This study systematically integrates acoustic methods and machine learning (ML) into marine ecosystem management, developing a comprehensive ML framework that combines passive acoustic monitoring (PAM) data with ecological survey observations to predict key coral reef ecological indicators, including fish abundance, fish species richness, and live coral cover. The framework extracts features from multiple acoustic frequency bands and deploys a complete ML workflow covering seven algorithms across three categories: tree-based models (Random Forest, LightGBM, Gradient Boosting), neural networks (Multilayer Perceptron, Recurrent Neural Networks, Bayesian Neural Networks), and an ensemble strategy (Voting Regressor). Evaluated on ten coral reef sites in Sanya, China, the framework was comprehensively compared in terms of predictive accuracy and computational efficiency. Results indicate that the LightGBM model achieves the highest predictive performance of these biological indicators, providing a more efficient, scalable, and non-invasive solution for marine fish monitoring, which can support decision-making in marine ecosystem management. The proposed machine learning-based framework has the potential to be integrated into decision-support tools for ecosystem management, enabling more efficient monitoring of coral reef ecosystems worldwide.

The retreat of Arctic sea ice is driving an increase in vessel traffic and associated underwater noise, which interferes with the frequency bands used by Arctic marine mammals. Detecting co-occurring vessel noise and marine mammal vocalizations in passive acoustic monitoring (PAM) data can help to assess their adverse impacts and guide mitigation strategies. This paper proposes two ship noise detection techniques: a modified variant of the Frequency Amplitude Variation (FAV) method, MFAV, which integrates signal processing with a simple statistical threshold to enhance both interpretability and detection performance; and a convolutional neural network (CNN) model specifically trained to advance ship detection in the Canadian Arctic. Comparative analysis of our PAM test dataset from the western Canadian Arctic, based on peak F1-scores, demonstrates that the CNN model generalizes well to unseen sites and, with one exception, consistently outperforms both MFAV and FAV by 1%-8%, maintaining scores above 91%. Furthermore, MFAV improves the detection of boats by up to 22% and of larger ships by 6%. The developed methods are publicly available as an open-source tool on GitHub, contributing to the advancement of acoustic vessel monitoring techniques in Canadian Arctic waters in support of conservation efforts aimed at protecting Arctic marine mammal habitats.

Detailed ecological information on sperm whales (Physeter macrocephalus) in the South China Sea (SCS) remains scarce despite its potential as a key nursing ground. Based on passive acoustic monitoring data obtained during vessel-based surveys (2019-2025), this study provides the characterization of the population's coda patterns, which reflect cultural social structure. Using a machine learning-based algorithm to analyze inter-pulse intervals from 2163 regular clicks, estimated total lengths ranged from 7.74 to 12.48 m, with 90% of individuals under 10 m. This result confirms the presence of females and immature individuals in the region, forming social units typical of nursing areas. Unsupervised clustering of 641 codas identified 20-21 discrete types, primarily consisting of 4-7 clicks. The predominance of two "4 + 1 ++ 1" patterns identifies the SCS population as part of the widespread Pacific "four-plus" vocal clan, which is found in the Eastern Pacific (e.g., Galápagos, Peru, and Chile) as well as the Western Pacific (e.g., Tonga and Papua New Guinea). These findings provide critical baseline acoustic data for sperm whales in the SCS and highlight the potential importance of long-term monitoring of this region for the species conservation.

Forests are at the frontline of both climate and biodiversity crises: forests are fundamental to the regulation of Earth's atmospheric carbon dioxide levels, and tropical forests alone harbour over half of all terrestrial species. Consequently, forests are at the centre of global efforts to both prevent extinctions of species and sequester carbon to mitigate climate change. Most carbon sequestration initiatives do not measure how effective they are in terms of biodiversity protection, largely because it is expensive, complex and often not required. Compounding this problem is the lack of reliable biodiversity baselines, against which positive changes due to conservation and negative changes due to human pressure could be measured. The Soundscape Baselines Project is a global initiative to collect and analyse bioacoustic baselines in some of the world's remaining intact forests. The baseline data are collected using a modular system of passive acoustic monitoring and ancillary data that can be expanded and integrated with other technologies. It is managed by local teams of scientists, conservation practitioners and community members. Emphasis is on the usability of data and equipment beyond the project, community engagement and building an interconnected network of scientists across forested countries, particularly historically underexplored partnerships between tropical forest countries. In this paper, we describe the initiative and share initial results to demonstrate how the project could be used for biodiversity conservation purposes in the future and outline steps towards operationalization. We also draw a list of priority areas for additional baseline locations across remaining intact forests. We discuss how bioacoustic baselines established now and in the future will become invaluable in a rapidly changing climate and the resulting unprecedented change in Earth's forests. Establishing rigorous biodiversity baselines using bioacoustics and other technologies and approaches through collaborative science is fundamental to ground biodiversity conservation in evidence, equity and transparency.

Estimating the traffic rates of bats migrating across the North and Baltic Seas to develop efficient mitigation measures at offshore wind energy facilities

Measurement error in multinomial data is a well-known and well-studied inferential problem that is encountered in many fields, including engineering, biomedical and omics research, ecology, finance, official statistics, and social sciences. Methods developed to accommodate measurement error in multinomial data are typically equipped to handle false negatives or false positives, but not both. We provide a unified framework for accommodating both forms of measurement error using a Bayesian hierarchical approach. We demonstrate the proposed method’s performance on simulated data and apply it to acoustic bat monitoring and official crime data.

Sperm whale acoustic ecology around two sub-Antarctic islands.

Passive acoustic monitoring (PAM) is a powerful tool for studying marine biodiversity, but large-scale analysis of underwater recordings is constrained by noise, overlapping signals, and limited labeled data. Here, we present a scalable, unsupervised contrastive learning framework for marine soundscapes. Using a large PAM dataset spanning multiple biogeographies, we show that the proposed approach organizes recordings into clusters with well-defined internal structure, as assessed using intrinsic clustering metrics and within-cluster similarity. The resulting clusters reveal recurring acoustic patterns that correspond to broad sound-source categories, including biological sounds such as fish calls and choruses, and anthropogenic sounds such as vessel noise, without explicitly enforcing these distinctions during training. Compared with established approaches, including cepstral features, variational autoencoders, and supervised pipelines, the proposed framework produces embeddings that support more compact and stable unsupervised clustering while preserving fine-scale acoustic variation beyond predefined species labels. By learning a shared representation across recordings from multiple sites and years, we examine the reproducibility of acoustic patterns across locations and identify both site-shared and site-specific sound signatures. Although the method is not designed to recover coarse species labels, it enables label-efficient analysis by reducing reliance on manual annotation and supporting exploratory characterization of complex marine soundscapes. Together, these results highlight multi-positive contrastive learning with a teacher network and acoustically informed augmentations as an effective strategy for scalable, discovery-driven analysis of passive acoustic monitoring data.

In high-latitude ecosystems, insectivorous bats must make use of short nights during the summer, during which temperature and light conditions interact in non-linear ways. These abiotic dynamics directly influence insect availability and, in turn, bat activity. We examined the influence of night-time temperature and night length on the nightly and seasonal activity patterns of the northern bat (Eptesicus nilssonii) across Finland (60°–66°N), using passive acoustic monitoring data collected over 9,000 detector-nights across seven years. We predicted that bat activity would be constrained by low temperatures, strong winds or heavy rain, and that the nightly pattern would depend on an interaction between night length and weather conditions. Our results show activity peaked 1–2 h after sunset under most conditions, with ambient temperature strongly influencing overall activity levels. On short, warm nights, activity was relatively evenly distributed across the few hours of darkness, whereas long, warm nights produced a bimodal activity pattern with peaks after sunset and before sunrise. This dual-peaked pattern occurred only under thermal conditions that likely support sustained insect activity. In contrast, on colder or windier nights, activity was sharply concentrated to the first hours after sunset. Seasonal and latitudinal trends revealed that activity was most restricted in spring, particularly in northern regions, while the progressing summer expressed more evenly distributed patterns. In autumn, activity patterns diverged across latitudes, reflecting interactions between temperature and night length. The results of our study demonstrate that E. nilssonii may dynamically adjust its foraging behaviour in response to interacting abiotic constraints, optimizing energy gain while minimizing predation risk. Our findings quantify the environmental thresholds facilitating various activity patterns and provide rare evidence of bimodal nocturnal activity at high latitudes. Our study highlights the importance of large-scale, long-term monitoring for understanding how changing climatic conditions influence species behaviour in boreal ecosystems.

Abstract Dryocopus pileatus (Pileated Woodpecker) landscape use is widespread in the Pacific Northwest, especially in forested locations with structurally complex canopies, but declines at higher latitudes, higher elevations, and in areas of greater topographic complexity. High-use sites featured high conifer canopy cover and high Abies grandis (grand fir) basal area. These findings emerged from our analyses of ∼1.7 million hours of acoustic recordings collected from 3,948 sites distributed across 10 million hectares of forest lands in the Pacific Northwest, USA. Data were processed using a deep learning model to detect species-specific vocalizations. Landscape use rates were high overall, with detections at 83% of surveyed locations. Detection probability decreased significantly over the field season and was negatively influenced by environmental noise and precipitation, emphasizing the need to account for these factors when interpreting acoustic monitoring data. The deep learning model detected D. pileatus vocalizations with high precision (>97% true positives among apparent detections). We corrected encounter histories based on precision, thereby improving the accuracy and reliability of ecological inferences. Passive acoustic monitoring, combined with machine learning and remotely sensed landscape variables, effectively uncovered nuanced patterns of occupancy and landscape use by D. pileatus at broad spatial scales. By elucidating patterns of landscape use, occupancy, and detectability, our results can help to guide conservation efforts and future research on this species. Given the role that D. pileatus plays as an ecosystem engineer, its distribution and habitat associations bear strongly on forest ecosystem management and biodiversity conservation in the context of changing forest conditions. This broad-scale assessment highlights the ecological importance of D. pileatus in shaping forest communities, influencing habitat availability for cavity-dependent species, and maintaining biodiversity across forested landscapes.

ABSTRACTLatitudinal gradient can influence ecosystem dynamics and species distribution, yet the influence on some aspects, such as intra‐species diversity, is less well understood. The large‐scale distribution of harbour porpoises (Phocoena phocoena) in the Northeast Atlantic, indicates that ecological adaptations within the species might be greater than currently recognised. This study investigates variation in foraging behaviour using long‐term passive acoustic monitoring data collected between 2009 and 2023 from Iceland, Sweden, Ireland, and France. In each area, Generalised Additive Models (GAMs) were used to investigate the influence of large‐scale environmental variables (water temperature, salinity, primary production, diel phase and daylength) on foraging behaviour. Our results show variability and complexity in foraging and differences in temporal patterns between areas. In the northernmost regions, with the largest variation in daylight, foraging behaviour was related to diel phase, with primarily nocturnal foraging recorded during the year, but with predominantly diurnal foraging in Iceland during late fall. In the southernmost regions, less effect of diel phase on foraging was found. Harbour porpoises in Sweden and Iceland exhibited increased day‐round foraging during calving periods, highlighting the role of reproductive energetics in behavioural adaptations, and the complexity and importance of foraging during different seasons. Our findings underscore the influence of environmental drivers in shaping foraging strategies, supporting the concept that harbour porpoises optimise these based on local conditions and prey availability. Using long‐term datasets, spanning broad geographical and temporal scales, this study contributes to the wider ecological understanding of animals with extensive latitudinal distributions, highlighting intra‐species variance and the need for region‐specific conservation and management.

Population-specific acoustic features are vital for using passive acoustic monitoring to study marine mammal populations in remote regions. Southern Hemisphere fin whale (Balaenoptera physalus) songs include region-specific high-frequency components, with the 86- and 99-Hz high-frequency components present in the Atlantic Sector of the Southern Ocean. Using long-term passive acoustic monitoring data, we show that, despite gradual interannual and intra-annual variabilities, these features remain distinct and recognizable across regions and years. Our findings support their use as reliable acoustic markers for monitoring fin whale populations, providing valuable insights into distribution and population structure.

ABSTRACTUnderstanding of the spatio‐temporal occurrence of cetaceans post the whaling era is essential for protecting and improving management strategies of these marine mammals. To determine the monthly and diel acoustic occurrence of four baleen whale species relative to environmental conditions off the west coast of South Africa, we collected passive acoustic monitoring data within Child's Bank marine protected area in January and May through October 2024 at various water depths. Burst tonal calls of the southern African Bryde's whale offshore population were detected in January and May through July with the highest occurrence in January. Humpback whale songs and southern right whale gunshot sounds were detected from May through October with high occurrence in September and with smaller modes in other months. Antarctic minke whale bioduck calls were also found in June through October, showing high occurrence in August through October. Calls from an unknown source with similar characteristics to Antarctic minke whale bioduck calls were present in May, July, and August with the highest occurrence in August. Diel acoustic occurrence of Bryde's, southern right, Antarctic minke, and minke‐like whale calls indicated that these animals vocalised more during the day while humpback whales were more vocally active at night. Sea surface height and sea surface temperature, either separately or in combination, were the most important predictors of whale acoustic occurrence, highlighting the influence of environmental conditions on the distribution, habitat selection, and ecology of these whales. Overall, this study advances our understanding of the movement, occurrence, and behavioural patterns of several baleen whales relative to environmental conditions. It also provides the first description of the southern African Bryde's whale offshore population's call characteristics, which will be useful at guiding future studies to acoustically differentiate between it and the inshore population.

Computational study on non-contact acoustic wave monitoring data processing method based on BIM technology and its calculation in the reinforcement design of human defense project

Growing concern over anthropogenic underwater noise, highlighted by initiatives like the Marine Strategy Framework Directive (MSFD) and its Technical Group on Underwater Noise (TG Noise), emphasizes regions like the Western Black Sea, where increasing activities threaten marine habitats. This region is experiencing rapid growth in maritime traffic and resource exploitation, which is intensifying concerns over the noise impacts on its unique marine habitats. While machine learning offers promising solutions, a research gap persists in comprehensively evaluating diverse ML models within an integrated framework for complex underwater acoustic data, particularly concerning real-world data limitations like class imbalance. This paper addresses this by presenting a multi-faceted framework using passive acoustic monitoring (PAM) data from fixed locations (50–100 m depth). Acoustic data are processed using advanced signal processing (broadband Sound Pressure Level (SPL), Power Spectral Density (PSD)) for feature extraction (Mel-spectrograms for deep learning; PSD statistical moments for classical/unsupervised ML). The framework evaluates Convolutional Neural Networks (CNNs), Random Forest, and Support Vector Machines (SVMs) for noise event classification, alongside Gaussian Mixture Models (GMMs) for anomaly detection. Our results demonstrate that the CNN achieved the highest classification accuracy of 0.9359, significantly outperforming Random Forest (0.8494) and SVM (0.8397) on the test dataset. These findings emphasize the capability of deep learning in automatically extracting discriminative features, highlighting its potential for enhanced automated underwater acoustic monitoring.

ABSTRACTAutomated detection of acoustic signals is crucial for effective monitoring of sound‐producing animals and their habitats across ecologically relevant spatial and temporal scales. Recent advances in deep learning have made these approaches more accessible. However, few deep learning approaches can be implemented natively in the R programming environment; approaches that run natively in R may be more accessible for ecologists. The “torch for R” ecosystem has made deep learning with convolutional neural networks (CNNs) accessible for R users. Here, we evaluate a workflow for the automated detection and classification of acoustic signals from passive acoustic monitoring (PAM) data. Our specific goals include (1) present a method for automated detection of gibbon calls from PAM data using the “torch for R” ecosystem, (2) conduct a series of benchmarking experiments and compare the results of six CNN architectures; and (3) investigate how well the different architectures perform on data sets of the female calls from two different gibbon species: the northern gray gibbon (Hylobates funereus) and the southern yellow‐cheeked crested gibbon (Nomascus gabriellae). We found that the highest‐performing architecture depended on the species and test data set. We successfully deployed the top‐performing model for each gibbon species to investigate spatial variation in gibbon calling behavior across two grids of autonomous recording units in Danum Valley Conservation Area, Malaysia and Keo Seima Wildlife Sanctuary, Cambodia. The fields of deep learning and automated detection are rapidly evolving, and we provide the methods and data sets as benchmarks for future work.

Fish choruses are a significant biological component of Australian underwater soundscapes. Due to the links between fish chorus production and the life functions of the fishes producing them, passive acoustic monitoring of fish choruses can provide important ecological data that can be applied to the management of fish populations and their respective habitats. However, little is known regarding the distribution, behaviour, drivers, and functions of fish choruses produced in offshore environments. This study utilised long-term passive acoustic monitoring data to investigate the environmental drivers of a fish chorus recorded along the southern Australian continental shelf, off Portland, Victoria. Generalised Additive Mixed Models were applied to acoustic recordings alongside a range of environmental remote sensing data which captured local conditions such as primary productivity, water temperature, current, sea surface height, nutrient availability, salinity, moonlight levels, and wind speed. This study revealed a significant relationship between the Portland fish chorus and periods of high local primary productivity resulting from the Bonney Coast upwelling system. This upwelling system is predicted to be significantly impacted by climate change in the future. As the Portland fish chorus may be produced as a function of feeding or reproduction, this study has provided a baseline for tracking the response of the chorus source species to future environmental and anthropogenic change.

House mice (Mus musculus domesticus) are among the most widely studied laboratory models of mammalian social behaviour, yet we know relatively little about the ecology of their behaviours in natural environments. Here, we address this gap using radiotelemetry to track social interactions in a population of wild mice over 10 years, from 2013 to 2023, and interpret these interactions in the context of passive acoustic monitoring data collected from August 2022 to November 2023. Using automated vocal detection, we identify 1.3 million individual vocalizations and align them in time with continuously collected telemetry data recording social interactions between individually identifiable mice. We find that vocalization is seasonal and correlated with long-term dynamics in features of social groups. In addition, we find that vocalization is closely associated in time with entrances to and exits from those groups, occurs most often in the presence of pups, and is correlated with how much time pairs of mice spend together. This work identifies seasonal patterns in the vocalizations of wild mice and lays a foundation to investigate the social role of acoustic communication in wild populations of an important laboratory model organism.

Seismic station coverage in the oceans is limited due to high costs and logistical challenges, leading to insufficient earthquake data from oceanic regions. Ocean drones, with quiet operation, buoyancy-driven mechanics, and autonomous underwater profiling, provide a promising alternative for near-real-time data acquisition. We evaluated an oceanic seismological platform using 6 years (2015–2021) of passive acoustic monitoring data from ocean drones in the Santos Basin, southeastern Brazil, originally not designed for earthquake monitoring. Our analysis identified 12 potential earthquake signals, characterized by low-frequency seismic energy and emergent patterns. These findings demonstrate that ocean gliders are highly effective for earthquake monitoring, offering significant advantages for long-term, targeted seismic observations in coastal and marginal areas where conventional methods often face operational limitations.

Distinctive Features of the Development of the Tonga Underwater Volcano Eruption According to Acoustic Monitoring Data