Artificial intelligence helping researchers hear the call of the Black-throated Finch

Bravus Mining and Resources has partnered with Queensland scientists to use cutting-edge artificial intelligence technology to improve remote monitoring of the Black-throated Finch at the Carmichael mine near Clermont in central Queensland.

Researchers from E2M created an automated recogniser to capture the birds’ calls, which will support best practice monitoring of the species and open the door for better surveillance of other rare birds.

In a new scientific article published in the international journal Ecological Informatics, the researchers detailed the use of machine learning to make major advances in the technology used to analyse bioacoustics data.

Bravus Mining and Resources Chief Operating Officer Mick Crowe said this latest research and innovation built on the company’s comprehensive work to protect the local finch population.

“We developed a targeted Management Plan to protect local Black-throated Finches and their habitat as part of the strict environment conditions our Carmichael mine operates under,’’ Mr Crowe said.

“Research undertaken over many years now shows those plans are working and the finches are thriving.

“However, our work is also contributing a new understanding of the species and this exciting development to unlock new monitoring technologies will help to improve the management of finch populations more broadly in Queensland.

“Innovations like this help ensure our Management Plan remains world's best practice and that we continue to mine in a way that is responsible and creates jobs and business opportunities for regional Queenslanders for generations to come.”

E2M Senior ecologist John van Osta said the research was developed in consultation with the Queensland Department of Environment and Science and would support the development of automated recognisers for other rare and difficult-to-survey bird species.

“The publication of this work in a respected scientific journal shows our commitment to scientific rigour and supports research and management of Black-throated Finch throughout their range,’’ he said.

“This research will support best practice monitoring for the species and will provide valuable insights for others working to study and protect this species around Queensland.”

Machine learning is a type of artificial intelligence and computer science where computers use data and algorithms to improve the accuracy of their performance, imitating human learning.

The innovation comes only months after world-first research into the Black-throated Finch found populations of the bird are thriving at Bravus Mining and Resources’ Carmichael mine.

To support the management of the finch population, Bravus uses bioacoustics recordings to help track bird movements and to identify individual bird's home ranges, providing insights into their day-to-day behaviour.

While bird calls can be recorded and later manually analysed, automated recognisers can instantly detect the target bird call and are a more accurate way to detect bird species than even visual surveys.

However, they typically require many examples of bird vocalisations to accurately train.

To produce a more accurate automated recogniser for the Black-throated Finch, researchers used more than 9000 hours of audio recordings of the species in the Carmichael mine conservation area.

Using machine learning methods more common in medical imaging analysis and natural language processing, the program was taught how to target data to improve the finch model and manage ambiguous bird calls.

The result was an automated recogniser with a library of more than 1000 Black-throated Finch calls and a model that can successfully identify the birds as well as human experts in the field.

ENDS

Download PDF