Executive summary
Introduction and who this work is aimed at
Read More
Executive Summary >> Part One >> Part Two >> Part Three >> Part Four >> Glossary >> References
In Part One, we explore the state of the UK HE sector’s land (specifically, the core estate). Risks, opportunities and the value of UK HE’s ecosystems will be covered in Parts Two, Three and Four.
The online tool which we developed to accompany this report allows users to explore the land use data at institutions and at sector level.
Where possible, we have tested and cross-referenced our geospatial data against other public datasets, including the HESA Estates Management Record (EMR) and the UK Natural Capital Accounts. These have also been manually sense-checked across the sector map.
Our dataset on the UK HE estate is drawn from Open Street Map (OSM) [1]. By creating custom queries for land parcels tagged as ‘university’ and/or ‘college’ lands[2], we were able to create a representative dataset spanning 174 universities and providers [3] and covering 6,390.1 hectares (ha). Each institution is delineated with a distinct polygon.
As with any similarly large-scale exercise, our dataset inevitably contains a degree of “noise”. Our testing indicates that our land coverage aligns to an 80% degree with the extent of “total grounds area” in the 2022/23 HESA Estates Management Record, a (voluntary) sector-level dataset on HE estates[4].
“Total grounds area” (recorded as 7,293 ha for 135 reporting institutions in the Estates Management Record) generally does not cover farms and other ancillary lands outside the “core” estates of institutions. “Total site area”, which does include these land holdings, is recorded as 12,565 ha for the 135 institutions.
Our data therefore covers a little over half of institutions’ total reported lands.
That said, it is also important to note that the area under investigation does not include landholdings of other HE entities which in some cases (for example, colleges of the Universities of Oxford and Cambridge) may be substantial; nor does it cover the impacts of UK institutions’ supply chains on land use change inside or outside the UK, which tend to be substantially greater than the impacts of their onsite operations[5].
While it is not possible at present to undertake a comprehensive analysis of these wider dimensions due to the limitations of available data, we believe that the dataset on universities’ total grounds area is sufficiently robust to allow for grounded discussion of the land use of core estates at sector level. Where we offer estimates for individual institutions, those institutions will want to take this data as an outline guide and test it against their own records.
We have mapped our records of the UK HE estates onto other geospatial datasets in our sector mapping tool. The granularity of resolution for geospatial datasets can vary substantially: while some are detailed enough to provide information across different parts of a single site, others are at resolutions of several square kilometres – valuable for providing a sense of the characteristics of the general area, but not for detailed site analysis.
As typifies large-scale geospatial analysis, much richness and details around individual cases are smoothed over or cancelled out[6]. For example, we have not accounted for factors such as the condition of the ecosystems for carbon flux calculations, even though condition does have an appreciable impact on carbon flux.
Nor have we been able to explore in detail the full range of opportunities associated with the built estate (for example, urban nature-based solutions), or the value of nature-based solutions and land use for adaptation to climate change.
The social and cultural value of land for human recreation, health and wellbeing is another dimension which we have not been able to explore in detail, but university estates can clearly provide offer value here for their local communities and the general public, as well as their staff and students.
Similarly, while we note the economic dimensions of the opportunities considered in regional ecosystems and their links to education, skills and research, we do not include a detailed discussion.
We see further exploration of these areas as an exciting agenda for future work.
In our dataset, the estates of UK universities cover an area in the region of 6,390.1 hectares (ha).
The lands recorded are held across a variety of land tenures, including freehold, leasehold and in some cases Crown lands. They cover the “core” estate (or “grounds area”) of each institution: peripheral and ancillary holdings, which are substantial for some institutions, are not included in this dataset[7].
Therefore, within this work, terms such as “university lands” and “the higher education estate” generally refer to core estates unless stated otherwise.
Some of the data appears surprising at first glance. For example, Imperial College London – which we assumed to be an essentially urban estate – has a large campus outside of London and one of the highest tree covers of any institution. Even if some landholdings are less well-known than flagship campuses, they still hold options and opportunities for individual institutions and the sector.
While the HE estate is a small portion of total land area in the UK (around 0.026%), it is significant in national terms due to the economic outputs it generates (as noted in the Executive Summary – £43.9 billion income in 2022/23, equating to around 1.7% of UK GDP).
Risks related to land and location explored in this work therefore may cause not only potential damage to higher education infrastructure and associated costs, but disruption to important economic activities.
The sector’s land also offers opportunities to support sustainability objectives for institutions and their local areas, which we will explore later.
We used geospatial tools to map the surface cover (alternatively, the “land use”) of the HE estate.
What covers the earth’s surface, including how humans use land, has a fundamental role in the earth’s systems. A such, it is a critical dimension of many issues around environmental sustainability[8].
We mapped the sector’s land used into four categories which are useful for this exercise: built, grass, trees, water.
These categories represent generalised types of land cover. The ecologies of specific sites vary according to their locality.
Each land cover type is made up of several land categorisations which are collapsed from the reference dataset. The data comes from the UK Centre for Ecology and Hydrology (UKCEH), which lists 21 unique land cover types covering the whole UK at a 10-metre resolution. UKCEH offers parent categories for these subcategories, and this grouping was used as a template for the classification we used in this report. For easy reference, the parent categories with all subcategories are as follows:
UK HE land use categories
At the aggregate level, the UK HE sector’s estate’s surface cover breaks down as below[11].
We set these against equivalent figures for the whole UK, drawn from the UK Natural Capital Accounts 2024. The Natural Capital Accounts distinguish 8 land cover categories, which are mapped to our categories the final column of the table. Some categories do not map perfectly (for example, our dataset treats floodplains as land while the Natural Capital Accounts groups them with water), but we believe they are close enough to support comparisons between HE land use and land cover at the country level. See Part Four for further discussion of category mapping.
UK HE and UK-level land use: aggregate data
A schematic representation brings out clearly the highly built nature of the HE sector’s lands. Built land constitutes nearly 60% of the HE estate land surface, while for the whole UK it is 8%. HE lands have more grass (30%) than UK as a whole (11%), and a broadly comparable amount of tree-covered land (10% against 13%).
The other main differences are firstly that core university estates critically do not contain croplands which cover nearly 50% of the UK land surface (although some institutions do own such lands in their ancillary holdings), and generally do not contain areas such mountains, bog and fenland (the “other” category) which cover around 12.5% of the UK.
Individual institutions can see the breakdown of their land use through the online tool which accompanies this report.
We used this data to develop a typology of university estate profiles, using k-means clustering[12]. The typology is based on the relative proportion of the three core land use types (built, grass, trees) at each institution.
Water surface cover, although not technically land use, was also tested for significance but we found its influence on the typology relatively minor. 37 universities in our dataset have water bodies within their polygons. For those 37 institutions, water area accounts for less than 1.5% of land area and therefore had a negligible influence on the typology.
We therefore decided to exclude water cover from the typology classification, but we have still included it in the total land cover for the universities. This compromise allows the presence of water to be considered, while improving the statistical strength of our clustering methodology.
The typology identifies three clusters of institutions, each of which stands out for possessing a higher proportion of one of the three core land use types than the other two clusters.
Land use clusters: key points
Here we show the land cover breakdown of a typical member of clusters 1, 2 and 3, based on the mean land use data across each cluster.
Land use clusters: typical (mean) land cover for a member of each cluster
These categories are heuristic and developmental with this work. But we believe they are useful in terms of opportunities which are applicable to institutions and their estates (discussed in more detail in Part Three) and may encourage different ways of thinking about the sector and institutional profiles.
Notes and references
[1]Open Street Map (OSM) is a collaborative, open-source mapping project that represents the archetype in community driven information sharing. Maintained by its users, OSM differs from proprietary mapping services by encouraging community mappers to edit, update and share geographic information. The result is a rich dataset that is free to use, publicly available and easily downloadable. As such, this project is a case-study example of the importance of open-source information for sustainability. The dataset represents a snapshot in time of the data that was held in OSM on 20 December 2024. If users spot errors in the boundaries of their institution, we encourage them to engage with OSM to update this.
[2] Our methodology bears many similarities to a recent (more ambitious!) research project which mapped the UK’s farms. See Sheikh et al. 2025. A Field-Level Asset Mapping Dataset for England’s Agricultural Sector. Scientific Data 12: 1240.
[3] 129 higher education providers which submit data to HESA are not included in our dataset as they did not have data available on Open Street Map. The full list of mapped institutions is providedin the mapping tool that accompanies this report.
[4] Available at: https://www.hesa.ac.uk/data-and-analysis/estates/table-1.
[5] See Bull, J. et al. 2022. The biodiversity footprint of the University of Oxford. Nature 604, 420-424.
[6] See Costanza, R. et al. 2014. Changes in the global value of ecosystem services. Global Environmental Change 26: 152–158 for discussion of the characteristics of analysis at different spatial scales.
[7] This is helpful firstly for clarity and ease of processing (and to our knowledge, no comparably comprehensive spatial dataset of universities’ peripheral landholdings exists). Additionally, the opportunities for peripheral holdings such as commercial land or farmland are substantively different, and carry different kinds of opportunity costs, from the opportunities which we examine for core estates.
[8] International Panel on Climate Change (IPCC). 2019. Climate Change and Land: an IPCC special report on climate change, desertification, land degradation, sustainable land management, food security, and greenhouse gas fluxes in terrestrial ecosystems; United Nations Convention to Combat Desertification (UNCCD). 2022. The Global Land Outlook, second edition.
[9] Arable land is included in the grasslands of university polygons given the limitations with land use data resolution and pixel misclassification. In most cases, arable pixels represented grass areas. Higher education core estates (“total grounds area” in the Estates Management Record) also exclude commercial farmland).
[10] Classes included in the Other/NA category were not present in this analysis. Land types of this group accounted for less than 0.45 ha and were a result of misclassification, so they were manually reclassified to fit the proper categories within the university polygons. These classes are present elsewhere in the UK, however.
[11] These figures are an aggregation of pixels provided in the 2023 Land cover dataset from the UK Centre for Ecology and Hydrology (UKCEH).
[12] K-means clustering is an unsupervised machine learning algorithm (executed in Python) that groups similar data points into distinct clusters (k). We manually chose three groups to align with the key land cover types. The three resulting clusters are based on the institutions’ campus land cover percentage of the three core land use types which determines the distance from the cluster centroids.
