Researchers at the University of Cambridge have proposed a new way of tracking hedgehog habitats. Because the animals themselves are too small to be spotted by satellite, the team focused on blackberry bushes, where they usually find shelter and food.
Data from the European Space Agency’s (ESA) Sentinel satellites combined with machine learning algorithms were used for the analysis.
The developed model combines methods of logistic regression, nearest-neighbor classification, and the TESSERA system for processing satellite images. Additionally, observational data from citizen scientists collected through the iNaturalist platform was taken into account.
According to the scientists, this hybrid approach allowed them to build a map of suspected hedgehog habitats across the UK.
To test the accuracy of the model, the researchers conducted field trials in Cambridge. They checked the AI’s predictions against real terrain and verified that the model reliably identified large open thickets. Smaller bushes under trees were worse, due to the limitations of satellite imagery.
Despite the fact that the project is still at an early stage, its potential is already evident, according to the authors of the study. Unlike time-consuming nighttime observations, satellite analytics can cover large areas simultaneously, making it useful for national conservation programs.
The scientists emphasize that this is still a proof of concept, and the model itself has not yet been fully peer-reviewed. Nevertheless, they plan to continue the research by expanding testing and developing an active learning system that can be used in the field via mobile devices.
University officials emphasized that the potential of the method goes far beyond hedgehog protection. Similar algorithms could be used to monitor invasive plants, agricultural pests or track changes in ecosystems.
The Cambridge project shows how relatively simple AI tools can solve the problems of biodiversity conservation and complement traditional methods of field research, experts believe.