AI powered acoustic tools expanded for monitoring mammals
- Crop Innovations
- Jun 30
- 2 min read
A research team at James Cook University has shown how existing AI sound-recognition tools can be repurposed to monitor mammals at scale, not just birds and insects.
In a study published in Methods in Ecology and Evolution, researchers analysed more than 300,000 hours of wildlife audio collected across eastern Australia. Their goal was to test whether passive acoustic monitoring, already widely used for birds, could also work for vocal mammals such as koalas and other native species.
The innovation is not just the scale, but the reuse of established AI. The team adapted the open-source BirdNET deep learning model, originally built for bird calls, to recognise mammal vocalisations. Instead of building a new system from scratch, they trained and tuned an existing model to scan huge sound archives for target mammal calls. According to lead researcher Sebastian Hoefer, the system performed extremely well, especially for long-term monitoring.
This matters because Australia has one of the highest mammal extinction rates in the world, and many species are hard to track with traditional methods. Camera traps and field surveys are useful but expensive and time-intensive. Acoustic sensors can run continuously and cover large areas with less field effort.
AI-assisted acoustic monitoring can reveal when species are active, how their presence changes over time, and where management should focus. It also frees up time and funding to concentrate on non-vocal or harder-to-detect species.
There are still limits. Not all mammals vocalise clearly, and AI models need good reference data to stay accurate. But this study shows a promising shift. With the right training data, today’s acoustic AI tools can be expanded beyond birds and insects to support broader biodiversity monitoring at landscape scale.
Comments