Damien Chevallier

Monitoring animal behaviour provides critical insights into species ecology and offers essential information for guiding management and conservation efforts. Among the various approaches used to study behaviour, bio-logging—the use of animal-borne data recorders—has emerged as a valuable tool for observing animals in their natural habitats while minimising human disturbance. A major advancement in bio-logging has been the integration of accelerometers, which enable high-resolution analysis of movement and activity in free-ranging animals. By identifying movement and postural patterns captured by accelerometers, it becomes possible to associate specific accelerometric sequences with distinct behaviours. This task—automatically classifying repetitive accelerometric patterns linked to behaviours—can be achieved using machine-learning algorithms. The first ecological study applying machine learning to identify animal behaviours from accelerometer data was published in 2009. Since then, numerous studies have expanded this approach across a wide range of species, employing diverse methodological frameworks that make it challenging for new practitioners to identify best practices. The aim of this review is to provide a comprehensive overview of accelerometer-based behavioural identification in ecology using machine learning. We summarise the range of species and applications investigated, highlight key methodological trends, and offer practical guidance for researchers seeking to apply this approach. Based on 125 studies, we show that current practices largely rely on a general framework combined with species-specific adaptations. This has led to a predominance of species-focused methodological studies and limited generalisation across taxa. More importantly, the high diversity of datasets, combined with highly variable validation procedures—sometimes insufficient to assess model generalisation to novel data—prevents robust comparisons, limiting the identification of broadly applicable approaches. We therefore seek to re-establish a clear methodological framework that enables meaningful cross-study comparisons. Although methodological advances have been relatively modest until recently, 2024 marks a turning point, with a growing number of studies applying deep learning approaches that hold promise for improving model generalisation. While deep learning remains less widely adopted in ecology than in human or livestock behaviour recognition, leveraging advances from these fields and fostering interdisciplinary collaboration will be essential to accelerate progress. In particular, developments in real-time monitoring offer strong potential to enhance conservation efforts, an important next step in bio-logging. Nevertheless, despite increasing automation and generalisation, reliable behavioural classification models will continue to depend on robust ground-truth data and strong expertise in the natural history and biology of the study organisms.

Martinique sponge fauna was largely undocumented until 2017, when the first inventory of Porifera colonizing coral reefs, mangroves and caves around the island was published. We performed an integrative classification of sponges in the foraging area of hawksbill turtle (Eretmochelys imbricata) in Martinique. Sponge specimens were retrieved as direct or indirect diet items consumed by hawksbill turtles after video observations, and the feeding behaviors of these predators were tracked. Morphology was supplemented with molecular identification (DNA barcoding) based on a multi-locus approach using COI, 28S and ITS genetic markers. Seventeen different species were identified, belonging to seven orders: Poecilosclerida, Dictyoceratida, Verongiida, Agelasida, Haplosclerida, Clionaida, and Tetractinellida. Haplosclerida exhibited the greatest diversity and species abundance, followed by Verongiida. The 28S marker provided the highest confidence in species identification. We provided new barcode records for Hyattella cavernosa and Amphimedon caribica. Among the cataloged sponges, only four of them had been previously reported as food items of E. imbricata (Xestospongia muta, Iotrochota birotulata, Spirastrella coccinea and Cinachyrella kuekenthali). The rest represent newly documented items that are potentially preyed upon by this turtle predator. The characterization of sponges as being part of the feeding habitat of hawksbill turtles underpins management and protection plans for this critically endangered species, and the benthic community on which they feed, by providing criteria for generating networks of Marine Protected Areas (MPAs) in the Caribbean regions.

Contrary to expectations based on their higher cell numbers, larger and longer-lived species do not face dramatically increased risk of cancer. This strongly suggests that evolution has fashioned natural cancer resistance mechanisms, yet our knowledge remains limited on what these mechanisms might be. The cancer immunological surveillance hypothesis, proposed by Burnet and Thomas in the 1950s, highlights immunity as a key factor determining species-specific cancer resistance. Here we address the original, evolutionary interpretation of this hypothesis by investigating the relationship between cancer mortality risk and markers of efficient antigen presentation. Our results show that the expansion of the MHC class I gene complex, as well as increased selection for diversity at these genes is associated with sharply decreasing cancer mortality risk across mammals. This suggests that the efficient presentation of diverse peptides in somatic cells is important for cancer suppression across mammals, providing pioneering evidence that supports the cancer immunosurveillance hypothesis across species.

Marine megafauna species are affected by a wide range of anthropogenic threats. To evaluate the risk of such threats, species’ vulnerability to each threat must first be determined. We build on the existing threats classification scheme and ranking system of the International Union for Conservation of Nature (IUCN) Red List of Threatened Species by assessing the vulnerability of 256 marine megafauna species to 23 at‐sea threats. The threats we considered included individual fishing gear types, climate‐change‐related subthreats not previously assessed, and threats associated with coastal impacts and maritime disturbances. Our ratings resulted in 70 species having high vulnerability ( v > 0.778 out of 1) to at least 1 threat, primarily drifting longlines, temperature extremes, or fixed gear. These 3 threats were also considered to have the most severe effects (i.e., steepest population declines). Overall, temperature extremes and plastics and other solid waste were rated as affecting the largest proportion of populations. Penguins, pinnipeds, and polar bears had the highest vulnerability to temperature extremes. Bony fishes had the highest vulnerability to drifting longlines and plastics and other solid waste; pelagic cetaceans to 4 maritime disturbance threats; elasmobranchs to 5 fishing threats; and flying birds to drifting longlines and 2 maritime disturbance threats. Sirenians and turtles had the highest vulnerability to at least one threat from all 4 categories. Despite not necessarily having severe effects for most taxonomic groups, temperature extremes were rated among the top threats for all taxa except bony fishes. The vulnerability scores we provide are an important first step in estimating the risk of threats to marine megafauna. Importantly, they help differentiate scope from severity, which is key to identifying threats that should be prioritized for mitigation.

The critically endangered hawksbill turtle (_Eretmochelys imbricata_) is a keystone species that, through its decline, has a profound effect on marine ecosystems such as coral reefs. This study aimed to characterize the activity rhythm and dietary preferences of hawksbill turtles in Martinique to develop appropriate conservation strategies at both local and global scales. Six individuals (one female and five males) were monitored using on-board cameras. Findings show that hawksbills spend more than half of their time feeding at a depth of 14.43 ± 9.72 m, especially in the morning. Video recordings revealed consumed prey, including plants and animals, and these were identified to the family level. Habitat sampling revealed potential prey from six distinct families. Video recordings (selected prey) and habitat samples (potential prey) were compared to indicate prey selection. Aplysinidae was the most frequently observed family both as consumed and potential prey. Aplysinidae and Irciniidae, both marine sponge families, represent new records in the hawksbill diet. Spongin content and brominated metabolites are proposed among the plausible factors that could define turtle food preferences and strategies. Further combined analyses could help to better define feeding strategies as an integrative and collaborative approach, which is essential for the long-term protection of this critically endangered species.

Habitat degradation induced by human activities can exacerbate the spread of wildlife disease and could hinder the recovery of imperiled species. The endangered green turtle Chelonia mydas is impacted worldwide by fibropapillomatosis (FP), a neoplastic infectious disease likely triggered by the Scutavirus chelonidalpha5 with coastal anthropogenic stressors acting as cofactors in disease development. Here, we studied fibropapillomatosis dynamics and its demographic consequences using an 11-year capture-mark-recapture dataset in Anse du Bourg d’Arlet/Chaudière (ABAC) and Grande Anse d’Arlet (GA), two juvenile green turtle foraging grounds in Martinique, French West Indies. Afflicted turtles had similar mortality and permanent emigration rates to the non-afflicted ones. Fibropapillomatosis was commonly observed in large individuals and disease recovery may take several years. Consequently, permanent emigration before full recovery from the disease is suspected and might affect the developmental migration success. Additionally, the results revealed that the FP had higher prevalence and severity, and progressed two times faster in ABAC than in GA despite the proximity (< 2 km) and the similarity of the two foraging grounds. The reasons for these differences remain unidentified. Locally, further studies should be focused on the determination of the external and internal cofactors related to the observed FP dynamics. Finally, the investigations should be extended at a global regional scale to determine potential deleterious effect of the FP on the adult life-stage. These perspectives improves upon our overall understanding on the interplay between wildlife diseases, hosts and environmental factors.

Green sea turtles Chelonia mydas have the ability to hear and produce sounds under water, with some of them potentially involved in social communication. To investigate the potential biological function of these sounds, we used a combination of acoustic, video and multi-sensor recordings of 23 free-ranging juvenile green turtles and we examined the co-occurrences of sounds with behaviours or external events. Our study revealed that most of the sounds were produced when the sea turtles were resting or swimming. However, four sound types were produced in more specific contexts. Long sequences of rumbles were recorded after sunset and mainly during resting. All these rumbles appear to have been produced by several individuals recorded simultaneously, suggesting that rumbles may be used for social interactions. The frequency modulated sound was highly associated with scratching behaviour. The grunt that was produced occasionally when green turtles were vigilant or approaching a conspecific. The long squeak was produced significantly by a small number of individuals in the presence of humans. The grunt and the long squeak may be the first evidence of an alarm or warning signal for intra-specific communication in green turtles. Our results mark a significant milestone in advancing the understanding of sound production in the behavioural ecology of sea turtles. Further experimental investigations (i.e., playback experiments) are now required to test the hypotheses suggested by our findings. Warning signals could be used to prevent sea turtles of a danger and may contribute to their conservation.

The accelerometer, an onboard sensor, enables remote monitoring of animal posture and movement, allowing researchers to deduce behaviors. Despite the automated analysis capabilities provided by deep learning, data scarcity remains a challenge in ecology. We explored transfer learning to classify behaviors from acceleration data of critically endangered hawksbill sea turtles (Eretmochelys imbricata). Transfer learning reuses a model trained on one task from a large dataset to solve a related task. We applied this method using a model trained on green turtles (Chelonia mydas) and adapted it to identify hawksbill behaviors such as swimming, resting and feeding. We also compared this with a model trained on human activity data. The results showed an 8% and 4% F1-score improvement with transfer learning from green turtle and human datasets, respectively. Transfer learning allows researchers to adapt existing models to their study species, leveraging deep learning and expanding the use of accelerometers for wildlife monitoring.

Accurate scientific information is critical for undertaking appropriate conservation and management practices for imperiled species. One source of concern is that research findings might vary for non-biological reasons, including experimental design and analytical methods. To illustrate, we provide detailed modern analysis of reproductive data for leatherback turtles (Dermochelys coriacea). This species exhibits significant fluctuations in nesting densities across different regions, possibly driven by local rather than global factors. Key factors influencing these changes include hatching success and sex determination, both sensitive to incubation temperatures (e.g., lower temperatures yield more males, higher temperatures yield females). This study updates the understanding of temperature-dependent sex determination (TSD) in this species using Bayesian statistics. Growth rate data from the West Pacific and Northwest Atlantic populations show a similar, monotone increase with temperature, affirming the reliability of the models used. The analysis of TSD patterns indicates that observed differences are more likely due to study methodologies and clutch-specific factors rather than regional differences. These findings challenge previous assumptions, showing that leatherback TSD does not conform to a simple on/off pattern but is influenced by multiple, interacting environmental factors. Population dynamics models must account for these complexities, recognizing that both sex ratios and hatching success are critical to understand the rapid changes observed in some leatherback populations.