Can AI detect hedgehogs from space? Maybe if you find brambles first. (arstechnica.com)

Researchers at the University of Cambridge are using satellite imagery and machine learning to map potential hedgehog habitat across the UK by detecting bramble patches—thorny shrubs hedgehogs use for shelter, nesting and foraging—rather than trying to spot the animals themselves. European hedgehog populations have fallen roughly 30–50% in the last decade, and traditional nighttime surveys are costly and hard to scale. By identifying habitat features that correlate with hedgehog presence, the team hopes to provide conservationists with a scalable, continuous way to prioritize monitoring and restoration. Technically, the project combines Sentinel satellite imagery (processed via TESSERA earth-representation embeddings) with ground-truth labels from iNaturalist and on-the-ground GPS checks in Cambridge. The bramble detector uses relatively lightweight ML—logistic regression and k-nearest neighbors classifiers—rather than large language models. The group validated predictions by walking local sites with smartphones and GPS to compare model outputs to reality. The approach highlights a practical trade-off: you can’t see individual hedgehogs from space, but you can detect habitat proxies at scale. If refined, the pipeline could be extended to other key vegetation features, improving large-scale conservation planning while still depending on careful ground validation to manage false positives and resolution limits.