AI determines biodiversity based on sounds

According to a study, automated analysis of animal sounds can be used to draw conclusions about the development of biodiversity in the area. The evaluation was tested using artificial intelligence (AI) on tropical reforestation areas.

The sounds of birds, amphibians and mammals were taken into account, as a team led by forestry scientist Jörg Müller from the University of Würzburg reported in the journal “Nature Communications”.

Studies have previously shown that the presence of animal species can be monitored through soundscapes when forests are destroyed or restored. The research team has now installed recording equipment in northern Ecuador in South America on pastures and cocoa plantations that are no longer managed and where forests are gradually being re-established. The recordings of the various animal sounds were then analyzed using AI models.

“The research results show that the good data excellently reflects the return of biodiversity in the abandoned agricultural areas,” says Müller. “It is mainly the sound of the species communities that represents the repopulation very well.” These communities have a very characteristic composition in the forest and differ significantly from those in agricultural areas that are still active.

From the researchers’ perspective, the AI ​​models used could provide a basis for studying biodiversity in other reforestation areas. In a next step, the scientists want to improve and expand their models to register more species. The system will also be used in the Bavarian Forest National Park in the future.

“Our AI models can form the basis for a very universal tool for monitoring biodiversity in reforested areas,” Müller is convinced. In reforestation projects, the system could also help monitor whether a diverse forest actually develops in an area as indicated – or just a species-poor monoculture.

(sda/dpa)