# Can you derive group size from a recording of dolphins?

I have some recordings from an array of 5 ST300 stationary recorders, which recorded overnight during calm weather. Our recorders were set up in a square, with 25 m between each corner and one logger in the center, 2 m above the bottom in 10m deep water. We have 3 complete encounters of vocal activity, whistles and clicks, that were recorded on all 5 sound loggers. Our sampling rate was 96 kHz.

Are there methods to determine group size, or at least number of vocalizing animals, based on our set up?

Yes (sort of)! Especially if you are using clicks. You can set this up in PAMGuard with the array structure as you described and it can calculate bearings and bearing crosses for localization. From localizations to number of animals requires a tracking algorithm, P. Gruden has a few out.

EDIT: Here's a link to some of P. Gruden's publications with the Journal of the Acoustical Society of America (JASA).

• Not sure if PamGuard can do that when the ST300s are not synchronized. You would need much more recorders to also estimate the time difference.
– WMXZ
Commented Jul 4, 2022 at 5:12
• Do you have a link available for P. Guden's algorithms? Commented Jul 4, 2022 at 12:45
• @WMXZ ahh, right. Plus they drift a lot. Commented Jul 4, 2022 at 13:39
• Here's a link to P. Gruden's publications in JASA: asa.scitation.org/author/Gruden%2C+Pina Commented Jul 5, 2022 at 10:33
• @fabio_hirono can you add that link into the answer text as an edit? Sometimes things get lost down in the comments. Thanks for adding it :) Commented Jul 7, 2022 at 15:24

If localisation needs to be done: Figuring out the inter-receiver offset is the main challenge. This problem has been dealt with in this paper by Burgess et al. 2013. In the paper the authors study how to acoustically localise unsynchronised receivers and sources using time-difference-of-arrivals.

If localisation isn't a priority: Not a dolphin person here - please account for this :). With bats, one can assume a call rate of ~10 Hz ('normal' orientation echolocation), and thus seeing about 50 calls a second on a single-channel recording, one could arrive at a ballpark of 5 bats. Perhaps a similar calculation could be done with dolphin clicks. Since there are 5 'phones, perhaps the estimate will end up being better?