COS 47-5 - Vocal activity of bird and bat species changes along gradients of available habitat in intensely used agricultural landscapes of Central Europe

Tuesday, August 8, 2023

9:00 AM – 9:15 AM PDT

Location: B110

Presenting Author(s)

MB

Michael Beckmann

Helmholtz Centre for Environmental Research - UFZ
Leipzig, Germany

Coauthor(s)

JE

Julián Equihua

Helmholtz Centre for Environmental Research - UFZ
Leipzig, Germany
FZ

Felix Zichner

Helmholtz Centre for Environmental Research - UFZ
Leipzig, Germany
JE

Jan Engler

AviCon - Research and Planning, Germany
AP

Anne Paulus

Helmholtz Centre for Environmental Research - UFZ
Leipzig, Germany

Abstract: The presence of semi-natural habitats (e.g. small woody features) in agricultural landscapes is known to positively affect bird and bat species, but so far, very few landscape-scale comparative analyses investigate the contribution of natural habitat density and land-use to the presence, vocal activity and abundance of multiple bat and bird species. With novel monitoring techniques and AI assisted species classification, landscape ecological research has unlocked new directions in the monitoring of biodiversity at larger scales, over longer periods of time and at multiple locations at once.
Using such tools, we conducted two field surveys in the agricultural landscapes in Central Germany using passive acoustic monitoring devices (PAM) to investigate birds and bats. For the first survey we recorded birdsong in 62 sites selected along a gradient of available forest/shrub edge and proportion of grassland vs. cropland within a 500m radius around the small woody features within which PAMs were placed. For the second survey, we investigated 40 sites in agricultural landscapes of contrasting intensity to detect the presence and vocal activity of bat species. We recorded appr. 4500hrs of birds and 1500hrs of bats and used the AI based classifiers BirdNET and BatDetect2 to identify species and vocal activity. To estimate whether vocal activities vary along habitat gradients, we tested whether single sites hold a higher (or lower) vocal activity than would be assumed under random conditions. In addition, we estimated the center of mass (i.e. the weighted mean of vocal activity along both habitat gradients) and compared it against a null model which randomized sites across the site matrix.
For most species we found significant deviations from null expectations in at least 60% of studied sites. Of those, more sites had significantly lower than higher vocal activity, pointing to a high level of absences across the sites. Potential explanations include, competitive exclusion, disturbance, habitat suitability or a higher number of other audible species. Center of Mass analysis for vocal activity across sites revealed that most species were not biased towards presence in one particular set of habitat conditions.
By exploring the usefulness of PAM and AI assisted species detection tools for landscape scale analysis, our study highlights the benefits and drawbacks of these approaches. Further, we show how these new tools open up novel and unique research questions that might shape landscape ecological research for the years to come.