Biology subjects

Mathevon, N.

Publications and source records attributed to Mathevon, N..

4 recordsLinked to original sources

bioRxiv2026

Acoustic remote sensing with deep learning enables non-invasive estimation of seabird nest density

Passive Acoustic Monitoring (PAM) has advanced ecological research by enabling non-invasive recordings of wildlife vocalizations that provide insight into species presence, behavior, and reproductive activity. This remote-sensing approach is particularly valuable for species breeding in concealed habitats or remote areas where visual surveys are challenging. The Critically Endangered African Penguin, burrow-nesting seabird, exemplifies this challenge. Its highly vocal breeding behavior makes it an ideal case study for evaluating passive acoustic monitoring as a low-disturbance approach to estimating nest density. To evaluate this, we deployed Autonomous Recording Units at multiple sampling points across the Stony Point Penguin Colony, capturing soundscapes spanning different nest densities and environmental conditions. We then developed an automated detector for Ecstatic Display Songs (EDS), the species characteristic territorial song, using a Convolutional Neural Network trained on a multi-source dataset covering several breeding seasons, diverse acoustic environments, and both in situ and ex situ recordings. The model achieved high recall and precision and remained robust across diverse environmental conditions, supporting the use of heterogeneous training datasets for reliable bioacoustics detection. Using the automated EDS detections, we then investigated how vocal activity peaks relate to local nest density. A Generalized Additive Model revealed that EDS peaks strongly predicted nest density, with a nonlinear increase that plateaued at high calling rates. Importantly, models trained in one breeding season generalized well to the next. In conclusion, by integrating PAM with deep learning, this study provides a scalable, low-disturbance framework for estimating penguin nest density from soundscape data, supporting colony rangers in monitoring penguin colonies. HighlightsO_LILarge-scale acoustic monitoring captured the soundscape of a critically endangered seabird species. C_LIO_LIDeep-learning-based acoustic detection reliably identified key breeding vocalizations in field recordings. C_LIO_LIPeak vocal activity strongly and nonlinearly predicted active nest density across sampling points. C_LIO_LIThe vocal-nest relationship increases rapidly and plateaued at high levels of vocal activity. C_LIO_LIThis approach enables scalable, low-disturbance monitoring of seabird nest density. C_LI

zoology↗

bioRxiv2024

A cooperatively breeding mouse shows flexible use of its vocal repertoire according to social context

Mice exchange information using chemical, visual and acoustic signals. Long ignored, mouse ultrasonic communication is now considered to be an important aspect of their social life, transferring information such as individual identity or stress levels. However, whether and how mice modulate their acoustic communications is largely unknown. Here we show that a wild mouse species with a complex social system controls its vocal production both qualitatively and quantitatively, depending on social context. We found that the African striped mouse Rhabdomys pumilio, a cooperatively breeding species, has a vocal repertoire consisting of seven call types, which it uses differently depending on whether the individuals encounter another mouse that is familiar, unfamiliar, of the same or different sex. Familiar individuals, whether of the same or different sex, vocalize more than two unfamiliar same-sex individuals. The greatest diversity of vocalisations is recorded when a female and a male first encounter, suggesting that certain calls are reserved for courtship. Our results highlight that familiar mice alternate their vocalisations (turn-taking) while unfamiliar individuals tend to overlap one another. These observations suggest that African striped mice control the production and temporal dynamics of their vocalisations, addressing targeted information to specific receivers via the acoustic channel.

animal behavior and cognition↗

bioRxiv2024

Visualization and quantification of coral reef soundscapes using CoralSoundExplorer software

Despite hosting some of the highest concentrations of biodiversity and providing invaluable goods and services in the oceans, coral reefs are under threat from global change and other local human impacts. Changes in living ecosystems often induce changes in their acoustic characteristics, but despite recent efforts in passive acoustic monitoring of coral reefs, rapid measurement and identification of changes in their soundscapes remains a challenge. Here we present the new open-source software CoralSoundExplorer (https://sound-scape-explorer.github.io/docs/CSE/), which is designed to study and monitor coral reef soundscapes. CoralSoundExplorer uses deep learning approaches and is designed to eliminate the need to extract conventional acoustic indices. To demonstrate CoralSoundExplorers functionalities, we use and analyze a set of recordings from three coral reef sites, each with different purposes (undisturbed site, tourist site and boat site) located on the island of Bora-Bora in French Polynesia. We explain the CoralSoundExplorer analysis workflow, from raw sounds to ecological results, detailing and justifying each processing step. We detail the software settings, the graphical representations used for visual exploration of soundscapes and their temporal dynamics, along with the analysis methods and metrics proposed. We demonstrate that CoralSoundExplorer is a powerful tool for identifying disturbances affecting coral reef soundscapes, combining visualizations of the spatio-temporal distribution of sound recordings with new quantification methods to characterize soundscapes at different temporal scales. Author summaryTodays scientists are faced with the challenge of analyzing large amounts of data, such as those generated by passive acoustic monitoring of ecosystems. We have built CoralSoundExplorer (https://sound-scape-explorer.github.io/docs/CSE/), an efficient tool for analyzing large datasets of sound recordings, which transforms coral reef soundscape recordings into visual representations in 2D or 3D spaces. By spreading them across easy-to-explore acoustic spaces, CoralSoundExplorer enables the observer to quickly grasp the characteristics of soundscapes, their differences and similarities, and their organization on different temporal scales. These acoustic spaces and their temporal dynamics can be quantified, for example to account for the speed at which soundscapes change over time. In this study, we take the example of reef soundscapes from the island of Bora-Bora to illustrate the features and possibilities offered by CoralSoundExplorer. CoralSoundExplorer is open source and easy to use, even for non-specialists, thanks to an interface that requires no coding skills. We provide detailed instructions for installing and using CoralSoundExplorer to help users get started easily. CoralSoundExplorer needs to be installed on a computer to perform the analyses and calculations based on sound recordings. There is also an online interface (https://sound-scape-explorer.github.io/coral-sound-explorer/) enabling users to visualize data that have already been processed by CoralSoundExplorer.

ecology↗

bioRxiv2022

Improving the workflow to crack Small, Unbalanced, Noisy, but Genuine (SUNG) datasets in bioacoustics: the case of bonobo calls

Despite the accumulation of data and studies, deciphering animal vocal communication remains highly challenging. While progress has been made with some species for which we now understand the information exchanged through vocal signals, researchers are still left struggling with sparse recordings composing Small, Unbalanced, Noisy, but Genuine (SUNG) datasets. SUNG datasets offer a valuable but distorted vision of communication systems. Adopting the best practices in their analysis is therefore essential to effectively extract the available information and draw reliable conclusions. Here we show that the most recent advances in machine learning applied to a SUNG dataset succeed in unraveling the complex vocal repertoire of the bonobo, and we propose a workflow that can be effective with other animal species. We implement acoustic parameterization in three feature spaces along with three classification algorithms (Support Vector Machine, xgboost, neural networks) and their combination to explore the structure and variability of bonobo calls, as well as the robustness of the individual signature they encode. We underscore how classification performance is affected by the feature set and identify the most informative features. We highlight the need to address data leakage in the evaluation of classification performance to avoid misleading interpretations. Finally, using a Uniform Manifold Approximation and Projection (UMAP), we show that classifiers generate parsimonious data descriptions which help to understand the clustering of the bonobo acoustic space. Our results lead to identifying several practical approaches that are generalizable to any other animal communication system. To improve the reliability and replicability of vocal communication studies with SUNG datasets, we thus recommend: i) comparing several acoustic parameterizations; ii) adopting Support Vector Machines as the baseline classification approach; iii) explicitly evaluating data leakage and possibly implementing a mitigation strategy; iv) visualizing the dataset with UMAPs applied to classifier predictions rather than to raw acoustic features.

animal behavior and cognition↗