• Charles Sturt University researchers are helping teach AI to identify bird calls
• The research aims to assist with tracking and conservation of Australian woodland bird species
• The research is funded through the Australian Government’s Innovative Biodiversity Monitoring scheme

A new artificial intelligence (AI) tool is transforming woodland bird conservation in Australia with the help of a Charles Sturt University research team.

As Australia’s woodland bird species face alarming declines, one of the biggest challenges to their conservation is simply detecting them.

The Charles Sturt Gulbali Research Institute for Agriculture, Water and Environment is helping train AI to recognise the specific calls of bird species in order to improve tracking and conservation efforts.

Professor of Ecology with the Charles Sturt Gulbali Institute Dave Watson (pictured, inset) is helping lead the research. He said current recording technologies were effective, but sifting through the thousands of hours’ worth of audio was overwhelming.

“Acoustic recorders are able to capture bird songs over extended periods, offer a cost-effective and non-invasive way to monitor wildlife across vast landscapes and seasons and allow for the diverse behaviours of species, such as Golden Whistlers, Kookaburras, Red-tailed black cockatoos and Masked Owls,” Professor Watson said.

“Then, this is where AI comes in, as we annotate thousands of Australian woodland bird calls into a customised ‘BirdNET’ model so it can do the analysis work for us down the track.”

With funding from the Australian Government’s Innovative Biodiversity Monitoring scheme, researchers are using years of archived sound recordings from the Australian Acoustic Observatory to train the system.

The first phase of research is nearing completion, with models built for 50 of the 130 candidate species.

Professor Watson said this will allow BirdNET to identify thousands of bird calls in minutes as opposed to the potential months or years it would take human researchers.

“When you’re in the conservation technology space, it’s amazing to see these big jumps and breakthroughs where the floodgates open and innovation follows,” he said.

“The first breakthrough was machine-based monitoring with digital recordings of imagery and sound, and now with AI we’re able to automate processes that would have taken highly trained people years to accomplish.

“What makes this AI model especially powerful is its ability to learn from site-specific bird songs, because much like human dialects can vary from region to region, birds within the same species often have unique songs depending on their location.”

The research is co-led by the University of New England and collaborators from leading land management groups across Queensland University of Technology, University of Queensland, NRM South and Bush Heritage Australia.

With this cutting-edge tool, scientists are one step closer to ensuring that Australia’s unique woodland bird species are not only reliably detected but also better understood and protected for future generations.

“There’s lots of talk about the ‘Nature Positive’ concept lately, which relies on repeatable and trustworthy ways to estimate biodiversity – what is there, how is it travelling, are things getting better or worse,” Professor Watson said.

“‘Ecoacoustics’ can deliver this information and, with the work being done by our teams,  we’ll soon be able to scale it for national reporting.”