Image by Wolfgang Hasselmann on Unsplash
When researchers at the University of Cambridge set out to help hedgehogs, they didn’t start by tracking the animals themselves. They looked for brambles instead.
Yes, brambles—the same thorny bushes that scratch your arms on nature walks. These dense shrubs are a favorite hideout for European hedgehogs, offering them both shelter and food. And here’s the clever bit: while hedgehogs are too tiny to spot from space, brambles are big enough for satellites to see.
So instead of searching for hedgehogs the hard way—late night fieldwork, motion cameras, or waiting on citizen sightings—the Cambridge team decided to think bigger. Literally. They turned to satellites in orbit and trained an AI model to find patches of brambles across the UK.
Why brambles?
Brambles are like hedgehog apartment complexes. They offer thick cover to hide from predators, cozy spots for nesting, and a buffet of berries and insects—basically Airbnb and Uber Eats all rolled into one.
European hedgehogs have been declining by 30 to 50 percent over the past decade. So the logic goes: find the brambles, and you find potential hedgehog neighborhoods.
The tech behind the bramble hunt
The AI model, developed by researcher Gabriel Mahler, doesn’t rely on anything fancy like ChatGPT. Instead, it sticks to classic machine learning approaches—logistic regression and k-nearest neighbors. It’s lightweight and efficient, which makes it a strong candidate for real-time use in the field one day.
Here’s what the Cambridge team used:
- Satellite images from the European Space Agency’s Sentinel satellites
- TESSERA, a tool that processes and represents that satellite imagery
- Ground-truth data from iNaturalist, a citizen science platform loaded with wildlife sightings
And then they went out to see how well it all worked.
Image by Abdullah Madawi on Unsplash
Field testing: Can we actually find brambles?
To put their model to the test, Mahler and a few colleagues—Sadiq Jaffer, Anil Madhavapeddy, and Shane Weisz—packed smartphones and GPS gear and hit the ground around Cambridge.
One of the first stops was the parking lot at Milton Community Centre, flagged by the model as likely to have brambles. Sure enough, within twenty seconds, they found a big patch.
At Milton Country Park, every high-confidence spot predicted by the AI was packed with brambles. In a quiet residential area, the team found a seemingly empty lot that had been overrun by prickly shrubs.
Most fittingly, the model pointed them to a local reserve in North Cambridge. Its name? Bramblefields. And yes—the brambles were thriving.
What the model does well—and where it’s limited
The satellite-based model nailed the large, uncovered bramble patches. But it struggled with brambles hiding under tree canopies. That’s understandable—satellites don’t have x-ray vision.
As researcher Sadiq Jaffer explained, since the model is trained on overhead imagery, anything obscured from above naturally scores lower in confidence.
Image by Jacek Ulinski on Unsplash
Still early, but promising
This is all still a work-in-progress. The results haven’t been peer-reviewed yet, and their field test was more of an afternoon stroll than a formal study. But the team is upfront about that. They’re treating it as a proof-of-concept with room to grow.
And there’s exciting potential here:
- The model is lightweight enough to one day run on mobile phones
- It could support real-time field validation by researchers or volunteers
- Paired with citizen science platforms, it could scale faster than traditional methods
Looking ahead, approaches like this could help track not just hedgehogs, but invasive plants, pest outbreaks, or ecosystem changes linked to climate stress.
For hedgehogs, those quill-covered introverts trying to survive in a fast-changing world, every bramble patch spotted from above could be a step toward giving them safe ground below.
Keywords: AI wildlife tracking, hedgehog habitat, satellite imagery, conservation technology, machine learning biodiversity, remote sensing for nature, bramble detection AI, European hedgehog decline, TESSERA, iNaturalist data mapping.
