Upsweep: The Sound the Pacific Was Already Making When We Started Listening
The sound was already there before anyone was listening for it.
In August 1991 the Pacific Marine Environmental Laboratory, a research arm of the United States National Oceanic and Atmospheric Administration based in Newport, Oregon, began recording from SOSUS. SOSUS, the Sound Surveillance System, was a Cold War asset: a network of hydrophones laid on the seafloor by the US Navy to track Soviet submarines by the noise of their machinery. As the Cold War wound down, parts of the network were opened to civilian science, and a small group of oceanographers suddenly had something no research budget could have bought, a continuous ear on an entire ocean basin.
The first unexplained thing PMEL found in the Pacific data was not a submarine and not a whale. It was a repeating pattern nobody could name. They called it Upsweep, and more than three decades later that is still all it has: a name.
Upsweep is not a single noise. NOAA describes it as a long train of narrow-band upsweeping sounds, each lasting several seconds. Within every sweep the frequency climbs from low to high, and then the next sweep begins, and the next, in a sequence that can run on and on. People who have listened to the sped-up recordings tend to reach for the same comparisons, a distant siren or the rise and fall of an ambulance, though at its true speed and pitch it is slower and heavier than any of that. What matters technically is that the sweeps are narrow-band, meaning the acoustic energy is concentrated in a tight frequency range rather than smeared across the spectrum. Natural broadband events, such as ice breaking, do not usually sound like that.
The other technical fact is the loudness. Upsweep's source level is high enough that it can be recorded across the entire Pacific. That is a very large amount of acoustic energy for something no one has ever seen.
Locating a sound in the deep ocean is a matter of arithmetic. Sound in the sea travels through a natural waveguide, the deep sound channel, sometimes called the SOFAR channel, a layer at intermediate depth where temperature and pressure combine to give sound its minimum speed. Low-frequency energy that enters the channel is refracted back toward the middle instead of escaping upward or downward, so it can travel for thousands of kilometres with very little loss. If a signal arrives at three or more widely separated hydrophones, the differences in arrival time can be turned into a position. With only one instrument, there is no position at all.
Upsweep arrived on enough instruments to be placed. NOAA gives its source as roughly 54 degrees South, 140 degrees West. That is one of the emptiest places on the planet: open South Pacific, far south of the shipping lanes, well east of New Zealand and roughly 2,500 miles west of the southern tip of South America. There is no island, no research station and no permanent human presence anywhere near it.
There is, however, something on the maps. NOAA notes that the position falls near the location of inferred volcanic seismicity. That word inferred is doing real work. It does not mean anyone has photographed an erupting vent at that spot. It means the seismic record for the region is consistent with volcanic activity somewhere in the area. The laboratory's own summary of the case ends with a sentence it has not revised: the origin of the sound is unresolved.
Then there is the calendar. Upsweep is seasonal. Its amplitude rises and falls through the year, generally peaking in spring and autumn. NOAA is explicit that it does not know why. The peaks could mean the source itself is more active at those times, or they could mean the source is steady and the ocean between the source and the hydrophones changes with the seasons. The deep sound channel is not a fixed pipe. Its depth and its efficiency shift with water temperature, with the position of ocean fronts, and in the far south with the advance and retreat of sea ice. A signal that never changes can still get louder and quieter at a listening post thousands of kilometres away.
And Upsweep is fading. The overall source level has been declining since 1991. It has not stopped. The sounds are still detected on NOAA's equatorial Pacific autonomous hydrophone arrays, the purpose-built civilian instruments the laboratory deployed after it stopped depending on Navy hardware. Whatever is making the noise has been getting quieter for more than thirty years without switching off.
Upsweep did not stay alone for long. Through the late 1990s the same arrays picked up a small catalogue of strange recordings that would later become famous on the internet. Bloop, detected in 1997 across a range of more than 5,000 kilometres, was briefly discussed as possibly biological before NOAA attributed it to cryoseisms, the cracking and calving of large ice masses, a conclusion the agency had settled on by 2012. Slow Down, recorded on 19 May 1997, is a roughly seven-minute descent in frequency and is attributed to an iceberg running aground off the Antarctic Peninsula. Train, from the same year, is a steady hum thought to come from an iceberg dragging its keel across the seafloor in the Ross Sea. Julia, recorded in 1999, was traced to a large grounded iceberg off Antarctica. Whistle, captured on 7 July 1997 at 8 degrees North, 110 degrees West, resembles the signals produced by erupting submarine arc volcanoes, but it registered on only one hydrophone, so it could never be located at all.
The pattern in that list is easy to miss. Almost every celebrated NOAA mystery sound has been solved, and most of them turned out to be ice. Upsweep is the one that has not been solved, and it is also the only one that has been running continuously since before the recordings began.
No research vessel has ever gone to 54 South, 140 West to look. Ship time in the Southern Ocean is expensive and scarce, the weather at that latitude is among the worst in the world, and a sound that has been politely repeating itself for decades has never generated enough urgency to win a berth on a schedule.
Conclusions and Open Questions
The volcanic interpretation is the leading one, and it comes from NOAA's own acousticians. Submarine volcanism produces exactly the family of signals in question. When seawater reaches molten rock it flashes to steam in repeating pulses, and magma moving through a conduit produces harmonic tremor, a narrow-band tone that can climb or fall in frequency as pressure changes. The laboratory that made the recordings has spent decades using hydrophones to detect eruptions on the Juan de Fuca Ridge and elsewhere, so it knows what volcanic acoustics look like. The weakness is that no volcano has been confirmed at the site. There is no observed eruption, no mapped edifice, no plume, no fresh lava sample. A single volcanic source holding essentially the same acoustic signature for more than thirty years would also be unusual.
The propagation interpretation is raised by NOAA itself, and it applies to the seasonality rather than to the sound. It proposes that the source may be roughly constant while the ocean's ability to carry the signal varies through the year. Its weakness is that it is not an explanation of Upsweep at all. It explains the calendar and leaves the source exactly as unidentified as before.
The biological interpretation belongs to the early years. Repeating call trains are what large baleen whales produce, and fin whale sequences in particular were considered. It was set aside because the repetition is too mechanically regular and the source level too high for an animal. This one is effectively closed rather than open.
The ice interpretation is offered by analogy, on the reasonable grounds that Bloop, Julia, Slow Down and Train all resolved to ice. Its weakness is twofold. The location at 54 South is well north of the zone where large icebergs ground, and cryogenic signals are broadband, irregular and event-like, while Upsweep is narrow-band and relentlessly repetitive.
Then there is an open theory worth naming as open: that the source is a seafloor feature that simply is not on any chart. Most of the deep ocean at that latitude has never been surveyed by ship. What exists is satellite altimetry, which infers the shape of the seafloor from tiny bumps in the sea surface and resolves features only at kilometre scale. An active vent field or a small volcanic cone could sit at 54 South and be invisible to every map currently in use. This is a hypothesis about a gap in coverage, not a claim about an object, and it should be read that way.
What remains genuinely unexplained is the decline. Nobody has published a persuasive account of why a source would grow steadily quieter for more than three decades and not stop. A cooling volcanic system would do that. So, in principle, would a slow change in the ocean's sound channel, or a change in the sensitivity and placement of the instruments listening. Those possibilities have not been separated.
The open questions are unusually tractable for a mystery of this age. A single season of ocean-bottom seismometers deployed at 54 South, 140 West would almost certainly distinguish a volcanic source from anything else, because volcanic tremor has a seismic signature and a drifting iceberg does not. A multibeam bathymetric survey of a few hundred square kilometres would show whether there is an unmapped edifice there. Correlating the seasonal amplitude curve against measured sound-channel conditions and Southern Ocean ice extent would test the propagation idea directly, and that work could be done with archived data and no ship at all. Finally there is the question nobody likes: Upsweep has been fading since the first day it was recorded, and if it goes silent before anyone goes to look, the case closes with no answer in it.