Bird Detection Research Articles

Point count offsets for estimating population sizes of north American landbirds

Bird monitoring in North America over several decades has generated many open databases, housing millions of structured and semi‐structured bird observations. These provide the opportunity to estimate bird densities and population sizes, once variation in factors such as underlying field methods, timing, land cover, proximity to roads, and uneven spatial coverage are accounted for. To facilitate integration across databases, we introduce NA‐POPS: Point Count Offsets for Population Sizes of North American Landbirds. NA‐POPS is a large‐scale, multi‐agency project providing an open‐source database of detectability functions for all North American landbirds. These detectability functions allow the integration of data from across disparate survey methods using the QPAD approach, which considers the probability of detection (q) and availability (p) of birds in relation to area (a) and density (d). To date, NA‐POPS has compiled over 7.1 million data points spanning 292 projects from across North America, and produced detectability functions for 338 landbird species. Here, we describe the methods used to curate these data and generate these detectability functions, as well as the open‐access nature of the resulting database.

Answers blowing in the wind: Detection of birds, mammals, and amphibians with airborne environmental DNA in a natural environment over a yearlong survey

AbstractAnalysis of environmental DNA (eDNA) from passively collected airborne dust has demonstrated broad success for sensitive and robust detection of plants. Recent experiments at small spatial scales have suggested that animals can also be detected using airborne eDNA. However, airborne eDNA analysis has never been used for a long‐term whole‐community assessment of a natural terrestrial community or with passive dust collectors. We conducted a metabarcoding survey targeting vertebrate eDNA from dust carried in the air on an approximately 130‐acre shortgrass prairie passively collected over the course of a year. Our survey detected a wide variety of animal forms including an amphibian species, several bird species, and both small and large mammals. We found that airborne eDNA signals changed with known patterns of animal activity, wind speed, and rainfall. Overall, we demonstrate that passively collected airborne dust carries eDNA from terrestrial animals and could be used to detect a wide variety of terrestrial vertebrate species in a natural environment with minimal effort. To develop this as a valuable monitoring tool, research needs to focus on the ecology of eDNA carried in the air, which includes the origin, state, transport, dispersal, and fate of eDNA in the environment.

Preliminary assessment of thermal imaging equipped aerial drones for secretive marsh bird detection

Rails are a highly secretive group of marshland obligate species that are difficult to consistently survey and detect. Current survey efforts utilize either call–playback or autonomous recording devices, but the low detection probabilities for rails create challenges for long-term systematic monitoring. Between 8 April and 16 May 2022, we flew a small aerial drone equipped with a thermal camera to survey for six species of rail (Black Rail ( Laterallus jamaicensis), Yellow Rail ( Coturnicops noveboracensis), Sora ( Porzana carolina), Virginia Rail ( Rallus limicola), Clapper Rail ( Rallus crepitans), and King Rail ( Rallus elegans)) along the Gulf Coast of Texas to assess the feasibility of long-term drone monitoring. We successfully conducted 33 flights at 15 m above ground level and detected rails on the first visit at 42% of known occupied points. We achieved 27 total rail detections, including 12 Black Rail/Yellow Rail detections. Of the birds detected, 81% exhibited no response to the drone’s first approach. We intend for this preliminary data to shape future survey protocol for secretive species occupying difficult to navigate terrain.

Bird Detection and Species Classification: Using YOLOv5 and Deep Transfer Learning Models

Bird Detection and Species Classification: Using YOLOv5 and Deep Transfer Learning Models

Using Breeding Bird Survey and eBird data to improve marsh bird monitoring abundance indices and trends

The elusive nature of many marsh-breeding birds presents a challenge for effective population monitoring. The Great Lakes Marsh Monitoring Program (GLMMP), delivered by Birds Canada, addressed these challenges by concentrating survey efforts in marsh bird habitats and by using survey protocols aimed at maximizing marsh bird detections. GLMMP data suggest that numerous marsh bird species are declining. Here we consider the value of other avian monitoring programs to support our understanding of marsh bird population trends. Our goal was to compare the GLMMP, North American Breeding Bird Survey (BBS), and eBird with each other and with a combined survey, by evaluating frequency of detection, annual indices of abundance, and trend estimates. Using 23 years (1997–2019) of GLMMP, BBS, and eBird data, we calculated annual indices of abundance and trends for each survey for 18 marsh-breeding species across southern Ontario, Canada. We found that the GLMMP had more frequent detections, greater counts, and/or more precise trends for 8 species that breed almost exclusively in marshes, whereas 10 species with more variable habitat preferences had more frequent detections, greater counts, and/or more precise trends based on eBird and/or BBS. We found that combining counts from the GLMMP, BBS, and eBird increased the precision around trend estimates for 11/18 (61%) species; however, trend estimates for combined data tended to be positively biased relative to GLMMP trends for species that also frequent non-marsh habitats. We, therefore, provide evidence that combining citizen science data from multiple sources could increase the power to detect changes in marsh-dependent bird populations. Integrated datasets thus provide a promising avenue for future marsh bird conservation and management.

Solving the problem of biodiversity analysis of bird detection and classification in the video stream of camera traps

The work is devoted to solving the problem of assessing the comparative efficiency of several common architectures of convolutional neural networks for monitoring birds in a natural environment. The problem was solved by detecting birds recorded by video traps installed on feeders in several regions of Panama by different architectures. Then a comparison was made between the recognition quality metrics – IoU and mAP, and based on the values of the metrics, a conclusion was made about the effectiveness of the architectures. Experiments have shown that the YOLO architecture of the Tiny version with comparative modules wins in the accuracy table. In the future, it is planned to improve the application of neural network architectures by finalizing the dataset with the involvement of expert bird watchers and open ornithological ontologies.

Automatic Wild Bird Repellent System that is Based on Deep-Learning-Based Wild Bird Detection and Integrated with a Laser Rotation Mechanism

Automatic Wild Bird Repellent System that is Based on Deep-Learning-Based Wild Bird Detection and Integrated with a Laser Rotation Mechanism

To improve existing marsh bird survey protocols, we need to evaluate closure assumptions

AbstractOur collective understanding of secretive marsh birds has increased in the past decades due to the development and implementation of the North American Standardized Marsh Bird Monitoring Protocol (hereafter, Protocol). The Protocol proposes call broadcast surveys to increase vocalization and detection rates within 3 standardized survey periods aimed at surveying peak breeding activity for a suite of secretive marsh birds. We noted a trend in the literature linking occupancy modeling with the survey design from the Protocol, despite some evidence that vocalizations decline across survey periods, which could indicate lack of population closure. An underlying assumption of occupancy modeling is closure, and the Protocol was designed to focus on only birds which will remain in an area throughout the breeding season and not migrants that may only be present in the first survey period. Including migrating marsh birds, especially if a large percentage of marsh bird detections are migrants, can bias occupancy estimates and lead to erroneous density and population size estimates that may affect conclusions about habitat resource and bird associations. We urge researchers and managers to carefully consider the analytical and field techniques when designing studies for marsh birds and to not simply pair the biweekly survey design within the Protocol with occupancy modeling and ignore closure assumptions, turnover rates, and potential differences in resource use by migrating and breeding marsh birds. Specifically, we suggest that researchers consider short survey interval times (e.g., ~2 days rather than 2 weeks) or continuous call monitoring using automated recorded devices deployed for 1–2 weeks per survey location when using occupancy analysis on secretive marsh birds. We also call for future study of turnover rates, stopover duration, and vocalization rates during migration and breeding periods to better inform study designs and increase the appropriateness of statistical analysis.

Seasonality modulates habitat cover effects on avian cross‐boundary responses and spillover

Species cross‐boundary response is a key mechanism affecting species spillover into agricultural fields. However, temporal changes in edge permeability, which may depend on the seasonal availability of resources in both matrix and native habitats, remain poorly understood. Here we tested how edge crossing behavior and the associated spillover of birds into sun coffee plantations respond to landscape structure and seasonality. We monitored the movement of six insectivorous understory bird species (four forest specialists and two forest generalists) using an automated telemetry system along a gradient of forest cover (7–60%) during two seasons (dry versus wet) at nine sampling sites at the Brazilian Atlantic forest. We monitored 116 individuals and obtained a total of 15 129 bird detections across seasons. Bird resistance to crossing edges was strongly driven by an interaction between seasonality and forest cover in the surrounding landscape, with higher resistance to crossing edges along the dry season and in landscapes with lower amount of forest cover. Furthermore, spillover patterns in plantations were driven by an interaction between forest cover and distance from forest edges, but this was most pronounced for forest‐specialist bird species. Forest specialists moved more intensively and farther from edges in more forested landscapes, whereas forest generalists showed similar patterns of occupation regardless of forest cover and isolation. Our study contributes to a better understanding of avian cross‐boundary responses and spillover in response to landscape structure across seasons, and the factors driving bird movement decisions in anthropogenic landscapes. There are a myriad of possible mechanisms governing movement decisions, and these mechanisms may interact in complex ways and remain important foci for research within the fields of tropical ecology and evolution.

Detection of Endemic Sri Lankan Birds Using AI-Based Software

Detection of Endemic Sri Lankan Birds Using AI-Based Software

Design and Development of an Energy-efficient Audio-based Repellent System for Rice Fields

Bird infestation is one of the major limiting factors affecting the production of indigenous rice in Nigeria. Conventional audio-based systems used in repelling birds deteriorate in effectiveness, cause noise pollution and consume excess power. This research therefore, developed an energyefficient audio-based repellent system for rice fields incorporated with a convolutional neural network (CNN) model for bird detection. A CNN algorithm was developed using inception V3, built on transfer learning technique in google colab. The CNN model was trained with bird dataset of 275 bird species obtained from Kaggle database. The core of the system was the Raspberry pi 4, a low-powered microcomputer adopted to run the CNN model with information from a camera. The mechanical framework for the system was designed to support its sustainability in rice fields. The estimated coverage area was 42m2. The system used maximum power of 4.10W. The system supported multiple production of various sounds to prevent bird adaptability to fixed sounds, which could sustain its effectiveness. The incorporated camera aided in the detection of birds with consequent reduction in noise pollution inherent in conventional audio devices. The developed system used 53.52% of the average power reported for conventional audio systems. The developed audio-based repellent system is better in terms of bird detection, power consumption and less noise pollution. Further development on this system can be carried out with other detection algorithms which may require less power for processing. Keywords— Bird repellent system, convolutional neural network, audio repellent, rice field, energy-efficient

Large-scale distribution models for optimal prediction of Eastern black rail habitat within tidal ecosystems

Eastern black rails (Laterallus jamaicensis jamaicensis) are among the rarest and least-studied birds in North America and were recently listed as threatened under the U.S. Endangered Species Act. Spatial models that predict habitat quality across the subspecies range are therefore needed to inform conservation, recovery, and monitoring efforts for this rare bird. We used data from 47,585 call-broadcast surveys collected at 7906 sites over a 3-decade period (1990s, 2000s, 2010s; 23 total years) to build species distribution models for eastern black rails. We used hierarchical Bayesian occupancy models and predictive model selection to develop multi-scale models that optimally predict habitat suitability for eastern black rails within tidal wetlands while also accounting for imperfect detection of these cryptic birds during field surveys. We also used raster regression techniques to translate model predictions into 30-m resolution maps of habitat suitability for eastern black rails within tidal wetlands along the eastern seaboard of the United States. The model predicted suitability of breeding habitat as a function of wetland attributes (e.g., cover of high marsh and terrestrial border), hydrologic modification, and disturbance from human development measured over multiple spatial scales. We also found differences in habitat relationships for eastern black rails when compared to models that included both North American subspecies of black rail. Important results included negative effects of shrub-scrub wetlands, and strong positive effects of high marsh, terrestrial border, and impoundments on breeding season occupancy. Our study provides an example of integrating detection-non-detection data and modern statistical methods to build predictive distribution models for an extremely rare species, while also providing rigorous predictions of breeding habitat quality for the eastern black rail within tidal wetlands. These models will facilitate optimal monitoring, habitat conservation, and recovery planning efforts for eastern black rails and provide a foundation for future research and conservation of this imperiled bird.

Drivers of avian habitat use and detection of backyard birds in the Pacific Northwest during COVID-19 pandemic lockdowns

Birds living in developed areas contend with numerous stressors, including human disturbance and light, noise, and air pollution. COVID-19 pandemic lockdowns presented a unique opportunity to disentangle these effects during a period of reduced human activity. We launched a community science project in spring 2020 to explore drivers of site use by and detection of common birds in cities under lockdown in the U.S. Pacific Northwest. Our goals were twofold: (1) consider how intensity of urbanization, canopy cover, and availability of bird feeders and bird baths influenced avian habitat use; and (2) quantify how daily changes in weather, air pollution, and human mobility influenced detection of birds. We analyzed 6,640 surveys from 367 volunteers at 429 monitoring sites using occupancy models for 46 study species. Neither land cover nor canopy cover influenced site use by 50% of study species, suggesting that backyard birds may have used a wider range of habitats during lockdowns. Human mobility affected detection of 76% of study species, suggesting that birds exhibited species-specific behavioral responses to day-to-day changes in human activity beginning shortly after initial lockdown restrictions were implemented. Our study also showcases how existing community science platforms can be leveraged to support local monitoring efforts.

A general deep learning model for bird detection in high-resolution airborne imagery.

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.

Adulteration Detection of Edible Bird's Nests Using Rapid Spectroscopic Techniques Coupled with Multi-Class Discriminant Analysis.

Edible bird’s nests (EBNs) are vulnerable to adulteration due to their huge demand for traditional medicine and high market price. Presently, there are pressing needs to explore field-deployable rapid screening techniques to detect adulteration of EBNs. The objective of this study is to explore the feasibility of using a handheld near-infrared (VIS/SW-NIR) spectroscopic device for the determination of EBN authenticity against the benchmark performance of a benchtop mid-infrared (MIR) spectrometer. Forty-nine authentic EBNs from the different states in Malaysia and 13 different adulterants (five types) were obtained and used to simulate the adulteration of EBNs at 1, 5 and 10% adulteration by mass (a total of 15 adulterated samples). The VIS/SW-NIR and MIR spectra collated were subsequently processed, modelled and classified using multi-class discriminant analysis. The VIS/SW-NIR results showed 100% correct classification for the collagen and nutrient agar classes in authenticity classification, while for the other classes, the lowest correct classification rate was 96.3%. For MIR analysis, only the karaya gum class had 100% correct classification whilst for the other four classes, the lowest rate of correct classification was at 94.4%. In conclusion, the combination of spectroscopic analysis with chemometrics can be a powerful screening tool to detect EBN adulteration.

Detectability of birds under different sampling efforts and during the breeding season: a case study from Central Europe

During the years 2019 and 2020, I conducted a bird survey transect in the Bohemian Forest. I did not record any changes in habitat structure or weather conditions between the two years. The two surveys differed in sampling effort, which was significantly lower in 2020 (n = 5 visits) than 2019 (n = 14 visits). I found that sampling effort affected the assessment of avian community diversity but did not affect the total number of individuals recorded. I also recorded a similar pattern in the cumulative number of species between the two breeding seasons, but 80% of species were recorded ten days earlier with the higher sampling effort. In the year with the lower sampling effort, I recorded fewer species than in the year with higher sampling effort. In both study periods, avian community diversity peaked during May and June. These results suggest that even a sampling effort three times lower is still sufficient to detect most species if the minimal number of visits are conducted. The pattern of detectability during the breeding season differed significantly among species. Most species (n = 24) showed a decreasing linear detectability throughout the summer months (e.g. Turdidae or Muscicapidae), most probably due to their breeding activities. In two species (willow tit Poecile montanus and European goldfinch Carduelis carduelis), this linear relationship was reversed, probably due to singing of young birds from the previous breeding season and the effect of the autumn equinox on birdsong activity. Many species (n = 21) did not show any trend and the rest, mainly migratory species, showed non-linear relationships with the peak in the middle of the breeding season. The differences in trends of detectability (i.e. song activity) among bird species are therefore directly linked with their life history.

West Nile Virus Occurrence and Ecological Niche Modeling in Wild Bird Species and Mosquito Vectors: An Active Surveillance Program in the Peloponnese Region of Greece.

West Nile Virus (WNV) is maintained in nature in a bird-mosquito cycle and human infections follow a seasonal pattern, favored by climatic conditions. Peloponnese Region, located in Southern Greece, initiated an active WNV surveillance program to protect public health during 2019–2020. The project included monitoring of avian hosts and mosquito vectors, while sampling locations were prioritized after consideration of WNV circulation in birds, mosquitos and humans during previous seasons. Biological materials were collected from 493 wild birds of 25 species and 678 mosquito pools, which were molecularly screened for WNV presence. In this case, 14 environmental variables were associated with WNV detection in wild birds and mosquitos by using two separate MaxEnt models. Viral RNA was not detected in the target species during 2019, although in 2020, it was reported on 46 wild birds of ten species and 22 mosquito pools (Culex pipiens and Aedes albopictus). Altitude and land uses were significant predictors for both models and in fact, suitable conditions for virus occurrence were identified in low altitude zones. Bird- and mosquito-based surveillance systems yielded similar results and allowed for targeted vector control applications in cases of increased virus activity. Human cases were not reported on Peloponnese in 2020.

Birds Detection in Natural Scenes Based on Improved Faster RCNN

To realize the accurate detection of small-scale birds in natural scenes, this paper proposes an improved Faster RCNN model to detect bird species. Firstly, the model uses a depth residual network to extract convolution features and performs multi-scale fusion for feature maps of different convolutional layers. Secondly, the K-means clustering algorithm is used to cluster the bounding boxes. We improve the anchoring according to the clustering results. The improved anchor frame tends toward the real bounding box of the dataset. Finally, the Soft Non-Maximum Suppression method is used to reduce the missed detection of overlapping birds. Compared with the original model, the improved model has faster effect and higher accuracy.

Infrared Bird Target Detection Based on Temporal Variation Filtering and a Gaussian Heat-Map Perception Network

Flying bird detection has recently attracted increasing attention in computer vision. However, compared to conventional object detection tasks, it is much more challenging to trap flying birds in infrared videos due to small target size, complex backgrounds, and dim shapes. In order to solve the problem of poor detection performance caused by insufficient feature information of small and dim birds, this paper suggests a method of detecting birds in outdoor environments using image pre-processing and deep learning, called temporal Variation filtering (TVF) and Gaussian heatmap perception network (GHPNet), respectively. TVF separates the dynamic background from moving creatures. Using bird appearance features that are brightest at the center and gradually darker outwards, the size-adaptive Gaussian kernel is used to generate the ground truth of the region of interest (ROI). In order to fuse the features from different scales and to highlight the saliency of the target, the GHPNet network integrates VGG-16 and maximum-no-pooling filterer into a U-Net network. The comparative experiments demonstrate that the proposed method outperforms those that are state-of-the-art in detecting bird targets in real-world infrared images.

Highly pathogenic avian influenza virus incursions of subtype H5N8, H5N5, H5N1, H5N4, and H5N3 in Germany during 2020-21.

From October 2020 to July 2021, five different subtypes (H5N8, H5N5, H5N1, H5N4, and H5N3) and seven genotypes of highly pathogenic avian influenza viruses (HPAIV) belonging to clade 2.3.4.4b were detected in a broad array of avian hosts in Germany. Initial incursion by wild birds with an unprecedented involvement of charadriiforme species at the Wadden Sea coast only carrying subtype H5N3, lateral spread between poultry with detection of novel reassortants and mixed infections in poultry holdings, suspected spillback of HPAIV from poultry to wild birds, and detection of HPAIV-infected wild birds during the following summer in 2021 were hallmarks of this epizootic. Local reassortment events with low pathogenic AIV strains were detected by phylogenetic analyses, with a dominating HP H5N8 and later HP H5N1 strain responsible for most cases. In addition, the first-ever described HPAIV strain of subtype H5N4 could be genetically characterized.