"Wind farm developers are repeatedly thwarted by lawsuits. In around 70 percent of cases, the plaintiffs raise nature conservation concerns, around half of which relate to birds or bats. With our AI-supported system, companies can create high-quality, meaningful expert reports on species populations very efficiently. This not only reduces their time and costs, but also increases legal certainty - a major advantage with regard to approval procedures and potential lawsuits," explains project manager Dr. Christoph Scholz from Fraunhofer IEE, who also works for the University of Kassel.

Fraunhofer IEE is working on the "Deep Bird Detect" (DBD) project alongside the University of Kassel, the Christian-Albrechts-Universität zu Kiel, Chemnitz University of Technology, and several partners from the field. The research project, which was launched at the beginning of the year, will have a duration of three years. The German Federal Ministry for the Environment (BMUV) is funding the development of the system with almost two million euros as part of its "AI Lighthouses for the Environment, Climate, Nature and Resources" initiative to tackle ecological challenges with artificial intelligence.

Seamless analysis of audio signals

The automated AI system to be developed by Fraunhofer IEE and its partners, on the other hand, fully analyzes the signals. This provides sufficient data, both quantitatively and qualitatively, to be able to professionally assess the impact of the interference with nature associated with the construction of a wind farm. "The data is clearly traceable, which provides additional legal certainty," emphasizes project manager Scholz.

The uniform recording methodology of the "Deep Bird Detect" system also makes it possible to draw comparisons with other ecosystems. This provides information about long-term developments in these areas. It will even be possible to set up an entire monitoring network with which geographical species-specific changes can be detected automatically and at an early stage.

The researchers also want to design the DBD methodology in such a way that it can be transferred to other species groups such as bats, amphibians, or insects in order to make the inventory of ecosystems even broader.

Real-time processing on the recording devices

The signals are recorded by compact, robust recorders that are powered by solar cells. As they are self-sufficient and require very little maintenance, recording causes virtually no disturbance to the birds and other animals in the areas.

The automated real-time evaluation of the signals is also carried out on these devices. The project team uses deep learning methods such as FewShot Learning, Contrastive Learning and Active Learning. The researchers are further developing these methods so that they meet the requirements of edge computing devices. "From a technical perspective, our project is essentially about transferring existing technologies and processes to a field that has not yet been addressed with AI," says Prof. Dr. Sven Tomforde from Kiel University.

In order to make the use of AI transparent and comprehensible for all stakeholders involved in planning and approval, the scientists also want to develop an easy-to-understand app to present the results.

"In order to achieve the energy transition and climate protection goals, the expansion of wind energy must be significantly accelerated. We want to contribute to this with our research project: we are making the preparation of expert reports much more efficient, while increasing legal certainty - and at the same time advancing nature conservation with better recording," explains Scholz.                                         

Uni Kassel: Prof. Dr. Bernhard Sick

TU Chemnitz: Prof. Dr. Maximilian Eibl

CAU Kiel: Prof. Sven Tomforde