AI analyzes 300,000 hours of mammal calls to improve wildlife monitoring

A study published in Methods in ecology and evolutionused artificial intelligence (AI) to analyze more than 300,000 hours of voice recordings of mammal species from far north Queensland to southern New South Wales and discovered a new approach to mammal monitoring that will have a revolutionary impact on conservation.

“Australia is the worst place in the world for mammal extinction,” said Dr Sebastian Hoefer, lead researcher and postdoctoral researcher at JCU.

“Many of our mammals are endemic, with very restricted ranges and important ecological roles.

“Monitoring their populations is vital for effective conservation, but traditional techniques are difficult to apply to large areas.”

Dr Hoefer explained that their idea was inspired by similar successes in bird monitoring.

“AI processing of bird recordings has worked very well across entire ecoregions, so we wanted to see if the same technique could be applied to vocalizing mammals,” he said.

In a first, the team has fed 36 years’ worth of land mammal records from national parks, conservation areas and reserves across eastern Australia into open source BirdNET AI software.

“You can imagine it’s a bit like ChatGPT,” Dr. Hoefer said.

“It’s a machine learning recognition system that was trained on birds, but we can tell it what we want to find.

“For example, we can enter a single example of a mammal call, say a male koala bellowing, and then ask the system to locate all similar sounds in the recordings.”

Dr Hoefer explained that the strength of this new approach lies in its ability to process massive amounts of data collected over large areas and over vast periods of time, something traditional methods, such as camera traps or field surveys, struggle to achieve.

“It worked extremely well, especially for long-term monitoring,” he said.

“We can now rely on acoustic monitoring and AI for vocal mammal species, allowing us to devote more time and funding to monitoring mammals that do not vocalize.”

In terms of conservation, research expands understanding of the presence and activity patterns of mammals in the wild, which can help guide management decisions.

“It was amazing to discover how effective using mammal calls can be to detect and monitor species on such a large scale,” Dr Hoefer said.

“This approach gives us new information about where, when and how species are active in the landscape. For example, deer have very specific vocalization periods during the rutting season.

“By continuously recording throughout the year, we can identify these periods of peak activity and plan targeted management or monitoring during these periods. »

Moving forward, the researchers plan to develop a species-specific recognition tool for these mammals, trained on the validated calls collected during this study.

“This will make it even easier to detect and monitor mammal species across Australia in the future,” Dr Hoefer said.