Scotland’s first ever nationwide high-resolution habitat map has been launched.

It's hoped the satellite images can provide insight into how the country’s landscapes are changing over time, while examining the potential of artificial intellicence to make positive change for the environment.

The maps were developed by collecting over 500,000 data samples across Scotland for 22 types of landcover, before AI was used to analyse tens of thousands of satellite images from a range of different sensors. 

The end product, which has been in the making since early last year, is aimed at helping landowners and policy makers with their land-use decision making, including working out where to restore habitats.

The work is part of the Can Do Innovation Challenge Fun, managed by Scottish Enterprise, for which six applicants were chosen in January 2020 to develop AI-enabled technology.

The Herald:

One of those successful applicants was Space Intelligence, who worked alongside partners NatureScot and the Scottish Wildlife Trust to develop AI that can interpret satellite images and identify different land cover types such as forests, meadows and crops.

These images, or maps, are hoped to harness the potential of helping address the climate emergency, due to the way some types of vegetation and ways of using land store more carbon and are more beneficial to biodiversity than others.

The Herald: Two example heat maps showing competing factors that could be considered in prioritising habitat restoration: the opportunity for sequestering carbon, and the accessibility of the site.Two example heat maps showing competing factors that could be considered in prioritising habitat restoration: the opportunity for sequestering carbon, and the accessibility of the site. The Edinburgh-based Satellite data company now says it can provide government agencies and land managers with accurate, up-to-date information on what type of vegetation is growing where.

Ed Mitchard, Chief Scientific Officer and Co-Founder of Space Intelligence, said mapping Scotland’s landcover at this high resolution and to 22 different classes was a significant challenge for the team for a number of reasons.

He said: “Firstly, Scotland’s habitats are diverse and often exist in small patches and along subtle gradients, needing us to integrate multiple types of satellite data in our advanced AI system to perform the classification.

The Herald: Two small subsections of the maps, showing the detail of the classes and changes picked up.Two small subsections of the maps, showing the detail of the classes and changes picked up.

“Secondly, Scotland is one of the cloudiest countries in the world, meaning we had to go through tens of thousands of images to obtain multiple cloud-free views of each habitat patch, as well as bring in other satellite data types that can see through clouds.

“And finally, applying our detailed mapping to a whole country involved scaling our cloud-based technology to deal with datasets in the billions rather than tens of millions of pixels.”

He added: “We are delighted to have successfully overcome these challenges, producing highly accurate maps, and look forward to taking our new experience and technological developments to the next challenge.”

The Herald:

The Herald: Detail of the map around Loch Rannoch, Scottish Highlands, showing the Black Wood of Rannoch to the south of the loch, forest to the north, and steep hills and moorland all around. This is compared to a single season satellite image, showing that the detailed differences in classes and forest types are hard to see by eye, but picked up by Space Intelligence's advanced artificial intelligence, which looks at each habitat patch hundreds of times using different satellites throughout the year.Detail of the map around Loch Rannoch, Scottish Highlands, showing the Black Wood of Rannoch to the south of the loch, forest to the north, and steep hills and moorland all around. This is compared to a single season satellite image, showing that the detailed differences in classes and forest types are hard to see by eye, but picked up by Space Intelligence's advanced artificial intelligence, which looks at each habitat patch hundreds of times using different satellites throughout the year.

The Space Intelligence system is now available online here.