A stethoscope for the ocean: Unknownness-aware monitoring under false-positives-per-hour constraints in underwater soundscapes

Earth’s oceans cover 70% of the planet yet remain among the least observed regions. Underwater acoustics is a primary sensing modality, but recordings span biological, anthropogenic and natural sounds while labels remain sparse. We present an unknownness-aware monitoring framework for underwater acoustic sensing platforms (buoys, tags, mobile robots and cabled observatories): it adapts a pre-trained audio representation model to underwater soundscapes, defines an embedding-space unknownness score, and couples it with detector outputs in a Detect–Group–Promote–Union (DGPU) loop designed to surface candidate unknown events for expert review and to support operating-point design under false positives per hour (FP/h) constraints. Held-out test annotations are used only for quantitative evaluation of candidate surfacing and FP/h–recall trade-offs. Across three cetacean datasets, the framework achieves strong unknownness scoring performance (AUROC ≥ 0.99) on an out-of-distribution benchmark, improves event recall at near-constant FP/h on right-whale recordings, and yields species-wise operating points on survey data with Precision ≥ 0.9 and FP/h ≤ 0.5, achieving high recall for key taxa, supporting stethoscope-like ocean monitoring.