Date of Graduation
Fall 2025
Degree
Master of Science in Biology
Department
Biology
Committee Chair
Alexander Wait
Abstract
In the U.S., following colonization and agricultural expansion, giant cane (Arundinaria gigantea) has been reduced to approximately 2% of its original extent, potentially causing shifts in the abundance and distribution of over 70 associated animal species. Among those species is the Swainson’s Warbler (Limnothlypis swainsonii), which has been known to occupy canebrakes during the breeding season. Though there is little known about the historical range of the Swainson’s Warbler and its current population status, some have hypothesized that its population may have declined due to the reduction of its suitable breeding habitat. As this species is rare and cryptic, creative forms of monitoring are needed. Monitoring Swainson’s Warblers is particularly important in the western Ozarks, where sightings are rare, and the extent of its range is poorly known. Autonomous recording units (ARUs) offer a highly efficient way to monitor rare, vocal species. The recent application of convolutional neural networks (CNNs) and their improved accessibility with programs like BirdNET has made passive acoustic monitoring (PAM) via ARUs more effective for monitoring. In this study, I applied PAM to investigate the limits of the breeding range of the Swainson’s Warbler, and this species’ potential use of existing canebrakes in the western Ozarks. I assessed the performance of the core BirdNET model and compared it to the performance of a custom classifier trained with segments from local ARU recordings. Though extralimital sightings occur in the region, the secretive species was not detected at any canebrake site outside of the published breeding range during the study. The study did, however, generate insight into the application of PAM to detecting Swainson’s Warblers, including understanding which sounds present acoustic identification challenges for CNNs. Along these lines, I report analyses comparing the songs of the Louisiana Waterthrush and the Swainson’s Warbler, as the striking similarities between their songs pose great difficulty for PAM of the target species.
Keywords
passive acoustic monitoring (PAM), autonomous recording units (ARUs), BirdNET, Swainson’s Warbler, bioacoustics, transfer learning, convolutional neural networks (CNNs)
Subject Categories
Ecology and Evolutionary Biology
Copyright
© Gabriela Carroll
Recommended Citation
Carroll, Gabriela, "Giant Cane (Arundinaria Gigantea) and the Swainson’s Warbler (Limnothlypis Swainsonii): Implementing Innovative Monitoring Techniques for Species of Conservation Concern" (2025). Graduate Theses/Dissertations. 4115.
https://bearworks.missouristate.edu/theses/4115
