NABat ML: Utilizing deep learning to enable crowdsourced development of automated, scalable solutions for documenting North American bat populations

Ali Khalighifar; Benjamin S. Gotthold; Erin Adams; Jenny K. Barnett; Laura O. Beard; Eric R. Britzke; Paul A. Burger; Kimberly Chase; Zackary Cordes; Paul M. Cryan; Emily Ferrall; Christopher T. Fill; Scott E. Gibson; G. Scott Haulton; Kathryn Irvine; Lara S. Katz; William L. Kendall; Christen A. Long; Oisin Mac Aodha; Tessa McBurney; Sarah McCarthy-Neumann; Matthew W. McKown; Joy O’Keefe; Lucy D. Patterson; Kristopher A. Pitcher; Matthew Rustand; Jordi L. Segers; Kyle Seppanen; Jeremy L. Siemers; Christian Stratton; Bethany R. Straw; Theodore J. Weller; Brian E. Reichert

doi:10.1111/1365-2664.14280

Home/Archive/USGS

USGS · 70238032

NABat ML: Utilizing deep learning to enable crowdsourced development of automated, scalable solutions for documenting North American bat populations

View original source ↗Export metadata ↓

Abstract

Bats play crucial ecological roles and provide valuable ecosystem services, yet many populations face serious threats from various ecological disturbances. The North American Bat Monitoring Program (NABat) aims to use its technology infrastructure to assess status and trends of bat populations, while developing innovative and community-driven conservation solutions. Here, we present NABat ML , an automated machine-learning algorithm that improves the scalability and scientific transparency of NABat acoustic monitoring. This model combines signal processing techniques and convolutional neural networks (CNNs) to detect and classify recorded bat echolocation calls. We developed our CNN model with internet-based computing resources (‘cloud environment’), and trained it on >600,000 spectrogram images. We also incorporated species range maps to improve the robustness and accuracy of the model for future ‘unseen’ data. We evaluated model performance using a comprehensive, independent, holdout dataset. NABat ML successfully distinguished 31 classes (30 species and a noise class) with overall weighted-average accuracy and precision rates of 92%, and ≥90% classification accuracy for 19 of the bat species. Using a single cloud-environment computing instance, the entire model training process took <16 h. Synthesis and applications . Our convolutional neural network (CNN)-based model, NABat ML , classifies 30 North American bat species using their recorded echolocation calls with an overall accuracy of 92%. In addition to providing highly accurate species-level classification, NABat ML and its outputs are compatible with Bayesian and other statistical techniques for measuring uncertainty in classification. Our model is open-source and reproducible, enabling future implementations as software on end-user devices and cloud-based web applications. These qualities make NABat ML highly suitable for applications ranging from grassroots community science initiatives to big-data methods developed and implemented by researchers and professional practitioners. We believe the transparency and accessibility of NABat ML will encourage broad-scale participation in bat monitoring, and enable development of innovative solutions needed to conserve North American bat species.

Explore related subjects

Journal of Applied Ecology Article

Source-reported bounding extent: 25.29139626361966° to 60.48267077134361° latitude; -140.76638992797305° to -52.17263992797348° longitude. This indicates report coverage, not an exact sampling location. View area on OpenStreetMap.

Keep this discovery

Explore connections, maps & timelines

BibTeX RIS

Ali Khalighifar, Benjamin S. Gotthold, Erin Adams, Jenny K. Barnett, Laura O. Beard, Eric R. Britzke, Paul A. Burger, Kimberly Chase, Zackary Cordes, Paul M. Cryan, Emily Ferrall, Christopher T. Fill, Scott E. Gibson, G. Scott Haulton, Kathryn Irvine, Lara S. Katz, William L. Kendall, Christen A. Long, Oisin Mac Aodha, Tessa McBurney, Sarah McCarthy-Neumann, Matthew W. McKown, Joy O’Keefe, Lucy D. Patterson, Kristopher A. Pitcher, Matthew Rustand, Jordi L. Segers, Kyle Seppanen, Jeremy L. Siemers, Christian Stratton, Bethany R. Straw, Theodore J. Weller, Brian E. Reichert. 2022-09-20. NABat ML: Utilizing deep learning to enable crowdsourced development of automated, scalable solutions for documenting North American bat populations. https://doi.org/10.1111/1365-2664.14280

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related USGS reports

USGS2026

Solar energy development selectively filters avian taxonomic and functional diversity across a perennial agricultural landscape

Solar energy is rapidly expanding across agricultural landscapes, yet its consequences for avian communities remain poorly understood, particularly among heterogeneous land-use matrices. In 2024 and 2025, we surveyed birds across 107 sites in New York State, including conventional solar arrays with mowing or sheep grazing to manage vegetation, solar-adjacent working lands (e.g. hayfields, pastures) and reference working lands independently located from solar arrays. We integrated diversity metrics across four trait dimensions—taxonomic, morphological, ecological and conservation status—and examined interactions between species-level abundance and traits to understand avian community responses to solar energy development. Avian species in solar sites tended to be common, non-migratory (in behaviour and wing morphology) and associated with shrubland or forest. Compared to reference working lands, bird communities in solar and solar-adjacent sites were less taxonomically diverse and were composed of fewer declining and range-restricted species. However, solar and solar-adjacent sites retained comparable diversity of intrinsic traits, suggesting that ecological function may persist despite apparent species filtering and functional reshuffling. While solar energy development can support avian communities with unique functional signatures, our findings suggest that conventional solar arrays sited on agricultural working lands in New York maintain low bird diversity and are unlikely to support conservation-priority species—such as grassland birds characterized by declining populations, migratory tendencies and open-habitat dependence. Synthesis and applications . Our study provides a pathway to leverage functional diversity to quantitatively characterize solar-associated avian communities and the ecological mechanisms that shape them. Our results suggest that meeting biodiversity conservation goals while achieving solar energy targets requires ecologically informed siting and processes, particularly across regions in which co-prioritization of conservation, agriculture and energy presents sustainability trade-offs.

New York↗

USGS2026

Predictability creates vulnerability: Long-term acoustic telemetry reveals hidden threat windows for endangered Atlantic sturgeon

1.Individual movement strategies govern access to important habitats but can expose certain animals to spatially and temporally variable threats, producing consistent, phenotype-specific differences in risk. For endangered Atlantic sturgeon ( Acipenser oxyrinchu s), migration timing shapes exposure risk to vessel strikes, dredging activity, and shifting estuarine conditions, yet management to minimize exposure to these threats is predominantly based on population-average patterns of river occupancy. 2. We analysed a decade of acoustic telemetry data from 195 Atlantic sturgeon in the Hudson River to quantify individual variation in migration timing. We clustered individuals into distinct behavioural phenotypes and evaluated the consistency of these phenotypes across years to determine whether certain individuals may be present in spatiotemporal areas that are not represented in current conservation frameworks. 3. Results identified two persistent phenotypes: the short-residency phenotype exhibiting synchronized occupancy typical of the prescribed Atlantic sturgeon spawning run and an extended-residency group that migrates into the river later, displays prolonged river occupancy and diffuse outmigration into the fall. Extended-residency individuals were last detected approximately 88 days after short-residency individuals and occupied habitats downriver of the presumed spawning grounds, extending the period of sturgeon presence into riverine habitats outside the current spatiotemporal area of conservation focus. 4. Individual phenotypic membership was repeatable across nearly a decade, indicating that these behaviours reflect stable, consistent phenotypes rather than stochastic variation. 5. Synthesis and applications . Our results reveal persistent intrapopulation movement diversity in the largest remaining Atlantic sturgeon population and highlight predictable periods during which a unique phenotype may experience disproportionate exposure to in-river threats. These findings indicate that environmental reviews, permitting decisions and seasonal work windows may benefit from incorporating phenotype-specific occupancy patterns and percentile-based estimates of river residency rather than relying solely on historical spawning dates or population averages. Expanding telemetry and monitoring designs to include broader spatial and seasonal sampling may further improve detection of overlooked movement strategies and reduce potential biases in demographic and habitat use assessments. More broadly, recognizing behavioural diversity in recovery planning may provide a more complete basis for evaluating habitat use, threat exposure, and conservation priorities across sturgeon populations.

New York↗

USGS2026

Testing the efficacy of industrial mitigation measures for caribou in the Arctic

1. Mitigation measures are commonly employed to reduce the negative effects of industrial development on wildlife but are not often evaluated for their efficacy. For example, oil fields in the Arctic typically incorporate design features intended to increase permeability for migratory, barren-ground caribou ( Rangifer tarandus) , even though there is limited empirical evidence of the effectiveness of some of these features. 2. Given expected increases in energy development in the North American Arctic, we examined whether two mitigation measures commonly used for migratory caribou, elevating pipelines and separating roads and pipelines, were effective at increasing the probability caribou would cross infrastructure. 3. We conducted our investigation on adult female caribou in the Central Arctic Herd of Alaska during summer, analyzing movement data from telemetry collars (2015-2020) in conjunction with spatial data on oil field infrastructure. To evaluate whether caribou would cross infrastructure as a function of the mitigation measures, we employed a generalized additive modeling framework capable of detecting non-linear and threshold responses. 4. We found that caribou were more likely to cross a pipeline when the nearest pipeline was elevated (≥1.9-m), a result that supports current mitigation recommendations. We also found that caribou were more likely to cross both a road and pipeline when they were directly adjacent to one another, as opposed to being spatially separated, a result that contradicts recommended mitigation strategies. 5. Synthesis and applications . As new energy projects are designed and implemented in environments around the globe, it is important to ensure that mitigation efforts for wildlife are scientifically validated for their efficacy. In the North American Arctic, such efforts will be critical for minimizing the impacts of expanding industrial development on migratory caribou and on the human communities that rely on them for subsistence.

Alaska↗