So far, monitoring the biological diversity has been a complex and cost-intensive task requiring highly specialized knowledge. With the help of artificial intelligence (AI), however, changes in biodiversity can also be detected easily and at an early stage using audio recordings. Researchers at the Karlsruhe Institute of Technology (KIT) are developing a method for large-scale assessment of habitat conditions. For their research in the Bio-O-Ton-2 project they rely on the active participation of citizen scientists. In the future, this method is intended to be used by public authorities as an early‑warning system.
To identify habitats worthy of protection and preserve them in the long term, early detection of changes in natural environments is crucial. “A quick and cost-effective method that enables large-scale visibility of changes in Germany would be the perfect monitoring solution,” said Dr. Susanne Benz of KIT’s Institute of Photogrammetry and Remote Sensing (IPF). “Audio recordings made by citizen scientists work great for this. Using smartphones with geotagging functions, everybody can provide data from a large number of habitats today.” The researchers are going to use this approach in their Bio-O-Ton-2 project to support public authorities when monitoring biotopes in the future.
Collecting Biodiversity Data with Smartphones
A round table panel composed of specialists from public authorities, nature conservation organizations, mapping agencies, and research institutions is supervising the project. “We started out by jointly designing a space assessment classification system, which was then used by the project partner ci-tec as the basis for training an AI system,” said Benz. “We found out that sound recordings are the best measuring method.”
One-minute audio recordings capture the typical sounds of a place, such as animal and nature sounds, wind, or even distant traffic noise. They provide data suitable for precise scientific evaluation that can be recorded extensively and enable an initial assessment of the space without the need to enter it.
The classification system is based on the habitat types defined in the European Habitats Directive, which describe natural areas in Europe worthy of protection. Additional audio recordings as well as satellite and weather data contribute to creating a pool of reference data from areas that have already been assessed, combining data obtained by remote sensing with details on the state of each habitat. AI experts from ci-tec are using this data basis to develop methods for consolidating the heterogeneous data and for training AI models to detect relationships between acoustic patterns and the various types of habitats. “Our aim is to develop a procedure that lends itself to practical application in public authorities and facilitates monitoring,” said Benz. The Institute for Social-Ecological Research (ISOE) is conducting a concomitant acceptance study and an evaluation of potential use cases.
Citizen Scientists Equipped with Smartphones Wanted
“To make the project a success, we need the help of the public,” said Benz. “We therefore collaborate with Dawn Chorus, a citizen science project inviting people to record and share nature sounds with their smartphones.“ For recording, basically any place can be chosen. Benz emphasized that areas already charted as valuable spaces in terms of nature conservation are particularly important for training the AI models because they allow easy tracking of changes in biodiversity. “We invite everyone who spends time outdoors to record nature sounds using the Dawn Chorus app. But remember that many of those spaces are protected and therefore particularly valuable for biodiversity. So be sure not to leave the official trails for doing the sound recordings. This is all you need to know up front.”
About Bio-O-Ton-2
The idea for Bio-O-Ton-2 came up in a project on the diversity of species and habitat monitoring funded by the Federal Ministry of Research, Technology and Space (BMFTR). After a one-year preparation phase, Bio-O-Ton-2 has entered the second project phase in January 2025. It is scheduled for three years and is funded with EUR 833,000 by the BMFTR. KIT’s project partners are ci-tec GmbH and the ISOE.
Naturraeume_Tonaufnahmen