AI Spots Hedgehogs from Space by Tracking Their Bramble Homes
▼ Summary
– Cambridge researchers are using satellite imagery and AI to map UK hedgehog habitats by identifying bramble patches, which serve as their shelter.
– This approach addresses the difficulty of tracking hedgehogs directly, as their populations have declined by 30-50% in the past decade.
– Brambles provide essential resources for hedgehogs, including daytime shelter, nesting sites, protection from predators, and food sources like insects and berries.
– The AI model offers a scalable alternative to traditional, labor-intensive survey methods, enabling large-scale conservation planning.
– The team validated their bramble-detecting AI model, which uses simple machine learning techniques, through ground-truthing with GPS and smartphone checks in Cambridge.
Finding innovative ways to monitor declining wildlife populations is a critical challenge for conservationists. A team at the University of Cambridge is tackling this problem for hedgehogs by using a clever technological workaround. Since these small animals are nearly impossible to see directly from orbit, scientists are training artificial intelligence to locate them indirectly by pinpointing their preferred homes. The research focuses on identifying bramble patches, which are the thorny shrubs that offer European hedgehogs essential shelter and foraging grounds.
The sharp decline in hedgehog numbers across the UK, estimated between 30 and 50 percent in recent years, makes effective monitoring more important than ever. Traditional survey methods, which involve night-time fieldwork and public sightings, are difficult to implement on a national scale. Researcher Gabriel Mahler proposed a more efficient solution: using satellite imagery to map potential habitats. His AI model is designed to scan vast areas from space, searching not for the elusive animals themselves, but for the distinctive signatures of the brambles they depend on.
These dense, prickly shrubs serve as a vital resource for hedgehog survival. They provide safe daytime resting spots, secure locations for building nests, and protection from predators. Furthermore, brambles support the ecosystem by attracting insects and producing berries, which in turn sustains the invertebrate populations that form the core of the hedgehog diet. By accurately mapping these key plant clusters, conservationists can identify priority areas for protection and targeted surveys without the need for exhaustive ground-level searches.
It is important to clarify the type of AI being used in this project. The system is not a complex large language model but relies on more straightforward machine learning approaches. The model combines logistic regression with k-nearest neighbors classification to analyze the data. It processes satellite imagery from the European Space Agency’s Sentinel missions, using TESSERA earth representation embeddings, and cross-references this information with verified ground observations contributed by users on the iNaturalist platform.
To validate the accuracy of their bramble-detecting AI, the research team conducted a practical field test. Mahler, along with colleagues Sadiq Jaffer, Anil Madhavapeddy, and Shane Weisz, spent a day walking through various areas around Cambridge. Equipped with smartphones and GPS devices, they physically checked locations that the model had identified as likely bramble patches, comparing the AI’s predictions against what they actually found on the ground.
(Source: Ars Technica)