The exodus of flying animals from their roosting locations is often visible as expanding ring‐shaped patterns in weather radar data. The NEXRAD network, for example, archives more than 25 years of data across 143 contiguous US radar stations, providing opportunities to study roosting locations and times and the ecosystems of birds and bats. However, access to this information is limited by the cost of manually annotating millions of radar scans. We develop and deploy an AI‐assisted system to annotate roosts in radar data. We build datasets with roost annotations to support the training and evaluation of automated detection models. Roosts are detected, tracked, and incorporated into our developed web‐based interface for human screening to produce research‐grade annotations. We deploy the system to collect swallow and martin roost information from 12 radar stations around the Great Lakes spanning 21 years. After verifying the practical value of the system, we propose to improve the detector by incorporating both spatial and temporal channels from volumetric radar scans. The deployment on Great Lakes radar scans allows accelerated annotation of 15 628 roost signatures in 612 786 radar scans with 183.6 human screening hours, or 1.08 s per radar scan. We estimate that the deployed system reduces human annotation time by ~7×. The temporal detector model improves the average precision at intersection‐over‐union threshold 0.5 (APIoU = .50) by 8% over the previous model (48%→56%), further reducing human screening time by 2.3× in its pilot deployment. These data contain critical information about phenology and population trends of swallows and martins, aerial insectivore species experiencing acute declines, and have enabled novel research. We present error analyses, lay the groundwork for continent‐scale historical investigation about these species, and provide a starting point for automating the detection of other family‐specific phenomena in radar data, such as bat roosts and mayfly hatches.
- Award ID(s):
- 1661259
- NSF-PAR ID:
- 10162015
- Date Published:
- Journal Name:
- Proceedings of the AAAI Conference on Artificial Intelligence
- Volume:
- 34
- Issue:
- 01
- ISSN:
- 2159-5399
- Page Range / eLocation ID:
- 378 to 385
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
more » « less
Abstract -
more » « less
Abstract In this study, we combined a machine learning pipeline and human supervision to identify and label swallow and martin roost locations on data captured from 2000 to 2020 by 12 Weather Surveillance Radars in the Great Lakes region of the US. We employed radar theory to extract the number of birds in each roost detected by our technique. With these data, we set out to investigate whether roosts formed consistently in the same geographic area over two decades and whether consistency was also predictive of roost size. We used a clustering algorithm to group individual roost locations into 104 high‐density regions and extracted the number of years when each of these regions was used by birds to roost. In addition, we calculated the overall population size and analyzed the daily roost size distributions. Our results support the hypothesis that more persistent roosts are also gathering more birds, but we found that on average, most individuals congregate in roosts of smaller size. Given the concentrations and consistency of roosting of swallows and martins in specific areas throughout the Great Lakes, future changes in these patterns should be monitored because they may have important ecosystem and conservation implications.
-
more » « less
Abstract The spatial organization of populations determines their pathogen dynamics. This is particularly important for communally roosting species, whose aggregations are often driven by the spatial structure of their environment.
We develop a spatially explicit model for virus transmission within roosts of Australian tree‐dwelling bats (
Pteropus spp.), parameterized to reflect Hendra virus. The spatial structure of roosts mirrors three study sites, and viral transmission between groups of bats in trees was modelled as a function of distance between roost trees. Using three levels of tree density to reflect anthropogenic changes in bat habitats, we investigate the potential effects of recent ecological shifts in Australia on the dynamics of zoonotic viruses in reservoir hosts.We show that simulated infection dynamics in spatially structured roosts differ from that of mean‐field models for equivalently sized populations, highlighting the importance of spatial structure in disease models of gregarious taxa. Under contrasting scenarios of flying‐fox roosting structures, sparse stand structures (with fewer trees but more bats per tree) generate higher probabilities of successful outbreaks, larger and faster epidemics, and shorter virus extinction times, compared to intermediate and dense stand structures with more trees but fewer bats per tree. These observations are consistent with the greater force of infection generated by structured populations with less numerous but larger infected groups, and may flag an increased risk of pathogen spillover from these increasingly abundant roost types.
Outputs from our models contribute insights into the spread of viruses in structured animal populations, like communally roosting species, as well as specific insights into Hendra virus infection dynamics and spillover risk in a situation of changing host ecology. These insights will be relevant for modelling other zoonotic viruses in wildlife reservoir hosts in response to habitat modification and changing populations, including coronaviruses like SARS‐CoV‐2.
-
more » « less
Abstract Aircraft collisions with birds span the entire history of human aviation, including fatal collisions during some of the first powered human flights. Much effort has been expended to reduce such collisions, but increased knowledge about bird movements and species occurrence could dramatically improve decision support and proactive measures to reduce them. Migratory movements of birds pose a unique, often overlooked, threat to aviation that is particularly difficult for individual airports to monitor and predict the occurrence of birds vary extensively in space and time at the local scales of airport responses.
We use two publicly available datasets, radar data from the US NEXRAD network characterizing migration movements and eBird data collected by citizen scientists to map bird movements and species composition with low human effort expenditures but high temporal and spatial resolution relative to other large‐scale bird survey methods. As a test case, we compare results from weather radar distributions and eBird species composition with detailed bird strike records from three major New York airports.
We show that weather radar‐based estimates of migration intensity can accurately predict the probability of bird strikes, with 80% of the variation in bird strikes across the year explained by the average amount of migratory movements captured on weather radar. We also show that eBird‐based estimates of species occurrence can, using species’ body mass and flocking propensity, accurately predict when most damaging strikes occur.
Synthesis and applications . By better understanding when and where different bird species occur, airports across the world can predict seasonal periods of collision risks with greater temporal and spatial resolution; such predictions include potential to predict when the most severe and damaging strikes may occur. Our results highlight the power of federating datasets with bird movement and distribution data for developing better and more taxonomically and ecologically tuned models of likelihood of strikes occurring and severity of strikes. -
Selection of habitat is a key determinant of reproductive success, and the process of finding and choosing these sites is often influenced by the presence of conspecifics. Many bats frequently switch roosts, and some bats repeatedly find new roosts. To find roosts with conspecifics or group members, bats can use social cues. However, most research on how bats use social cues for roost-finding has focused on acoustic cues. Here, we review and discuss the evidence for bat roost selection using scent cues from guano and urine stains, which are present at most bat roosts. We outline reasons why bats might, or might not, use scent in roost detection and selection, and we review evidence on the possible use of guano and urine in roost-finding from eight studies with 12 bat species (across four families). Overall, the sparse evidence that exists indicates that scent cues from guano and urine are not a strong and consistent lure in the species and situations that were tested. Most studies had unclear results or found no effect. Two of the eight studies found weak experimental evidence for bats using guano or urine to select a roosting site. Even if guano and urine can indicate the presence of bats at a roost, it is possible that the resulting olfactory cues do not contain sufficient social information to be used in roost selection, in contrast to olfactory cues from scent marking. Studies of how bats use sensory cues beyond sound could contribute to a better understanding of bat social behavior and roosting ecology.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Journal Article:
- https://doi.org/10.1609/aaai.v34i01.5373
