I have learned about an interesting mathematical model for time series and I am curious to know if it can find an application in bioacoustics.
The model is named "Hawkes process" and is a generalization of Poisson process. The key idea is that each acoustic event would be represented by a "point" on the time line and that the expected call rate could vary in time as a function of past events. https://en.wikipedia.org/wiki/Hawkes_process
It seems to me that the feedback loop between past events and current intensity could serve as an appropriate (if very simplistic) model for certain vocal interactions in social species. Intuitively, soon after one individual calls, conspecifics join in, triggering a chain reaction of calls. Dog barks and avian alarm calls come to mind.
I am aware of the work by Stowell et al on GLM point processes:
Stowell, D., Gill, L., & Clayton, D. (2016). Detailed temporal structure of communication networks in groups of songbirds. Journal of the Royal Society Interface, 13(119), 20160296.
and work on vocal turn-taking
Demartsev, V., Strandburg-Peshkin, A., Ruffner, M., & Manser, M. (2018). Vocal turn-taking in meerkat group calling sessions. Current Biology, 28(22), 3661-3666.
My research angle is slightly different since I am less interested in discovering the interaction network between individuals. Meanwhile, I would like to focus the study on measuring their cognitive reaction time (via the Hawkes kernel). In the context of PAM, I would also like to show the limitations of measuring vocal activity as a stationary quantity and the benefits of integrating feedback loops.
Thus, I am looking for open bioacoustic datasets, either single-channel or multi-channel, which could allow me to test the goodness of fit of Hawkes process models. The vocal events must be sparse enough to be represented by infinitesimal points on the timeline: multi-syllable songs would not fit that assumption. Calls must be clearly attributable to individuals. I would prefer to work with natural habitats but captive animals are potentially OK for me. The dataset does not need to be very large: a few hundred events would suffice. That being said, I would prefer larger datasets, of course.
Modeling interactions between different species (e.g., predator–prey or anthropophony–biophony) would make the study even more original. Perhaps along the lines of the "landscape of fear" in ecology.
Miller, P. J., Isojunno, S., Siegal, E., Lam, F. P. A., Kvadsheim, P. H., & Curé, C. (2022). Behavioral responses to predatory sounds predict sensitivity of cetaceans to anthropogenic noise within a soundscape of fear. Proceedings of the National Academy of Sciences, 119(13), e2114932119.
This idea came up while brainstorming towards the workshop on Vocal Interactivity in-and-between Humans, Animals, and Robots (VIHAR). https://vihar-2024.vihar.org/ If you have data which suit my needs, are eager to test this "mutual excitation" hypothesis with a mathematical model, and have some time in the next six weeks, may I suggest we work together! But also, I feel like your answer could serve the bioac.SE community at large
