Luke Rendell

Passive acoustic monitoring (PAM) generates large datasets that are costly to analyze manually, creating the need for accurate automated detection methods. While deep learning (DL) has shown promise for PAM‐related tasks, the transferability of trained models to unseen datasets remains a critical challenge. To address this, we evaluated four methodological approaches for maximizing the transferability of a DL sperm whale ( Physeter macrocephalus ) click detector: (1) cross‐environment evaluation (training on five datasets, testing on an unseen sixth), (2) limited target‐domain data training (model retraining using only a small, site‐specific dataset), (3) pretraining with random fine‐tuning , and (4) pretraining with active fine‐tuning (selecting fine‐tuning samples based on model uncertainty). While trained models saw a drop in performance when tested on a new dataset, those pretrained on large diverse datasets outperformed limited target‐specific training even when deployment conditions were not represented during training, confirming training diversity improves transferability. Fine‐tuning with 500 target recordings effectively mitigated performance drops when deploying a trained model to new datasets. Active fine‐tuning consistently outperformed all other approaches, although the improvement over random fine‐tuning was marginal. For new studies, we recommend using models pretrained on diverse datasets and fine‐tuning them with minimal target data.

Individual feeding success can change with the size of a social group through competition, cooperation, recruitment and other processes. Female sperm whales (Physeter macrocephalus) live in long-term social units that themselves can form larger, temporary groups, but the individual-level impacts of group formation are unknown. We measured feeding success on 88 days following groups off the Galápagos Islands and Chile using defecation rates. Feeding success increased about fourfold as group sizes rose from one social unit (about 10 animals) to 4–5 units. Our data did not suggest recruitment into successful groups, and the whales’ prey are generally relatively small and slow, arguing against cooperative capture. Instead, it seems most plausible that individuals receive valuable information about patchy prey distributions from foraging group members over scales of 100 to 1000 m. Feeding success fell for the largest groups, suggesting a foraging optimum of about 40–50 animals off the Galápagos Islands and 30–40 animals off Chile. These are close to the mean typical group sizes in these areas, which indicates that efficient foraging on patchy prey may largely drive group size. Considerable differences between oceans in the group sizes of female sperm whales are probably linked to variation in the patchiness of resource availability.

Bioacoustic data from passive acoustic monitoring (PAM) generates large datasets where obtaining detailed auditing and labelling is often impractical, resulting in weak annotations (e.g., presence/absence of species over several minutes of recording). In order to effectively capture the complex temporal patterns and key features of long audio segments, we propose a framework comprising dataset standardisation, feature extraction, and classification via temporal convolutional networks (TCN). This approach eliminates the necessity for setting heuristic decision rules or creating time-consuming strong labels. To demonstrate the effectiveness of our approach, we use sperm whale (Physeter macrocephalus) click trains in 4-min recordings as a case study, from a dataset comprising diverse sources and deployment conditions to maximise generalisability. Our TCN classifiers achieve recall rates exceeding 0.83 at a 0.13 false positive rate, comparable to agreement rates between expert annotators. We compare two methods of feature extraction, variational autoencoders (VAEs) and traditional handpicking of features, and found them to yield similar performance results, with the VAE-based classifiers seeing a more stable performance across datasets and recording conditions. These results offer a way forward in leveraging numerous existing annotated bioacoustic datasets to train automatic classification models, effectively overcoming previous limitations associated with weak labels.

Since first being scientifically recognised almost 150 years ago, bowhead whale (Balaena mysticetus) song has been documented as a remarkably diverse form of communication. While songs in other taxa play fundamental roles in sexual selection and social organisation, the function of bowhead whale song remains poorly understood, in part due to a lack of studies that quantifiably explore variation in song and its fine-scale structure. Here, variation in the songs of Bering–Chukchi–Beaufort bowhead whales was quantified across three years (2011, 2014, 2020) using unit-level analyses and similarity metrics (Levenshtein distance and Dice's similarity). In each year, songs formed distinct song types, some of which remained stable for several weeks, while others changed rapidly within minutes. This variation was uneven across song components, revealing that songs contain a combination of persistent and variable features. By showing that stability and plasticity co-occur in an acoustic display, these findings indicate that multiple, possibly interacting selective pressures, including ecological constraints, social organisation, and sexual selection, may shape song structure. Overall, this study introduces a reproducible framework for song analysis, which provides a foundation for the cross-year and cross-population comparisons essential in understanding the role and evolution of bowhead whale songs.

Population isolation and bottlenecks can affect cultural evolution. This has been shown in bird vocal dialects but is less studied in mammals. The isolated and weakly structured population of sperm whales (Physeter macrocephalus) in the Mediterranean Sea provide an opportunity to address this knowledge gap. We studied dialect variation in this population using the species’ primary social vocalization, stereotyped patterns of clicks called codas. We analysed 5291 codas recorded over 112 days between 2003 and 2021 from both the western and eastern deep-water basins of the Mediterranean Sea. We confirmed previous studies showing the dominance of codas with a 3+1 pattern but also identified two distinct dialect groups. The 2+1 and slow 3+1 coda types identified the western dialect while the 8I and fast 3++1 coda types identified the eastern dialect. Notably, groups recorded in the east occasionally produced the ‘western’ slow 3+1 coda types but the converse was not true, suggesting a directionality in dialect formation. The similarities between the apparent effects of isolation on sperm whale dialects and isolation effects reported in studies of the cultural evolution of both bird song dialects and human languages suggest convergences in vocal dialect evolution across social taxa.

Preserving biodiversity requires the genetic monitoring of wild populations, but traditional invasive sampling techniques can impact animal welfare. For cetaceans, a promising non-invasive approach is the collection of exhaled respiratory vapour, or ‘blow’, to generate individual genetic profiles. Here, we demonstrate the feasibility of generating whole genomes from the blow of humpback whales with data of sufficient quality for population genomics. A total of 58 blow samples from 26 Northeast Pacific humpback whales ( Megaptera novaeangliae ) were collected in Gitga’at First Nation territory using a commercially available drone. The high endogenous content at 84% on average allowed us to generate low-coverage nuclear genomes (mean 2.3×, range 0.2×-3.5×) and high coverage mitochondrial genomes (mean 94×, range 7.7×-151.3×) for inference of population structure, diversity and phylogenomics. The reliability of the blow-derived genomes was demonstrated by direct comparison between replicate blow samples, as well as paired tissue samples from a subset of individuals.

Animal culture, information and behaviours acquired and shared by social learning are a form of biodiversity with intrinsic and practical value. Cooperative foraging, a mutualistic resource acquisition behaviour observed across diverse taxa, is strongly connected to social networks via behavioural states, cues and often social learning, as it typically involves high interaction rates. Understanding the distribution, diffusion and learning mechanisms of such cooperative behaviours is an important but understudied aspect of nonhuman culture. Bubble net feeding (‘bubble netting’) is a specialized foraging technique practised by certain humpback whale (Megaptera novaeangliae) populations globally. Over 20 years in the northern Canadian Pacific, we observed the diffusion of two forms: cooperative group and independent (or ‘solo’) bubble netting. Network-based diffusion analysis—a tool to test for social learning—finds strong evidence for social learning of bubble netting when the overall social network is used, even after accounting for traits such as site fidelity and sex (10.6 × 103 to 35.4 × 103 times more support for social versus asocial learning; p < 0.0001). A homophily check using pre-acquisition association data returned ambiguous results, likely due to the inherent sociality of this cooperative foraging behaviour. Nonetheless, the rapid diffusion of bubble netting is clearly important for population viability and should inform conservation planning for this threatened population.

Culture, defined as information or behaviours shared within a population and acquired from conspecifics through social learning, acts as a second inheritance system which has important implications for species' ecology and evolution. Understanding the influence of social learning and culture in animals' lives is essential to planning and predicting outcomes of conservation actions. Culture plays a key role in cetaceans’ lives, and some of the best evidence for social learning has come from baleen whales (Mysticetes). The Convention on the Conservation of Migratory Species of Wild Animals (CMS), an international treaty operating under the United Nations Environment Programme (UNEP), has recently integrated animal culture into their conservation initiatives. An essential first step in such a conservation framework is to identify key indicators of social learning in well-studied species, which could be applied to other more elusive species. Here, as part of the UNEP CMS Expert Group on Animal Culture and Social Complexity, we first evaluate key evidence for social learning and culture in baleen whales through several case studies. We then suggest key indicators by behavioural context to assist in identifying potential cases of social learning in more elusive species generating a practical guide for future conservation assessment and management. This article is part of the theme issue ‘Animal culture: conservation in a changing world’.