TL;DR

The largest woodland bat survey ever carried out in Britain would have been impossible without AI, according to the scientists behind it, the BBC reported on 30 October. Over three years, acoustic recorders in 310 woodlands logged 98 million bat calls, a volume the team says would have taken more than three decades to sort by hand. The results include hotspots for the rare, red-listed Barbastelle in south-west England.

How it worked

Researchers fixed recording devices to poles among the trees and left them to capture bats’ echolocation calls night after night. That produced 90,000 hours of audio. Open-source AI tools then picked out and classified the calls, letting the team spot patterns and flag species across the whole country.

“AI has fundamentally changed the scale at which we can monitor wildlife,” said Kate Jones of University College London, who called the project “a nice story of AI for good”. The Bat Conservation Trust and Forest Research also worked on the survey.

What it found

The Barbastelle lives in ancient woodland and is notoriously hard to track. The survey recorded it, along with other rare species, in Somerset, Devon, Cornwall and Dorset, and its busiest sites clustered in the South West. The Bat Conservation Trust’s Lia Gilmour described the project as “unprecedented” and singled out Somerset as a stronghold. “There was no way we would be able to do this manually,” she said.

Why it matters

In our view, the survey shows AI being used for what it does well: handling volumes of routine classification that no human team could get through. It is a contrast with other recent wildlife stories, in which AI-edited photos have contaminated citizen science records used to track species.

Looking Forward

We think the survey points to a practical use for public bodies and conservation groups across the UK: cheap sensors combined with open-source models can turn monitoring that was once too costly into routine national data. The harder questions are about funding the fieldwork and keeping the recorders running, rather than the analysis.