Who Is Heard? A New Bioacoustic Indicator to Assess Climate Change Impacts on Alpine Bird Communities

Climate change is rapidly altering alpine ecosystems, creating an urgent need for indicators capable of detecting biological responses of animal communities. While impacts of climate change on animal biodiversity are often assessed using few or single-species approaches, responses of entire animal communities remain poorly explored.

Here, we propose and test a new bioacoustic indicator to quantify community-level phenological responses of birds to climatic variability using passive acoustic monitoring. The indicator integrates three complementary components: (i) community composition, (ii) seasonal patterns of occurrence for non-resident species, and (iii) vocal activity phenology, quantified through the timing, duration, and intensity of diel and daily vocal activity. The main aim is to assess if this approach can be used as a tool for long-term monitoring of climate-driven changes in bird communities, including both community composition and phenological responses.

Within the Agile Arvier project (Next Generation EU), passive acoustic monitoring was conducted in the Aosta Valley region (north-western Italian Alps) at three sites between 1800 and 2100 m a.s.l.: an abandoned pasture, a larch and a spruce dominated forests. Three autonomous recorders (Song Meter 4, Wildlife Acoustics), one per site, were deployed to record continuously. To validate the proposed indicator, we used the first year of acoustic data and performed species identification using a deep learning classifier (BirdNET), followed by expert manual validation. Verified detections were used subsequently to quantify the intensity of vocal activity for phenological analyses. Climatic variables (air temperature, solar radiation, wind speed, and precipitation) were included in the analyses. Generalized Additive Models were used to quantify the effects of climatic variables on vocal activity, while Linear Mixed Models were applied to analyse shifts in the daily start and end times of vocalizations at the community level.

Across the three sites, 72 bird species were detected, representing all species expected to occur at the monitored sites. The arrival and departure dates of non-resident species were clearly detected, and vocal activity consistently described the phenology of the analysed periods. Solar radiation and air temperature emerged as the primary drivers of vocal activity, while increased wind speed significantly reduced it. Community-level phenological patterns differed among habitats: birds in the abandoned pasture began vocal activity later and ended earlier than in forested sites, resulting in a narrower daily vocal window. In contrast, larch and spruce forests exhibited highly similar phenological patterns despite differences in elevation, suggesting a potential buffering effect of forest structure.

As vocal activity metrics were tightly linked to environmental variables across fine and multiple temporal scales, the proposed bioacoustic indicator can be considered a robust and scalable approach for monitoring long-term variation in both the composition and phenology of mountain bird communities in response to climate change.