Ecological studies often require observations of animals and their behaviour. Motion‐activated cameras (camera traps) based on passive infrared detection (PIR) are a popular solution for recording ...animal activity in situations when it is impractical for humans to make sufficient observations. However, the reliability of these cameras for recording smaller vertebrates remains uncertain. We assessed the reliability of two widely used PIR camera traps (Bushnell 119740 and Moultrie 13068) for detecting small vertebrates. Specifically, we tested the effects of (1) camera trap model, (2) camera–subject distance and (3) animal class and size on the probability of detection. Brown rats moving across the ground were detected by the Bushnell camera with >80% probability at camera–subject distances ranging from 60 cm to 2 m, while the Moultrie camera was less efficient at distances >1 m. The probability of the Bushnell camera detecting birds feeding on flowers decreased from c. 80% at distances of 40–60 cm to <10% at 2 m and beyond. Larger birds (>20 g) were more likely to be detected than smaller birds (<15 g). Close‐focusing lenses on the Bushnell camera readily allow identification of individual bird species. These results help to establish guidelines for camera trap selection and placement in ecological studies of small terrestrial mammals and birds.
Camera traps allow animal activity to be recorded when it is impractical for humans to make observations. Close‐focusing camera traps allow identification of small vertebrates, but the triggering reliability of these cameras has not been thoroughly tested. We found that close‐focusing camera traps can reliably record activity of rats at distances of up to 2 m, but should be no more than 40–60 cm from small nectar‐feeding birds for triggering to be reliable.
Few records of Lagidium wolffsohni exist in both Chile and Argentina. Through camera traps, we obtained a new record of L. wolffsohni in northern Patagonia, Chile. We also describe its feeding ...activity and other behavioral aspects.
Our study was aimed to establish the circadian activity of the Red fox (Vulpes vulpes) and the Stone marten (Martes foina) in the protected area "Zlatiyata", situated in an agricultural landscape. ...The study was conducted from September 2021 to March 2022 using camera traps. It provides new data on the behavioral ecology of the target species contributing to the knowledge of their co-existence in croplands.
Camera traps serve as a valuable tool for wildlife monitoring, generating a vast collection of images for ecologists to conduct ecological investigations, such as species identification and ...population estimation. However, the sheer volume of images poses a challenge, and the integration of deep learning into automated ecological investigation tasks remains complex, particularly when dealing with low-quality images in long-term monitoring programs. Existing approaches often struggle to strike a balance between image enhancement and deep learning for ecological tasks, thereby overlooking crucial information contained within low-quality images. This research introduces a pioneering adaptive image processing module (AIP) that seamlessly incorporates image processing into camera trap ecological tasks, elevating the performance of wildlife monitoring activities. Specifically, a differentiable image processing (DIP) module is presented to enhance low-quality images, with its parameters predicted by a Non-local based parameter predictor (NLPP). Additionally, an end-to-end approach based on hybrid data containing both original and synthetic data is proposed, encompassing adaptive image processing methods and downstream tasks for camera traps, adaptable to various scenarios. This approach effectively reduces the manual labor and time required for professional image processing. When applied to real-world camera trap images and synthetic image datasets, our method achieves an accuracy of 92.26% and 86.65% in classifying wildlife, respectively, demonstrating its robustness. By outperforming alternative methods under harsh conditions, the application of the adaptive image processing module instills greater confidence in deep learning applications within complex environments.
Display omitted
•A differentiable image processing module is proposed to tackle the low-quality environments.•A non-local layer is incorporated to extract crucial features for parameter prediction.•The adaptive image processing module can be seamlessly integrated into pre-existing models.•A joint training method is proposed to ensure the image processing effect.•A complex and variable test baseline is built for testing generalization.
Deep learning (DL) algorithms are the state of the art in automated classification of wildlife camera trap images. The challenge is that the ecologist cannot know in advance how many images per ...species they need to collect for model training in order to achieve their desired classification accuracy. In fact there is limited empirical evidence in the context of camera trapping to demonstrate that increasing sample size will lead to improved accuracy.
In this study we explore in depth the issues of deep learning model performance for progressively increasing per class (species) sample sizes. We also provide ecologists with an approximation formula to estimate how many images per animal species they need for certain accuracy level a priori. This will help ecologists for optimal allocation of resources, work and efficient study design.
In order to investigate the effect of number of training images; seven training sets with 10, 20, 50, 150, 500, 1000 images per class were designed. Six deep learning architectures namely ResNet-18, ResNet-50, ResNet-152, DnsNet-121, DnsNet-161, and DnsNet-201 were trained and tested on a common exclusive testing set of 250 images per class. The whole experiment was repeated on three similar datasets from Australia, Africa and North America and the results were compared. Simple regression equations for use by practitioners to approximate model performance metrics are provided. Generalizes additive models (GAM) are shown to be effective in modelling DL performance metrics based on the number of training images per class, tuning scheme and dataset.
Overall, our trained models classified images with 0.94 accuracy (ACC), 0.73 precision (PRC), 0.72 true positive rate (TPR), and 0.03 false positive rate (FPR). Variation in model performance metrics among datasets, species and deep learning architectures exist and are shown distinctively in the discussion section. The ordinary least squares regression models explained 57%, 54%, 52%, and 34% of expected variation of ACC, PRC, TPR, and FPR according to number of images available for training. Generalised additive models explained 77%, 69%, 70%, and 53% of deviance for ACC, PRC, TPR, and FPR respectively.
Predictive models were developed linking number of training images per class, model, dataset to performance metrics. The ordinary least squares regression and Generalised additive models developed provides a practical toolbox to estimate model performance with respect to different numbers of training images.
•Deep Network classifier sample size requirements investigated for long-term wildlife monitoring sites.•Training sample size was linked to model performance metrics such as accuracy.•Logarithmic trends observed between training sample size and accuracy, precision, recall and false positive rate.•Model performance asymptotes with 150–500 images per class providing good accuracy.•The effects of data set, samples per class, model architectures and tuning strategies were investigated and compared.
In southern Mexico, Voluntarily Designated Conservation Areas (VCA) represent a biological conservation strategy wherein governance and management are entrusted to the territory. Within the VCAs of ...the La Chinantla region in the state of Oaxaca, Mexico, community monitoring utilizing camera traps has been conducted with the assistance of government programs. This initiative has yielded a substantial number of records for medium and large mammals. Nevertheless, the available information has not undergone systematic analysis, constraining its utility in strategic planning and the evaluation of biodiversity conservation endeavors. This study seeks to highlight the impact of community monitoring in 18 VCAs on understanding the altitudinal distribution of mammal diversity in La Chinantla. The analysis incorporates data from a community monitoring covering 129 camera trap stations (4,384 camera days) strategically positioned along an elevation gradient ranging from 50 to 2000 m above sea level, over the period 2011–2014. We assessed alpha and beta diversity, as well as the community structure of medium and large mammals within three distinct elevation zones. A total of 26 species of medium-sized mammals were documented, revealing distinct mammal assemblages in each zone. However, 15 species were common across all zones. We found that the highest species richness was observed below 400 m, where tropical rainforest vegetation predominates. We also found that the species turnover component had a significant impact on the total beta value. Despite the considerable involvement of local residents in the monitoring program and their acquisition of social, technical, and ecological knowledge, there is still a need to strengthen their capabilities to enhance community monitoring. Finally, fostering collaboration between local communities, academic institutions, and governmental initiatives is essential for the successful conservation of mammals in La Chinanlta.
1. A central goal of animal ecology is to observe species in the natural world. The cost and challenge of data collection often limit the breadth and scope of ecological study. Ecologists often use ...image capture to bolster data collection in time and space. However, the ability to process these images remains a bottleneck. 2. Computer vision can greatly increase the efficiency, repeatability and accuracy of image review. Computer vision uses image features, such as colour, shape and texture to infer image content. 3. I provide a brief primer on ecological computer vision to outline its goals, tools and applications to animal ecology. 4. I reviewed 187 existing applications of computer vision and divided articles into ecological description, counting and identity tasks. 5. I discuss recommendations for enhancing the collaboration between ecologists and computer scientists and highlight areas for future growth of automated image analysis.
Environmental DNA (eDNA) metabarcoding can identify terrestrial taxa utilising aquatic habitats alongside aquatic communities, but terrestrial species' eDNA dynamics are understudied. We evaluated ...eDNA metabarcoding for monitoring semi-aquatic and terrestrial mammals, specifically nine species of conservation or management concern, and examined spatiotemporal variation in mammal eDNA signals. We hypothesised eDNA signals would be stronger for semi-aquatic than terrestrial mammals, and at sites where individuals exhibited behaviours. In captivity, we sampled waterbodies at points where behaviours were observed (‘directed’ sampling) and at equidistant intervals along the shoreline (‘stratified’ sampling). We surveyed natural ponds (N = 6) where focal species were present using stratified water sampling, camera traps, and field signs. eDNA samples were metabarcoded using vertebrate-specific primers. All focal species were detected in captivity. eDNA signal strength did not differ between directed and stratified samples across or within species, between semi-aquatic or terrestrial species, or according to behaviours. eDNA was evenly distributed in artificial waterbodies, but unevenly distributed in natural ponds. Survey methods deployed at natural ponds shared three species detections. Metabarcoding missed badger and red fox recorded by cameras and field signs, but detected small mammals these tools overlooked, e.g. water vole. Terrestrial mammal eDNA signals were weaker and detected less frequently than semi-aquatic mammal eDNA signals. eDNA metabarcoding could enhance mammal monitoring through large-scale, multi-species distribution assessment for priority and difficult to survey species, and provide early indication of range expansions or contractions. However, eDNA surveys need high spatiotemporal resolution and metabarcoding biases require further investigation before routine implementation.
Beavers (Castor canadensis and C. fiber) build dams that modify catchment and pond water balances, and it has been suggested that they can be a nature-based solution for reducing flood hydrographs, ...enhancing low flow hydrographs and restoring hydrological functioning of degraded streams. How water moves through a beaver dam is determined by its flow state (e.g., overflow, underflow). However, current conceptual models only consider flow state as changing over the beaver site occupation-abandonment cycle. To assess whether flow state changes at shorter timescales and identify possible triggers (e.g., rainfall, animals), we integrated camera trap imagery, machine learning, water level measurements, and hydrometeorological data at beaver dams in a montane peatland in the Canadian Rocky Mountains. Contrary to current models, we found that flow states changed frequently, changing a maximum 12 times during the 139-day study period, but that changes had limited synchronicity amongst the dams in the same stream. More than two-thirds of the changes coincided with rainfall events. We observed no changes in flow state in response to beaver activity or wildlife crossings perhaps due to the camera positioning. Our findings augment the long-term oriented framework, which links changes to the occupancy cycle of a beaver pond and frequent and hydrological-driven changes. To develop realistic predictions of hydrological impacts of beaver dams, ecohydrological models should update their representation of the influence of beaver dams to include short-term dynamism of flow states and potential triggers. Our study advances the understanding of the important, yet understudied, role of beaver dams in stream restoration and climate change initiatives.
Display omitted
•Understanding streamflow through beaver dams (flow state) aids predictability.•Dam flow state change was understood as a long-term process driven by beavers.•We found a short-term dynamic, which is fast and has limited predictability.•Hydrological drivers were predominant in triggering flow state changes.•Camera traps are a key tool for assessing effects of beaver dams on streamflow.
The European hare Lepus europaeus is an exotic mammal with a wide distribution potential, suggesting the presence of highly flexible anti-predation mechanisms. Through the use of camera traps, the ...activity pattern of this species was studied for two years at a site of the coastal foothills in southern Chile. During the second year it was observed that the hare maintained a non-random pattern of activity in relation to the light periods of the day. This fact suggests the adoption of anti-predator strategies, considering the variation in the presence of their predators between both years. The effect of predation on the observed temporal patterns is discussed, as well as its population fluctuations in the long term, to describe the role of this lagomorph in this type of ecosystem.
PDF
