MCA Pilot PUI: Leveraging machine learning to better understand biodiversity patterns measured through passive acoustic sampling

Furman University, Greenville SC

Investigators

Abstract

New technologies are increasing the amount of data available that measures and monitors biological systems, helping to answer long-standing questions in biology and opening new research directions. For example, acoustic data collected across a range of ecosystem types can be used to address questions in animal communication, the role of human noise in soundscapes, and population dynamics. However, these data often require new skills to process and understand. This mid-career advancement project will partner the principal investigator with a leader in the bioacoustics analysis field to leverage new machine learning tools to address these questions. Conversely, the rich acoustic data being processed will also be used to validate and improve the machine learning tools. Broader impacts of the research will include increasing the number of individuals trained to use next-generation analytical tools like machine learning for ecological research, via research-based coursework, undergraduate research experiences, and workshops at professional conferences. The research will be shared via traditional scholarly outputs as well as a unique percussion production for diverse audiences. Ultimately, this project will facilitate alignment of multiple lines of research and scholarship for the principal investigator, shaping a clear trajectory of increasingly impactful research. This project will contribute to advancements in the application of passive acoustic sampling and machine learning to address fundamental biological questions. This research will use a rich acoustic dataset from diverse ecosystems including forests, farmland, and urban environments. The project will advance, with improved analyses and precise classification of sounds and noise, three lines of inquiry: 1) how noise affects bird vocalizations and communications, but with increased consistency in measuring bird songs as compared to past research; 2) quantification of acoustic indices as proxies for biodiversity, and in particular how to define and filter anthrophony; and 3) how occupancy modeling for conservation applications can be improved by passive acoustic sampling. This project will also train the next generation of students and researchers in machine learning techniques for biodiversity research via curriculum advancement and workshops. This project is jointly funded by the Population and Community Ecology Cluster in the Division of Environmental Biology, and the Established Program to Stimulate Competitive Research (EPSCoR) program. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.