Title: Scaling up individual animal behavioral monitoring with automated acoustic localization
Abstract:
Passive acoustic monitoring is improving biologists’ ability to collect data on the presence and absence of sound-producing animals. Beyond presence and absence, determining individual animals’ identities is important for many applications in ecology and evolution, such as censusing, studying dispersal, observing individual animals’ behaviors and interactions, and estimating apparent survival. While most sound-producing animals are individually identifiable from the characteristics of their sounds, collecting individually identified sound data even at small scales remains labor-intensive, and very few studies apply these methods to large-scale passive acoustic monitoring data. A new method for automated acoustic localization - estimation of animals’ positions by “triangulating” their sounds - enables large-scale collection of individually identified sounds of territorial animals.
This talk will discuss an application of automated acoustic localization to collect individually identified sound data, as well as to estimate demographic measurements and behavioral traits of several passerine birds. We deployed 49 synchronized passive acoustic recorders in a 240 x 240 m plot of boreal mixedwood forest in Alberta, Canada. We simultaneously collected four types of in-person ground truth data for several species of passerine birds: abundance, territory locations, singing rate, and movement rate. I then applied a recently-developed automated acoustic localization pipeline to estimate the same four measurements and compared these to estimates made by in-person observation. Localization-based estimates of abundance, territory locations, and movement rate match those taken by human observers, whereas song rate is not estimated well by the localization-based technique. Applications of individual identification by localization include estimating trait data for passive abundance estimation techniques and rapidly generating large samples to train algorithms to individually identify animals by their sounds.
Kitzes Lab
Wednesday, November 15th, 2023
12:00PM
Langley A219B