Periodic Reporting for period 1 - SOUNDCHANGE (Listening to change: acoustic monitoring of multi-trophic community dynamics in the face of increasingly variable and extreme environmental conditions)
Berichtszeitraum: 2023-09-01 bis 2025-08-31
Species are often predicted to lag behind the unprecedented pace of ecological changes imposed by current human activities. Such changes result not only in modification of the mean environmental values in an ecosystem but also in increasingly extreme and variable environmental conditions. These extreme and variable conditions operate at fine spatio-temporal scales. Passive Acoustic Monitoring (PAM) is a rapidly growing field of research that allows for an unparalleled temporal resolution and a wide taxonomic diversity of sound producing species in terrestrial and aquatic environments. In this context, I will use PAM to understand fine scale spatio-temporal community dynamics in response to climate change. For that I plan to study elevational gradients along mountain ranges. This will allow to cover a variety of environmental conditions along different elevations from low altitude forests to high altitude meadows. Moreover moutain environments are warming up twice as fast as the rest of the earth so they are under particularly strong constraints. With this research, I aim to understand community dynamics at unprecedented spatio-temporal resolution.
The overarching objective of SOUNDCHANGE was to understand fine scale spatio-temporal dynamics of multi-tropic communities in response to increasingly extreme and variable conditions.
With this project, we planned to develop a robust conceptual and methodological community ecology framework that allows to access the fine scale temporal information required to understand multi-trophic community dynamics in a changing world.
We decided to pursue our research with known species instead. This resulted in the selection of a simple multi-trophic network of 20 insectivorous birds and 10 orthoptera. We managed to work on the dynamics of 9 bird species that we detected with BirdNET for which we have developed models of acoustic activity in function of environmental conditions. We plan to expand this to the 30 species selected when the classification models are fitted and validated.
We have then looked at the co-occurrence between the 9 species of bird studied and found all species shared at least half of their spatial distribution with other species.
Finally we have fitted acoustic SDMs on the 9 species studied combining data from our recordings and data from GBIF but have yet to look at the impact of changes on the distributions.
With this project, we planned to develop a robust conceptual and methodological community ecology framework that allows to access the fine scale temporal information required to understand multi-trophic community dynamics in a changing world.
We decided to pursue our research with known species instead. This resulted in the selection of a simple multi-trophic network of 20 insectivorous birds and 10 orthoptera. We managed to work on the dynamics of 9 bird species that we detected with BirdNET for which we have developed models of acoustic activity in function of environmental conditions. We plan to expand this to the 30 species selected when the classification models are fitted and validated.
We have then looked at the co-occurrence between the 9 species of bird studied and found all species shared at least half of their spatial distribution with other species.
Finally we have fitted acoustic SDMs on the 9 species studied combining data from our recordings and data from GBIF but have yet to look at the impact of changes on the distributions.
Rain and wind are a major source of noise in acoustic recordings, especially in environments fully exposed to the elements such as moutain tops. However, up to now, rain and wind are not processed in a standard way in PAM. The recordings are either included but they may bias the analysis or removed altogether, potentially removing occurences of species of interest. We have developed a rain and wind detector that performs better than current state of the art detectors. This detector will be applied to all our recordings, enabling to monitor rain and wind presence and intensity. In the future, we would like to improve species detection in rain and wind.
We have also applied a new generative language model to our recordings, BioLingual. This state-of-the-art model was trained on bioacoustics recordings. We found that it works really well as a way to characterize mountain soundscapes. Specifically is allows to distinguish night vs. day as well as forest vs. open environments.
We have also applied a new generative language model to our recordings, BioLingual. This state-of-the-art model was trained on bioacoustics recordings. We found that it works really well as a way to characterize mountain soundscapes. Specifically is allows to distinguish night vs. day as well as forest vs. open environments.