Vocal correlates of arousal in bottlenose dolphins (Tursiops spp.) in human care

Individual identification in acoustic recordings

Recent advances in bioacoustics combined with acoustic individual identification (AIID) could open frontiers for ecological and evolutionary research because traditional methods of identifying individuals are invasive, expensive, labor-intensive, and potentially biased. Despite overwhelming evidence that most taxa have individual acoustic signatures, the application of AIID remains challenging and uncommon. Furthermore, the methods most commonly used for AIID are not compatible with many potential AIID applications. Deep learning in adjacent disciplines suggests opportunities to advance AIID, but such progress is limited by training data. We suggest that broadscale implementation of AIID is achievable, but researchers should prioritize methods that maximize the potential applications of AIID, and develop case studies with easy taxa at smaller spatiotemporal scales before progressing to more difficult scenarios.

Study of repertoire use reveals unexpected context-dependent vocalizations in bottlenose dolphins (Tursiops truncatus)

Dolphins are known for their complex vocal communication, not least because of their capacity for acoustic plasticity. Paradoxically, we know little about their capacity for flexible vocal use. The difficulty in describing the behaviours performed underwater while vocalizing makes it difficult to analyse the contexts of emissions. Dolphins' main vocal categories are typically considered to be used for scanning the environment (clicks), agonistic encounters (burst pulses) and socio-affiliative interactions (whistles). Dolphins can also combine these categories in mixed vocal emissions, whose use remains unclear. To better understand how vocalizations are used, we simultaneously recorded vocal production and the associated behaviours by conducting underwater observations (N = 479 events) on a group of 7 bottlenose dolphins under human care. Our results showed a non-random association between vocal categories and behavioural contexts. Precisely, clicks were preferentially emitted during affiliative interactions and not during other social/solitary contexts, supporting a possible complementary communicative function. Burst pulses were associated to high arousal contexts (agonistic and social play), pinpointing on their use as an "emotively charged" signal. Whistles were related to solitary swimming and not preferentially produced in any social context. This questions whistles' functions and supports their potential role as a distant contact call. Finally, mixed vocalizations were especially found associated with sexual (bust pulse-whistle-click), solitary play (burst pulse-whistle) and affiliative (click-whistle) behaviours. Depending on the case, their emission seems to confirm, modify or refine the functions of their simple counterparts. These results open up new avenues of research into the contextual use of dolphin acoustic signals.

Antillean manatee calves in captive rehabilitation change vocal behavior in anticipation of feeding

Captive animals typically develop anticipatory behaviors, actions of increased frequency done in anticipation of an event such as feeding. Anticipatory behaviors can be an indicator of an animal's welfare. However, for rehabilitating animals that are expected to be reintroduced into the wild, these behaviors need to be extinguished to ensure successful release. Scheduled activities such as feeding occur daily and vocalizations could potentially be used to identify anticipatory behavior. Here, we tested the hypothesis that manatee calves modify their vocal production rate as a form of anticipatory behavior. Vocalizations of two Antillean manatee (Trichechus manatus manatus) calves were recorded for 10 min before, during, and after feeding sessions at Wildtracks, a manatee rehabilitation center in Belize. The number of calls were counted across recording sessions and three acoustic parameters were measured from calls including duration, frequency modulation, and center frequency. A repeated measures ANOVA comparing the number of calls across sessions indicated manatees produced significantly more calls before feeding sessions than during and after sessions. In addition, manatees increased the duration and lowered the frequency of calls before feeding sessions. This information can give further insight on ways to improve rehabilitation protocols and manage human interactions to increase the overall survival rate of rehabilitated manatees when released back into the wild.

AI and the Doctor Dolittle challenge

Talking to animals is a fundamental human desire. The emergence of powerful AI algorithms, and specifically Large Language Models, has driven many to suggest that we are on the verge of fulfilling this wish. A few large scientific consortia have been formed around this topic and several commercial entities even offer such services. We frame the task of communicating with animals as 'The Doctor Dolittle challenge' and identify three main obstacles on the route to doing so. First, although generative AI models can create novel animal communication samples, it is very difficult to determine their context, and we will forever be biased by our human umwelt when doing so. Second, using AI to extract context in an unsupervised manner must be validated through controlled experiments aiming to measure the animals' response. This is difficult, and moreover, AI algorithms tend to cling on to any available information and are thus prone to finding spurious correlations. And third, animal communication focuses on a restricted set of contexts, such as alarm and courtship, highly limiting our ability to communicate regarding other contexts. Nevertheless, using the tremendous power of novel AI methods to decipher and mimic animal communication is both fascinating and important. We thus define the criteria for passing the Doctor Dolittle challenge and call upon scientists to take on the mission.

Molecular Assessments, Statistical Effectiveness Parameters and Genetic Structure of Captive Populations of Tursiops truncatus Using 15 STRs

Simple Summary

The bottlenose dolphins are one of the most used species in entertainment, assisted therapy, education, and research on welfare. However, their maintenance in captivity requires powerful and sensitive tools for preserving their diversity. The number of genetic markers for this purpose remains controversial, restraining the marine species’ genetic diversity determination. We aimed to select 15 hypervariable molecular markers whose statistical parameters were made in 210 captive dolphins from 18 Mexican centers to support their usefulness. The proposed set of markers allowed us to obtain a genetic fingerprint of each dolphin. Additionally, we identified the structure of the captive population, analyzing the groups according to the capture location. Such characterization is key for maintaining the captive species’ biodiversity rates within conservation and reintroduction programs. However, these 15 genetic markers can also be helpful for small- isolated populations, subspecies and other genera of endangered and vulnerable species.

Abstract

Genetic analysis is a conventional way of identifying and monitoring captive and wildlife species. Knowledge of statistical parameters reinforcing their usefulness and effectiveness as powerful tools for preserving diversity is crucial. Although several studies have reported the diversity of cetaceans such as Tursiops truncatus using microsatellites, its informative degree has been poorly reported. Furthermore, the genetic structure of this cetacean has not been fully studied. In the present study, we selected 15 microsatellites with which 210 dolphins were genetically characterized using capillary electrophoresis. The genetic assertiveness of this set of hypervariable markers identified one individual in the range of 6.927e¹³ to 1.806e¹⁶, demonstrating its substantial capability in kinship relationships. The genetic structure of these 210 dolphins was also determined regarding the putative capture origin; a genetic stratification (k = 2) was found. An additional dolphin group of undetermined origin was also characterized to challenge the proficiency of our chosen markers. The set of markers proposed herein could be a helpful tool to guarantee the maintenance of the genetic diversity rates in conservation programs both in Tursiops truncatus and across other odontocetes, Mysticeti and several genera of endangered and vulnerable species.

Convolutional Neural Networks for the Identification of African Lions from Individual Vocalizations

The classification of vocal individuality for passive acoustic monitoring (PAM) and census of animals is becoming an increasingly popular area of research. Nearly all studies in this field of inquiry have relied on classic audio representations and classifiers, such as Support Vector Machines (SVMs) trained on spectrograms or Mel-Frequency Cepstral Coefficients (MFCCs). In contrast, most current bioacoustic species classification exploits the power of deep learners and more cutting-edge audio representations. A significant reason for avoiding deep learning in vocal identity classification is the tiny sample size in the collections of labeled individual vocalizations. As is well known, deep learners require large datasets to avoid overfitting. One way to handle small datasets with deep learning methods is to use transfer learning. In this work, we evaluate the performance of three pretrained CNNs (VGG16, ResNet50, and AlexNet) on a small, publicly available lion roar dataset containing approximately 150 samples taken from five male lions. Each of these networks is retrained on eight representations of the samples: MFCCs, spectrogram, and Mel spectrogram, along with several new ones, such as VGGish and stockwell, and those based on the recently proposed LM spectrogram. The performance of these networks, both individually and in ensembles, is analyzed and corroborated using the Equal Error Rate and shown to surpass previous classification attempts on this dataset; the best single network achieved over 95% accuracy and the best ensembles over 98% accuracy. The contributions this study makes to the field of individual vocal classification include demonstrating that it is valuable and possible, with caution, to use transfer learning with single pretrained CNNs on the small datasets available for this problem domain. We also make a contribution to bioacoustics generally by offering a comparison of the performance of many state-of-the-art audio representations, including for the first time the LM spectrogram and stockwell representations. All source code for this study is available on GitHub.