E&E seminar will be from Pat Richard Lyon from the Kitzes lab will be presenting their research titled:
Title: Changing the Game: How bioacoustics and machine learning can be used to better guide avian population monitoring and habitat management
Friday, February 19, 2025
A219B Langley Hall
12:00 PM
To combat global patterns of avian population declines, conservation agencies and land managers must employ large-scale habitat restoration efforts. Particularly for forested habitats in the eastern United States, silvicultural restoration treatments, such as shelterwood harvesting, can increase the structural diversity of vegetation layers, providing a mosaic of habitat types that many forest birds rely on. To determine if these habitat restoration practices are working to bring back bird populations and stem declines, in-person point count surveys are traditionally used to monitor avian species before, during, and after treatments. However, point count surveys can be incredibly time consuming and require many trained personnel, are logistically infeasible at the large-scales required to mitigate avian declines, and often only identify species presence instead of more detailed demographic information. Therefore, easily deployed methods that can monitor many points across a landscape simultaneously, such as autonomous recording units (ARUs), may be the ideal tool for aiding conservation ecologists in their efforts to monitor the effects of habitat restoration.
In this seminar, I highlight several ways in which advanced bioacoustic techniques using ARUs and machine learning classifiers can uncover patterns of avian occupancy, habitat preference, territory settlement, and successful reproduction. Using paired datasets generated by either point count surveys or ARU data for two imperiled forest bird species, I found that ARU surveys provide more precise estimates of occupancy, or the probability that a species is present at a point given imperfect detection. Additionally, only ARU surveys provided the necessary power to identify two habitat covariates known to be important for nesting habitats for both species. From these findings, I suggest that large-scale forest management projects should strongly consider using ARUs to best monitor avian community response at large scales. In the second part of my seminar, I investigate how long-term acoustic data can be used to examine the timing and distribution of habitat settlement for a species known to respond directly to silvicultural management techniques. Using acoustic data recorded continuously throughout the entire avian breeding season for three consecutive years, I created detection histories at a scale and duration that would be impossible to generate using point count surveys alone. With these detection histories, I identified specific areas and habitats within forest management zones that are preferentially selected, or have earlier arrival, compared to other areas. Additionally, I used these same long-term acoustic data to find signals of successful reproduction, specifically fledgling begging calls, which may indicate areas of managed or unmanaged habitat that are of the highest quality. Updating avian survey methods to include ARUs allows ecologists to go beyond species presence when determining restoration success and can facilitate population monitoring that matches the scale of the habitat restoration needed to reverse avian declines.