HM Land Registry wants to make more use of satellite data – and plans to spend £72 million to do so, with a contract notice set to land in October.

The non-ministerial government department was created in 1862 and oversees British land and property ownership valued at some £8 trillion.

It set out in its 2022-2025 business plan to make “our data more accessible and interoperable” – but wants to do more with geospatial.

That’s according to a market engagement notice published on July 28.

HM Land Registry says that it has an “exciting and ambitious vision to transform our data into interoperable, machine interpretable, geospatially enabled information to better serve our customers with digital services…”

It wants “wider strategic guidance with particular focus on geospatial expertise” as it works on four programmes focused on “delivering radical, permanent changes to the way we structure, store, extract and use data.” 

Getting this right, HM Land Registry said in the market notice, will have “profound enabling effects for… informed policymaking across Government; and unlocking wider economic growth and resilience.”

It estimated that it will publish a contract notice by October 6. 

HM Land Registry works closely with the government’s Geospatial Commission – which commissioned the Alan Turing Institute in 2022 to explore how land use decision making can be supported by geospatial data, including earth observation data, and artificial intelligence.

The institute published its findings on that in November 2024, after spinning up a demo “state-of-the-art geospatial AI” tool called DemoLand.

It warned that the “LLMs which power the DemoLand agent incur nontrivial costs as they must access the ChatGPT API.” (Why they “must” access the ChatGPT API rather than using an on-premises open-source model at a fraction of the cost is anybody’s guess; answers on a postcard.)

“The training required to harness the power of the foundational satellite model, likewise, requires significant GPU resources. These costs could constitute a challenge to the wide scale adoption of this technology…

“ An additional challenge exists in licensing the data required for retraining these models. [We used] publicly available Sentinel-2 data.

“While this low-resolution imagery works better than expected, other use cases are likely to require higher resolution imagery which can be expensive and come with licensing restrictions,” The Institute noted. 

It concluded that there is an “opportunity for a central body to help organise both licensed access to data and the compute infrastructure required to develop geospatial AI tools” – could this be the Land Registry?

The link has been copied!