Waiting.
This is Node 01's instrument panel: what our listening station hears, second by second. Until the microphone is in the creek, it replays a real research recording of wild bottlenose dolphins through our software, so everything here is marked Simulation.
The big strip is a picture of sound: time runs left to right, pitch runs bottom to top, and brighter means louder. People can only hear the very bottom slice, up to about 20 kHz. Dolphin clicks show up as thin bright stripes far above that. A dashed outline marks a sound the scientists who made the recording labelled; a glowing outline marks one our software noticed on its own. Press Listen to hear the part people can hear.
Answers come from the node, live NOAA and weather-service data, our research on 59 animals, and a plain-English glossary. It matches keywords; it isn't AI, so when it doesn't know, it says so.
Node
No heartbeat has arrived. Start the node process to bring this panel to life.
Last 60 seconds
Quiet, or nothing annotated in this window.
Counts above are the dataset's own annotations replayed as ground truth. Ocean's detections are scored against them on the simulator page.
Recent sounds
All events →Nothing yet.
Known means the dataset authors annotated it. Detected means an Ocean detector found it. Species on known events is the dataset's label (whistles were classified by the authors; clicks are grouped by inter-click interval).
How this page gets its numbers
- 01Waterreplaced by an authentic research recording
- 02Hydrophonereplaced by an authentic research recording
- 03Ocean Noderecorder and computer on the dock: 192,000 samples a second, 384,000 planned
- 04Analysisspectrograms, previews, detectors run on the node
- 05Eventknown, detected, reviewed or confirmed
- 06Webthis page, a second or two behind
What's happening on this page
From a dolphin's click to your speakers, in seven steps.
- 1250,000measurements a second
It starts with a real recording
Eight minutes of wild bottlenose dolphins, recorded by marine scientists off the coast of Spain on an underwater microphone that measures the water's pressure a quarter of a million times a second.
The recording → - 21second at a time
Our node plays it back, live
A small computer (in Google's cloud for now, on the dock later) feeds the recording through our software one second at a time, exactly as if it were coming out of the creek. It's the same code the real hydrophone will use.
From creek to screen → - 36×higher than you can hear
Every second becomes a picture
The node turns each second into a picture of sound, from the lowest rumble up to 125 kHz, about six times higher than human ears go. That's the big moving strip, and why dolphin clicks show up even though you can't hear them.
What the picture shows → - 42detectors on watch
Software listens for animals
One detector hunts for whistles (curving lines), the other for bursts of clicks (bright stripes). Anything they flag appears here as Detected, a machine's guess, never a confirmed sighting.
How we label sounds → - 521,854clicks labelled by hand
The scientists keep score
The researchers labelled their recording sound by sound. Those labels play alongside as Known, so every detection can be checked against the truth, and we publish how often our software is right and how often it isn't.
Detector scores → - 648,000a second, for your ears
You hear the part people can hear
For the Listen button, the node makes a copy at the rate a normal audio player uses, keeping the sounds within human hearing, and streams it to your browser a second or two behind.
Why you can't hear clicks → - 7NextBohicket Creek
Then the creek takes over
When our hydrophone goes in the water at the dock, only the first step changes: the same pipeline will be listening to Bohicket Creek, live, and this page will stop saying Simulation.
The hardware plan →
Go deeper
- Every sound →Every sound the node has logged, known and detected, to filter and page back through
- How it works →From a microphone in the creek to this page, in plain English
- Detector scores →How often our software is right, and how often it isn't
- The station →How the recorder is doing, on the app's Station screen
- What's down there →The animals a microphone in the water could pick up, coast by coast
- Simulator console →The controls we run the simulation with. Anyone can watch; only we can change it.