Imagine leaving microphones in 310 woodlands and coming back with a sound library so huge that sorting it by hand would take more than 30 years. That is the scale of Britain’s largest woodland bat survey — and AI bat call identification is what made the results usable.

Researchers at University College London (UCL), working with the Bat Conservation Trust, Forest Research, and funding through Defra’s Natural Capital and Ecosystem Assessment programme, helped turn roughly 90,000 hours of nighttime audio into a national picture of how bats use English woodlands. This is an official university and partner release about the survey, not a peer-reviewed journal paper, and the findings are already useful for conservation.

Why bats are so hard to count

Bats are nocturnal, fast, and easy to miss. You rarely see them well enough to identify a species on the wing. What you can hear — if you have the right gear — is their echolocation: the rapid clicks they use to navigate and hunt.

Acoustic sensors catch those clicks. The National Woodland Bat Survey placed three sensors at each of 310 one-hectare woodland sites across England and recorded for nine nights at a stretch between 2021 and 2024. The haul: over 98 million bat calls, about 217 terabytes of audio, and nearly 10,000 recording nights.

Analysing this volume of acoustic data would simply not have been feasible without AI, the teams behind the survey have said — and the BBC reported that doing it manually would take more than 30 years.

That is not a metaphor. It is why wildlife monitoring has always been stuck at a smaller scale than biologists would like.

How AI bat call identification worked

The partners used open-source tools developed with UCL and the University of Edinburgh, including a system known as BatDetect2, to detect bat echolocation calls in the recordings and help classify them.

Speed alone is not enough. A network of 19 bat-call experts from the Bat Conservation Trust’s Verifier Network checked a sample of the AI’s identifications. Those experts had already shown their skills through a standardised assessment. The result is a hybrid workflow: machines dig through the mountain of audio; people spot-check so the maps can be trusted.

Professor Kate Jones, director of the UCL People and Nature Lab, put the shift plainly: AI has changed the scale at which wildlife can be monitored. Open tools let organisations flag species patterns across thousands of nights of recording — while expert review keeps the science honest.

What the microphones found

The survey detected every bat species or species group found in Britain except the serotine, a less common species that is especially hard to pin down from sound alone.

It also caught some of Britain’s rarest bats, including the grey long-eared bat and the barbastelle, both on the UK Red List, plus near-threatened species such as Nathusius’ pipistrelle and Leisler’s bat.

Barbastelles are elusive forest specialists. Hotspots of their activity showed up across southern England (with much of the South East quieter), and the survey recorded them in parts of Cornwall, Devon, Dorset, and Somerset where earlier records were sparse. That fills real gaps on the map — the kind of detail woodland managers and conservation planners actually need.

Common and soprano pipistrelles, Britain’s most abundant bats, turned up almost everywhere the sensors went. Common pipistrelles were detected in all 310 woodlands. Broadleaved woodlands showed the highest average activity, underlining how important those habitats are for foraging, roosting, and moving through the landscape.

For several rare species, this survey is currently the only national-scale monitoring of its kind away from known roosts. That baseline matters as climate and land use change.

Why this is a fun — and serious — AI win

It is easy to joke that AI finally earned its keep by listening to bats all night. The serious point is simpler: AI bat call identification turned an impossible backlog into a living dataset.

• Conservation groups get a clearer view of where rare bats are active.
• Forest researchers can link bat activity to woodland structure, surrounding landscapes, and climate.
• Anyone curious about “AI for good” gets a concrete story: open tools, expert checks, and maps that can guide how woodlands are created and managed.

The data on species occurrence and activity are available to explore online through the National Woodland Bat Survey Data Explorer — another sign the team wants the work used, not locked away.

The takeaway on AI bat call identification

Britain’s biggest woodland bat survey did not happen because someone invented a magical bat detector. It happened because sensors could listen for years, and AI bat call identification — with expert verification — could keep up.

More than 98 million calls. More than 30 years of hand-sorting, avoided. Rare bats showing up in places we barely knew to look. That is a woodland success worth celebrating.

Read the UCL release: AI helps reveal Britain’s bats in largest-ever national woodland bat survey

Also covered by the BBC: Largest woodland bat survey a ‘nice story of AI for good’