Why land use?
Meadows and pastures cover around a third of Germany’s agricultural land and are among the most species-rich habitats in Central Europe. The number of species living there depends heavily on how the land is managed: how often it is mown, how heavily it is fertilised and how intensively it is grazed. The Biodiversity Exploratorien, one of the largest research platforms of its kind in Europe, are investigating this relationship. Since 2006, researchers from a wide range of disciplines have been working on the same plots of land in three regions of Germany: the Swabian Alb, the Hainich-Dün and the Schorfheide-Chorin. These areas consist of ordinary meadows, pastures and forests, which are managed with varying degrees of intensity by farmers and foresters. This allows researchers to compare, under real-world conditions, the impact of land use on biodiversity.
Our sub-project: Automatically recording insects
Insects pollinate plants, provide food for birds and other animals, and break down dead organic matter. In many places, their populations are declining. To understand why, we need to know when, where and which insects are present. This is precisely what has proved difficult so far: traditional methods such as net catches or traps are labour-intensive, usually kill the insects and provide only snapshots on a few days of the year.
However, a meadow changes within hours. After mowing, flowers, food and cover disappear from one day to the next. What happens to the insects then – how sharply their activity declines and how quickly they return – cannot be captured through individual sampling dates. In this sub-project, we therefore ask:
Can insects be reliably and automatically recorded and identified using cameras and artificial intelligence?
How do insect communities react to mowing, and how long does this effect last?
What role do the season and weather play in insect activity?
How we work
We are developing camera traps that continuously photograph insects in the field without catching them. We have started by optimising the design of the traps, including the camera technology, positioning and coloured surfaces that attract insects. Experts identify the insects in the images, and we use these images to train AI models that can independently recognise, count and classify insects. At present, this works reliably down to the order level – for example, beetles, flies or bugs; we are working on more detailed identifications.
The traps are located across all three Exploratoriums on 27 grassland sites, nine per region. In 2024, one trap was operational per site; in 2025, there were three. In 2025 alone, this resulted in over 17 million images of insects and spiders being collected in 102 days. Such a volume of data would be impossible to achieve using traditional trapping methods, and it allows us to distinguish the effects of mowing, season and weather from one another.
Ecology and computer science are working closely together on this project: our department contributes expertise in entomology and ecological issues, whilst Patrick Mäder’s group at the Technical University of Ilmenau provides image analysis and AI models.
Initial findings
Insect activity plummets following mowing. Our data show that this effect can persist for over eight weeks – significantly longer than previously assumed. Publication
Key facts
Lead: Sebastian T. Meyer in collaboration with Patrick Mäder (Ilmenau University of Technology) · PhD student: Robert Künast (Technical University of Munich) · Funding: DFG Infrastructure Priority Programme ‘Biodiversity Exploratories’ (SPP 1374) · Duration: 2023–2027 · More: Sub-project page