Who Owns the Glen?

Picture a village hall on a wet Tuesday evening. The community wants to buy the woodland above the village, or the field behind the school, or the old pier that nobody seems to look after. Somebody has brought the guidance from Community Land Scotland (communitylandscotland.org.uk), and the very first step on its roadmap is a simple question: "Who owns it?" As the guidance points out, the answer will often decide which route the community takes to acquire it.

It sounds like the easiest question in the world. In Scotland, it often isn't. The Land Register, the modern map-based register of titles, covered a little over half of Scotland's land area at the start of 2025. Much of the rest is recorded only in the "General Register of Sasines" (nrscotland.gov.uk), which dates from 1617 and is often described as the oldest national land register in the world. That is a fine thing to have, but it is a register of deeds, and a deed has to be found, read and interpreted before it tells you anything. And even once you know the name of the owner, that owner may be a company registered in another country, a trust or a partnership, and the people who actually make the decisions appear in yet another register, if they appear at all.

Andy Wightman (linkedin.com) has spent the best part of thirty years answering the question at the scale of the whole country, from his 1996 book Who Owns Scotland to the "Who Owns Scotland" (whoownsscotland.org.uk) project and the analysis he publishes at andywightman.scot. His recent figures (andywightman.scot) put the number of owners holding half of Scotland's privately owned rural land at 421. Scotland is often described as having one of the most concentrated patterns of land ownership in the developed world, and the Land Reform (Scotland) Act 2025 (legislation.gov.uk), which received Royal Assent in December 2025, is the latest attempt to do something about it.

In a previous article (Joining the Dots) I looked at how a family of technologies known as knowledge graphs can help investigative journalists and think tanks. Land is, if anything, an even better fit. Land questions are questions about connections: between parcels and titles, titles and owners, owners and the people behind them, and all of these and what the land is actually used for. They stretch back centuries. The sources often disagree. And many of the outcomes that matter are decided by rules written into law. Those are exactly the things knowledge graphs are good at.

The supply chain behind "who owns Scotland?"

It helps to think of land research as a supply chain, much like a manufacturer's. Raw materials come in from many suppliers: the Land Register and its map, the Sasines, the Register of Persons Holding a Controlled Interest in Land (legislation.gov.uk) (in force since April 2022), Companies House, the councils' registers of common good property, estate agents' sale particulars, grant and subsidy records, planning applications, old estate maps, burgh charters and council minutes, and a great deal of local knowledge. These are developed into individual facts: this title covers this piece of ground, this company owns it, it bought it on this date. The facts are marshalled into a lexicon of the places, owners, rights and relationships involved. From the lexicon, arguments are assembled: ownership is becoming more concentrated; this estate falls within the scope of the new Act; this park is common good land. And the arguments go into finished products: a report, a community's application to buy, a council's register, a policy paper, a news story.

A manufacturer keeps a bill of materials for every product: a list of every part it contains and where each part came from. It is what lets them answer the question "what is this made of?" and, when a supplier turns out to have shipped a faulty batch, "which of our products contain it?"

Land research badly needs the same thing, and the figures themselves show why. The headline statistic about concentration has moved over the years. The well-known figure of 432 owners for 2012 was later corrected to 440, and a preliminary count of 433 for 2024 became 421 in the final analysis. That is exactly how research work behaves as the underlying data improves. It is also exactly why you want the bill of materials: so that when a part changes, you can see which numbers move. Or take community ownership. The Scottish Government's "Community Ownership in Scotland 2024" (gov.scot) counts 213,803 hectares, 2.7% of Scotland's land area, while "Who Owns Scotland 2024" from Andy Wightman counts 220,458 hectares. Neither is necessarily wrong. They are built from different parts, and the Government's own publication says that there is no definitive source of assets in community ownership. A bill of materials lets you see exactly where two such figures part company.

The technologies I want to describe map neatly onto that supply chain, so the rest of this article follows it.

Parcels, titles and owners as sentences

The basic parts of a knowledge graph are short sentences with three parts: a subject, a verb and an object. The standard for writing them is called RDF, the Resource Description Framework. For land, they might look like this:

Title QED00001 — covers —> Parcel 17

Glenmeadhanach Estates Ltd — is the registered proprietor of —> Title QED00001

Glenmeadhanach Estates Ltd — is registered in —> Jersey

The Àrd-ìre Trust — has a controlled interest in —> Glenmeadhanach Estates Ltd

(The estate and its title are invented, though readers who have done this kind of research will recognise the shape.) Each sentence stands on its own, and each comes from a different register. Because the company is the same thing in all of them, they join up by themselves into a chain running from a piece of ground to the people who decide what happens on it.

The important detail is that each thing in the graph has its own identifier, rather than being known by its name, a move often summed up as going from "strings to things". Land needs this more than most subjects. An "estate", in everyday speech, is a name on a map and on the gate. It is not a legal person and not a single title. One estate can be made up of dozens of titles held by several related companies, the same estate name can be used by unrelated landholdings at opposite ends of the country, and the owner's name may be spelled three different ways in three registers. Giving each title, parcel, company and person its own identifier, and then recording "these are the same" or "these belong together" as deliberate, sourced statements, is what turns a pile of records into something you can reason with.

A lexicon for land

Which brings us to meaning. What does it mean to own land? In Scotland the answer might be the registered proprietor, the beneficial owner behind a nominee, the person with a controlled interest recorded in the new register, an agricultural tenant, a crofter, a shareholder in a common grazing, or, in a different sense entirely, everyone, through the right of responsible access. A graph that records all of these as "owns" will produce confident nonsense.

An ontology is the agreed lexicon that prevents this: the kinds of things that matter (parcels, titles, holdings, owners, tenancies, rights, uses) and the relationships between them, each with a definition people have signed up to. This may sound like a job for lawyers and surveyors, and partly it is. But as I have argued before (semanticarts.com), building an ontology is a social process as much as a technical one, and it is a very good way of finding out where people do not actually agree. A community group, a landowner, a council and a researcher who share a lexicon can at least argue about the same thing.

There is no need to start from scratch. Internationally, ISO's Land Administration Domain Model (ISO 19152) sets out a common way of describing the parties, the land, and the rights, restrictions and responsibilities that connect them. Scotland's own Land Rights and Responsibilities Statement speaks the same language of rights and responsibilities. The Land Portal Foundation coordinates LandVoc (fao.org), a published vocabulary of land governance concepts built on the UN Food and Agriculture Organization's AGROVOC thesaurus, and Europe's INSPIRE programme defines a common model for cadastral parcels. A shared lexicon matters even within Scotland. Since 2018, Scotland's councils have been required to keep a register of their common good property (legislation.gov.uk), and each has done so in its own way. Comparing them is currently a job of hand work; with a common lexicon, it becomes a question you can ask.

Where is it? Questions with a map attached

Once the sentences are in a database built for them, called a triplestore, you can ask questions in a language called SPARQL. A SPARQL question is a pattern with gaps in it, and the database finds every way of filling in the gaps.

Land adds a dimension that most subjects don't have: place. A companion standard called GeoSPARQL (ogc.org) lets the graph hold boundaries and locations, so that questions can be about geography as well as connections. In plain English:

Find every holding over a given size that has changed hands in the last five years, lies within ten kilometres of a community that has registered an interest in buying land, and is ultimately controlled from outside the UK.

Wightman's analysis of every large rural land sale in 2023 (andywightman.scot) is exactly this kind of exercise, done painstakingly by hand. Once the parts are in a graph, the next year's version is a question rather than a project.

Land use belongs in the same graph as land ownership. The same holdings can be linked to what happens on them: forestry, farming, sporting estates, peatland restoration, renewable energy, and the public grants and private carbon credits that increasingly come with them. A 2022 study for the Scottish Land Commission (landcommission.gov.scot) found that carbon and natural capital values were an increasing influence on land prices, alongside timber. Following the line from a carbon credit, to the land it comes from, to the company that owns that land, to the people who control the company, is precisely the kind of question that is hard to answer from separate registers and straightforward in a graph.

Inspecting the records

Before anything is built from the parts, they need checking. A standard called SHACL, the Shapes Constraint Language, lets you write down the checks once and apply them to the whole graph. Every title must have a proprietor. Every parcel must have a boundary and an area. A holding's stated area should be roughly the sum of its parcels. Two sole owners of the same title cannot overlap in time. Every entry in a common good register must cite the deed, charter or minute it rests on. Run the checks and you get a report of what is missing or contradictory, which is where any researcher wants to look first.

It is worth adding a word about silence. Knowledge graphs work on an open world assumption: if the graph does not record an owner, that does not mean there isn't one. The land may simply not be in the Land Register yet, its owner sitting in a deed in the Sasines. Genuinely ownerless land is a different matter, with its own process, and a careful graph keeps "we have not found it yet" and "there is no owner" well apart.

Rules from the statute book

Land is governed by rules, and many of them are exactly the kind of "if this, then that" logic a computer can apply. Knowledge graphs can hold rules alongside facts, often written in a small rules language called Datalog, and apply them repeatedly until no new conclusions appear.

The simplest example follows control through a chain of companies:

If A has a controlled interest in B, and B owns C, then A has an interest in C.

Apply that through as many layers as the corporate structure has, and the graph will tell you who stands behind a piece of ground.

A more interesting example is counting. Any law that applies above a size threshold raises a question that sounds simple and isn't: what counts as one holding? One title? All the titles held by one company? All the land held by a group of connected companies and trusts? Wightman's analysis of the landholdings within the scope of the Land Reform Bill (andywightman.scot) is, among other things, a careful answer to that question. Written down as rules, a counting method becomes something everyone can see and argue about, and something a think tank can vary. Change the rule, run it again, and see which holdings move in or out of scope. That is a policy discussion conducted with the bill of materials on the table.

Rules cannot settle a legal argument, but they can lay one out. Whether a park or a building is common good land depends on its history and on legal tests developed by the courts. When Wightman reported in 2011 (andywightman.scot) that Union Terrace Gardens in Aberdeen was common good land, that conclusion was an assembly built from historical documents and legal reasoning. Recorded in a graph, with each step pointing to its evidence, the same conclusion can be inspected by the council, the community and, if it comes to it, a court, and each of them can see exactly what it rests on.

Four centuries of change, and the bill of materials

Land records have a depth of time that few other subjects can match. Titles are split and merged. Companies change their names and their owners. Councils come and go: when Scotland's burghs were abolished in 1975, their common good passed to new councils, and in 1996 to newer ones again, and every handover was a chance for records to go astray. A useful graph has to know not just who owns something, but who owned it on a particular date, and when that fact was recorded, which are not always the same thing.

Knowledge graphs can record both. Each source's statements can be kept in their own compartment (a named graph, in the jargon), so you always know which register said what. Ownership can carry the dates it began and ended. And the W3C has a standard vocabulary, called "PROV" (w3.org), for recording what each piece of information was derived from, what produced it, and who was responsible.

Put together, that is the bill of materials. A published figure points to the analysis that produced it, the analysis to the rules and facts it used, and each fact to the register entry, deed or document it came from. Then the manufacturer's recall question can be asked of land. If a title is corrected, a company's controller is updated, or a deed turns out to have been misread, which facts change, which holdings change, and which published figures and applications relied on them? The same chain can be followed for ideas as well as facts: how a proposal moves from a campaigner's pamphlet, to a review group's recommendation, to a section of an Act. For anyone who studies land reform, that is a history worth being able to query.

Beyond Scotland

Scotland is unusual in the depth of its concern with land, but it is not alone. In England, Guy Shrubsole's "Who Owns England?" (whoownsengland.org) used freedom of information requests and Land Registry data to estimate that less than 1% of the population owns half the land, and HM Land Registry publishes its house price data as linked data. In the Netherlands, the land registry, Kadaster (kadaster.nl), has gone furthest. It publishes its base registers of land, addresses, buildings and topography as a single linked knowledge graph, checked with SHACL and open to SPARQL queries. That is good evidence that a national land registry can work this way.

Further afield, the Land Matrix (landmatrix.org) tracks large-scale land acquisitions around the world, recording deals that are intended, concluded and failed. The Voluntary Guidelines on the Responsible Governance of Tenure, endorsed by the UN Committee on World Food Security in 2012, set out internationally agreed principles for land rights. The particular laws differ from country to country, but the questions are the same everywhere. Who owns this? Who really controls it? What is it used for, and for whose benefit? What has changed, and when? A shared lexicon and a common set of standards mean that work done in one country can be compared with, and built on by, work done in another.

Getting started

For a community group, a campaigning organisation, a researcher or a think tank, the advice is much the same as for any other knowledge graph. Start with the finished product and work backwards. Pick one question: one glen, one year of sales, one council's common good. Write down what you need to know to answer it, and let that decide what goes in. Reuse existing vocabularies rather than inventing your own. Most of the tools are free and open source, and a graph covering a single area will run on an ordinary laptop. Find one technically minded partner to help with the plumbing, while the people who know the land keep hold of the meaning.

Two cautions. First, a graph is only as good as its parts, and much of the most important material about land is in deeds and archives that someone has to read. The graph does not do the reading; it makes sure the reading is not lost. Second, land registers are public, but a graph that joins them together is a new thing. Where it concerns individuals rather than companies, data protection law applies, and it should be designed with that in mind from the start.

Back to the village hall

"Who owns it?" is the first question in the village hall, and behind it sits a supply chain of records stretching back more than four hundred years. Knowledge graphs will not make the answer simple. Some of it will always need a person reading a deed. What they can do is make the answer visible: every fact with its source, every rule written down, and every conclusion with its bill of materials. Then the community, the landowner, the council and the researcher can at least be sure they are arguing about the same thing, which, on the land question, would be no small achievement.