Explore the vast range of titles available.

Automated 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. Our paper presents a workflow utilizing deep learning with convolutional neural networks in the “torch for R” ecosystem for automated detection of gibbon calls from passive acoustic monitoring data. We compare six CNN architectures, assessing their performance on data sets of female calls from two gibbon species. The top‐performing models are deployed to analyze spatial variation in gibbon calling behavior across two conservation areas. Our study provides valuable insights and methods for ecologists working on automated detection and deep learning applications in wildlife monitoring.

We describe a case of lumpy skin disease in an endangered banteng in Cambodia and the subsequent initiation of a vaccination campaign in domestic cattle to protect wild bovids from disease transmission at the wildlife-livestock interface. Lumpy skin disease virus (LSDV) was first detected in domestic cattle in Cambodia in June of 2021 and rapidly spread throughout the country. In September 2021, a banteng was seen in Phnom Tnout Phnom Pok wildlife sanctuary with signs of lumpy skin disease. Scab samples were collected and tested positive for LSDV. Monitoring using line transect surveys and camera traps in protected areas with critical banteng and gaur populations was initiated from December 2021-October 2022. A collaborative multisector vaccination campaign to vaccinate domestic livestock in and around priority protected areas with banteng and gaur was launched July 2022 and a total of 20,089 domestic cattle and water buffalo were vaccinated with Lumpyvax TM . No signs of LSDV in banteng or gaur in Cambodia have been observed since this initial case. This report documents the first case of lumpy skin disease in wildlife in Cambodia and proposes a potential intervention to mitigate the challenge of pathogen transmission at the domestic-wildlife interface. While vaccination can support local livestock-based economies and promote biodiversity conservation, it is only a component of an integrated solution and One Health approach to protect endangered species from threats at the wildlife-livestock interface.

Dhole (Cuon alpinus) is threatened with extinction across its range due to habitat loss and prey depletion. Despite this, no previous study has investigated the distribution and threat of the species at a regional scale. This lack of knowledge continues to impede conservation planning for the species. Here we modeled suitable habitat using presence‐only camera trap data for dhole and dhole prey species in mainland Southeast Asia and assessed the threat level to dhole in this region using an expert‐informed Bayesian Belief Network. We integrated prior information to identify dhole habitat strongholds that could support populations over the next 50 years. Our habitat suitability model identified forest cover and prey availability as the most influential factors affecting dhole occurrence. Similarly, our threat model predicted that forest loss and prey depletion were the greatest threats, followed by local hunting, non‐timber forest product collection, and domestic dog incursion into the forest. These threats require proactive resource management, strong legal protection, and cross‐sector collaboration. We predicted <20% of all remaining forest cover in our study area to be suitable for dhole. We then identified 17 patches of suitable forest area as potential strongholds. Among these patches, Western Forest Complex (Thailand) was identified as the region's only primary stronghold, while Taman Negara (Malaysia), and northeastern landscape (Cambodia) were identified as secondary strongholds. Although all 17 patches met our minimum size criteria (1667 km2), patches smaller than 3333 km2 may require site management either by increasing the ecological carrying capacity (i.e., prey abundance) or maintaining forest extent. Our proposed interventions for dhole would also strengthen the conservation of other co‐occurring species facing similar threats. Our threat assessment technique of species with scarce information is likely replicable with other endangered species. Populations of dhole (Cuon alpinus) are declining in Southeast Asia, but no previous study has evaluated populations at a regional scale. We modeled suitable habitat using presence‐only camera trap data of dhole and prey in mainland Southeast Asia and assessed threat levels to identify dhole long‐term strongholds. Seventeen habitat patches were identified as having at least some potential as strongholds. The Western Forest Complex in Thailand was estimated to be a primary stronghold, with two secondary strongholds in Taman Negara Malaysia and north‐eastern Cambodia. High quality, but smaller patches identified would require habitat restoration. Our habitat suitability model revealed forest cover and prey availability as the most influential factors affecting dhole occurrence. Similarly, our threat model predicted forest loss and prey depletion were the greatest threats facing dhole, followed by hunting, non‐timber forest product collection, and domestic dog incursion. Our threat assessment technique for a species with scarce information might be applicable to other endangered species.

Automated 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 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 datasets of the female calls from two different gibbon species: the northern grey 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 dataset. 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 datasets as benchmarks for future work.

Recent advances in deep and transfer learning have revolutionized our ability for the automated detection and classification of acoustic signals from long-term recordings. Here, we provide a benchmark for the automated detection of southern yellow-cheeked crested gibbon (Nomascus gabriellae) calls collected using autonomous recording units (ARUs) in Andoung Kraleung Village, Cambodia. We compared the performance of support vector machines (SVMs), a quasi-DenseNet architecture (Koogu), transfer learning with pretrained convolutional neural network (ResNet50) models trained on the ‘ImageNet’ dataset, and transfer learning with embeddings from a global birdsong model (BirdNET) based on an EfficientNet architecture. We also investigated the impact of varying the number of training samples on the performance of these models. We found that BirdNET had superior performance with a smaller number of training samples, whereas Koogu and ResNet50 models only had acceptable performance with a larger number of training samples (>200 gibbon samples). Effective automated detection approaches are critical for monitoring endangered species, like gibbons. It is unclear how generalizable these results are for other signals, and future work on other vocal species will be informative. Code and data are publicly available for future benchmarking.

Automated detection of acoustic signals is crucial for effective monitoring of vocal animals and their habitats across ecologically-relevant spatial and temporal scales. Recent advances in deep learning have made these approaches more accessible. However, there are few deep learning approaches that 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 the use of transfer learning with convolutional neural networks accessible for R users. Here, we evaluate a workflow that uses transfer learning for the automated detection 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) compare the results of transfer learning for six pretrained CNN architectures; and 3) investigate how well the different architectures perform on datasets of the female calls from two different gibbon species: the northern grey gibbon (Hylobates funereus) and the southern yellow-cheeked crested gibbon (Nomascus gabriellae). We found that the highest performing architecture depended on the test dataset. We successfully deployed the top performing model for each gibbon species to investigate spatial of 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 datasets as benchmarks for future work.