AI Helps Log 98 Million Bat Calls Across 310 English Woodlands

England’s largest woodland bat survey has used artificial intelligence to pick out more than 98 million bat calls from recordings made in 310 woodlands, giving conservationists their first national picture of several rare species. Results from the National Woodland Bat Survey were published on 30 October by the Bat Conservation Trust (BCT) and Forest Research, which led the work.

The scale is the story. Between 2021 and 2024, acoustic detectors recorded on 9,734 nights and produced 217 terabytes of audio, about 90,000 hours. BCT says that is more than ten years of continuous listening.

How AI made the woodland bat survey possible

Bats are hard to count. They fly at night, move quickly and are rarely seen, but they give themselves away through echolocation, high-pitched calls mostly beyond human hearing. Detectors left in a wood can record those calls, and each species has its own acoustic signature.

According to the survey’s dataset record on Defra’s data platform, the programme samples up to 210 one-hectare National Forest Inventory woodland sites in England, with AudioMoth sensors recording from sunset to sunrise. The calls are then classified with BatDetect2, a machine learning model developed at University College London and the University of Edinburgh.

The AI was not left to work alone. A subset of its identifications was checked by 19 bat call experts from BCT’s Verifier Network, whose members have passed a standardised identification test. The BBC reported that the analysis would have taken more than 30 years without AI.

“AI has fundamentally changed the scale at which we can monitor wildlife,” said Professor Kate E Jones of UCL. “These efficient open-source AI tools allow organisations like the BCT to identify millions of bat calls across thousands of nights of recordings, identify patterns and flag species at a national scale, while BCT’s expertise ensures the results can be carefully verified.” She told the BBC it was a “nice story of AI for good”.

What the survey found in England’s woods

Woodlands matter to every British bat species, and broadleaved woods recorded the highest average activity. Common pipistrelles turned up in all 310 woodlands. Soprano pipistrelles were found at 99% of sites, Myotis bats at 98%, noctules at 94% and brown long-eared bats at 90%, Forest Research says.

Every bat species or species group found in Britain was detected except the serotine, which is hard to identify from sound alone. The survey also recorded the grey long-eared bat and the barbastelle, both on the UK Red List, along with the near-threatened Nathusius’ pipistrelle and Leisler’s bat.

The geography is new. Common and soprano pipistrelle activity was highest in the west of England. Barbastelle hotspots were spread across southern England, apart from much of the South East, and the survey logged barbastelles in parts of Cornwall, Devon and Dorset where few records existed before.

“By harnessing AI technology, we have produced one of the most comprehensive datasets ever assembled on bats in England’s woodlands,” said Dr Chloe Bellamy, senior spatial scientist at Forest Research. “This extraordinary data will improve how we create and manage woodland, to support bat populations and wider nature recovery.”

The survey was funded by the UK government through Defra’s Natural Capital and Ecosystem Assessment (NCEA) programme. David Jones, its senior responsible owner, said the survey “gives a full picture of bats across England’s woodlands, using technology to analyse millions of recordings at a scale that wouldn’t have been possible by people alone.”

A baseline that fills gaps in bat monitoring

For seven species, including both horseshoe bats, the barbastelle, both long-eared bats, Leisler’s bat and Nathusius’ pipistrelle, the survey is now the only national-scale survey in England away from known roosts. For three of them, the barbastelle, Leisler’s bat and Nathusius’ pipistrelle, it is the only active national monitoring in England of any kind, according to Forest Research.

That turns a one-off count into the start of a trend line. Dr Lia Gilmour, research manager at BCT, said bats “can indicate the quality of the habitats they rely on for survival, whether feeding, breeding or just getting from A to B”, and called the datasets “a powerful resource”. The team plans to model how climate, woodland cover, hedgerows and habitat connections shape where bats live. BCT says baseline data collection will continue until 2027.

Open data is what makes this AI result useful

The headline number is 98 million calls, but the more lasting result is that the data is public. Species occurrence and activity figures can be downloaded as CSV files under the Open Government Licence, and explored through an online Data Explorer. Any woodland manager, local authority or researcher can now check how bats use a type of wood without paying for a fresh survey.

The method is also worth copying. The AI does the volume work and trained people check a sample, which keeps errors visible rather than buried in the output. Jones’s point that the tools are open source matters for the same reason: a model others can inspect and reuse travels further than a one-off contract.

There are limits. The survey covers England only, so it says nothing direct about Scotland or Wales, despite some coverage describing it as British. Up to 210 sites a year is a sample, not a census, and the serotine gap shows that sound-based AI still struggles with some species.

It fits a wider pattern on our Tech for Good hub of AI being used to read the natural world at scale, from Ai2’s open satellite embeddings for every 10m pixel on Earth to Mila’s faster wildfire hazard mapping and ICEYE and OroraTech’s combined radar and thermal wildfire monitoring. The test for the bat survey comes in later years, when repeat recordings show whether these populations are growing or shrinking.

Sources

• Bat Conservation Trust: AI reveals 98 million bat calls in Britain’s largest woodland bat survey, 30 October 2026
• Forest Research: AI helps reveal Britain’s bats in largest-ever national woodland bat survey, 30 October 2026
• Forest Research: The NCEA National Woodland Bat Survey (project page)
• Defra Data Services Platform: National Woodland Bat Survey dataset record
• BBC News: AI makes Britain’s largest woodland bat survey possible, scientists say, 30 October 2026

Related stories

• Ai2 Releases Open Satellite Embeddings for Every 10m Pixel on Earth
• Mila and NRCan AI Maps Wildfire Hazard Hundreds of Times Faster
• ICEYE and OroraTech Pair Radar and Thermal Satellites on Wildfires