To understand how wild animals are doing, scientists try to listen

This article has been reviewed in accordance with the editorial process and policies of Science X. The editors have emphasized the following attributes while ensuring the credibility of the content:

verified facts

peer reviewed publication

reputable news agency

corrected


A computer model aided by artificial intelligence was able to distinguish bird calls from recordings made in the Choco region of Ecuador.

× near


A computer model aided by artificial intelligence was able to distinguish bird calls from recordings made in the Choco region of Ecuador.

Reed duet and high trill duet against low insect drone. Their symphony is the sound of a forest and is monitored by scientists to assess biodiversity.

The recording from the Ecuadorian forest is part of new research looking at how artificial intelligence can track the lives of animals in recovering habitats.

When scientists want to measure reforestation, they can survey large areas of land with instruments such as satellite and lidar.

But determining how quickly and abundantly wildlife is returning to an area presents a more difficult challenge—sometimes requiring an expert to sift through recordings and pick out animal calls.

Jörg Müller, professor and field ornithologist at the Biocenter of the University of Würzburg, wondered if there was a different way.

“I saw the gap that we needed, especially in the tropics, better methods to quantify the enormous diversity … to improve conservation action,” he told AFP.

He turned to bioacoustics, which uses sound to learn more about the lives and habitats of animals.

It’s a long-standing research tool, but recently it’s been combined with computer learning to process large amounts of data more quickly.

Muller and his team recorded audio at locations in the Chocó region of Ecuador, ranging from recently abandoned cacao plantations and pastures, to farmland recovering from use, to old-growth forests.


Sound of an old forest. Credit: Jörg Müller

They first had experts listen to the recordings and pick out birds, mammals and amphibians.

They then performed an acoustic index analysis, which provides a measure of biodiversity based on broad soundscape metrics such as noise volume and frequency.

Finally, they conducted two weeks of recordings by an AI-assisted computer program trained to distinguish 75 bird calls.

More entries needed

The program was able to select the calls it was trained on in a consistent manner, but could it correctly identify the relative biodiversity at each location?

To test this, the team used two baselines: one from the experts who listened to the audio recordings, and a second based on insect samples from each site, which offer a proxy for biodiversity.

While the library of sounds available to train the AI ​​model meant it could only identify a quarter of the bird calls that experts could, it was still able to correctly estimate biodiversity levels in each location, the study said.

“Our results show that soundscape analysis is a powerful tool for monitoring the recovery of faunal communities in a hyperdiverse tropical forest,” said the study, published Tuesday in the journal Nature Communications.


The tool has some limitations, including the relatively few available bird sounds to train the computer model with.

× near


The tool has some limitations, including the relatively few available bird sounds to train the computer model with.

“Soundscape diversity can be quantified in a cost-effective and robust manner across the full gradient from active agriculture to regenerating and old-growth forests,” he adds.

There are still drawbacks, including a lack of animal sounds to train AI models on.

And the approach can only catch species that announce their presence.

“Of course (no) information about plants or silent animals. However, birds and amphibians are very sensitive to ecological integrity, they are a very good proxy,” Müller told AFP.

He believes the tool could become increasingly useful given the current push for “biodiversity credits” – a way to monetize the protection of animals in their natural habitat.

“The ability to directly quantify biodiversity, rather than relying on proxies such as growing trees, encourages and enables external evaluation of conservation actions and promotes transparency,” the study said.

More info:
Jörg Müller, soundscapes and deep learning enable tracking of biodiversity recovery in rainforests, Nature Communications (2023). DOI: 10.1038/s41467-023-41693-w. www.nature.com/articles/s41467-023-41693-w

Log information:
Nature Communications

Leave a Comment

Your email address will not be published. Required fields are marked *