TESSERA - the AI model helping to protect hedgehogs from space
Hedgehogs are under threat in this country. Government surveys suggest their numbers have fallen by up to 75% in rural areas since the turn of the century, and sightings of them in our gardens are rare. But now these much-loved animals are getting help from an unlikely new ally: AI.
Researchers here at the University of Cambridge have been developing a powerful new AI tool called TESSERA. It analyses the data from thousands of measurements of Earth taken from satellites to generate exceptionally detailed scientific maps of our planet.
After learning about the plight of UK hedgehogs, TESSERA researchers have teamed up with the Cambridge Conservation Initiative to map hedgehog habitats in the UK.
By connecting the habitat maps with other research into hedgehog habits and movements, the idea is to better understand the environmental factors behind their decline.
The Department of Computer Science and Technology's Professor of Planetary Computing Anil Madhavapeddy co-leads the development of TESSERA . He explains that TESSERA could be especially helpful for identifying pressures on rural hedgehogs, which have seen the steepest decline in numbers.
In these environments, hedgehog habitats are increasingly being replaced by large open fields and roads. These also break connectivity between existing habitats, putting additional pressure on remaining populations of the animal.
Mapping and protecting habitats
TESSERA can help map the extent of the problem and perhaps guide efforts to join up or protect remaining habitats, Madhavapeddy suggests. And he adds that "if we take action now, we can figure out how to stop this dangerous decline in their numbers."
TESSERA works by churning through petabytes of data to create information-rich maps of the Earth. In these maps, each pixel is a stack of data about a 10m x 10m square of the Earth's surface. The AI churns through so much data that it can fill in gaps caused by factors such as cloud cover.
Hedgehogs frequently live in bramble patches (which hide them from predators and shelter them from harsh weather). So for the hedgehog project, researchers trained TESSERA to look for patterns in the data that matched how bramble bushes scatter light from above.
To validate TESSERA's output, the researchers generated a few local maps around Cambridge – and then decided to go out and see how well the AI model had performed. The answer was, very well. In every location on their map that TESSERA flagged, the researchers found brambles, according to Sadiq Jaffer, Assistant Research Professor in CST.
Surprise findings
There were even some surprises. On one occasion, TESSERA flagged up what seemed like an unlikely hotspot for brambles as it was in the middle of a built-up residential area.
"We thought that can’t be right," Sadiq says. But when the researchers visited it, they discovered that it was in fact an overgrown building site. "It turned out that a couple of houses had been knocked down and while they were waiting to rebuild something, the area had overgrown with brambles." (See more details about Sadiq's quest to find out whether AI tools can 'see' bramble patches from space here.)
The researchers are still working through the hedgehog habitat maps but are hoping to publish their results later in the year.
- The first version of TESSERA went online in June 2025 as an open model available to any researcher for free. The computational infrastructure required to compute TESSERA’s maps, known as embeddings, was provided by the University of Cambridge and industrial supporters including Vultr and AMD.
- Learn more about TESSERA and how it’s supporting research, at geotessera.org
- And see Satellites and AI used to track UK hedgehogs in bid to slow decline — BBC News
Read the original article by Rachel Gardner on the Department of Computer Science and Technology's website here.