Researchers are using AI to help them locate rare birds, using the technology to 'listen' through hundreds of hours of recordings and boosting conservation efforts for the endangered species.
James Cook University postdoctoral fellow, Dr Slade Allen-Ankins, is the lead author of a study where listening devices were deployed in the habitat of the endangered black-throated finch.
"These birds are estimated to have lost 88 % of their former distribution in the past four decades, with two main strongholds remaining – one of which is Queensland's Desert Uplands bioregion where we conducted the study," said Dr Allen-Ankins.
He said acoustic monitoring is rapidly becoming a common survey method but effective analysis of long-duration audio recordings is still challenging.
"An audiorecorder can stand and 'listen' for weeks, while visits to sites are short and may miss rare birds. But there's just too much sound collected to listen to it all, and yet we must find rare sounds – it's like finding a needle in a haystack," said Dr Allen-Ankins.
He said in the study, convolutional neural networks were trained to recognise the birds' call from amongst all of the background noise, and the results were compared to on-ground bird surveys, with the AI call recognisers as successful as on-ground surveys at finding the birds.
"In addition, listening to and checking the top 200 detections per site for all 3 years of recordings took only around 8 hours, representing an almost 250-fold reduction in time compared to listening to the entire audio dataset," said Dr Allen-Ankins.
Professor Lin Schwarzkopf, co-author of the study, said the rapid loss of biodiversity worldwide demands urgent conservation action, but knowledge of the distribution of endangered and rare species is often limited, hampering efforts.
"Acoustic monitoring should be considered as a valuable tool to be used alongside manual surveys to allow effective monitoring and conservation of this and other endangered species."
Link to paper here.
