AR games mingle with underground assets in the data plan for 200-year-old Ordnance Survey
Feature Britain's Ordnance Survey (OS), founded in 1791, is interloping in the digital age. Minecraft, AR gaming, and EV charger locations have all become part of its portfolio, alongside the paper-based maps beloved by the nation's legion of cagoule-clad outdoor types.
But the digital mission goes deeper, CTO Manish Jethwa tells The Register, pointing to roof types as an example.
"OS has secretly, behind the scenes, been working on it. It's an embedded part of our production pipeline. The datasets that we produce today are already utilizing machine learning, including understanding roof types and roof materials automatically from aerial imagery. We use machine vision to get 3D models of the roofs and understand aspect and area. When you can then take those two datasets together, you can start to very quickly understand areas for solar potential of buildings, we can also detect where buildings already have solar panels," he says.
OS has been putting together its buildings dataset over five years. In September 2024, its basement presence was released and in March this year OS released a suite of roof data to provide information on 28 key attributes for more than 41 million buildings across the UK.
The data is used by banks and insurers to provide highly accurate information about the built environment which they can trust when making risk and lending decisions, identifying fraud or making sustainability assessments.
We use machine vision to get 3D models of the roofs and understand aspect and area
OS developed the Unique Property Reference Number (UPRN), which can be integrated with other third-party datasets such as flood-risk data which financial services institutions need to make critical decisions. The UPRN provides the definitive link at a building level which is far more accurate than using postcode data.
Now a publicly owned limited company, the nation's cartographer was founded in the wake of Scotland's Jacobite rebellion of 1745 and won backing during the Napoleonic wars. It first launched an electronic OS MasterMap database back in 2001, containing more than 600 million geographic features.
Earlier this year, OS began a search for a tech supplier to help it obtain and manage data from utility companies for a project that aims to avoid damage to subterranean infrastructure, which costs around £2.4 billion a year. Now in its beta phase, the National Underground Asset Register (NUAR) provides a digital map of underground pipes and cables in England, Wales, and Northern Ireland. The government said the project would aid different groups that install, maintain, operate, and repair buried infrastructure.
Built over the last 20 years, the data stack is diverse, Jethwa says. "From a legacy perspective, our databases are built on traditional architecture, using PostgreSQL to develop capabilities in-house."
In terms of data management at scale, OS predominantly relies on Microsoft Azure's data stack internally, but also works with vendors such as Databricks to run spatial queries on large-scale datasets and create catalogs.
But the size of the datasets OS is able to exploit can be a burden as well as an asset, Jethwa explains. "They are going to persist for at least a decade, if not more, and so the methodologies that we develop typically have to be enduring capabilities. It's a challenge for us," he says.
He explains that large-scale business transformation projects to develop new services can't address the whole data lifecycle, because of the longevity of the datasets. OS is moving from a "project management or program management perspective much more onto like a product-led [one], the idea of creating enduring teams, both on the platforms that we build and in terms of the services that we offer, means that we can continuously iterate rather than thinking about them as isolated projects," he says.
While it has deep roots in machine learning, OS is also adopting the most recent trend in genAI and large language models. Internally the team is exploring how they can help open up datasets to non-expert users, Jethwa says.
"Building spatial queries goes into the reserves of geographic information system (GIS) experts. If you want to all of the buildings that are within 100 meters of a water feature, because maybe you're trying to identify buildings that might be susceptible to flooding, you now need to build very specific queries, which are buffering the existing features by 100 meters and then taking buildings that exist within that buffer zone.
"Those kind of spatial data queries can be quite hard to be able to create by hand. But large language models help you to be able to translate a natural language expression, through an API request, to be able to actually pull that data out. It opens up the realm to a domain expert, but not necessarily a GIS expert," he says.
OS has to balance its mission to provide usable geographic information to businesses and other organizations with its duty to the public. "In addition to kind of providing these, these datasets out to the markets, we're just trying to get more people outside," Jethwa says.
In 2021, OS partnered with game developer PRELOADED to build an AR game to encourage people to get outside and explore the UK.
"People are adopting new ways to be able to interact with our data. They want to be able to find and share route information with one another. That's essentially the 'get outside' community aspect of what we do, whether it is through games or now through the OS maps app," Jethwa says.
The sources of data upon which OS can build is ever expanding. It can extract features directly from drone and aerial imagery by using deep learning algorithms and offer it in an ever more up-to-date fashion.
However it is not just about accuracy. Jethwa says OS wants to make the data fit for purpose so it can be adopted in more places. "We want to hit more destinations in the sense that we want to get our data into more hands. We know already that our data is touched by the average user 42 times every day. But even that doesn't feel like it's enough." ®