Passive Acoustic Monitoring will transform biodiversity assessment and large-scale ecological surveys through continuous ecoacoustic data collection. Yet, dataset growth has outpaced existing segmentation methods, many of which are tuned to narrow taxonomic subsets lacking abiotic sounds, limiting ecological realism. Natural soundscapes contain overlapping biophony, geophony, anthrophony, and technophony, making resource-efficient and ecologically valid segmentation an ongoing challenge. As such, we introduce an unsupervised framework designed to reveal meaningful acoustic structures without predefined labels. The pipeline integrates systematic sampling, sound event detection, Mel-Frequency Cepstral Coefficient extraction, dimensionality reduction via Uniform Manifold Approximation and Projection (UMAP), and clustering with Hierarchical Density-Based Spatial Clustering of Applications with Noise (HDBSCAN). We evaluate the framework across six biodiverse Australian soundscapes comprising of (n=19,230) 4.5-second non-overlapping segments. Results indicate that clusters are internally coherent, with Voronoi tessellations over the UMAP space showing distinct spatial boundaries. External validation with (n=2000) manually annotated samples demonstrates strong alignment with ecologically meaningful sound types. F1-scores ranged from 83.7% to 98.3%, with precision and recall exceeding 91% across most sites. By combining unsupervised clustering with ecological validation, our framework offers a practical and resource-efficient solution for organising unlabelled ecoacoustic data while reducing manual effort and preserving ecological integrity.

ABSTRACT Frog species worldwide are particularly vulnerable to the effects of climate change and are the most endangered vertebrate group. To identify and mitigate impacts, it is necessary to have a method to monitor activity, particularly breeding activity, to detect changes over time. To this end, this study examined the breeding phenology of the poorly known magnificent broodfrog (Pseudophryne covacevichae) across its disjunct distribution in upland sclerophyll forests of the Wet Tropics of Australia. Using a species‐specific acoustic recogniser, we processed near continuous recordings from November 2022 to November 2023, across ten sites, spanning parts of the species' disjunct ranges, the Atherton Tablelands region and Paluma Range. Recorders sampled nightly for 5 min of every hour and detected ~350,000 advertisement calls using a 95% accuracy threshold. Despite a geographic separation of under 200 km, the timing and duration of calling varied markedly among sites and regions. Atherton Tablelands populations exhibited relatively synchronous, wet season calling, whereas Paluma Range populations had delayed and more prolonged calling periods, extending into the mid to late dry season. Climate window analyses indicated that calling was influenced by antecedent environmental conditions, but the time scale and effect differed between regions. In the Atherton Tablelands, calling intensity was positively associated with high rainfall accumulation over multiple weeks, high minimum temperatures and low humidity over the preceding month. In contrast, in the Paluma Range, calling intensity declined with higher rainfall over the previous 2 weeks and was associated with cooler, but more humid nights in the week prior. The intraspecific variation in calling activity suggests that local hydrological conditions may mediate reproductive timing and duration, with rainfall either facilitating or suppressing calling, depending on the region. Understanding such fine‐scale intraspecific variation in breeding may assist in identifying populations more at risk to the effects of future climate change.

Monitoring threatened species is crucial for conservation, providing the necessary information on species distribution and abundance required to detect declines and evaluate the impact of conservation actions. Passive acoustic monitoring networks, such as the Australian Acoustic Observatory (A2O), have significant potential to aid the monitoring of vocal threatened species by collecting continuous data across many locations in a cost-effective manner. We analysed over 2 million hours of acoustic recordings collected as part of the A2O from 63 sites between 2019 and 2022 to determine the site presence of 74 threatened species. Using the existing deep-learning model BirdNET to detect species belonging to classes within the model, and embeddings search for species outside the model, we successfully detected 41 out of the 74 threatened target species at a minimum of one A2O location, with some broadly distributed threatened species detected at more than ten sites. Threatened species belonging to all three threatened categories (i.e., Vulnerable, Endangered, Critically Endangered), and all three broad taxonomic groupings (i.e., birds, frogs, mammals) were found within A2O recordings, and the majority of sites searched (83%) detected at least one threatened species. We provide information on the distribution of detections for all target species, an evaluation of classification performance for all species within the existing BirdNET model, as well as easy-to-use linear classifiers trained on BirdNET embeddings for all threatened species detected, which have improved performance over the standard BirdNET model. We conclude that acoustic monitoring networks such as the A2O, powered by deep-learning models, can serve as an important tool for monitoring threatened species.

Technological advances over the last two decades have seen a large uptake in passive acoustic monitoring (PAM) to supplement traditional biodiversity surveys. To address the growing backlog of acoustic recordings these projects create, multispecies acoustic classifiers are now widely used in favour of manually processing data. Despite the uptake of these technologies, the efficacy of these recognisers needs to be evaluated to ensure they are detecting levels of diversity and community structure similar to traditional surveys. This study compared metrics of avian diversity across eastern Australia, between traditional bird surveys (both dawn, and dawn + nocturnal), and PAM processed using BirdNET, a widely used multi-species classifier. On average, PAM and multi-species classifiers returned higher values for all assessed diversity metrics (Species Richness, Chao2 estimators, Petchey's Functional Diversity, Rao's Q and Phylogenetic Distance), however traditional surveys supplemented with nocturnal surveys returned intermediate values. The efficacy of classifiers varied considerably across the study locations, with the number of incorrect identifications in the tropics substantially higher than those in temperate areas. Both methodologies failed to detect certain taxonomic groups, some threatened species, and detected significantly different avian communities. While these results provide a promising outlook for the future of PAM, it underscores the importance of maintaining traditional surveys as part of biodiversity monitoring, and relying on skilled ornithologists to ensure recorded acoustic data is appropriately interrogated. Furthermore, this study cautions against relying solely on automated classifiers in regions where training data for models is depauperate, such as Australia's tropical and subtropical woodlands.

Invasive populations are managed most effectively when surveillance aligns with periods and places of maximal activity. Understanding the conditions influencing breeding of invasive species may help provide information for studies of population dynamics and range expansion. Despite their continued successful advances, the patterns and drivers of invasive cane toad (Rhinella marina) calling activity, as a signal of potential breeding activities, remain poorly resolved at broad scales in Australia. Using continent-wide passive acoustic recording data paired with a machine-learning acoustic classifier, we describe patterns in cane toad choruses across most of their current range in Australia, spanning three States and climate zones (~ 15° latitude and 27° longitude). A total of 163,701 minutes of cane toad calls from 33,799 nights of recordings showed that they chorused nearly year-round, with seasonal peaks. Cane-toad calling probability and intensity increased on dark, calm, humid nights and near lentic systems or temporary water, with low- to medium- canopy cover and decreased with increasing wind and moonlight. Social cues appeared important: prior-week calling activity strongly elevated both probability and intensity of calling. Interestingly, temperature effects were context-dependent, without a prior chorus, calling probability and intensity peaked at moderate temperatures; following chorusing, both metrics remained high across a broader thermal range. However, variable importance differed amongst climate zones: atmospheric and lunar factors were more influential for initiating choruses in subtropical and semi-arid sites, whereas habitat and hydrology contributed more to sustaining choruses in the Tropics. Together, these results reveal a detailed picture of cane toad calling ecology in Australia, identifying when, where and under what conditions chorusing is favoured. More broadly, this study demonstrates the value of ecoacoustics for investigating vocal invasive species at large spatial scales and provides a useful basis for future monitoring and modelling. Graphical abstract

Passive acoustic monitoring has emerged as a powerful tool for conducting biodiversity monitoring, with acoustic indices offering a potentially quick way of predicting traditional biodiversity metrics such as species richness without the need to determine species identity. However, key to using these indices is validating their relationship with biodiversity metrics of interest using on-ground surveys, understanding how best to aggregate indices and build predictive models for monitoring use, and understanding how these relationships and models generalize across space and time. Over a four-year period, acoustic recordings collected alongside repeated on-ground surveys at 50 sites were used to investigate the relationship between 14 acoustic indices and two measures of avian diversity (i.e., bird species richness and total bird count). Specifically, we looked at how the number of recording days and different diel periods impact single index correlations, and whether multi-index models are able to provide more accurate predictions than single-index models. We found correlations between both measures of avian diversity and most acoustic indices improved with increasing amount of audio recording days used, with the dawn period having the strongest correlations for species richness (Acoustic Complexity Index, Mid-Frequency Cover, and Spectral Density), and day having the strongest correlations for total bird count (Acoustic Complexity Index, NDSI, and Mid-Frequency Cover). Multi-index models were generally similar in performance to the best performing single index models for predicting species richness. We also found that model predictive performance decreased more strongly when used as generalised predictors for new locations, in comparison to new time periods. Our results highlight the need to validate acoustic index relationships with diversity metrics before applying them, and cautioning the use of acoustic indices for this purpose in new locations based on prior relationships.

Passive acoustic monitoring has emerged as a useful technique for monitoring vocal species and contributing to biodiversity monitoring goals. However, finding target sounds for species without pre-existing recognisers still proves challenging. Here, we demonstrate how the embeddings from the large acoustic model BirdNET can be used to quickly and easily find new sound classes outside the original model’s training set. We outline the general workflow, and present three case studies covering a range of ecological use cases that we believe are common requirements in research and management: monitoring invasive species, generating species lists, and detecting threatened species. In all cases, a minimal amount of target class examples and validation effort was required to obtain results applicable to the desired application. The demonstrated success of this method across different datasets and different taxonomic groups suggests a wide applicability of BirdNET embeddings for finding novel sound classes. We anticipate this method will allow easy and rapid detection of sound classes for which no current recognisers exist, contributing to both monitoring and conservation goals.

Ecoacoustics has emerged as a pivotal discipline in the conservation and monitoring of ecosystems, offering insights into species’ behaviour and ecosystem health through soundscape analysis. Central to this is the need for accurate annotations of environmental audio recordings, which underpin the computational models used in ecological monitoring. However, due to the increasingly large scale of datasets, annotation using existing tools and techniques cannot be performed at feasible speeds or with the necessary accuracy required for real-world application. The LEAVES (Large-scale Ecoacoustics Annotation and Visualisation with Efficient Segmentation) platform addresses this gap by leveraging unsupervised clustering techniques optimised for the high-throughput annotation of large-scale ecoacoustics datasets. Our evaluation across six real-world datasets shows that LEAVES improves annotation efficiency by up to 7.12 times compared to manual annotation while maintaining 79%–90% label similarity to validated data. We expect that our proposed tool will greatly accelerate the annotation process when generating high-quality labelled datasets, supporting larger-scale studies with broader community engagement in ecoacoustics research.

Invasive species pose a significant threat to global biodiversity and ecosystem health, necessitating effective monitoring tools for early detection and management. Here, we present the development and assessment of a user-friendly and transferable monitoring tool for the invasive cane toad (Rhinella marina) using passive acoustic monitoring (PAM) and machine learning algorithms. Leveraging a continental-scale PAM dataset (Australian Acoustic Observatory), we trained a cane toad classifier using the BirdNET algorithm, a convolutional neural network architecture capable of identifying acoustic events. We validated thousands of BirdNET predictions across Australia, and our classifier achieved over 90 % accuracy even at many sites outside the areas from which the training data were obtained. Additionally, because cane toads typically call for long periods, we significantly enhanced detection accuracy by incorporating contextual information from time-series data, essentially checking if other calls occurred around each detection (an optimized threshold approach using conditional inference trees). This method substantially reduced false positives and improved overall performance in cane toad detection at sites across Australia. Overall, our method will allow others to develop accurate and precise automated acoustic monitoring tools tailored to their situation, with minimal training data, addressing the critical need for accessible solutions in biodiversity monitoring, control of invasive species and conservation.

Mammals play vital roles in ecological communities, but many are in rapid decline worldwide. Comprehensive monitoring of mammal populations is crucial for effective conservation, but large-scale monitoring presents significant challenges. Remote sensing techniques such as passive acoustic monitoring offer viable and effective solutions for surveying animal communities. While passive acoustic monitoring has shown promising results for birds, its application in mammal biodiversity assessments has received little testing. In this study, we compared passive acoustic monitoring (combined with BirdNET embeddings) to traditional observer-based monitoring and camera trapping for assessing terrestrial mammal biodiversity over multiple years across an extensive spatial scale in eastern Australia. Using embeddings from the BirdNET deep learning model, we efficiently analysed a cumulative total of 317,410 h of continuous audio data, recorded at a sampling rate of 22 kHz (effective up to 11 kHz), from six sites, successfully detecting all 17 target vocal mammal species. Given the inherent inability of passive acoustic monitoring to detect non-vocalising species or those vocalising outside this frequency range, we considered the mammal community in two ways: (i) entire mammal community and (ii) vocal mammals only. For detecting species in the entire mammal community, observer-based monitoring performed the best, followed by camera trapping and then passive acoustic monitoring. However, when focusing on vocal mammals only, all methods showed comparable performance for the same 28-day survey period, with passive acoustic monitoring demonstrating significantly better performance when leveraging long-term audio data. Additionally, we found a significant positive correlation between species richness of vocal mammals and that of the entire mammal community, suggesting that vocal mammal richness may serve as a proxy for overall mammal biodiversity. Furthermore, PAM's effectiveness was not influenced by common life-history traits, suggesting PAM may be a broadly applicable survey tool for vocal mammals. Despite the benefits, the inability of passive acoustic monitoring to detect non-vocal mammals necessitates a combined approach with other survey methods such as camera trapping or observer-based monitoring to achieve a comprehensive assessment of terrestrial mammal biodiversity. This combined approach is likely to enhance future biodiversity monitoring, offering a more detailed understanding of ecosystems and supporting effective conservation practices.

BACKGROUNDInvasive species are a major threat to biodiversity on a global scale. Control strategies for these species could be improved by understanding and exploiting life history vulnerabilities. For example, most invasive anurans require waterbodies with specific characteristics for spawning; therefore, modifying these characteristics could influence spawning success. Asian toads (Duttaphrynus melanostictus) were accidentally introduced to the east coast of Madagascar around 2010, and have since established and spread across an area exceeding 850 km2. To determine if Asian toads select spawning sites with specific characteristics within their invaded range, we surveyed habitat characteristics at 30 waterbodies used by toads for spawning, and compared these to characteristics at 30 adjacent, unused waterbodies, in urban and rural areas in eastern Madagascar.RESULTSToads selectively oviposited in small waterbodies with gently sloping banks, while the structure of surrounding vegetation, water chemistry (salinity, water temperature), presence of other species, depth of the waterbody, and substrate of the waterbody bank did not appear to influence spawning site selection.CONCLUSIONOur results provide a pathway to examine modification of potential Asian toad spawning sites as a management strategy for these invasive pests. (c) 2025 Society of Chemical Industry.

Morphological adaptations facilitate effective movement within habitats. Claws are among the most common adaptations enabling organisms to use inclined and vertical surfaces. However, some taxa have evolved adhesive pads in addition to claws, with claws suggested to be more effective at gripping coarse surfaces, while pads attach better to fine-grained surfaces. Using test surfaces that represented the range of surface roughness used by six species of diplodactylid geckos in nature, we quantified the role of claws and pads acting together, and of pads alone. We examined two functional traits, attachment (on inclines, 45° and vertical surfaces, 90°) and clinging ability (on inclines only). Claws were critical to attachment on vertical surfaces, and attachment declined linearly with decreasing surface roughness. Although attachment was lowest on fine-grained surfaces, this was where claws had the greatest functional contribution. Clinging ability also declined linearly with decreasing surface roughness, where claws played an additive role. Our study highlights novel results describing the function of gecko adhesive systems on different surfaces and suggests a clade-specific interaction of claws and pads. Specifically, we highlight that pads alone can be capable of attachment on rough surfaces, with claws contributing more on fine-grained surfaces.

Ecoacoustic methods provide opportunities for ecological studies of vocalizing species within the context of the natural habitats and communities in which they occur. Continuous acoustic monitoring of species assemblages can reveal patterns in breeding phenology, behavior, and interactions. We used long-duration false-color spectrograms derived from acoustic indices to detect the nightly chorusing of a community of anurans in a tropical savanna in north Queensland. We described the chorusing patterns of each species over two wet seasons at three breeding sites, and used conditional random forest analysis to investigate the influence of various environmental factors. Frogs in these habitats form multispecies aggregations at water bodies during breeding periods when males form large choruses to attract females. The chorusing patterns revealed the species have different breeding periods, which could be broadly categorized as explosive or prolonged. While rain events were often a trigger for the commencement of the breeding period, species responded differently to environmental conditions. Choruses of explosive breeding species occurred only on the night of, or night after, the first high rainfall event of the wet season. The prolonged breeding species showed idiosyncratic patterns of chorusing, which were generally consistent across sites. Fine-grained nightly data on patterns of chorusing and the relationship with environmental conditions allow us to understand the detectability of the presence, or absence, of the frog species in these habitats, and provide baseline data for monitoring and management programs.

Context Biodiversity is declining worldwide, with many species decreasing in both number and range. Acoustic monitoring is rapidly becoming a common survey method in the ecologist’s toolkit that may aid in the conservation of endangered species, but effective analysis of long-duration audio recordings is still challenging. Aims The aims of this study were to: (1) develop and test call recognisers for the endangered southern black-throated finch (Poephila cincta cincta) as well as the similar sounding, but non-endangered, double-barred finch (Taeniopygia bichenovii); and (2) compare the ability of these recognisers to detect these species with that of on-ground bird surveys at under-surveyed locations in the Desert Uplands bioregion which is at the edge of the known range of the black-throated finch. Methods A range of convolutional neural network call recognition models were built and tested for both target species, before being deployed over new audio recordings collected at 25 sites during 2020, 2021 and 2022, and compared with the results of on-ground bird surveys at those same sites. Key results Call recognisers for both species performed well on test datasets from locations in the same area as the training data with an average area under the precision-recall curve (PRAUC) of 0.82 for black-throated finch and 0.87 for double-barred finch. On-ground bird surveys in the Desert Uplands bioregion detected black-throated finches at two locations in different years, and our call recognisers confirmed this with minimal post-validation of detections. Similar agreement between methods were obtained for the double-barred finch, with site occupancy in the Desert Uplands bioregion confirmed with audio recognition in all nine surveys with on-ground detections, as well as during four additional surveys that had no on-ground detections. Conclusions Using call recognisers to survey new locations for black-throated finch presence was equally successful as on-ground surveys, and with further refinements, such as retraining models with examples of commonly misclassified vocalisations added to the training data, minimal validation should be required to detect site presence. Implications Acoustic monitoring should be considered as a valuable tool to be used alongside manual surveys to allow effective monitoring and conservation of this endangered species.

Reptiles play a crucial role in maintaining biodiversity and ecosystem health, yet many species face increasing threats because of various anthropogenic factors. To enhance our understanding of reptile diversity and habitat use, evaluation of the effectiveness of diverse survey techniques is necessary. The relative efficacy of different methods may vary significantly across regions or communities, highlighting the need for a comprehensive approach using multiple survey methods over extensive spatial and temporal scales. In this study, we compared the effectiveness of seven survey methods-pitfall traps, funnel traps, spotlighting, arboreal cover boards, incidental encounters, camera traps, and passive acoustic monitoring (PAM)-for assessing reptile biodiversity over several years across an extensive spatial range in open eucalypt woodlands in eastern Australia. Pitfall and funnel traps were the most effective methods for detecting reptiles across all sites and latitudes. A combination of pitfall and funnel traps accumulated species most quickly, had high detection probabilities, and accounted for nearly 90% of all different reptile species detected in this study. However, with a decrease in latitude reptile diversity increased and other survey methods became necessary to document the full extent of the reptile communities. Reptile assemblages captured by different survey methods varied significantly, except for the communities captured by pitfall and funnel traps. No single method captured all species, and no species was detected by every method. PAM failed to detect any reptiles and may not be viable for assessing reptile biodiversity in Australia. Pitfall and funnel traps proved highly effective for detecting terrestrial reptiles within open eucalypt woodlands in Australia; however, the selection of methods for evaluating reptile biodiversity depended on the objectives and target fauna. When possible, to maximize species richness, survey designs should incorporate an array of concurrently deployed methods, particularly in regions with higher overall species richness. Nevertheless, if resources and time are limited, pitfall and funnel traps, combined with incidental encounters, should capture the majority of species.

Computational ecoacoustics has seen significant growth in recent decades, facilitated by the reduced costs of digital sound recording devices and data storage. This progress has enabled the continuous monitoring of vocal fauna through Passive Acoustic Monitoring (PAM), a technique used to record and analyse environmental sounds to study animal behaviours and their habitats. While the collection of ecoacoustic data has become more accessible, the effective analysis of this information to understand animal behaviours and monitor populations remains a major challenge. This survey paper presents the state-of-the-art ecoacoustics data analysis approaches, with a focus on their applicability to large-scale PAM. We emphasise the importance of large-scale PAM, as it enables extensive geographical coverage and continuous monitoring, crucial for comprehensive biodiversity assessment and understanding ecological dynamics over wide areas and diverse habitats. This large-scale approach is particularly vital in the face of rapid environmental changes, as it provides crucial insights into the effects of these changes on a broad array of species and ecosystems. As such, we outline the most challenging large-scale ecoacoustics data analysis tasks, including pre-processing, visualisation, data labelling, detection, and classification. Each is evaluated according to its strengths, weaknesses and overall suitability to large-scale PAM, and recommendations are made for future research directions.

Large-scale natural soundscapes are remarkably complex and offer invaluable insights into the biodiversity and health of ecosystems. Recent advances have shown promising results in automatically classifying the sounds captured using passive acoustic monitoring. However, the accuracy performance and lack of transferability across diverse environments remains a challenge. To rectify this, we propose a robust and flexible ecoacoustics sound classification grid search-based framework using optimised machine learning algorithms for the analysis of large-scale natural soundscapes. It consists of four steps: pre-processing including the application of spectral subtraction denoising to two distinct datasets extracted from the Australian Acoustic Observatory, feature extraction using Mel Frequency Cepstral Coefficients, feature reduction, and classification using a grid search approach for hyperparameter tuning across classifiers including Support Vector Machine, k-Nearest Neighbour, and Artificial Neural Networks. With 10-fold cross validation, our experimental results revealed that the best models obtained a classification accuracy of 96% and above in both datasets across the four major categories of sound (biophony, geophony, anthrophony, and silence). Furthermore, cross-dataset validation experiments using a pooled dataset highlight that our framework is rigorous and adaptable, despite the high variance in possible sounds at each site.

Effective monitoring tools are key for tracking biodiversity loss and informing management intervention strategies. Passive acoustic monitoring promises to provide a cheap and effective way to monitor biodiversity across large spatial and temporal scales, however, extracting useful information from long-duration audio recordings still proves challenging. Recently, a range of acoustic indices have been developed, which capture different aspects of the soundscape, and may provide a way to estimate traditional biodiversity measures. Here we investigated the relationship between 13 acoustic indices obtained from passive acoustic monitoring and biodiversity estimates of various vertebrate taxonomic groupings obtained from manual surveys at six sites spanning over 20 degrees of latitude along the Australian east coast. We found a number of individual acoustic indices that correlated well with species richness, Shannon’s diversity index, and total individual count estimates obtained from traditional survey methods. Correlations were typically greater for avian and total vertebrate biodiversity than for anuran and non-avian vertebrate biodiversity. Acoustic indices also correlated better with species richness and total individual count than with Shannon’s diversity index. Random forest models incorporating multiple acoustic indices provided more accurate predictions than single indices alone. Out of the acoustic indices tested, cluster count, mid-frequency cover and spectral density contributed the greatest predictive ability to models. Our results suggest that models incorporating multiple acoustic indices could be a useful tool for monitoring certain vertebrate groups. Further work is required to understand how site-specific variables can be incorporated into models to improve predictive capabilities and how to improve the monitoring of taxa besides avians, particularly anurans.

Passive acoustic monitoring (PAM) has become increasingly popular in ecological studies, but its efficacy for assessing overall terrestrial vertebrate biodiversity is unclear. To quantify this, its performance for species detection must be directly compared to that obtained using traditional observer-based monitoring (OBM). Here, we review such comparisons across all major terrestrial vertebrate classes and identify factors impacting PAM performance. From 41 studies, we found that while PAM-OBM comparisons have been made for all major terrestrial vertebrate classes, most comparisons have focused on birds (65%) in North America (52%). PAM performed equally well or better (61%) compared to OBM in general. We found no statistical difference between the methods for total number of species detected across all vertebrate classes (excluding reptiles); however, recording period and region of study influenced the relative performance of PAM, while acoustic analysis method and which method sampled for longer overall showed no impact. Further studies comparing PAM performance in non-avian vertebrates using standardised methods are needed to investigate in more detail the factors that may influence PAM performance. While PAM is a valuable tool for vertebrate surveys, a combined approach with targeted OBM for non-vocal species should achieve the most comprehensive assessment of terrestrial vertebrate communities.