling permission must say when the land will be restocked" ] .

Other checks need arithmetic across a whole unit: the 15%, the 65%, the 10% in five years. These are written as SPARQL queries inside SHACL. Counting each area only once is exactly the kind of rule that is easy to state, tedious to check by hand and simple to check in a graph, provided every area has its own identifier.

The two kinds of check suit two rhythms. Record-level checks can run on every change, as each permission or survey arrives, so that nothing incomplete gets in. Whole-unit checks run at audit, over the unit as a whole. In HOLOS terms, the first kind belongs in a holon's boundary and the second in a scheduled validation run, which is why the limitation above matters less than it sounds. I tried both on a small invented version of Glenmeadhanach Forest, and two different validators found the same two problems: a felling permission with no restocking date, and a unit with 14.8% managed for biodiversity, just short of the 15%.

SHACL's results carry a severity (violation, warning or information), much as an auditor separates major non-conformities from minor ones and from observations. And the validation report is itself data. It records which shapes, from which edition of which standard, were checked against which records, when, and with what result. That matters, because standards have editions. Scottish Forestry still applies the 75% limit to woodland creation applications assessed against the 2017 edition of the Forestry Standard, and 65% to the rest (ruralpayments.org). In a graph, each shape can say which edition it belongs to and which applications it applies to, so that two applications in the same week can be checked against two different rules and both be right.

The same approach serves the other schemes that visit the hillside:

• the conditions of a Forestry Grant Scheme contract;
• the Woodland Carbon Code, whose projects are validated and verified by bodies accredited by the UK Accreditation Service (woodlandcarboncode.org.uk), and where a shape could flag an application to fell inside a project that promised never to clearfell;
• the information an EUDR due diligence statement needs;
• the contents of a land management plan, once regulations set them.

One survey can serve many certificates.

Shapes check records, not woods. An auditor still has to walk the ground, look at the deer fence and talk to the forest manager. What shapes change is how that time is spent: on the ground and on the judgement calls, rather than on reconciling spreadsheets.

The rules are moving

The food article described how regulation is turning paperwork into data. Forests are in the middle of it, and the direction is the same: more facts, tied to places, checked by others.

• The EU Deforestation Regulation covers seven commodities, and wood is one of them, from logs and sawn timber to pallets, paper and furniture. From 30 December 2026, large and medium companies placing them on the EU market, and small ones in the timber sector, must show that each product is legal and deforestation-free, with the geolocation of every plot of land it came from. Plots over four hectares need a polygon, not a point. Other small businesses follow on 30 June 2027 (lewissilkin.com). The UK is rated low risk, so Scottish timber faces simplified due diligence (gov.uk), but the geolocation still has to travel with the wood.
• Great Britain's own rules. Due diligence for timber has been required across the UK since 2013. The UK Government now plans a wider regime for forest-risk commodities that will include wood, apply to businesses with a turnover over £1 million and require geolocation data, with legislation expected in 2027 (mayerbrown.com).
• Plant health. The UK–EU agreement on sanitary and phytosanitary rules, agreed in principle at the start of November (foodnavigator.com), is about plant health as well as food (that is what "phytosanitary" means). Seedlings, Christmas trees and timber with bark all move under plant health rules, and dynamic alignment means those rules will follow the EU's.
• Product passports. The EU's ecodesign working plan for 2025 to 2030 lists furniture among its first priority products (green-forum.ec.europa.eu). Ecodesign rules are the route by which digital product passports arrive, so a wooden chair may one day carry a link to the forest it came from.
• Carbon and nature. The EU's framework for certifying carbon removals adopted a methodology for afforestation in July 2026 (climate.ec.europa.eu). Under the Nature Restoration Regulation, member states owed the Commission draft national restoration plans by 1 September 2026, and twenty of them missed the date (eunews.it). Meanwhile, the proposal that would have standardised forest data across the EU, the Forest Monitoring Law, was rejected by the European Parliament in October 2025 (eustafor.eu). Europe wants the data, but has not agreed how to collect it.
• Scotland. At home, the Natural Environment Act brings statutory biodiversity targets and new deer powers, the Land Reform Act brings land management plans, and the Climate Change Plan published in March 2026 carries the planting targets.

Read across the list and the same facts keep coming up: where the land is, what grows on it, when it was planted and felled, who owns it, what lives there and what it stores. Each rule wants them in its own shape, for its own reader. That is the data-centric argument in regulatory form. A business that holds those facts once, in an open standard and with their provenance, can meet each new rule by writing new shapes and a new projection, rather than buying a new system. A business that holds them in a dozen applications will be re-keying them for each regulator in turn.

What others have learned

Scotland is not the first country to try to see its forests whole, and others' experience is instructive.

• Finland: every transaction improves the record. The Finnish Forest Centre builds its forest resource data from field plots, aerial photography and airborne laser scanning on a six-year cycle, and publishes it as open data without owners' identities (metsakeskus.fi). The notifications that owners must file before felling, and their grant applications, are among the sources that keep the data current. That is a data-centric loop, run by a state.
• Norway: one map of the forest. NIBIO's SR16 forest resource map combines the national inventory's field plots with laser scanning and satellite images into a 16-metre grid of tree species, biomass, stand age and more, published in its Kilden map service (nibio.no).
• Sweden: more than one right on the same land. Reindeer herding by 51 Sámi communities extends across roughly half of Sweden, much of it the same land that is used for forestry and mining, and the two sides argue over the grazing that development takes (arctictoday.com). A graph that recorded only ownership would miss the most contested right on the ground.
• Brazil: check boundaries at the door. Brazil's rural environmental registry is self-declared, and its property boundaries overlap. A 2024 study in PNAS (pmc.ncbi.nlm.nih.gov) found more deforestation on land with multiple claims, and estimated that the overlaps overstate forest reserves by up to 9.7 million hectares. A registry that tested each new boundary against existing ones as it arrived, as a shape using GeoSPARQL could, would catch much of that at entry.
• Indonesia: one map is hard. In December 2010 Indonesia's president was shown strikingly different maps of the country's forests, and the One Map Policy began the following year. In 2013 Forest Watch Indonesia found 14.7 million hectares under overlapping licences for logging, plantations and mining (opengovpartnership.org), and customary land has been among the hardest to reconcile. A single map for everyone is a tall order. A shared model with many keepers may be easier.
• Nepal: communities as keepers. Some 2.2 million hectares of Nepal's forests have been handed to more than 22,000 community forest user groups, and forest cover has risen (myrepublica.nagariknetwork.com).
• Paying for forests. Since 2024 Japan has collected a national forest environment tax of 1,000 yen a year from every resident, to fund the upkeep of its forests (city.inagi.tokyo.jp). At COP30 in Belém, Brazil launched the Tropical Forest Forever Facility, which aims to pay countries about $4 a year for each hectare of standing tropical forest, verified by satellite, with deductions for each hectare lost or degraded by fire (eco-business.com). It raised $6.7 billion of sponsor capital against the $25 billion sought (news.mongabay.com). Payment by the hectare is payment by data.
• Competition for land. Since 31 October 2025, New Zealand has restricted the registration of new exotic forests on better farmland in its emissions trading scheme, after whole farms were sold for carbon forestry (rmla.org.nz). Ireland planted 2,527 hectares of new forest in 2025, against a target of 8,000 (farmersjournal.ie).
• Measuring from above. The European Space Agency's Biomass satellite, launched in April 2025, carries the first radar in orbit that can see through forest canopies to weigh trunks and branches, and its data has been open to all since January 2026 (esa.int). The FAO's latest global assessment (fao.org) puts the world's forests at 4.14 billion hectares, and gross deforestation at 10.9 million hectares a year between 2015 and 2025: slower than before, but still vast.

The lessons for Scotland:

• make the data open where it can be, and protect it where it must be;
• let every transaction improve the shared record, as Finland does;
• check boundaries when they arrive, not years later;
• represent every right on the land, not just ownership;
• expect money to follow measurement.

A proposal for Scotland

The food article proposed a federation of holons for Scotland's food. Land needs the same, and the two should share a lexicon, because a field of oats and a forest compartment are pieces of the same land.

The hooks are already there:

• the Land Reform Act's land management plans;
• the Natural Environment Act's biodiversity targets and deer plans;
• the regional land use partnerships, which are working towards a national framework due for approval in spring 2027 (southofscotlandenterprise.com);
• the Climate Change Plan's planting targets;
• the EUDR's demands on anyone selling wood into Europe.

All of them need the same facts.

So does science. As this article is published, Forest Research is presenting the findings of its 2021 to 2026 research programme at SciFor 26 in Edinburgh (forestresearch.gov.uk), in more than 40 sessions across 13 themes, from carbon modelling and nature markets to growth and yield, species choice, timber prices and climate risk. Findings like these are most useful when they can be joined to the ground they describe: a yield model to the compartments it predicts, a pest risk to the stands it threatens, a carbon model to the projects that rely on it.

What I would propose is data-centric in McComb's sense and federated in Koestler's, with five elements:

1. A single, simple, shared model. gist would sit at the top, with a small Scottish extension for what only Scottish land has: compartments and forest plans, the Forestry Standard's terms, woodland management units, felling permissions, carbon and peatland units, designations, deer management groups, crofts and common grazings. Beneath it would come INSPIRE, Darwin Core, a mapping from StanForD, the ecosystem accounts, GeoSPARQL for place and PROV for provenance. The aim is a few hundred concepts, not thousands.

2. Holons for those who keep the facts: estates and forests, Scottish Forestry's permissions, Forestry and Land Scotland's national forests, Forest Research's inventories and models, certification bodies, the carbon registry, NatureScot, deer management groups, community trusts and councils. Each keeps its own data under its own rules, and refers to others' data by identifier.

3. Standards and rules as shapes. The Forestry Standard, UKWAS, the Woodland Carbon Code, grant conditions, EUDR evidence and land management plans would each be a versioned set of shapes that visits the data.

4. Projections for each audience:

   • geolocation and legality for the EU buyer;
   • carbon evidence for the carbon buyer;
   • the land management plan for the community;
   • progress against targets for NatureScot;
   • land for housing, and jobs, for the council;
   • aggregated figures for researchers.

   The locations of sensitive species stay hidden from those who do not need them.

5. Pilots before platforms. Five would test most of the ideas:

   • One landscape. All the overlapping holons in one regional land use partnership's area, to see whether they can be joined without being merged.
   • Seedling to stud. One forest's timber traced through a sawmill into the frames of affordable homes, with the certification and EUDR evidence produced as projections rather than paperwork.
   • One deer management group. Counts, culls, damage assessments and venison across several estates, in support of the new deer powers.
   • One community forest. Woodland crofts, housing and the forest plan in one graph, somewhere like Ardochy.
   • One audit. A UKWAS audit run with shapes alongside the auditor, to see what the shapes catch and what they miss.

None of this needs a national platform first. If you hold data about land:

• give each thing its own identifier;
• keep the facts separate from the applications that use them;
• write your rules down as shapes;
• reuse an existing vocabulary rather than inventing one;
• keep the bill of materials as you go.

Back to the lorry

The lorry pulls out of the passing place and heads for the main road. The trees on its trailer were planted in 1989, the year after the tax breaks ended, and they have outlived every system that ever recorded them. Some of the wood may stand in the frame of a house for another century, in a glen that needs people as much as it needs trees. Whether the next crop is spruce or oak, and whether the open hill above it is planted for timber, for carbon or for both, it will be decided with data held by a dozen different keepers. A knowledge graph cannot make that decision. What it can do is let all of them see the same hill: the wood and the trees, the deer and the capercaillie, the carbon and the timber, and the people who live among them.