Ecology and computer science unite to protect the environment

They grew up thousands of kilometres apart with a common love for nature. Now, the ecologist and the computer scientist are working together in Australia to monitor and save our environment.

Professor Paul Roe, Head of QUT’s School of Computer Science and the Lead Researcher at the Australian Acoustics Observatory (A20) and Dr Daniella Teixeira, ecologist, and Research Fellow at QUT, are currently engaged on a partnership with Google Australia to develop an AI model to automatically identify and monitor bird species.

The project is focussed on the threatened Glossy Black Cockatoo but their work will extend in future to other wildlife, including koalas, as well as invasive species like cane toads, Asian house geckos and Indian Myna birds.

Professor Roe completed his Master of Engineering at the University of York in the UK, then obtained his PhD in Philosophy at the University of Glasgow before arriving at QUT in 1993.

“Coming to Australia was a dream with its awe-inspiring biodiversity. The more I have worked with ecologists the more I have realised how amazing and unique the environment is here, but also how little we understand it,” Professor Roe said.

“We need to understand the environment so we can protect it. Bringing ecology and computer science is the key.”

“I am interested in making ecology a big data science. I want to scale environmental observation by working on the whole problem and not just one aspect or another. For terrestrial fauna, acoustics is really the only way for scale monitoring; the only way we can ‘listen’ to the Australian bush.”

Dr Teixeira, who recently featured in the David Attenborough film for Netflix – Breaking Boundaries: The Science of Our Planet (2021), spent her childhood in South Africa spending much of her time exploring national parks and diving in oceans.

While completing her PhD in 2016, she used bioacoustics to monitor black-cockatoos, in particular the Kangaroo Island glossy black-cockatoo and the south-eastern red-tailed black-cockatoo.

“The bushfires of 2019-2020 impacted a lot of the habitat of the glossies in Queensland, NSW, Victoria and Kangaroo Island,” Dr Teixeira said.

“When I finished my PhD, I turned my attention to Glossies in Queensland because they don’t get as much attention here.  In 2021, I joined QUT and also applied for bushfire recovery threatened species funding in collaboration with Griffith University and the University of the Sunshine Coast.

“We used this to put sound recorders in locations across south-east Queensland to monitor the glossies. I was already working with Paul and his team and then began a collaboration with computer scientist Dr Lance De Vine in the School of Computer Science who built the AI for the project with Griffith and USC.”

As part of Google Australia’s Digital Future Initiative, the project has so far resulted in the development of A20 Search, a new AI-powered search engine for matching animal recordings across a vast database, allowing researchers to draw insights about the health of Australia’s treasured species and ecosystems.

A20 Search will enable nonprofits, universities, and governments to easily search millions of hours of audio from the Australian Acoustics Observatory and will be open sourced to the broader research community to help influence decisions about land and wildlife management.

Researchers will be able to upload audio recordings of a species to find similar sounds across the database, filter by location and date, and download results for other systems.

Professor Roe describes A20 Search as a huge leap in environmental monitoring.

“In A20 Search we have built a search tool to liberate the data collected in the field. Instead of trying to manually sift through what amounts to hundreds of years of data that we could not live long enough to go through, AI does it for us,” Professor Roe said.

“The reduced costs of the technology is transforming environmental monitoring which is consequently transforming environmental conservation. Partnering with Google Australia allows us to share this technology with other scientists and researchers, which in turn will help land managers make informed decisions about conservation, management, and biodiversity protection.”

By seeing how AI was being used in other areas like astronomy, Professor Roe said he knew big data could transform other sciences.

“We already had some sensors in place to survey the Lewin’s Rail near the airport and I realised AI could revolutionise how we monitor fauna, especially considering how much of Australia is inaccessible,” he said.

“Sound recorders can be placed all over the country to make recordings with no humans around which is about as authentic as you can get.

“Through the Australian Acoustic Observatory, we can then store, manage, and analyse those sounds. After it was set up, the ABC contacted me for its Hoot Detectives series to look for owls as a citizen scientist project. The team from Google saw this and got in touch with me and it has grown from there.”

Since 2019, the A20 project has captured more than 17 million hours of raw audio, collected through a national network of recorders.

Dr Teixeira said the model automatically separates, enhances, and completely isolates the sounds of the Glossy Black Cockatoos, filtering out surrounding noise such as wind and insects. It even picks up very faint calls of the bird that would often be easily missed by more traditional methods of identification.

She is already working on similar projects for other species with Dr De Vine.

“Australia’s natural ecosystems are under threat due to the rise of invasive species, climate change, expanding human development, deforestation, bush fires, and many other factors,” Dr Teixeira said.

“As a nation, we have a responsibility to understand how our species and environments are impacted, and what else we can do to protect them. Birds are often considered indicators of ecosystem condition because many species respond to changes in the environment.

“A20 Search is just the beginning. The ultimate aim is to use it worldwide but realistically, if we can use AI to identify birds and other wildlife in Australia that’s a good start.