Session

Grassland birds are particularly sensitive to land-use change and intensification, which has led to declines in many species. Passive acoustic monitoring and automated identification are increasingly used to assess avian diversity. Open grasslands are particularly suitable for this method due to the lack of acoustic barriers and low background noise for the microphones. In spring 2024 we carried out a passive acoustic bird monitoring within the framework of the Biodiversity Monitoring South Tyrol, a long-term monitoring in a mountainous region. We selected 28 grassland sites that were divided into meadows and pastures, colline-montane and subalpine-alpine zones (based on the elevation) and their management intensity. At each site we placed a recording device (AudioMoth v.1.2.0) to record bird songs throughout the whole breeding season. To analyze the collected data, we used the bird song recognition software BirdNet Analyzer 2.4 and Raven Pro 1.6.5 for the analysis and batch processing. We focused on species richness and analyzed the grassland soundscape by using various acoustic indices. BirdNet identified most bird species correctly, but human verification is still needed. In particular, identifications with a low confidence score, a value given by BirdNet that describes the probability of correct detection, are prone to error. By building species-specific confidence thresholds, the performance of BirdNet can be significantly improved. A diverse soundscape is highly correlated with a diverse landscape and a richer bird community.