SPECIES DETECTION: BIRDS
Verstraeten W. W., Vermeulen B., Stuckens J., Lhermitte S., Van der Zande D., Van Ranst M., Coppin P. (2010): Webcams for bird detection and monitoring: A demonstration study. Sensors 10: 3480-3503.
FULL TEXT
Abstract
Better insights into bird migration can be a tool for assessing the spread of avian borne infections or ecological/climatologic issues reflected in deviating migration patterns. This paper evaluates whether low budget permanent cameras such as webcams can offer a valuable contribution to the reporting of migratory birds. An experimental design was set up to study the detection capability using objects of different size, color and velocity. The results of the experiment revealed the minimum size, maximum velocity and contrast of the objects required for detection by a standard webcam. Furthermore, a modular processing scheme was proposed to track and follow migratory birds in webcam recordings. Techniques such as motion detection by background subtraction, stereo vision and lens distortion were combined to form the foundation of the bird tracking algorithm. Additional research to integrate webcam networks, however, is needed and future research should enforce the potential of the processing scheme by exploring and testing alternatives of each individual module or processing step.
Ovaskainen O., Moliterno de Camargo U., Somervuo P. (2018): Animal Sound Identifier (ASI): software for automated identification of vocal animals. Ecology Letters 21: 1244-1254.
FULL TEXT
Abstract
Automated audio recording offers a powerful tool for acoustic monitoring schemes of bird, bat, frog and other vocal organisms, but the lack of automated species identification methods has made it difficult to fully utilise such data. We developed Animal Sound Identifier (ASI), a MATLAB software that performs probabilistic classification of species occurrences from field recordings. Unlike most previous approaches, ASI locates training data directly from the field recordings and thus avoids the need of pre-defined reference libraries. We apply ASI to a case study on Amazonian birds, in which we classify the vocalisations of 14 species in 194 504 one-minute audio segments using in total two weeks of expert time to construct, parameterise, and validate the classification models. We compare the classification performance of ASI (with training templates extracted automatically from field data) to that of monitoR (with training templates extracted manually from the Xeno-Canto database), the results showing ASI to have substantially higher recall and precision rates.
Ushio M., Murata K., Sado T., Nishiumi I., Takeshita M., Iwasaki W., Miya M. (2018): Demonstration of the potential of environmental DNA as a tool for the detection of avian species. Scientific Reports 8: 4493.
FULL TEXT
Abstract
Birds play unique functional roles in the maintenance of ecosystems, such as pollination and seed dispersal, and thus monitoring bird species diversity is a first step towards avoiding undesirable consequences of anthropogenic impacts on bird communities. In the present study, we hypothesized that birds, regardless of their main habitats, must have frequent contact with water and that tissues that contain their DNA that persists in the environment (environmental DNA; eDNA) could be used to detect the presence of avian species. To this end, we applied a set of universal PCR primers (MiBird, a modified version of fish/mammal universal primers) for metabarcoding avian eDNA. We confirmed the versatility of MiBird primers by performing in silico analyses and by amplifying DNAs extracted from bird tissues. Analyses of water samples from zoo cages of birds with known species composition suggested that the use of MiBird primers combined with Illumina MiSeq could successfully detect avian species from water samples. Additionally, analysis of water samples collected from a natural pond detected five avian species common to the sampling areas. The present findings suggest that avian eDNA metabarcoding would be a complementary detection/identification tool in cases where visual census of bird species is difficult.
Day K., Campbell H., Fisher A., Gibb K., Hill B., Rose A., Jarman S. N. (2019): Development and validation of an environmental DNA test for the endangered Gouldian finch. Endangered Species Research 40: 171-182.
FULL TEXT
Abstract
Detecting animals by identifying their DNA in water is a valuable tool for locating and monitoring species that are difficult to detect through other survey techniques. We developed a test for detecting the endangered Gouldian finch Erythrura gouldiae, a small bird endemic to northern Australia. Only 1 previous study has reported an environmental DNA (eDNA) test that unequivocally identifies a bird species using the water bodies from which they drink. In controlled aviary trials with a pair of Gouldian finches, first detection in 200 ml of water occurred after as little as 6 h, but the detection rate was higher at 30 h. DNA persisted in water exposed to the sun for <12 h and in the shade for 12 h. In trials with 55 finches, persistence was up to 144 h. The eDNA test for finches and the Gouldian finch-specific test were positive for waterholes where Gouldian and other finch species were observed each morning over 3 d. Importantly, where no Gouldian finches were observed for up to 72 h prior to water sampling, the Gouldian test was negative. Where other species of finch but no Gouldian finch were observed and counted, the finch test was positive, but the Gouldian finch test was negative. This approach could be developed for broad-scale monitoring of this endangered species, and potentially applied to a much broader range of terrestrial species that shed DNA into water bodies.
Stowell D., Wood M. D., Pamuła H., Stylianou Y., Glotin H. (2019): Automatic acoustic detection of birds through deep learning: the first bird audio detection challenge. Methods in Ecology and Evolution 10: 368-380.
FULL TEXT
Abstract
Assessing the presence and abundance of birds is important for monitoring specific species as well as overall ecosystem health. Many birds are most readily detected by their sounds, and thus, passive acoustic monitoring is highly appropriate. Yet acoustic monitoring is often held back by practical limitations such as the need for manual configuration, reliance on example sound libraries, low accuracy, low robustness, and limited ability to generalise to novel acoustic conditions. Here, we report outcomes from a collaborative data challenge. We present new acoustic monitoring datasets, summarise the machine learning techniques proposed by challenge teams, conduct detailed performance evaluation, and discuss how such approaches to detection can be integrated into remote monitoring projects. Multiple methods were able to attain performance of around 88% area under the receiver operating characteristic (ROC) curve (AUC), much higher performance than previous general-purpose methods. With modern machine learning, including deep learning, general-purpose acoustic bird detection can achieve very high retrieval rates in remote monitoring data, with no manual recalibration, and no pretraining of the detector for the target species or the acoustic conditions in the target environment.
Schütz R., Tollrian R., Schweinsberg M. (2020): A novel environmental DNA detection approach for the wading birds Platalea leucorodia, Recurvirostra avosetta and Tringa totanus. Conservation Genetics Resources 12: 529-531.
FULL TEXT
Abstract
Wading birds play an important role in coastal and wetland ecosystems. As a result of anthropogenic disturbances, numbers of wading birds have declined over the past years, and therefore, the monitoring of bird species is crucial to preserve their habitats and counteract this. Due to known limitations of conventional monitoring, molecular approaches are increasingly becoming complementary methods. Thus, we evaluated the potential of an environmental DNA (eDNA) monitoring approach for wading birds. We developed three species-specific primer sets for wading birds targeting the COI region. After examining the primers in silico and on extracted bird DNA, the primers were used to successfully amplify eDNA from water samples.
Kahl S., Wood C. M., Eibl M., Klinck H. (2021): BirdNET: A deep learning solution for avian diversity monitoring. Ecological Informatics 61: 101236.
FULL TEXT
Abstract
Variation in avian diversity in space and time is commonly used as a metric to assess environmental changes. Conventionally, such data were collected by expert observers, but passively collected acoustic data is rapidly emerging as an alternative survey technique. However, efficiently extracting accurate species richness data from large audio datasets has proven challenging. Recent advances in deep artificial neural networks (DNNs) have transformed the field of machine learning, frequently outperforming traditional signal processing techniques in the domain of acoustic event detection and classification. We developed a DNN, called BirdNET, capable of identifying 984 North American and European bird species by sound. Our task-specific model architecture was derived from the family of residual networks (ResNets), consisted of 157 layers with more than 27 million parameters, and was trained using extensive data pre-processing, augmentation, and mixup. We tested the model against three independent datasets: (a) 22,960 single-species recordings; (b) 286 h of fully annotated soundscape data collected by an array of autonomous recording units in a design analogous to what researchers might use to measure avian diversity in a field setting; and (c) 33,670 h of soundscape data from a single high-quality omnidirectional microphone deployed near four eBird hotspots frequented by expert birders. We found that domain-specific data augmentation is key to build models that are robust against high ambient noise levels and can cope with overlapping vocalizations. Task-specific model designs and training regimes for audio event recognition perform on-par with very complex architectures used in other domains (e.g., object detection in images). We also found that high temporal resolution of input spectrograms (short FFT window length) improves the classification performance for bird sounds. In summary, BirdNET achieved a mean average precision of 0.791 for single-species recordings, a F0.5 score of 0.414 for annotated soundscapes, and an average correlation of 0.251 with hotspot observation across 121 species and 4 years of audio data. By enabling the efficient extraction of the vocalizations of many hundreds of bird species from potentially vast amounts of audio data, BirdNET and similar tools have the potential to add tremendous value to existing and future passively collected audio datasets and may transform the field of avian ecology and conservation.
Shewring M. P., Vafidis J. O. (2021): Using UAV‐mounted thermal cameras to detect the presence of nesting nightjar in upland clear‐fell: A case study in South Wales, UK. Ecological Solutions and Evidence 2: e12052.
FULL TEXT
Abstract
Confirming the presence and location of European Nightjar Caprimulgus europaeus nests is a significant fieldwork challenge in ecological monitoring. Nest sites can be located through direct observation or capture and radio tracking of breeding individuals; however, such work is time consuming, disturbing and costly. Unmanned aerial vehicles (UAV) equipped with thermal sensors may enable rapid survey over large areas by detecting nest locations based on the contrast of relatively warm nests and the surrounding cooler ground. The application of this concept using UAV-mounted thermal sensors was trialled in two upland clear-fell forestry sites in South Wales, UK. Detection trials were undertaken at five known nightjar nest sites to assess optimal timing and flight height for surveys. Nest heat signatures were clear during dusk and dawn, but not during the daytime. Nests were identifiable at flight heights up to 25 m, but flight heights of 12–20 m were optimal for the numbers of pixels per nest. This approach was tested in a field trial of a 17-ha forestry site where the presence and position of nesting nightjars were unknown. An automated transect at dusk and dawn at 15 m flight elevation identified two active nightjar nests and four male nightjar roost sites. Without image analysis automation, the process of manual inspection of 2607 images for ‘hotspots’ of the approximate size and shape of nightjar nests was laborious. The UAV approach took around 18 h including survey time, processing and ground verification, whilst a nightjar nest finding survey would take 35 h for the same area. The small size of nightjars and the low resolution of the thermal sensors requires low altitude flight in order to maximize detectability and pixel coverage. Low flight elevation requires more consideration of the risk of collision with trees or posts. Consequently, the approach would not be suitable for covering areas of highly variable terrain.
Folliot A., Haupert S., Ducrettet M., Sèbe F., Sueur J. (2022): Using acoustics and artificial intelligence to monitor pollination by insects and tree use by woodpeckers. Science of the Total Environment 838: 155883.
FULL TEXT
Abstract
The collection and interpretation of field data is a prerequisite for informed conservation in protected environments. Although several techniques, including camera trapping and passive acoustic monitoring, have been developed to estimate the presence of animal species, very few attempts have been made to monitor ecological functions. Pollination by insects and wood use, including tree related foraging and intraspecific communication, by woodpeckers are key functions that need to be assessed in order to better understand and preserve forest ecosystems within the context of climate change. Here, we developed and applied for the first time an acoustic survey to monitor pollination by insects and wood use by woodpeckers in a protected Alpine forest in France. We deployed four autonomous recorders over a year, resulting in 2285 h of recordings. We trained a convolutional neural network (CNN) on spectrographic images to automatically detect the sounds of flying insects’ buzzing and woodpeckers’ drumming as they forage and call. We used the output of the CNN to estimate the seasonality, diel pattern, climatic breadth and distribution of both functions and their relationships with weather parameters. Our method showed that insects were flying (therefore potentially pollinating flowers) in bright, warm and dry conditions, after dawn and before dusk during spring and summer. Woodpeckers were mainly drumming around March at the time of pair formation in cool and wet conditions. Having considered the role of weather parameters, climate change might have contrasting effects on insect buzzing and woodpecker drumming, with an increase in temperature being favorable to pollination by insects but not to wood use by woodpeckers, and a concomitant increase in relative humidity being favorable to wood use but not to pollination. This study reveals that a systemic facet of biodiversity can be tracked using sound, and that acoustics provide valuable information for the environment description.
Fukuhara R., Agarie J., Furugen M., Seki H. (2022): Nesting habitats of free-ranging Indian peafowl, Pavo cristatus, revealed by sniffer dogs in Okinawa, Japan. Applied Animal Behaviour Science 249: 105605.
FULL TEXT
Abstract
The Indian peafowl, Pavo cristatus, is the largest of the pheasants and ground-nesting bird. Because this species was introduced in Okinawa, Japan about 40 years ago and is now increasingly recognized as a pest in the Yaeyama region, and the extermination of this species has been promoted for ecosystem conservation. It was thought that there is need to develop a method to reduce the peafowl population and dogs were trained to detect them. We faced two difficulties in developing sniffer dogs to detect peafowl nests as below; (i) it is difficult to obtain a sufficient number of eggs to train the detection dogs, (ii) another difficulty is that the nesting period of peahens and the preferred vegetation used for nesting in Okinawa have not been reported. At first, we screened and trained two dogs of Welsh Corgi Pembroke and one dog of Brittany Spaniel for peafowl nest detection dogs with peahen feathers instead of peahen eggs. These dogs demonstrated a 100% sensitivity rate and there were no false-positives of their precision test. In second step, display survey of peacock to predict the nesting period of peahen was conducted in Kohama (KHM) and Kuroshima (KRS) islands, and the earliest dates of trail displays were observed in late February between 2014 and 2019 in both Islands. We predicted that the eggs would begin hatching in early April. We surveyed peahen nests for a total of 837 h and 1153 km on 962 transects with sniffer dogs and detected 423 peahen’s nests in KHM and KRS islands between 2014 and 2019. Peahens made her nests mainly in Poaceae or Asteraceae plants in research areas. The peak hatching was between mid-April and early May. In this study, we were able to detect peahen nests in the field by dogs trained with peahen feather, and we also identified the exact nesting season and preferred vegetation in Okinawa to control of Indian peafowl as an invasive species.
Hane M., Thornton-Frost J., Springford A., Kroll A. (2022): Factors associated with automated detection of Northern Spotted Owl (Strix occidentalis caurina) four-note location calls. Avian Conservation and Ecology 17: 26.
FULL TEXT
Abstract
Automated signal detection of passive acoustic data produces enormous amounts of data that requires efficient processing. Furthermore, processed data requires assessment to ensure correct categorization of sounds to match field observations. Failure to compare data directly may lead to inaccurate estimates of occupancy state or population status and contribute to sub-optimal management decisions. We evaluated three automated detection methods for Northern Spotted Owl (Strix occidentalis caurina) four-note location calls at two sites representing vegetational and topographic conditions common to Northern Spotted Owl sites in western Oregon, United States. Our results indicated that the detection distance, resulting areal coverage, and occupancy status all varied with site, call broadcast direction, and software used for analysis. A machine learning algorithm (convolutional neural network) built specifically for detection of Northern Spotted Owl performed better at determining occupancy than two commercially developed software packages. At distances less than 250 m, the convolutional neural network correctly identified occupancy in more than 73% of trials and both commercial methods correctly identified occupancy in less than 60% of trials. Areal coverage was a function of distance from source to microphone, location of the source relative to the microphone, and method of call analysis. Calls broadcast toward the microphones were more likely to be detected than calls broadcast away from the microphones. Our results, although limited in scope, suggest that detection distance merits extended evaluation before autonomous recording units are deployed broadly as replacements for human observers.
Huang P. Y., Poon E. S. K., Chan L. Y., Chan D. T. C., Huynh S., So I. W. Y., Sung Y. H., Sin S. Y. W. (2022): Dietary diversity of multiple shorebird species in an Asian subtropical wetland unveiled by DNA metabarcoding. Environmental DNA 4: 1381-1396.
FULL TEXT
Abstract
Global declines in shorebird populations resulting from foraging habitat loss have been recently reported, and the situation within the East Asian-Australasian Flyway (EAAF) is particularly concerning. Despite previous studies that analyzed the foraging niches of shorebirds worldwide, the dietary niche dynamics of shorebirds coexisting in Asia are very poorly understood. This study is therefore among the early few that aim to unveil the trophic organization of shorebirds in a subtropical wetland within the EAAF which is vital for species conservation. Our study first determined the dietary spectra of more than 10 shorebird species, such as Calidris ferruginea (near threatened), Charadrius leschenaultii, and Pluvialis squatarola, by applying DNA metabarcoding with 18S and COI markers to fecal DNA. The diet of Tringa stagnatilis was also characterized, which was previously undescribed. Shorebirds that occurred in the wetland consumed a variety of food items, primarily a high abundance of malacostracans, mollusks, annelids, insects, and some arachnids. Different proportions of plant materials were also detected in many shorebird species. Using the data, we then revealed clear patterns of inter- and intraspecific variations between these shorebirds. Importantly, we specifically compared the similarities of the spring diets among seven sympatric shorebird species. We found that the dietary compositions of the seven species have segregated from each other to varying levels, but the many similar taxa we identified in the diets among these shorebirds imply that these populations of shorebirds could be competing at different levels. Thus, any reductions in the abundance and diversity of these important food resources would likely intensify their inter- and intraspecific competition, and simultaneously threaten the survival of multiple species. With these findings, conservation measures must be taken to protect and monitor the vital food resources for these energy-deprived shorebirds during migration.
Kim J. I., Baek J. W., Kim C. B. (2022): Image classification of Amazon parrots by deep learning: A potentially useful tool for wildlife conservation. Biology 11: 1303.
FULL TEXT
Abstract
Parrots play a crucial role in the ecosystem by performing various roles, such as consuming the reproductive structures of plants and dispersing plant seeds. However, most are threatened because of habitat loss and commercial trade. Amazon parrots are one of the most traded and illegally traded parrots. Therefore, monitoring their wild populations and global trade is crucial for their conservation. However, monitoring wild populations is becoming more challenging because the manual analysis of large-scale datasets of images obtained from camera trap methods is labor-intensive and time consuming. Monitoring the wildlife trade is difficult because of the large quantities of wildlife trade. Amazon parrots can be difficult to identify because of their morphological similarity. Object detection models have been widely used for automatic and accurate species classification. In this study, to classify 26 Amazon parrot species, 8 Single Shot MultiBox Detector models were assessed. Among the eight models, the DenseNet121 model showed the highest mean average precision at 88.9%. This model classified the 26 Amazon parrot species at 90.7% on average. Continuous improvement of deep learning models classifying Amazon parrots may support monitoring wild populations and the global trade of these species.
Ribeiro Jr J. W., Harmon K., Leite G. A., de Melo T. N., LeBien J., Campos-Cerqueira M. (2022): Passive acoustic monitoring as a tool to investigate the spatial distribution of invasive alien species. Remote Sensing 14: 4565.
FULL TEXT
Abstract
Invasive alien species (IAS) are a threat to biodiversity and ecosystem function worldwide. Unfortunately, researchers, agencies, and other management groups face the unresolved challenge of effectively detecting and monitoring IAS at large spatial and temporal scales. To improve the detection of soniferous IAS, we introduced a pipeline for large-scale passive acoustic monitoring (PAM). Our main goal was to illustrate how PAM can be used to rapidly provide baseline information on soniferous IAS. To that aim, we collected acoustic data across Puerto Rico from March to June 2021 and used single-species occupancy models to investigate species distribution of species in the archipelago and to assess the peak of vocal activity. Overall, we detected 16 IAS (10 birds, 3 mammals, and 3 frogs) and 79 native species in an extensive data set with 1,773,287 1-min recordings. Avian activity peaked early in the morning (between 5 a.m. and 7 a.m.), while amphibians peaked between 1 a.m. and 5 a.m. Occupancy probability for IAS in Puerto Rico ranged from 0.002 to 0.67. In general, elevation and forest cover older than 54 years were negatively associated with IAS occupancy, corroborating our expectation that IAS occurrence is related to high levels of human disturbance and present higher occupancy probabilities in places characterized by more intense human activities. The work presented here demonstrates that PAM is a workable solution for monitoring vocally active IAS over a large area and provides a reproducible workflow that can be extended to allow for continued monitoring over longer timeframes.
Weinstein B. G., Garner L., Saccomanno V. R., Steinkraus A., Ortega A., Brush K., Yenni G., McKellar A. E., Converse R., Lippitt C. D., Wegmann A. (2022): A general deep learning model for bird detection in high‐resolution airborne imagery. Ecological Applications 32: e2694.
FULL TEXT
Abstract
Advances in artificial intelligence for computer vision hold great promise for increasing the scales at which ecological systems can be studied. The distribution and behavior of individuals is central to ecology, and computer vision using deep neural networks can learn to detect individual objects in imagery. However, developing supervised models for ecological monitoring is challenging because it requires large amounts of human-labeled training data, requires advanced technical expertise and computational infrastructure, and is prone to overfitting. This limits application across space and time. One solution is developing generalized models that can be applied across species and ecosystems. Using over 250,000 annotations from 13 projects from around the world, we develop a general bird detection model that achieves over 65% recall and 50% precision on novel aerial data without any local training despite differences in species, habitat, and imaging methodology. Fine-tuning this model with only 1000 local annotations increases these values to an average of 84% recall and 69% precision by building on the general features learned from other data sources. Retraining from the general model improves local predictions even when moderately large annotation sets are available and makes model training faster and more stable. Our results demonstrate that general models for detecting broad classes of organisms using airborne imagery are achievable. These models can reduce the effort, expertise, and computational resources necessary for automating the detection of individual organisms across large scales, helping to transform the scale of data collection in ecology and the questions that can be addressed.
Jønsson K. A., Thomassen E. E., Iova B., Sam K., Thomsen P. F. (2023): Using environmental DNA to investigate avian interactions with flowering plants. Environmental DNA 5: 462-475.
FULL TEXT
Abstract
Animal pollination is an important and highly valued ecosystem function and the role of birds as pollinators is increasingly acknowledged. However, such interactions can be challenging to document and often require extensive field programs. Over the last decade, environmental DNA (eDNA) has been analyzed from several different contemporary sample types, such as water, soil, flowers, and air. The applications of these studies include biodiversity monitoring, detection of endangered species, community compositions, and more recently, flower–arthropod interactions. However, it remains unknown whether flower eDNA is applicable to other taxonomic groups interacting with plants, as well as the deposition and degradation of eDNA on flowers. Here, we test whether eDNA from flowers can be used for detecting bird pollinators. In a controlled environment (an aviary with great tits [Parus major]), we show that birds leave significant traces of DNA on the flowers without observed visits (airborne eDNA). We further show that when birds had been in contact with the flowers, DNA concentrations increased to levels significantly higher than airborne background DNA. Subsequently, we sampled five clusters of wild flowers in Papua New Guinea and detected four species of birds, two of which are nectar-feeders, and one that is an insectivorous species known to visit flowers. These four bird species were regularly seen in the area and caught in mist-nets in the days prior to sampling of the flowers. In total, 29 bird species were recorded (18 mist-netted) in the area and of these, eight are nectarivorous. Our quantitative approach suggests that it is possible to distinguish airborne background bird DNA deposited on flowers from actual flower visits of birds in the wild, although this might be highly context-specific. Our findings are of broad interest within research on ecosystem functioning, biotic interactions, and plant–animal mutualism.
Newton J. P., Bateman P. W., Heydenrych M. J., Kestel J. H., Dixon K. W., Prendergast K. S., White N. E., Nevill P. (2023): Monitoring the birds and the bees: Environmental DNA metabarcoding of flowers detects plant–animal interactions. Environmental DNA 5: 488-502.
FULL TEXT
Abstract
Animal pollinators are vital for the reproduction of ~90% of flowering plants. However, many of these pollinating species are experiencing declines globally, making effective pollinator monitoring methods more important than ever before. Pollinators can leave DNA on the flowers they visit, and metabarcoding of these environmental DNA (eDNA) traces provides an opportunity to detect the presence of flower visitors. Our study, collecting flowers from seven plant species with diverse floral morphologies, for eDNA metabarcoding analysis, illustrated the value of this novel survey tool. eDNA metabarcoding using three assays, including one developed in this study to target common bush birds, recorded more animal species visiting flowers than visual surveys conducted concurrently, including birds, bees, and other species. We also recorded the presence of a flower visit from a western pygmy possum; to our knowledge this is the first eDNA metabarcoding study to simultaneously identify the interaction of insect, mammal, and bird species with flowers. The highest diversity of taxa was detected on large inflorescence flower types found on Banksia arborea and Grevillea georgeana. The study demonstrates that the ease of sample collection and the robustness of the metabarcoding methodology has profound implications for future management of biodiversity, allowing us to monitor both plants and their attendant cohort of potential pollinators. This opens avenues for rapid and efficient comparison of biodiversity and ecosystem health between different sites and may provide insights into surrogate pollinators in the event of pollinator declines.